Human epithelial tooth organoids model disrupted amelogenesis following environmental and pharmacological exposures

Abstract

Background

Dental enamel formation is a tightly regulated developmental process that is particularly vulnerable to disruption during early life, leading to permanent enamel defects such as molar incisor hypomineralization. While epidemiological studies have linked certain environmental factors and medical treatments to developmental defects of enamel, direct causal evidence in humans remains limited due to the lack of suitable in vitro models.

Methods

Here, we use human epithelial tooth organoids to examine the effects of clinically relevant environmental and pharmacological exposures, including amoxicillin, bisphenol A, vinblastine, and vincristine on ameloblast differentiation and maturation. Organoids were exposed during defined stages of differentiation, followed by gene expression analysis, viability assays, transcriptomic profiling, and ultrastructural characterization.

Results

Exposure to amoxicillin, vinblastine, and vincristine impaired ameloblast maturation, as evidenced by reduced expression of maturation-stage enamel matrix proteins, including AMTN and ODAM, without markedly affecting overall cell viability or organoid morphology. Transcriptomic analysis of amoxicillin-treated organoids revealed selective downregulation of metabolic and secretory pathways, including oxidative phosphorylation, glycolysis, and unfolded protein response. Furthermore, epithelial tooth organoids recapitulated known mechanisms of action of chemotherapeutic agents, such as α-tubulin depolymerization. These findings indicate that the external exposures investigated can interfere with ameloblast maturation programs in a human-derived in vitro system.

Conclusions

Our study establishes human epithelial tooth organoids as a robust platform to model exposure-induced disruptions in amelogenesis and provides direct experimental evidence that commonly encountered environmental and pharmacological agents can impair enamel development. This model offers new opportunities to investigate the mechanisms underlying developmental defects of enamel in a human-relevant context.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s13287-026-05206-y.

Background

Dental enamel is the most mineralized tissue in the human body and serves as the outermost protective layer of the teeth [1, 2]. Amelogenesis, the process of enamel formation, is a complex, multi-step mechanism that begins in utero for the primary teeth and continues through early childhood for most permanent teeth [3, 4]. This process is divided into four well-defined stages: the pre-secretory stage, during which pre-secretory ameloblasts differentiate into secretory-stage ameloblasts; the secretory stage, where enamel matrix proteins (EMPs) such as ameloblastin (AMBN), enamelin (ENAM), and amelogenin (AMELX) are secreted by the secretory ameloblasts; the transition stage, marked by controlled cell death and structural and functional changes in ameloblasts; and finally, the maturation stage, during which EMP expression predominantly shifts to odontogenic ameloblast-associated protein (ODAM) and amelotin (AMTN), and where the enamel hardens through mineralization and the removal of organic components [1, 3, 5, 6]. Each stage is tightly regulated, and disturbances, mainly during the secretory and maturation stages, can lead to developmental defects of enamel (DDEs). Amelogenesis of permanent dentition occurs during specific windows of early development, from the prenatal period through early childhood. Exposure to disruptive factors during this time can result in permanent structural abnormalities of the enamel [7–9].

Several DDEs have been identified to date, encompassing a broad spectrum of structural and mineral abnormalities that arise when ameloblast function or enamel maturation is disrupted. These include conditions such as amelogenesis imperfecta, a group of inherited disorders affecting enamel structure and mineralization; dental fluorosis resulting from excess fluoride exposure during development; and hypomineralization conditions, such as molar-incisor hypomineralization (MIH) [9–11]. Affecting approximately 13–14% of children worldwide, MIH is one of the most prevalent DDEs, affecting the enamel of the first permanent molars, often in combination with the incisors [12–15]. Clinically, MIH is characterized by abnormal enamel translucency, often presenting as well-defined brown or yellow opacities, with affected molars displaying a soft, porous structure [16–18]. MIH results from disrupted enamel mineralization during the maturation phase of amelogenesis, although its precise etiology remains multifactorial and incompletely understood, with genetic, prenatal, and postnatal environmental contributors all implicated [19–21]. Early childhood illnesses, fever, and the use of amoxicillin (AMX) in the first year of life have been positively associated with MIH, though it remains difficult to determine whether the illness itself, the resulting fever, the subsequent use of antibiotics, or a combination of these factors drives this association [12, 21–23]. Additionally, exposure to bisphenol A (BPA), a known environmental pollutant and endocrine disruptor, has been associated with enamel hypomineralization [8, 24–26]. However, direct causal evidence remains limited, as most findings are based on epidemiological data, underscoring the need for human cell-based models to provide empirical evidence of causality.

Beyond environmental exposures and commonly prescribed medications, clinical treatments may also interfere with enamel development. Children treated with vinca alkaloid agents, such as vinblastine (VB) and vincristine (VC), frequently develop dental defects later in life, including enamel hypoplasia, characterized by localized enamel thinning and surface irregularities [27]. Representative clinical examples are shown in Supplementary Fig. 1, illustrating two distinct forms of developmental defects of enamel: amelogenesis imperfecta (Suppl. Fig. S1a) and chemotherapy-associated enamel hypoplasia (Suppl. Fig. S1b). As these treatments for pediatric cancers often coincide with critical windows of tooth development, understanding their potential impact on amelogenesis is crucial, particularly for childhood cancer survivors at risk of long-term oral health issues [27–31]. While there is abundant clinical evidence for dental anomalies in childhood cancer survivors, the mechanisms by which chemotherapeutic agents interfere with ameloblast function and enamel maturation remain incompletely understood [32, 33].

One of the primary challenges in the study of amelogenesis and enamel-related pathologies is the transient nature of ameloblasts. At the end of their lifecycle, ameloblasts either become part of the reduced enamel epithelium, which covers the enamel prior to tooth eruption, or undergo apoptosis. Consequently, enamel lacks the capacity for repair or regeneration, and stable (human) ameloblast cell models are exceedingly difficult to establish [5, 34]. Our research group recently succeeded in generating the first epithelial tooth organoids (ETO) using epithelial cell rests of Malassez (ERM) isolated from the dental follicle (DF) of extracted human third molars [35, 36]. While the ERMs were originally dismissed as cellular debris, they were found to be quiescent epithelial stem cells originating from the Hertwig epithelial root sheath (HERS), the apical extent of the internal and external enamel epithelia that plays an important role during tooth development. In line with the previous findings that ERM-derived cells differentiate into ameloblast-like cells when co-cultured with dental pulp stem cells, our ETO also recapitulate this differentiation process [37]. Upon stimulation with ‘transforming growth factor-β’ (TGF-β)-rich medium, they expressed high levels of mature ameloblast markers AMTN and ODAM [35]. Organoids derived from organs such as the brain, intestine, and lung have demonstrated their utility as reliable models for studying tissue-specific etiology and treatments [38]. Similarly, we aim to use the ETO as an in vitro model of the ameloblast. Consequently, this study is the first of its kind to investigate the effects of AMX, BPA, VB, and VC on amelogenesis, intending to gain mechanistic insights into how exposure to these compounds may disrupt enamel formation. Briefly, using our established protocol, ETO were differentiated toward an ameloblast-like phenotype in the presence of these agents, and the effects on cell viability, enamel markers, and intracellular processes were investigated. In addition, ETO treated with AMX were analyzed by RNA sequencing (RNA-seq). Transmission electron microscopy (TEM) was applied to samples exposed to AMX, VB, and VC. Taken together, our findings show that AMX, VB, and VC can influence ameloblast formation, thereby demonstrating that ETO can serve as a human-derived platform for studying the etiology of, and perhaps developing novel treatments for, enamel disorders.

Methods

Dental follicle isolation and dissociation

The dental follicle was isolated from unerupted third molars extracted at Ziekenhuis Oost-Limburg (Genk & Lanaken) from patients aged 13–19 years, reflecting the availability of unerupted third molars at routine extraction appointments (Patient list; Suppl. Table 1). This study was approved by the Medical Ethics Committee of Hasselt University (13/0104U). Dental follicles were removed from the teeth and rinsed in 70% ethanol, phosphate-buffered saline (PBS), and collection medium (Suppl. Table 2). The tissue was finely cut (~ 1 mm²) and further dissociated in dissociation medium (Suppl. Table 3) for 2 h at 37 °C. After incubation, medium A (Suppl. Table 4) was added, and the suspension was filtered through a 40 μm sterile cell strainer into single cells.

Establishment and passaging of organoid culture

The cells were resuspended in a mixture of serum-free defined medium (SFDM; Suppl. Table 5) and Matrigel (Growth Factor Reduced; Corning) at a ratio of 30:70, and 20,000 cells were seeded per 20 µL drop. After solidification of the Matrigel, tooth organoid medium (TOM, Suppl. Table 6) was added for routine culture and passaging. ROCK inhibitor (RI; 10 µM; Thermo Fisher Scientific) was added on the day of plating. Additionally, throughout the first passage, Amphotericin B was supplemented. Tooth organoids were incubated at 37 °C and 1.9% CO₂ with medium changes every 2–3 days. Organoid cultures were passaged every 7–10 days. Matrigel droplets were harvested using ice-cold SFDM, and organoids were dissociated with TrypLE (containing 5 µM RI; Gibco) combined with mechanical trituration. The remaining large organoid fragments were filtered out using a 40 µM cell strainer. Cells were passaged or seeded for experiments as described above.

In vitro differentiation of epithelial tooth organoids

After passaging, organoids were cultured for seven days in TOM to stimulate proliferation. On day seven, the medium was switched to mineralization-inducing medium (MIM; Suppl. Table 7) to stimulate ameloblast-like differentiation. An overview of the four external factors investigated, including the tested concentrations, analysis time points, experimental readouts, and control groups, is provided in Table 1. Three control conditions were included throughout all differentiation experiments: ETO cultured in TOM served as an undifferentiated baseline control; ETO cultured in MIM without any added compound served as the differentiation control (MIM control); and for BPA experiments, methanol (0.1% v/v) was additionally included as a vehicle control, as BPA was dissolved in methanol. Samples were analyzed after four and seven days of differentiation, as a previous time-lapse gene expression analysis showed that secretory-stage markers peaked on day five [35].

Table 1.

Tested compounds during ameloblast differentiation

Compound	Concentration	Supplier	Analyses	Controls
Amoxicillin (AMX)	0.1–4 mg/mL	Sigma-Aldrich (A8523)	Viability, qPCR, RNAseq, WB, TEM	TOM, MIM
Bisphenol A (BPA)	0.4–30 µg/mL	Merck (239658)	Viability, qPCR	TOM, MIM, MeOH (0.1%v/v)
Vinblastine (VB)	25–500 nM	Thermo Fisher Scientific (203450050)	Viability, qPCR, WB, TEM	TOM, MIM
Vincristine (VC)	25–500 nM	Thermo Fisher Scientific (203440050)	Viability, qPCR, WB, TEM	TOM, MIM

Cell viability assay

After four or seven days of differentiation, 10% AlamarBlue Reagent (Bio-Rad) was added to the cells. Following 24 h of incubation in the dark, fluorescence was measured by CLARIOstar PLUS (BMG LABTECH). Data is represented as relative viability, normalized against MIM. Conditions showing cell viability below 50% were excluded from further analysis.

Gene expression analysis by qRT-PCR

RNA was extracted using the Triazol/Qiazol reagent method, after which total RNA concentration and purity were assessed using a NanoDrop 2000 Spectrophotometer (Thermo Fisher Scientific). RNA was reverse transcribed with qScript cDNA SuperMix (Quantabio), and the resultant cDNA samples were analyzed with SYBR-green-based quantitative reverse transcriptase PCR (qRT-PCR) (ThermoFisher). Forward and reverse primers used for qRT-PCR amplification are listed in Supplementary Table 8. Cyclin A (CycA) was included as a housekeeping gene. Relative gene expression levels were calculated with the 2^(−ΔΔCt) method (Ct target—Ct housekeeping gene).

RNA sequencing analysis

Total RNA was isolated from ETO and ameloblast-like cells with or without AMX treatment (4 mg/mL) as described earlier (n = 4 biological replicates per condition). RNA library preparation (eukaryotic RNA-seq) and paired-end sequencing (150 bp) were performed by Novogene using a NovaSeq X Plus system (Illumina). Sequencing quality was assessed using FastQC (v0.11.9). Transcript quantification was performed using Salmon (v1.4.0) against the human reference genome (GRCh38, release 45) with GC and sequence bias correction enabled. Gene-level count matrices were generated using tximport (v1.18.0) and used for differential gene expression analysis with DESeq2 (v1.30.1). Principal component analysis was performed following the rlog transformation for quality assessment. Gene set enrichment analysis was performed using GSEA (v4.3.2) with Hallmark gene sets from MSigDB (v2025.1) [39, 40]. Log fold change estimates were shrunk using the apeglm method for visualization and gene ranking [41]. Full RNA-seq datasets are provided as additional files (Supplementary Table 10, Additional Files 1–4).

Western blot analysis

Organoids were lysed in RIPA buffer supplemented with protease inhibitor (Roche). Lysates were cleared by centrifugation at 14,000 x g for 15 min at 4 °C. Protein concentrations were determined using a Bicinchoninic Acid (BCA) assay (Thermo Fisher Scientific). Equal amounts of protein were separated by SDS-PAGE and transferred onto a polyvinylidene fluoride (PVDF) membrane. Membranes were blocked for 1 h in 5% BSA (Bovine Serum Albumin) in PBS 0.1% Tween and incubated overnight with primary antibody (Suppl. Table 9) diluted in blocking buffer, followed by HRP-labeled (Horseradish Peroxidase) conjugated secondary antibody (Suppl. Table 9) incubation for 1 h at RT. Proteins were detected using an Enhanced Chemiluminescence (ECL) substrate. Band intensities were quantified using ImageJ and normalized against housekeeping protein beta-actin.

Transmission electron microscopy

Organoid samples were fixed in glutaraldehyde followed by osmium tetroxide post-fixation, dehydrated through a graded ethanol series, and embedded in epoxy resin. Semi-thin sections were prepared for initial evaluation and imaged using a NIKON Eclipse Ci microscope equipped with a NIKON DS-Fi3 camera. Ultrathin Sections  (40–70 nm) were subsequently cut using an ultramicrotome, mounted on copper grids, and contrasted with uranyl acetate and lead citrate before imaging. TEM analysis was performed using a JEM1400 transmission electron microscope (JEOL) equipped with a Xarosa TEM CMOS camera (EMSIS) and corresponding RADIUS acquisition software.

Statistical analysis

Statistical analysis was performed using GraphPad Prism (v.10.4.1) software. Data are presented as mean ± SEM. The normality of the data distribution was assessed prior to statistical testing. For comparisons involving more than two groups, paired one-way ANOVA followed by Dunnett’s multiple comparisons test was used for normally distributed data. For non-normally distributed data, the Friedman test was applied. No data points were excluded from the analysis as variability between patients was considered biological in nature. Statistical significance was defined as P ≤ 0.05.

Results

Amoxicillin supplementation disrupts the maturation stage of amelogenesis in vitro

To investigate the causal impact of AMX (Fig. 1a) on amelogenesis, ETO were first cultured in TOM for one week to promote proliferation. Next, ameloblast-like cell differentiation was triggered by transitioning to MIM as previously described by our group [35], in the presence of different concentrations of AMX. Analyses were conducted after four and seven days of differentiation (Fig. 1b) to evaluate how AMX affected pre-secretory stage markers Dentin sialophosphoprotein (DSPP), Von Willebrand factor D and EGF domains (VWDE), and maturation stage markers AMTN and ODAM. To confirm successful differentiation of ETO toward an ameloblast-like state, PCA revealed a clear separation between TOM and MIM conditions (Suppl. Fig. S2a), with differential gene expression analysis highlighting upregulation of key amelogenesis-associated genes in MIM (Suppl. Fig. S2b). Building on this validated differentiation platform, RNA-seq was subsequently performed to investigate transcriptomic changes induced by AMX exposure. After four days of differentiation, morphological changes were observed in cells cultured in MIM compared to ETO maintained in TOM (Fig. 1c). Consistent with previous characterizations [35], MIM-cultured organoids displayed a more irregular, bubble-like outer contour. In contrast, TOM-cultured organoids exhibited a smoother and more well-defined outline. No apparent morphological differences were observed between AMX-treated and control MIM cultures, with comparable organoid morphology, cell shape, and density across conditions. Using Alamar Blue reduction as a read-out of metabolic activity and cell viability, no significant differences were detected between AMX-treated conditions and MIM controls, indicating preserved metabolic activity across the tested AMX concentrations (0.1-4 mg/mL, Fig. 1d). Similar morphological features and cell viability profiles were observed after seven days of differentiation (Fig. 1g, h). No significant differences in the expression of pre-secretory amelogenesis markers DSPP and VWDE were detected at either time point (Fig. 1e–i). In contrast, expression of the maturation stage marker AMTN was significantly reduced after seven days of differentiation in the presence of 4 mg/mL AMX, corresponding to an approximately 2.8-fold decrease relative to MIM controls (ΔCt: -4.21 ± 1.41 vs. -5.65 ± 1.82). In addition, a downward trend in ODAM expression was observed under the same conditions (Fig. 1j). Taken together, these results suggest that AMX exposure primarily impacts maturation stage amelogenesis while leaving early differentiation, morphology, and metabolic activity largely unaffected.

Fig. 1.

Amoxicillin decreases maturation stage EMP expression during amelogenesis in vitro. a Chemical structure of amoxicillin (AMX) generated using ChemSketch. b Timeline of the experimental setup. Image created with BioRender. c, g Morphological changes of ameloblast-like cells following AMX supplementation (0.1–4 mg/mL) after four (c) or seven (g) days of differentiation. d, h Cell viability after four (d) and seven (h) days of differentiation, assessed using the Alamar Blue assay. AMX-treated conditions were normalized to the MIM control (n = 4). e, f, i, j Gene expression levels of pre-secretory stage markers (DSPP and VWDE) and maturation stage markers (AMTN and ODAM) after four (e, f) and seven (i, j) days of differentiation (n = 6). The dotted line indicates the threshold for detectable expression. Points on or below this line are considered to have no expression. Data are presented as mean ± SEM. Statistical analyses were performed as described in Materials and Methods (P ≤ 0.05). Scale bars: 500 μm

AMX treatment induces selective transcriptomic and metabolic changes in MIM organoids

PCA revealed limited global transcriptomic separation between AMX-treated organoids and MIM controls (Fig. 2a). Despite the limited overall variance, differential expression analysis identified distinct gene expression changes between MIM and AMX-treated organoids. A heatmap of the top 100 differentially expressed genes (50 upregulated and 50 downregulated) illustrates these transcriptional differences (Fig. 2b). Among the downregulated genes, six candidates with potential relevance to ameloblast biology, cell adhesion, and cellular stress pathways were selected for further validation (DNAJC15, ITGB6, TGFBI, ADD2, CHL1, and MMP12). qRT-PCR analysis confirmed significant downregulation of MMP12 in AMX-treated organoids compared to MIM controls (Fig. 2c), whereas the remaining genes did not show significant differences upon AMX exposure (Suppl. Fig. S3). To gain functional insight into the transcriptional alterations induced by AMX treatment, Gene Ontology (GO) enrichment analysis was performed (Fig. 2d). This analysis revealed significant downregulation of terms related to extracellular space. While enrichment for canonical glycolysis did not reach the predefined significance threshold (FDR < 0.05), it showed a strong trend toward downregulation. Gene Set Enrichment Analysis (GSEA) further supported pathway-level modulation in response to AMX treatment (Fig. 2e). Notably, oxidative phosphorylation, glycolysis, protein secretion, and unfolded protein response (UPR) pathways showed significant negative enrichment in AMX-treated samples, supporting reduced metabolic and secretory pathway activity. Together, these data suggest that AMX induces limited global transcriptomic changes but selectively modulates metabolic and ER stress-related pathways during differentiation of ETO.

Fig. 2.

Transcriptomic changes in MIM organoids upon AMX treatment. a Principal Component Analysis (PCA) showing global transcriptomic differences between TOM, untreated MIM, and AMX-treated MIM organoids. b Heatmap of the top 100 differentially expressed genes (50 up- and 50 downregulated) across samples. Candidate genes selected for validation are shown with a red asterisk. c Gene expression levels of MMP12. Bars represent mean ± SEM; (n = 4 biological replicates). Statistical analyses were performed as described in Materials and Methods (P ≤ 0.05). d GO enrichment analysis of downregulated genes in AMX-treated organoids, plotted as -log10(FDR). The dashed line indicates the significance threshold (FDR = 0.05). e GSEA plots illustrating pathway-level enrichment of hallmark gene sets in AMX-treated vs. MIM control organoids

Amoxicillin supplementation reduces desmosomal abundance and increases ER stress

The observed downregulation of the maturation stage marker AMTN, in combination with the observed transcriptomic alterations, warrant further investigation into cellular organization and ultrastructural features. All subsequent analyses were conducted on ETO cultured for seven days in MIM with 4 mg/mL AMX, corresponding to the time point and concentration with the most pronounced effects. First, epithelial cell–cell junctions were investigated. Gene expression analysis revealed a significant downregulation of desmoglein-1 (DSG1), a crucial component of desmosomes, in MIM compared to TOM conditions (Fig. 3a). No significant differences in DSG1 expression were observed between MIM-cultured cells in the presence or absence of AMX. At the protein level, Western blot analysis confirmed a significant reduction of DSG1 expression in MIM-cultured cells compared to TOM-cultured ETO (Fig. 3b). In contrast to the gene expression data, a downward trend in DSG1 protein levels was observed in AMX-treated cells compared to MIM controls (Fig. 3b). To further characterize these findings, ultrastructural analysis was performed using TEM. Semi-thin sections revealed a less compact tissue organization in MIM-cultured organoids compared to those maintained in TOM. In TOM-cultured organoids, epithelial cells were tightly packed and organized into a dense epithelial structure, whereas MIM-cultured organoids exhibited a loose organization with.

Fig. 3.

Amoxicillin decreases desmosomal abundance and alters ER homeostasis. a Gene expression levels of desmoglein-1 (DSG1) (n = 6). b DSG1 protein expression and representative Western blot of one patient sample (n = 3). Full-length blots are presented in Suppl. Fig. S4. c Upper panels show semi-thin overview images of the organoids (red asterisks indicate intercellular spacing). Lower panels display ultrastructural images obtained by transmission electron microscopy (TEM); the boxed region indicates the area shown at higher magnification, and white arrows indicate desmosomal junctions. d Gene expression levels of Heat Shock Protein Family A (Hsp70) Member 5 (HSPA5) (n = 6). e HSPA5 protein expression and representative Western blots of one patient sample (n = 3). Each patient is represented by a different symbol to show patient-to-patient consistency across conditions in Western blot data. f Representative ultrastructural TEM images of the endoplasmic reticulum (ER). Boxed regions are enlarged in the lower panels. Yellow asterisks indicate electron-lucent vesicular compartments, whereas blue asterisks indicate membrane-rich structures with concentric or multilamellar morphology. Data are presented as mean ± SEM. Statistical analyses were performed as described in Materials and Methods (P ≤ 0.05). Scale bars: 50 μm (black) or 1 μm (white), as indicated

increased intercellular spacing (upper panels; Fig. 3c). At the ultrastructural level, desmosomal junctions were readily observed in TOM conditions, appearing as electron-dense complexes connecting adjacent epithelial cells. In contrast, desmosomal structures appeared less frequently in MIM-cultured organoids, including those exposed to AMX (lower panels; Fig. 3c).

Next, changes in the endoplasmic reticulum (ER) were assessed. Gene expression analysis revealed a significant upregulation of heat shock protein family A (Hsp70) member 5 (HSPA5), a key component of the UPR pathway, following AMX treatment (Fig. 3d). In contrast, Western blot analysis showed a non-significant decrease (p-value: 0.1920) in HSPA5 protein levels under the same conditions (Fig. 3e). Similarly to DSG1, this downward expression trend was consistently observed across individual patient-derived organoid lines. Ultrastructural analysis suggested alterations of intracellular membrane systems in MIM-treated cells, both in the absence and in the presence of AMX. These included enlarged, electron-lucent vesicular compartments predominantly localized in the perinuclear region. In addition, multiple membrane-rich structures with concentric or multilamellar morphology were observed, ranging from compact whorl-like assemblies to more irregular, layered membranous profiles. In contrast to MIM-cultured cells, the ER of cells maintained in TOM appeared as elongated, parallel cisternae with regular spacing, and no prominent multilamellar or whorl-like membranous structures were observed. Mitochondria exhibited intact morphology with well-defined cristae, and vesicular compartments were small and evenly distributed throughout the cytoplasm (Fig. 3f). Together, these observations indicate that AMX exposure is linked to altered epithelial junctional organization and ER homeostasis in differentiating ETO.

Effects of Bisphenol A exposure on amelogenesis-associated morphology and gene expression in vitro

To assess the effect of BPA (Fig. 4a) on amelogenesis in vitro, the same experimental setup used for AMX exposure was applied (Fig. 1b). ETO supplemented with 30 µg/mL BPA exhibited pronounced morphological alterations already after four days of culture in MIM, including organoid shrinkage, loss of structural integrity, and increased fragmentation (Fig. 4b). These morphological changes were more pronounced after seven days of culture in MIM (Fig. 4f). In contrast, other tested BPA concentrations (0.2–10 µg/mL) did not induce apparent morphological changes compared to MIM control organoids (same control conditions as shown in Fig. 1c–g). Cell viability, assessed by Alamar Blue reduction, remained above the predefined exclusion threshold of 50% after four days of 30 µg/mL BPA exposure but dropped below this threshold after seven days in MIM (Fig. 4c–g). Although metabolic activity remained above the exclusion criteria at the four-day time point, the severe morphological alterations led to the exclusion of 30 µg/mL BPA from subsequent analyses. The tested BPA concentrations (0.2–10 µg/mL) caused no significant differences in expression of amelogenesis markers (Fig. 4d, e–h, i). Methanol (0.1% v/v) was included as a vehicle control for BPA and did not result in observable changes in organoid morphology, cell viability, or gene expression. Taken together, these results demonstrate that BPA induces cytotoxicity at high concentrations but does not significantly alter expression of amelogenesis markers at non-toxic doses.

Fig. 4.

Bisphenol A does not influence amelogenesis in vitro. a Chemical structure of bisphenol A (BPA) generated using ChemSketch. b, f Morphological changes of ameloblast-like cells following BPA supplementation (0.4–30 µg/mL) after four (b) or seven (f) days of differentiation. The same TOM and MIM controls of Fig. 1 were used. c, g Cell viability after four (c) and seven (g) days of differentiation, assessed using the Alamar Blue assay. BPA-treated conditions were normalized to the MIM control (n = 4). Methanol (diluted 1:1000) was included as a vehicle control. BPA 30 µg/mL was excluded due to its toxicity (d, e, h, i). Gene expression levels of pre-secretory stage markers (DSPP and VWDE) and maturation stage markers (AMTN and ODAM) after four (d, e) and seven (h, i) days of differentiation (n = 6). The dotted line indicates the threshold for detectable expression. Points on or below this line are considered to have no expression. Data are presented as mean ± SEM. Statistical analyses were performed as described in Materials and Methods (P ≤ 0.05). Scale bars: 500 μm

Epithelial tooth organoids recapitulate the intrinsic effects of vinblastine and vincristine in vitro

The mechanism of action of the chemotherapeutic agents VB and VC involves binding to free tubulin dimers, thereby destabilizing microtubule dynamics and preventing proper mitotic spindle formation. This ultimately results in mitotic arrest due to inhibition of the metaphase-to-anaphase transition [42] (Fig. 5a). To determine whether ETO recapitulate this mechanism, α-tubulin (TUBA) expression was evaluated at both the transcriptional and protein levels following treatment with 500 and 250 nM VB or VC for seven days of differentiation. Gene expression analysis revealed no significant alterations in TUBA expression compared to MIM controls (Fig. 5b). In contrast, Western blot analysis revealed a clear dose-dependent reduction in TUBA protein levels following treatment with both chemotherapeutic agents, with a significant decrease observed at 500 nM VC (Fig. 5c). To investigate whether VB and VC affect ER stress pathways, HSPA5 expression was evaluated at both gene and protein levels. Except for 500 nM VB, all other conditions induced a significant downregulation of HSPA5 expression compared to MIM controls (Fig. 5d). At the protein level, HSPA5 levels showed a similar downward trend across conditions, reaching statistical significance only in the 500 nM VB condition (Fig. 5e). To further characterize the effects of VB and VC at the ultrastructural level, TEM was performed. Again, TOM controls displayed a well-preserved ultrastructural organization with normal rough ER. In contrast, ETO differentiated in MIM contained electron-lucent vesicular compartments predominantly localized in the perinuclear region. In VB- and VC-exposed organoids, numerous multilamellar, myelin-like structures were observed, often resembling ovoid “oyster”-shaped vesicles. These structures were largely absent in TOM and MIM-control organoids (Fig. 5f).

Fig. 5.

ETO successfully replicate the working mechanism of chemotherapeutic agents vinblastine and vincristine in vitro. a Schematic showing the working mechanism of microtubule-disturbing chemotherapeutic agents vinblastine (VB) and vincristine (VC). Image created with BioRender. Chemical structures are generated with ChemSketch. b Gene expression levels of alpha-tubulin (TUBA) (n = 6). c TUBA protein levels and representative Western blot of one patient sample (n = 3). d Gene expression levels of endoplasmic reticulum (ER) stress marker HSPA5 (n = 6). e HSPA5 protein expression and representative Western blots of one patient sample (n = 3). Each patient is represented by a different symbol to show patient-to-patient consistency across conditions in Western blot data. Full-length blots are presented in Suppl. Fig. S5. f Ultrastructural (TEM) analysis. Boxed regions are enlarged to the right. Data are presented as mean ± SEM. Statistical analyses were performed as described in Materials and Methods (P ≤ 0.05). Scale bars: 1 μm, as indicated

Vinblastine and vincristine exposure alter the maturation stage of amelogenesis in vitro

To determine the direct causal effects of VB and VC on amelogenesis, organoids were exposed as described in the experimental framework above (Fig. 1b). Organoids exposed to 500 nM VB or VC displayed a reduction in overall organoid size after four days of differentiation in MIM, which persisted at seven days (Figs. 6a–e and 7a–e). Despite the reduced size, overall organoid morphology remained structurally preserved. Consistently, Alamar Blue analysis demonstrated relative viability values above 50% for both chemotherapeutic agents at both differentiation time points (Figs. 6b–f and 7b–f). Treatment with VB resulted in gene-specific and time-dependent alterations in enamel-associated marker expression. After four days of differentiation, VWDE expression was significantly 5.6-fold increased following exposure to 250 nM VB compared to MIM controls (ΔCt: -16.68 ± 1.34 vs. -14.20 ± 1.56) (Fig. 6c). At the same time point, ODAM expression was significantly downregulated at concentrations ranging from 100 to 500 nM VB (Fig. 6d). After seven days of differentiation, the reduction in VWDE expression was no longer observed (Fig. 6g). In contrast, AMTN and ODAM expression were significantly decreased at 500 and 250 nM VB. Notably, both AMTN and ODAM expression levels demonstrated a dose-dependent response at this time point (Fig. 6h). For VC, AMTN expression was significantly reduced after four days of differentiation at 500 nM (Fig. 7d). This effect became more pronounced after seven days, where both AMTN and ODAM expression were significantly downregulated at 500 and 250 nM VC, again displaying a dose-dependent pattern (Fig. 7h). Additionally, VWDE expression was significantly increased at seven days following treatment with 100 nM VC (Fig. 7g). Together, these data indicate that vinca alkaloids dose-dependently disrupt ameloblast differentiation, characterized by increased pre-secretory markers and reduced maturation-stage markers.

Fig. 6.

Vinblastine lowers AMTN and ODAM expression during amelogenesis in vitro. a, e Morphological changes of ameloblast-like cells following VB supplementation (25–500 nM) after four (a) or seven (e) days of differentiation. The same TOM and MIM controls of Fig. 1 were used. (b, f) Cell viability after four (b) and seven (f) days of differentiation, assessed using the Alamar Blue assay. VB-treated conditions were normalized to the MIM control (n = 4). c, d, g, h Gene expression levels of pre-secretory stage markers (DSPP and VWDE) and maturation stage markers (AMTN and ODAM) after four (c, d) and seven (g, h) days of differentiation (n = 6). The dotted line indicates the threshold for detectable expression. Points on or below this line are considered to have no expression. Data are presented as mean ± SEM. Statistical analyses were performed as described in Materials and Methods (P ≤ 0.05). Scale bars: 500 μm

Fig. 7.

Vincristine decreases mature ameloblast markers AMTN and ODAM in vitro. a, e Morphological changes of ameloblast-like cells following VC supplementation (25–500 nM) after four (a) or seven (e) days of differentiation. The same TOM and MIM controls of Fig. 1 were used. b, f Cell viability after four (b) and seven (f) days of differentiation, assessed using the Alamar Blue assay. VC-treated conditions were normalized to the MIM control (n = 4). c, d, g, h Gene expression levels of pre-secretory stage markers (DSPP and VWDE) and maturation stage markers (AMTN and ODAM) after four (c, d) and seven (g, h) days of differentiation (n = 6). The dotted line indicates the threshold for detectable expression. Points on or below this line are considered to have no expression. Data are presented as mean ± SEM. Statistical analyses were performed as described in Materials and Methods (P ≤ 0.05). Scale bars: 500 μm

Discussion

The study of amelogenesis and related enamel disorders remains poorly understood, mainly due to the lack of physiologically relevant (human) in vitro models that recapitulate the complex process by which ameloblasts mature and form enamel. Traditional approaches, such as primary ameloblast cultures, are limited due to short-term viability and ethical concerns, whereas immortalized cell lines may exhibit genomic instability and heterogeneous phenotypes, thereby reducing their translational relevance [5]. Over the past few years, significant progress has been made in developing more advanced models, including induced pluripotent stem cells and 3D organoid cultures, that hold greater promise for regenerative and personalized enamel research [5]. In this study, we focus on ETO as a novel human-derived platform for in vitro studies of amelogenesis. Specifically, ETO are employed as a patient-in-a-dish model to investigate the effects of external disease-relevant stimuli, including AMX and BPA, both associated with MIH, as well as the chemotherapeutic agents VB and VC, which are frequently administered to children during important time windows of amelogenesis. Through comparative analysis of cell morphology, viability, marker expression, and ultrastructure, we provide evidence that ETO not only recapitulate aspects of ameloblast differentiation but also model compound-specific disruptions in amelogenesis pathways.

The ETO were derived from the dental follicle of unerupted third molars, a tissue harboring ERM that retain epithelial characteristics and proliferative capacity. This organoid model was originally established and validated by our group [35], making the present study a logical extension of our previous work. Dental follicle tissue is routinely discarded during third molar extraction, providing an ethically accessible source of human dental epithelial cells. However, as the donor tissue originates from adolescent patients, the cells reflect a later developmental stage than the early childhood window during which most DDE-associated exposures typically occur. Nevertheless, these ETO are derived from endogenous somatic stem cells, they retain tissue-specific epigenetic memory, offering a more faithful and reliable model than induced pluripotent stem cell (iPSC)-derived counterparts [43]. Furthermore, as with all in vitro models, the system lacks the systemic hormonal and metabolic context of the in vivo environment, which may influence ameloblast behavior and compound sensitivity.

AMX is a commonly prescribed pediatric antibiotic used to treat bacterial infections of the respiratory, urinary, skin and soft tissue systems. As a broad-spectrum β-lactam antibiotic of the aminopenicillin class, it binds to penicillin-binding proteins and inhibits bacterial cell wall synthesis, ultimately leading to cell lysis [44]. Epidemiological studies have identified AMX exposure as a risk factor for MIH, highlighting the importance of investigating its potential impact on amelogenesis [23, 45]. In our model, AMX primarily affected the maturation stage marker AMTN, without altering cell viability or overall organoid morphology. Although previous studies have not specifically assessed AMTN expression, several in vivo studies have reported changes in other EMPs, including AMELX and AMBN [46, 47]. Importantly, these studies were mainly conducted in rat models, in which AMX was administered via food intake or intragastrically by oral gavage. AMTN expression increases during the maturation stage of amelogenesis, when the enamel matrix undergoes proteolytic processing and mineralization. In addition to its role as a maturation stage marker, AMTN has been suggested to contribute to the adhesion of maturation stage ameloblasts to the enamel surface, thereby supporting maintenance of the epithelial-enamel interface during late-stage enamel formation [48, 49]. The observed reduction in AMTN expression, therefore, suggests that AMX exposure may preferentially disrupt the maturation stage of amelogenesis.

To determine whether broader transcriptional changes accompanied these maturation stage alterations, we performed RNA-seq on differentiated MIM organoids exposed to AMX. Overall, AMX treatment induced only modest transcriptomic changes, indicating that global cellular identity was largely preserved. Instead of widespread transcriptional reprogramming, pathway-level analyses pointed toward coordinated modulation of metabolic and secretory processes. Given the high energetic and secretory demands of maturation stage ameloblasts, coordinated reductions in these programs, alongside decreased AMTN and MMP12 expression, suggest diminished maturation stage functionality [50–54]. Collectively, our data support a model in which AMX exposure does not trigger widespread stress or loss of identity but rather leads to partial impairment of maturation stage processes, consistent with key features of MIH, where enamel thickness is generally preserved while mineral quality is altered [14, 55].

In line with the pathway-level alterations observed in the transcriptomic analysis, we next examined markers of epithelial integrity and ER-associated homeostasis at the protein and ultrastructural levels. AMX exposure was associated with decreased DSG1 expression, suggesting impaired epithelial cell-cell adhesion. DSG1 is a key component of desmosomes required to maintain epithelial integrity [56]. While earlier studies have mainly focused on tight junction proteins rather than desmosomal components, AMX exposure in juvenile mouse models has been shown to downregulate tight junction proteins, including occludin and claudins 1 and 4 in mature amebloblasts [57]. Moreover, both tight junctions and desmosomes are known to play essential roles in maintaining ameloblast polarity, cohesion, and coordinated enamel matrix secretion [58–60]. Together, these findings suggest that AMX may also affect desmosomal integrity in a human-derived organoid model.

We further investigated whether AMX exposure influenced ER-associated homeostasis, as suggested by enrichment of secretion- and ER-related pathways in the RNA-seq analysis. HSPA5, a key regulator of the UPR pathway, showed increased gene expression, while protein levels were reduced. This inconsistency between mRNA and protein levels may indicate a compensatory transcriptional response to restore proteostatic balance at the protein level. Similar patterns have been reported under conditions of prolonged or dysregulated ER stress, where impaired translation, altered protein stability, or other post-transcriptional regulatory mechanisms may prevent protein levels from increasing despite elevated mRNA expression [61–63]. Given that ameloblasts are highly secretory cells that depend on tightly regulated ER function for the synthesis and processing of EMPs, disturbances in ER homeostasis may directly affect enamel maturation [64, 65]. Ultrastructural analysis suggested reorganization of intracellular membrane systems in MIM-treated organoids, both in the presence and absence of AMX, whereas TOM controls displayed preserved ER architecture. Although ultrastructural observations alone cannot demonstrate ER stress, the presence of membranous whorls, enlarged perinuclear vesicular compartments, and multilamellar structures is consistent with activation of adaptive stress-related pathways [66, 67]. Collectively, these findings suggest that AMX exposure may promote altered ER remodeling and proteostatic imbalance, thereby contributing to impaired enamel maturation.

BPA is a well-known endocrine-disrupting compound, widely used in polycarbonate plastics and epoxy resins. In contrast to AMX, BPA exposure did not significantly affect amelogenesis markers at non-toxic concentrations in our organoid model. While 30 µg/ml caused cytotoxicity, lower concentrations had no measurable effect on DSPP, VWDE, AMTN, or ODAM expression. These results suggest that, under the conditions tested, BPA does not directly disrupt ameloblast differentiation or the maturation processes critical for enamel formation. The potential role of BPA in enamel defects, including MIH, has been investigated, but epidemiological evidence is limited and inconsistent. For example, a cohort from the Utah Children’s Project found no clear association between prenatal exposure to phenol or phthalate and enamel hypomineralization [8]. Nevertheless, in vivo studies indicate that BPA can affect enamel development under certain conditions. Jedeon et al. reported that rat incisors exposed to BPA showed hypomineralized enamel and alterations in enamel-related gene expression, including increased ENAM and decreased KLK4, suggesting disrupted matrix maturation [25]. The inconsistency between these in vivo findings and our in vitro results may reflect differences in species, developmental stage, exposure duration, systemic endocrine regulation, or the absence of hormonal cues or metabolic enzymes in organoid cultures. Taken together, our data do not support a direct effect of BPA on amelogenesis, but they do not rule out more complex indirect effects that could occur in vivo.

We next investigated the impact of chemotherapeutics on amelogenesis, focusing on VB and VC, two vinca alkaloids commonly used in pediatric oncology. Children treated with these agents frequently develop dental defects later in life, including enamel hypoplasia and structural anomalies [31, 68]. Using our human-derived organoid model, we first studied whether the intrinsic cellular effects of VB and VC could be reproduced in vitro. Protein levels of TUBA were markedly reduced, consistent with microtubule destabilization and mitotic inhibition, which is a well-established effect of vinca alkaloids in proliferating cells [42]. Additionally, HSPA5 expression was downregulated at both the gene and protein levels, reflecting suppression of ER stress response pathways, a known mechanism by which these compounds prevent tumor cells from activating a protective UPR pathway. Ultrastructural analysis further revealed extensive intracellular remodeling in VB- and VC-exposed organoids, which displayed numerous multilamellar, myelin-like structures. These formations likely represent accumulated autophagic or lysosomal compartments, including ER or mitophagy vesicles. Disruption of microtubules by vinca alkaloids impairs intracellular transport, preventing normal trafficking of vesicles, proteins, and organelles. As a result, these membrane-rich structures accumulate, reflecting a stress response to disrupted protein processing and intracellular trafficking. Previous in vivo studies in rat incisors have reported similar cellular responses to vinca alkaloids. Early work by Moe et al.. showed that systemic VB disrupts the ultrastructure of ameloblasts, affecting microtubule and organelle organization [69, 70]. Nanci et al. further reported loss of ruffled border microtubules, disturbed cell polarity, and intracellular vesicle accumulation following VB treatment [71].

Finally, we assessed the effects of VB and VC on key markers of amelogenesis. To our knowledge, this is the first study to investigate the impact of vinca alkaloids on these markers in a human-derived organoid model, providing a novel perspective on how cytoskeletal disruption translates to molecular changes in amelogenesis. After seven days of differentiation, both chemotherapeutics induced apparent dose-dependent disruptions, with higher concentrations producing stronger effects on gene expression. Morphologically, exposed organoids appeared smaller, consistent with impaired cellular elongation and maturation. Typically, the transition from the pre-secretory to secretory stage involves polarization and elongation of ameloblasts, followed by shortening as cells enter the maturation stage [1]. Disruption of the cytoskeleton by vinca alkaloids likely prevents these coordinated morphological changes, leading to an accumulation of pre-secretory-like cells. This provides a mechanistic explanation for the observed increase in VWDE expression, a pre-secretory marker, alongside the decreased maturation stage markers AMTN and ODAM. Taken together, these results indicate that vinca alkaloids can directly interfere with amelogenesis, linking cytoskeletal destabilization to morphological and molecular changes that are likely to contribute to enamel defects observed in pediatric patients treated with these agents.

Several limitations of the present study warrant consideration. First, despite extensive optimization efforts with multiple commercially available antibodies, reliable Western blot detection of AMTN and ODAM protein levels could not be achieved, likely reflecting the known challenges of immunoblotting heavily glycosylated extracellular matrix proteins [72–74]. Gene expression analysis, therefore, served as the primary readout for these maturation stage markers. Second, the organoid model used here, while human-derived and physiologically relevant, does not fully recapitulate the complexity of amelogenesis in vivo. Given that amelogenesis is influenced by a broad range of genetic, epigenetic, hormonal, and systemic factors, the isolated in vitro exposure conditions studied here cannot fully model the multifactorial nature of DDE development. Third, donor-derived organoids were collected opportunistically from patients aged 13–19 years, allowing the model to reflect a specific developmental window and donor population. Finally, the exposure concentrations used were selected to bracket nonlethal ranges in the organoid system rather than to precisely model pediatric pharmacokinetics, and direct translation to in vivo exposure levels will require future pharmacokinetics-informed studies.

Conclusions

In conclusion, our study establishes human epithelial tooth organoids as a robust and physiologically relevant platform to model amelogenesis and its disruption by external factors. We demonstrate that clinically relevant exposures, including AMX and pediatric chemotherapeutics, can interfere with ameloblast differentiation and maturation, affecting intracellular homeostasis and enamel matrix gene expression. Importantly, this model captures compound-specific effects, highlighting its utility for studying diverse mechanisms of enamel disruption. These findings provide human-based experimental evidence supporting a potential link between environmental and therapeutic exposures and disrupted ameloblast maturation, and position epithelial tooth organoids as a valuable tool for investigating the pathogenesis of DDEs.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ADD2

Adducin 2

AMBN

Ameloblastin

AMELX

Amelogenin

AMTN

Amelotin

AMX

Amoxicillin

BCA

Bicinchoninic acid

BPA

Bisphenol A

BSA

Bovine serum albumin

CHL1

Cell adhesion molecule L1 like

CYCA

Cyclin A

DF

Dental follicle

DNAJC15

DnaJ heat shock protein family (Hsp40) member C15

DSPP

Dentin sialophosphoprotein

DSG1

Desmoglein-1

ECL

Enhanced chemiluminescence

EMP

Enamel matrix proteins

ENAM

Enamelin

ER

Endoplasmic reticulum

ERM

Epithelial cell rests of Malassez

ETO

Epithelial tooth organoids

GO

Gene ontology

GSEA

Gene set enrichment analysis

HERS

Hertwig epithelial root sheath

HRP

Horseradish peroxidase

HSPA5

Heat shock protein family A (Hsp70) member 5

iPSC

Induced Pluripotent Stem Cells

ITGB6

Integrin beta-6

MIH

Molar incisor hypomineralization

MIM

Mineralization-inducing medium

MMP12

Matrix metalloproteinase-12

ODAM

Odontogenic ameloblast-associated protein

PBS

Phosphate-buffered saline

PCA

Principal component analysis

PVDF

Polyvinylidene fluoride

qRT-PCR

Quantitative reverse transcriptase PCR

RI

Rock inhibitor

RNA-Seq

RNA sequencing

SFDM

Serum-free defined medium

TEM

Transmission electron microscopy

TGFBI

Transforming growth factor, beta-induced

TOM

Tooth organoid medium

TUBA

α-tubulin

VB

Vinblastine

VC

Vincristine

VWDE

Von Willebrand factor D and EGF domains

Author contributions

S.H. contributed to the conception and design of the study, data acquisition, analysis and interpretation, drafting, and critically revising the manuscript. F.H. contributed to the conception and design of the study, the analysis and interpretation, and the critical revision of the manuscript. J.M. performed all extractions of the human third molars and the associated tissues, provided clinical input, and critically revised the manuscript. H.V. and H.D. contributed to the interpretation of the data and critically revised the manuscript. E.W. provided the necessary TEM facilities for the study as well as critically revising the manuscript. I.L. and A.B. contributed to the conception and design of the study, the interpretation and analysis, and the critical revision of the manuscript. All authors provided final approval and agreed to be accountable for all aspects of the work.

Funding

This work was supported by a grant from the Bijzonder Onderzoeksfonds Universiteit Hasselt (BOF22KP12 to S.H. and A.B.) and by the Fund for Scientific Research (FWO) – Flanders (G083626N to A.B. and H.D.; G0A2E26N to F.H. and I.L.; 1226325 N to F.H.). Computational resources for all bulk transcriptome analyses were provided by the Vlaams Supercomputer Centrum (VSC), managed by FWO.

Data availability

The RNA-seq datasets generated during the current study are included as additional files (Additional Files 1-4), summarized in Supplementary Table 10. All other datasets are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study was initially approved by the Medical Ethics Committee of Hasselt University (reference 13/0104U) on 03 February 2014. An amendment to this approval was granted on 15 November 2023. The approved project is titled “Human dental stem cells as an alternative stem cell source for use in regenerative medicine and disease models.” All human tissue samples, including extracted teeth and dental follicle tissue, were obtained with written informed consent from the donors or their legal guardians.

Competing interests

Hugo Vankelecom, Ivo Lambrichts, and Annelies Bronckaers are inventors on a patent (Application number EP2023054422, Publication number 2023/161276, published on 31/8/2023) related to the generation of human epithelial tooth organoids. All other authors declare no competing interests.