Soluble Notch agonist enables human ameloblast maturation and enamel-like tissue formation for tooth regeneration

Abstract

Enamel, the hardest mineralized material in the human body, protects the underlying living tissues, the dentin and pulp of the tooth. However, over 90% of adults have lost or damaged enamel and cannot regenerate the protective structure due to lack of enamel-producing cells, ameloblasts. iPSC-derived secretory Ameloblasts (isAM) have promise in future regenerative dentistry. Today, it is not known why iAM maturation requires intimate contact with the dentin-producing cell type, odontoblast. Here, we reveal that one of the critical signaling ligands emanating from odontoblasts for ameloblast maturation is Delta, the ligand for Notch receptor. We showed that our designed, soluble Notch agonist can induce iAM organoid maturation in an unprecedented manner, without interactions with odontoblast layer. Notably, soluble Notch agonist induces the iAM maturation to a novel, WDR72-positive mature secretory AM stage (ismAM) in our ameloblast organoid model. When transplanted under the kidney capsule of NOD-SCID mice, these ismAM organoids generated enamel-like calcified material, as confirmed by microCT analysis, marking the first demonstration that Notch-activated iAM organoids can form such tissue in vivo. This novel maturation procedure enabled us to analyze the specific requirements of DLX3 function in ameloblasts, independent of its known function in odontoblasts. We now show that DLX3, a gene associated with Amelogenesis Imperfecta, is required on a cell-autonomous manner in human ameloblasts for the expression of Enamelin, MMP20, and WDR72, a role not previously demonstrated in mouse models.

Introduction

From the teeth of Krakens in the deep sea to the ivory tusks in the savanna, these living ectodermal organs are indispensable for animal survival. In cephalopods and mammals alike, teeth are protected by an outer calcified layer called enamel. While some animals can regenerate continuously this protective enamel layer, humans cannot. In human developing tooth the enamel secreting cells, ameloblasts undergo apoptosis at the time of tooth eruption resulting in lost regenerative capacity of the protective layer in our aging body. Consequently, any damage to enamel later in life cannot be repaired naturally.

This lack of regenerative potential is particularly detrimental in conditions such as Amelogenesis Imperfecta (AI), a genetic condition caused by defects in the ameloblast cells that produce enamel, resulting in fragile teeth with abnormal shape, thickness, and pigmentation.¹ Beyond AI, genetic studies have identified single nucleotide polymorphisms (SNPs) near dental development genes that are associated with risk for dental caries, including rs2278163 near DLX3 —a gene that, when deleted in the neural crest of mice, severely impairs odontoblast differentiation and dentin formation² and whose forced expression in mouse ameloblast lineage upregulates enamel matrix protein (EMP) genes such as Amelx, Enam, and Klk4 ameloblasts³—and rs198968 near KLK4, which is necessary for ameloblast maturation.^4,5 Genome-wide association studies (GWAS) of primary dentition caries have also implicated DLX3 and KLK4 as suggestive loci contributing to caries risk.⁶ Mutations in DLX3 are uniquely associated with defects in both enamel and dentin, as seen in human tricho-dento-osseous syndrome and certain AI subtypes, underscoring its central regulatory role across epithelial and mesenchymal compartments during tooth development. Despite this, the cell-autonomous contribution of DLX3 to ameloblast maturation remains poorly defined, largely due to the absence of experimental systems capable of decoupling enamel and dentin differentiation. Our iPSC-derived iAM organoid model provides such a platform to isolate DLX3 function specifically within the ameloblast lineage. However, the functional role of enamel- or dentin-related genes in both AI and caries susceptibility remains unclear, and addressing this gap requires an experimental system capable of producing functional ameloblasts to enable mechanistic studies and guide regenerative strategies.

Recent progress in regenerative dentistry has highlighted the application of human induced pluripotent stem cells (hiPSCs) to generate induced ameloblast cells (iAMs)⁷. These iAMs possess significant potential for upcoming approaches aimed at enamel repair and regeneration. If iAMs can be differentiated into functional ameloblasts, they could serve as a viable source of cells for enamel regeneration. However, today the maturation of iAMs remains a significant hurdle, primarily due to a necessity for understanding the specific mechanisms and signals required for their development. One particularly intriguing element is the crucial requirement for close interaction of AMs with odontoblasts (OBs), the dentin-producing cells.⁷ Odontoblast–ameloblast crosstalk has been investigated using the SP7 (Osterix) promoter, which drives expression specifically in odontoblasts and osteoblasts but not in ameloblasts, thereby enabling selective manipulation of odontoblast-derived signals that might influence ameloblast differentiation.⁸ Conditional knockout mouse models targeting SP7-expressing cells (e.g., Osx-Cre; Macf1^fl/fl, Osx-Cre; Stat3^fl/fl, Osx-Cre; Cbfβ^fl/fl)^9–15 frequently exhibit both dentin and enamel defects, suggesting that compromised SP7-expressing odontoblasts may indirectly affect ameloblast activity and pointing to potential crosstalk between these two cell populations.

The Notch pathway is a highly conserved cell signaling mechanism known to govern multiple stages in cellular differentiation.^16,17 It plays a pivotal role in maintaining the balance between progenitor cell proliferation and differentiation, ensuring proper tissue development and homeostasis.¹⁶ The Notch pathway operates through direct cell-cell communication, wherein Notch receptors (Notch1–4) on a responding cell interact with ligands (Jagged1/2 and Delta-like ligands DLL1, DLL3, and DLL4) presented on the surface of neighboring cells. Upon ligand-receptor engagement, the Notch receptor undergoes two sequential proteolytic cleavages. The first cleavage, mediated by ADAM-family metalloproteases, releases the extracellular domain of the receptor, while the second cleavage, facilitated by the γ-secretase complex, liberates the Notch intracellular domain (NICD). Once freed, the NICD translocate to the nucleus, forming a complex with the transcriptional regulators RBP-J/γ-CSL and MAML. These interactions drive the transcription of Notch target genes, including members of the Hes and Hey families, which play critical roles in cell differentiation, proliferation, and survival. During tooth development, both Notch ligands (such as DLL1 and Jagged2) and receptors (Notch1–3) are expressed in the dental mesenchyme and epithelium, supporting a potential signaling interaction between odontoblasts and ameloblasts.^18,19 This tightly regulated cascade ensures that Notch signaling is activated in response to direct cell-cell interaction, enabling specific and localized effects on tissue morphogenesis. Interestingly, recent GWAS have identified a significant link between a SNP within an intron of ADAMTS9 gene and amelogenesis imperfecta^20,21

We have previously used single-cell sequencing of human fetal ameloblasts (AM) to develop a human iPSC-based induced AM (iAM) maturation protocol^7,22 to generate induced secretory ameloblasts (isAM). In this protocol, ameloblasts require interactions with odontoblasts (co-culture) to exhibit mineralization and expression of Amelogenin (AMELX), and Enamelin (ENAM). In the present study, we identify the critical signal emanating from OB that activates Notch signaling resulting in maturation of ameloblasts.

We demonstrate that our computationally designed soluble Notch agonist (C3-DLL4)²³ can generate mature ameloblasts in the absence of odontoblasts, establishing an odontoblast-free induction system for human ameloblast differentiation. This advance enables mechanistic dissection of ameloblast-intrinsic pathways and provides a powerful model to study genes implicated in enamel pathologies. Using this platform, we uncover that the amelogenesis imperfecta–associated gene DLX3 functions cell-autonomously within human ameloblasts to regulate ENAM, MMP20, and WDR72 expression.

This is particularly relevant because prior mouse studies involving K14-Cre–mediated Dlx3 deletion reported hypomineralized enamel but did not show strong changes in enamel matrix genes,^24,25 suggesting that some aspects of DLX3’s role in human enamel development may differ or remain unresolved. Given the mixed but intriguing human genetic evidence linking DLX3 variants to enamel development and caries risk, our organoid model offers a controlled platform to define the cell-autonomous roles of DLX3 in both ameloblast versus odontoblast maturation.

Results

Notch pathway during ameloblast maturation

iPSC-derived iAM requires a co-culture with iOB for the maturation into secretory ameloblasts.⁷ To identify the critical signaling pathways emanating from odontoblasts for AM maturation, we performed TopPath algorithm analysis for previous single-cell-sequencing data from human fetal tissues.^7,26 This analysis revealed Notch as a candidate signaling pathway involved in the crosstalk between OB and AM during the AM maturation stage (Fig. 1a, b). While other pathways (FGF, WNT and EGF) were also identified as potentially important signaling pathways at this stage, their contribution from OB to sAM signaling was not predicted by the talklr R package²⁷ to be as prominent as Notch pathway (Figs. 1a, b and S1b–d).

Fig. 1.

Notch pathway activation is required for ameloblast maturation in co-culture with odontoblasts. a TopPath pathway analysis of single-cell sequencing data from fetal tissues, identifying key signaling pathways involved in odontoblast (OB) to ameloblast (AM) communication. The Notch pathway (10.6%) is highlighted as a major contributor to AM maturation, alongside other pathways such as FGF, WNT, EGF, and BMP. b Schematic model depicting Notch signaling interactions between OBs and AMs, showing ligand-receptor interactions focusing on incoming signal to eAM/sAM (red arrows). The width of the arrows is proportional to the number of interactions. Black arrow indicates other possible interaction not considered in the analysis of TopPath. c Heatmap representation of Notch ligands (DLL1, DLL4) and Notch receptors (NOTCH1, NOTCH2, NOTCH3, NOTCH4), illustrating ligand expression in OBs and receptor expression in AMs, supporting OB-to-AM signaling. d Differentiation and co-culture model for generating induced ameloblasts (iAMs) from human iPSCs. Induced odontoblasts (iOBs) are co-cultured with induced early ameloblast (ieAM) organoids to promote their maturation into secretory ameloblasts (isAMs), with and without DAPT-mediated Notch inhibition. e 3D confocal images of co-cultured iAM organoids showing ENAM (yellow) and DSPP (red) expression, confirming ameloblast maturation in co-culture. Merged images include DAPI (nuclei, blue) and Phalloidin (cytoskeleton, purple). f Schematic representation summarizing the observed expression patterns of DSPP (red) and ENAM (yellow) in differentiated OB-like and AM-like cells within the co-culture. g Comparative immunofluorescence staining of ENAM in Control (top) vs. DAPT-treated (bottom) conditions, revealing reduced ENAM expression and altered cellular morphology upon Notch inhibition. Scale bar = 100 μm. h Quantification of ENAM fluorescence intensity, normalized to control, showing a significant reduction in ENAM expression in the DAPT-treated group (^****P < 0.000 1)

To evaluate the expression of Notch ligands and receptors in OB and sAM, we analyzed the gene expression levels of Notch and Delta in these cell types. We generated a heatmap by averaging the expression levels of OB and AM clusters from the single-cell data at the late timepoints during human fetal development (20-22gw)⁷ (Fig. 1c). We found that the ligands (DLL1 and DLL4) were predominantly expressed in the OB, while the Notch receptors (NOTCH1, NOTCH2, and NOTCH3) were primarily present on the sAM side (Fig. 1c). A similar pattern was observed in previously curated single-cell mouse data²⁸ (Fig. S1e), reinforcing the human findings (File S1). This cross-species consistency suggests that Delta from odontoblasts might activate Notch in ameloblasts.

Notch inhibitor affects enamelin expression in AM and OB co-culture system

To study the interaction between OB and AM we optimized our previously published co-culture protocol²⁶ (Fig. S1a) by developing a serum-free medium (SFM) protocol for iOB generation for the first time. This process is enhanced by activating FGF signaling through a novel C6-mb7 FGFR1/2c agonist²⁹ and Hedgehog signaling via SAG (smoothened agonist), supporting their maturation into functional iOBs by day 25 (Fig. S1a). These mature iOBs are plated as a monolayer in Matrigel to create a physiological substrate for downstream experiments. To generate the co-culture system, we differentiated hiPSC-derived ameloblast (iAM) organoids as done before.⁷ Ameloblast differentiation is initiated by inducing oral epithelial lineage from hiPSCs under defined serum-free conditions from day 0 to day 10. At day 16, the early ameloblasts (ieAM) were directed to three-dimensional (3D) organoids using low-adhesion plates, and at day 24, the organoids are co-cultured with iOB monolayers in Matrigel, recreating the in vivo tooth development microenvironment (Figs. 1d and S1a). This co-culture system enables reciprocal interactions between iOB and ieAM, driving the latter’s progression to secretory ameloblast stage (isAM). This state is characterized by polarized cells secreting enamelin (ENAM), an essential protein for enamel formation, with polarity reversal of ieAM toward the iOB layer, indicative of interaction-driven maturation (Fig. 1e, f). We utilized the hiPSCs based ameloblast (iAM) and odontoblast differentiation (iOB) methods in a defined serum-free media approach to dissect Notch function in OB to AM signaling. To assess if Notch signaling is essential for ameloblast (AM) maturation, we used DAPT (N-[N-(3, 5-difluorophenacetyl)-l-alanyl]-s-phenylglycinet-butyl ester),³⁰ a well-characterized small molecule inhibitor of γ-secretase. γ-secretase, with Presenilin as its catalytic subunit, is responsible for cleaving the Notch receptor. This cleavage releases the NICD, allowing it to translocate to the nucleus and activate target gene transcription involved in cellular differentiation. By inhibiting γ-secretase, DAPT blocks the release of NICD, disrupting Notch signaling. Prior studies have shown that Notch signaling is essential for the survival and maintenance of epithelial stem cells in the continuously growing mouse incisor.³¹ Consistent with these findings, while iOB-iAM interaction drove ENAM secretion in the co-culture system, Notch inhibition via DAPT significantly reduced ENAM induction in this system (Fig. 1g, h), suggesting that active Notch signaling is crucial for AM maturation and ENAM protein synthesis in this model.

Soluble Notch agonist

Notch activity has been implicated in the differentiation of ameloblast support cells in dental tissue development.^32,33 Specifically, Notch signaling mediates the divergence of the stratum intermedium lineage from the ameloblast lineage, as demonstrated in developing tooth epithelium.³² Dissecting the function of Notch activity in ameloblast (AM) maturation requires spatial and temporal control over Notch activation. To specifically dissect the function of Notch activity in ameloblast (AM) maturation, we utilized our soluble, computationally designed Notch agonist C3-DLL4 scaffold.²³ DLL4 was selected as the ligand because it produced the strongest and most reproducible activation of Notch signaling among the canonical ligands evaluated during scaffold development, including DLL1 and Jagged1. This choice is consistent with prior structural and biophysical studies demonstrating that DLL4 binds Notch receptors with higher affinity and can support activation under lower mechanical tension than Jagged ligands,³⁴ and with quantitative analyzes showing that DLL4 signaling is largely independent of Lunatic Fringe–dependent glycosylation, unlike DLL1 and Jagged1.³⁵ These properties made DLL4 the most reliable ligand for constructing a geometry-defined soluble Notch agonist.

The Notch activator scaffold (C3-DLL4) is a computationally designed^36–38 multivalent, soluble protein complex capable of activating the Notch pathway (Fig. 2a).²³ This synthetic system features an engineered oligomeric structure^39–41 composed of three repeating subunits conjugated to the Delta-like 4 extracellular domain (DLL4). By arranging the DLL4 ectodomain in a trimeric scaffold, soluble C3-DLL4 mimics the natural ligand presentation required for Notch activation.²³ Specifically, C3-DLL4 configuration promotes interaction between neighboring cell surfaces, providing the necessary tethering and mechanical tension to activate the Notch pathway. Importantly, C3-DLL4 overcomes the limitations of generating mechanical tension by traditional immobilization methods, enabling effective Notch signaling by soluble ligand.

Fig. 2.

AI-designed Notch activator C3-DLL4 induces ameloblast maturation. a Structural model of the computationally designed C3-DLL4 protein complex, showing its trivalent configuration (C3(V = 3)), which clusters Notch receptors to simulate ligand-receptor interactions in cell communication. b Notch1-Gal4 reporter assay in U2OS cells treated with control or C3-DLL4, showing nuclear mCitrine expression exclusively in C3-DLL4-treated cells, confirming Notch pathway activation. Scale bar = 100 μm. c Quantification of mCitrine fluorescence intensity per cell, revealing a significant increase in Notch activation upon C3-DLL4 treatment (^****P < 0.000 1) compared to control. d Schematic of the experimental setup, illustrating the differentiation of iOBs and iAMs, followed by co-culture with or without C3-DLL4 and DAPT treatment to assess Notch activation in ameloblast maturation. e Immunofluorescence staining of iAM organoids treated with Control (top), C3-DLL4 (middle), and DAPT (bottom), showing ENAM (red) and MMP20 (red) expression. Merged images include DAPI (nuclei, blue), DSPP/DMP1 (odontoblast markers, green), ENAM (red), and Phalloidin (cytoskeleton, purple). Scale bar = 100 μm. f Quantification of ENAM fluorescence intensity, normalized to control, demonstrating a significant increase in ENAM expression in C3-DLL4-treated organoids (^***P < 0.001) and a significant reduction in DAPT-treated organoids (^**P < 0.01). g Quantification of MMP20 fluorescence intensity, showing a significant increase in C3-DLL4-treated organoids (^*P < 0.05) and a marked reduction in DAPT-treated organoids (^*P < 0.05)

To evaluate the effectiveness of the conjugated C3-DLL4 in activating Notch signaling, we used the established Notch1-Gal4 reporter system,⁴² which minimizes interference from endogenous Notch signals by replacing the NOTCH1 ankyrin repeat domain with Gal4 transcription activator. Upon binding to Delta, Notch receptor proteolytic processing releases a chimeric NICD1-Gal4 protein that can activate the engineered UAS-mCitrine gene. We showed that C3-DLL4 led to significant nuclear mCitrine expression, confirming Notch activation (Fig. 2b, c). Control cells treated with doxycycline did not display nuclear mCitrine, confirming that reporter activation was dependent on C3-DLL4 stimulation. This demonstrated the effectiveness of soluble, computer-designed C3-DLL4 to activate Notch receptor.

Notch activation accelerates isAM maturation in co-culture system

We applied C3-DLL4 in our co-culture system, to assess if Notch activation could accelerate induced ameloblast maturation to isAM stage. We differentiated iOB and iAM independently, combined the cultures and treated the co-cultured cells with C3-DLL4 (Figs. S1a and 2d). Treatment with C3-DLL4 enhanced the expression of mature EMP markers, including ENAM and MMP20, indicating that active Notch signaling promotes isAM maturation (Fig. 2e–g). These findings suggest that C3-DLL4-mediated Notch activation effectively enhances AM maturation in the co-culture model.

C3-DLL4 accelerates iAM maturation without co-culture

We further investigated if C3-DLL4 could accelerate iAM maturation without close proximity to iOB. Given that our initial results identified Notch as a candidate mediator of ameloblast maturation, particularly driven by interactions with iOB-derived Delta ligands, we hypothesized that direct activation of Notch in iAM would bypass the need for OB co-culture conditions. To assess this hypothesis, we generated early ameloblasts (ieAM) in 3D suspension cultures enabling self-organization into ieAM organoids (Fig. 3a). This 3D environment supports early ameloblast maturation, mimicking the spatial context of natural tooth development. We applied C3-DLL4 to ieAM organoids and cultured them for an additional 7 days to evaluate long-term effects on maturation (Fig. S2a).

Fig. 3.

Designed notch activator C3-DLL4 accelerates ameloblast maturation independent of odontoblasts. a Schematic representation of the experimental setup showing induced early ameloblast (ieAM) organoid formation from human iPSCs. ieAM organoids are treated with C3-DLL4 (Notch activator) or DAPT (Notch inhibitor), leading to their differentiation into induced secretory ameloblasts (isAM). b Heatmap of z-score normalized expression of stage-specific marker genes across ameloblast subtypes in response to soluble Notch agonist treatment. Gene expression profiles are shown for cells treated with C3-DLL4 (Notch agonist; C3-DLL4⁺) versus untreated controls (C3-DLL4⁻). Columns represent experimental conditions, and rows represent marker genes grouped by ameloblast differentiation stage: pre-ameloblasts (pA, green), early ameloblasts (eAM, pink), secretory ameloblasts (sAM, magenta), and mature ameloblasts (mAM, yellow). c Immunofluorescence staining of ieAM-derived organoids treated with Control (top), C3-DLL4 (middle), and DAPT (bottom), showing ENAM (yellow) and ZO-1 (green) expression. Merged images also include DAPI (nuclei, blue) and Phalloidin (cytoskeleton, red). Scale bar = 10 μm. d Quantification of ENAM fluorescence intensity, normalized to control, showing a significant increase in C3-DLL4-treated organoids (^****P < 0.000 1, (^***P < 0.001), and (^**P < 0.01) compared to control, while DAPT treatment significantly reduces ENAM expression. e Immunofluorescence staining of MMP20 (yellow) and ZO-1 (green) in control, C3-DLL4-, and DAPT-treated organoids. Scale bar = 10 μm. f Quantification of MMP20 fluorescence intensity, demonstrating a significant increase in C3-DLL4-treated organoids (^***P < 0.001) and a significant reduction in DAPT-treated organoids (^**P < 0.01). g Schematic summary illustrating the transition from ieAM to isAM organoid upon C3-DLL4 treatment, with the upregulation of ENAM and MMP20 and the establishment of an organized, polarized structure

RNA sequencing of ieAM organoids treated with or without the soluble Notch agonist C3-DLL4 revealed a clear shift in transcriptional profiles indicative of advanced ameloblast maturation. In the absence of Notch activation, organoids predominantly expressed early ameloblast (eAM) or undifferentiated epithelial markers such as COL11A2, SERPINF1, ARHGAP6, and TPI1, reflecting a stalled or immature transcriptional state. By contrast, C3-DLL4 treatment induced a dramatic transcriptional transition, characterized by the upregulation of secretory ameloblast (sAM) markers, including AMELX, ENAM, AMTN, and MMP20. Additionally, mature ameloblast (mAM) genes such as ODAM, KLK4, FAM83H, TUFT1, and STIM1 were robustly elevated, indicating progression to a more functionally differentiated state. This transcriptomic shift, visualized in the heatmap (Fig. 3b), confirms that Notch activation via C3-DLL4 is sufficient to drive ameloblast maturation independent of odontoblast-derived signals and supports the functional transition of ieAM organoids to the mature secretory stage.

To further relate our in vitro organoid findings to in vivo ameloblast states, we compared our bulk RNA-seq profiles to single-cell transcriptomes from developing mouse dental epithelium, as our human fetal single-cell dataset lacks maturation-stage ameloblasts (Fig. S3a, b). Pearson correlation analysis revealed that C3-DLL4–treated organoids correlated strongly with secretory and maturation-stage ameloblasts, whereas untreated organoids correlated primarily with early-stage ameloblasts (Fig. S3c). In parallel, pseudotime analysis of the mouse dental epithelium confirmed the sequential expression of hallmark ameloblast genes—Dspp in eAM, Enam in sAM, and Klk4 in mAM—mirroring the transcriptional trajectory observed in our human organoids (Fig. S3d). These cross-species comparisons reinforce that Notch activation in 3D ieAM organoids drives their progression toward a transcriptional and phenotypic profile characteristic of mature secretory ameloblasts.

After treatment of 3D ieAM organoids with the soluble Notch agonist C3-DLL4, we observed a robust upregulation of secretory ameloblast protein markers, including ENAM, MMP20, and AMELX (Figs. 3c–f and S4a, b). This upregulation is indicative of the transition of ieAM organoids to a more mature ameloblast state. Notably, the C3-DLL4-treated organoids also displayed enhanced cellular organization and polarization, which are characteristic features of ameloblast maturation in vivo. Since the ameloblast maturation to isAM stage was achieved in the absence of odontoblasts, we conclude that the C3-DLL4 scaffold-based activation of Notch is sufficient to mature ameloblasts. Furthermore, these data show that soluble C3-DLL4 can substitute for the natural Delta ligand typically provided by iOBs.

Conversely, to evaluate if Notch signaling is essential for iAM maturation in the absence of iOB co-culture, we employed again DAPT, a γ-secretase inhibitor³⁰ that prevents Notch receptor activation by blocking the release of the NICD. Treatment of isolated ieAM organoids with DAPT led to a marked reduction in the expression of key ameloblast maturation markers, such as ENAM, MMP20 and AMELX (Fig. 3c–f), indicating that autonomous Notch activation is sufficient for AM progression to a mature ameloblast state. Additionally, organoids exposed to DAPT did not display altered cell organization and polarity, further supporting the idea that active Notch signaling is essential for proper ameloblast maturation to secretory and mature stage (Fig. 3c–e).

Overall, our findings show that the soluble Notch agonist C3-DLL4 can accelerate the differentiation of iAM organoids into a more advanced ameloblast state, which we define as induced secretory mature ameloblasts (ismAM) (Fig. 3g). This ismAM state is characterized by high expression of key secretory and maturation markers AMELX, ENAM, MMP20, ODAM, KLK4, FAM83H, TUFT1, STIM1 and the acquisition of epithelial polarity and organization, all achieved in a defined culture condition without the need for odontoblast (OB) co-culture (Fig. 3f). These results indicate that activation of Notch signaling via soluble C3-DLL4 can recapitulate key molecular and structural features of OB-induced ameloblast maturation, suggesting that Notch-mediated signaling represents an important, but not necessarily exclusive, component of the odontoblast contribution to ameloblast development.

C3-DLL4–treated AM-organoids form enamel-like mineralized material in the kidney capsule

Since our sequencing analysis (Fig. 3b) showed that C3-DLL4–treated induced ameloblast organoids expressed key maturation markers such as KLK4, ODAM, TUFT1, FAM83H and ODAPH — a hallmark of mature ameloblasts associated with enhanced calcification—we next sought to determine whether these organoids could induce calcification and mineralization in vivo. We addressed this by transplanting the organoids under the kidney capsule of NOD-SCID mice, a well-vascularized site that supports graft growth and normally lacks calcification (Fig. 4a). MicroCT analysis revealed the presence of calcified material at the transplantation site, as shown in representative microCT slices and 3D reconstructions (Fig. 4b, c).

Fig. 4.

Transplanted isAM/ismAM organoids generate calcified enamel-like structures under the kidney capsule in vivo. a Schematic of the in vivo transplantation workflow. Induced secretory or maturation-stage ameloblast (isAM/ismAM) organoids were transplanted under the kidney capsule of immunocompromised mice. After 3 weeks, the kidneys were harvested and processed for micro-computed tomography (μCT) and histological analyses. b Representative 2D microCT scan image through a harvested kidney showing localized high-density mineralized foci within the graft (top: scale bar, 5 mm; bottom inset: scale bar, 1 mm). c 3D reconstruction of the kidney graft reveals discrete mineralized nodules within the transplanted tissue (rendered in blue). d Left: Immunofluorescence staining for human nuclear antigen (HuNu, green) and phalloidin (F-actin, red) highlights the transplanted region, showing clusters of human-derived cells with organized cytoskeletal polarity. Right: Masson’s Trichrome staining of the same grafted area reveals dense collagen deposition and structured epithelial organization surrounding the mineralized regions. Scale bars: 50 μm. e Immunofluorescence for HuNu (green), enamelin (ENAM, yellow), and phalloidin (F-actin, red) shows human-derived cells producing enamel matrix proteins within organized epithelial structures. The zoomed-out view (right) highlights the grafted region containing ENAM⁺ HuNu⁺ cells, and the zoomed-in inset (red box) reveals a central rosette with polarized actin and secretory morphology. Scale bars: zoomed-in, 10 μm; zoomed-out, 50 μm. f Co-staining for WDR72 (yellow), ZO1 (green), and phalloidin (red) demonstrates a secretory-stage ameloblast-like rosette structure. The wider field (right) shows the grafted epithelial zone, while the magnified inset (yellow box) highlights apical junctions and F-actin polarization surrounding a central lumen. Scale bars: zoomed-in, 10 μm; zoomed-out, 50 μm. g High-resolution imaging of WDR72⁺ ZO1⁺ cells with apical actin enrichment reveals a columnar epithelial architecture consistent with mature secretory ameloblast morphology. The zoomed-out view (right) provides spatial context of the graft, and the inset (red box) emphasizes the polarized alignment of ameloblast-like cells at the mineralizing front. Scale bars: zoomed-in, 20 μm; zoomed-out, 50 μm

The presence of human-derived organoids within the kidney capsule was confirmed by human nuclear antigen (HuNu) immunostaining, H&E (Fig. S5a, b) and Masson’s trichrome staining (Figs. 4d and 5Sb”’), with calcification further validated by Alizarin Red and von Kossa staining (Fig. S5a’-S5a”, S5b’-S5b”). Immunofluorescence demonstrated that the transplanted organoids maintained polarized epithelial structures and continued to express key ameloblast markers, including ENAM and WDR72 (Fig. 4e–g), as well as KLK4 and MMP20 (Fig. S5c, d).

Together, these data indicate that Notch-activated mature ameloblast organoids are capable of secreting enamel-like calcified material in vivo, representing an initial step toward functional enamel tissue formation.

DLX3 transcription factor is required cell autonomously for human ameloblast maturation

Mutations in the gene encoding a developmental transcription factor DLX3 cause amelogenesis imperfecta.^43–45 Interestingly, a GWAS identified a significant association between a SNP in DLX3 upstream region and Caries prevalence (Fig. S6a).⁶ Our analysis in developing human tissue revealed that DLX3 is expressed both in ameloblast and odontoblast lineage (Fig. S6b). Similar expression patterns can be observed in mouse dental tissue (Fig. S6c). These findings suggest that SNP-related defects in DLX3 expression either in odontoblasts or in ameloblasts could reduce enamel- or dentin-protein secretion, weaken the protective layers and increase susceptibility to caries during the individual’s lifetime. However, while it is known that DLX3 transcription factor is required to regulate the expression of DSPP in odontoblast differentiation, it is not clear today if, and at what stage, DLX3 is cell-autonomously essential for human ameloblast differentiation.

Since activation of Notch with designed proteins can generate mature AMs without OB, we now tested if DLX3 is autonomously required in the ameloblast lineage. We generated two independent DLX3 knockout (KO) mutant iPSC lines (KO-10 and KO-13) using CRISPR-Cas9 gene-editing technology (Fig. S6e). DLX3 protein includes transcriptional activation (TA) domains and the homeodomain. We generated stop codons prior to or on the homeodomain sequence resulting in lack of the key homeodomain helix 3 that facilitates the interaction with target DNA (Fig. S6e, f).

Since DLX3 is known to bind DSPP promoter and control DSPP expression in mature OB,² we differentiated wild-type and DLX3 mutant iPSC to neural crest cells (Fig. S6g–i) and subsequently into mature iOB using our previous protocol that activates the FGFR1/2c pathway.²⁶ In this study, however, we additionally introduced a Notch agonist in a novel serum-free (SFM) odontogenic medium (Fig. 5d). The previous analysis of human fetal single cell sequencing data predicted Notch pathway function in POB to OB differentiation/maturation (Fig. 5a).²⁶ To evaluate the expression of Notch ligands and receptors in pre-ameloblast (PA) and pre-odontoblast (POB), we analyzed the gene expression levels of Notch and Delta in these cell types. We generated a heatmap by averaging the expression levels of PA and POB clusters from the single-cell data at the late timepoints during human fetal development (20-22gw)⁷ (File S1). We found that the ligands (DLL1 and JAG1) were predominantly expressed in the PA, while the Notch receptors (NOTCH1, NOTCH2, and NOTCH3) were primarily present on the POB side (Fig. 5b) (File S1). A similar pattern with minor differences was observed in previously curated single-cell mouse data²⁸ (Fig. S6d), reinforcing the human findings. These cross-species comparisons highlighted conserved Notch–Delta signaling between pre-ameloblasts and odontoblast precursors, prompting us to test whether engineered Notch activation enhances iOB maturation in vitro. We therefore tested and showed that iOB maturation measured by DSPP expression was markedly increased with Notch activation using designed Notch agonist C3-DLL4 (Fig. 5d–h). With this matured iOB differentiation assay we observed a dramatic reduction of DSPP in DLX3 mutant OB (compared to wild type), both on protein and RNA level (Fig. 5e–h). These data confirm that our DLX3 mutants behave as expected showing significantly reduced expression of DSPP in OB lineage. This further confirms that DLX3 can activate DSPP transcription cell autonomously in mature OB.

Fig. 5.

DLX3 is required for odontoblast maturation and DSPP expression. a TopPath pathway analysis of ligand–receptor interactions between pre-odontoblasts (POBs) and odontoblasts (OBs) highlights Notch signaling. b Heatmap of z-score normalized expression shows Notch ligands (DLL1, DLL3, JAG1, JAG2) in pre-ameloblasts (PA) and receptors (NOTCH1–4) in POBs. c Distribution of Notch ligand–receptor interactions demonstrate ligands derived from PA and receptors from POBs. d Timeline illustrates odontoblast differentiation from WT and DLX3 KO iPSCs into induced neural crest cells (iNCs), POBs, and induced odontoblasts (iOBs). e Western blot analysis of DSPP, DLX3, and ACTIN is shown in WT and DLX3 KO iOBs treated with C3-DLL4. f Immunofluorescence staining of DSPP (green), Vimentin (red), and DAPI (blue) is shown in WT and DLX3 KO iOBs treated with C3-DLL4. g Quantification of DSPP fluorescence intensity normalized to WT control is presented (^**P < 0.01, ^***P < 0.001, ^****P < 0.000 1). qPCR analysis of DSPP expression normalized to β-Actin is shown (^**P < 0.01) (h)

Furthermore, to dissect the DLX3 function in Ameloblasts, we differentiated the DLX3 knockout (KO) iPSC lines into first induced early ameloblasts (ieAMs) in 2D model and then isAM in 3D organoid model using Notch activator, C3-DLL4 (Fig. 6a). Quantitative comparisons using Western blot analysis between differentiated DLX3 KO and wild type-iAM D16 cells revealed that the absence of DLX3 did not affect the expression of early ameloblast markers, including AMBN, SP6, or transient DSPP expression (Fig. 6b). These findings suggest that DLX3 is not required for the early ameloblast differentiation or regulation of DSPP transcription in ameloblasts. Similarly, when tested in matured isAM organoid system, the immunostaining analysis of ZO-1 revealed a normal cellular polarity of DLX3 mutant ameloblasts, with apical lumen marked with ZO-1 (Fig. 6d–f).

Fig. 6.

DLX3 is required for ameloblast terminal maturation and enamel matrix protein expression. a Schematic representation of the experimental setup, where wild-type (WT) or DLX3 knockout (KO) iPSCs are differentiated into induced early ameloblast (ieAM) organoids and further matured into induced secretory ameloblasts (isAM) organoids upon C3-DLL4 treatment. b Western blot analysis of AMBN, DLX3, SP6, H3, and DSPP protein levels in WT and DLX3 KO iAM organoids at day 16. c Heatmap of z-score normalized expression of stage-specific marker genes across ameloblast subtypes in response to soluble Notch agonist treatment. Gene expression profiles are shown for cells treated with Notch agonist; C3-DLL4⁺ control versus DLX3 KO. Columns represent experimental conditions, and rows represent marker genes grouped by ameloblast differentiation stage: pre-ameloblasts (pA, green), early ameloblasts (eAM, pink), secretory ameloblasts (sAM, magenta), and mature ameloblasts (mAM, yellow). d Immunofluorescence staining of ENAM (green), ZO-1 (red), Phalloidin (yellow), and DAPI (blue) in WT and DLX3 KO organoids (scale bar, 50 µm). e Quantification of mean ENAM fluorescence intensity per organoid in WT control versus DLX3 KO organoids (^*P < 0.05). f Immunofluorescence staining of MMP20 (green), ZO-1 (red), Phalloidin (yellow), and DAPI (blue) in WT and DLX3 KO organoids (scale bar, 50 µm). g Quantification of mean MMP20 fluorescence intensity per organoid in WT control versus DLX3 KO organoids (^****P < 0.000 1). h Western blot analysis of ENAM, WDR72, TUFT1, MMP20, AMELX, and KLK4 in WT and DLX3 KO organoids. i Quantification of mean WDR72 fluorescence intensity per organoid in WT control versus DLX3 KO-13 organoids (^***P < 0.001)

Having shown that DLX3 loss does not affect early ameloblast specification or polarity, we next asked whether DLX3 is required for terminal ameloblast maturation (Fig. 6a, c–f). We differentiated both wild-type (WT) and DLX3 KO iPSCs into ieAM organoids, followed by treatment with the engineered Notch agonist C3-DLL4. After 14 days of maturation, we harvested the resulting ismAM organoids for molecular and transcriptomic analysis (Fig. 6a). RNA-sequencing data revealed clear transcriptional divergence among the two groups treated with Notch agonists: control and DLX3 KO AM organoids, indicating that DLX3 loss selectively disrupts the transcriptional program associated with terminal ameloblast differentiation. To better understand the specific transcriptional transitions being impaired, we performed hierarchical clustering of genes based on previously annotated ameloblast lineage stages: pre-ameloblast (pA), early ameloblast (eAM), secretory ameloblast (sAM), and mature ameloblast (mAM) (Fig. 6c), as previously done for Fig. 3b. In WT isAM organoids, C3-DLL4 treatment successfully induced a subset of secretory-stage (sAM) genes, reflecting a transition toward a polarized, matrix-secreting state. In contrast, DLX3 KO organoids failed to upregulate most sAM- and mAM-enriched genes, instead maintaining a gene expression profile skewed toward PA and eAM identities. Genes critical for enamel matrix secretion and processing such as AMELX, MMP20, WDR72, TUFT1, and KLK4 were among the most suppressed in the KO group. This transcriptional arrest highlights that DLX3 function is required during AM maturation process. Notch signaling through C3-DLL4 that drives ameloblast maturation requires functional Dlx3, suggesting that Dlx3 acts prior to Notch during AM maturation process for the full complement of enamel-specific genes (Fig. 6c).

Gene Ontology (GO) enrichment analysis further supported these findings: WT isAM organoids displayed enrichment for biological processes associated with amelogenesis, biomineralization, and matrix secretion, while DLX3 KO organoids showed overrepresentation of more generic developmental pathways such as epithelial morphogenesis and organ formation, suggesting a failure to transition into a specialized enamel-producing state (Fig. S7c).

We then validated one of the top differentially expressed maturation markers, WDR72, which is essential for vesicle trafficking and matrix mineralization.⁴⁶ Immunofluorescence analysis of human fetal teeth confirmed that WDR72 is strongly expressed apically in mature ameloblasts, where it co-localizes with ZO-1 (Fig. S7a). Extending this to our organoid system, we demonstrate for the first time strong apical expression of WDR72 in WT ismAM organoids, co-localizing with ZO-1, a marker of apical tight junctions. In contrast, DLX3 KO-13 organoids showed dramatically reduced WDR72 expression, although epithelial polarity (as indicated by ZO-1 localization) remained intact (Fig. S7b; Fig. 6i Quantification). Decrease in WDR72 intensity in DLX3 KO samples (Fig. 6i), was consistent with defective ameloblast maturation.

To assess the broader impact of DLX3 loss on EMP production, we performed Western blot analysis for a panel of key maturation markers. In WT isAM organoids, we observed robust expression of ENAM, WDR72, MMP20, AMELX, TUFT1, and KLK4. In contrast, DLX3 KO isAM organoids showed a nearly complete loss of these proteins, confirming that DLX3 is required for activating the protein-level machinery of enamel secretion (Fig. 6h).

These data show that DLX3 is a cell-autonomous transcriptional regulator essential for terminal ameloblast maturation, acting in concert with Notch to complete enamel matrix production even without odontoblast-derived cues (Fig. 7).

Fig. 7.

Function of Notch and DLX3 in the ameloblast maturation. This schematic summarizes the stepwise maturation of induced ameloblast (iAM) organoids, highlighting the requirement of both Notch signaling and DLX3 activity for terminal differentiation. Early iAM (ieAM) organoids express markers of ameloblast lineage commitment (AMBN, DSPP, SP6, ZO-1) but lack secretory function. Notch pathway activation drives the transition to secretory iAMs (isAM), characterized by expression of enamel matrix proteins (ENAM, MMP20, AMELX). Full maturation into induced secretory mature ameloblasts (ismAM) requires DLX3 activity along with Notch activation and is marked by expression of terminal differentiation genes (WDR72, KLK4, TUFT1). The inset depicts co-localization of apical junction protein ZO-1 with WDR72, indicating establishment of epithelial polarity and terminal ameloblast functionality. The pseudotime trajectory, supported by mouse tooth single-cell data, illustrates sequential upregulation of stage-specific markers, including Dspp in early ameloblasts (eAM), Enam in secretory ameloblasts (sAM), and Wdr72 in mature ameloblasts (mAM). Together, the model emphasizes that Notch signaling is essential for progression to the secretory stage, while DLX3 is required for terminal ameloblast maturation and enamel matrix secretion. Our findings provide a tractable system to identify GWAS-nominated variants (SNPs) with functional effects in ameloblasts, accelerating translation toward therapies for dental caries

Discussion

We found that Notch signaling mediates key aspects of the interaction between odontoblasts (OBs) and ameloblasts (AMs) that promote AM maturation. Earlier developmental studies have demonstrated complementary expression of Notch receptors and ligands in the developing tooth, consistent with epithelial–mesenchymal signaling between odontoblasts and ameloblasts.⁴⁷ Our high-resolution single-cell analyzes extend these findings by resolving ligand–receptor expression at the individual cell-type level in both human and mouse fetal teeth, revealing conserved yet distinct distributions compared to earlier histological data. Specifically, we observe DLL1 and DLL4 enriched in odontoblasts and NOTCH1–3 in ameloblasts, suggesting a refined model of Delta–Notch signaling polarity during ameloblast maturation. To test whether Notch pathway activation is necessary and sufficient for AM maturation, we utilized small molecules and novel AI-designed (C3-DLL4) protein scaffold. Our results indicate that Notch inhibition (using DAPT) affects enamelin (ENAM) secretion in co-culture, demonstrating that Notch is required for AM maturation. Next, we utilized our designed soluble Notch scaffold that can activate Notch processing presumably due to the forces generated by the scaffold interacting with Notch in the surface of neighboring cells.²³ Notch activation with this soluble agonist enhanced the expression of mature isAM markers, ENAM, MMP20 and AMELX, even in the absence of co-culture with OB cells. These data suggest that Notch pathway activation is critical for AM maturation, and the designed Delta-scaffold can substitute for OB in this process. This newly developed maturation paradigm allowed us to reveal that the Amelogenesis Imperfecta gene DLX3 is required not only in OB but also in the AM cell lineage. Specifically, DLX3 is not necessary for the pre-ameloblast stage, but it is essential for the expression of ENAM and MMP20 during the ameloblast maturation stage.

In this paper, we utilized a designed scaffold that leverages the natural ligand-receptor interaction mechanism of Notch pathway by incorporating engineered extracellular domain of DLL4 tethered to computer designed scaffold protein. This design ensures stable ligand presentation and facilitates efficient engagement with Notch receptors on target cells. Moreover, the scaffold allows for precise spatial positioning and temporal control over Notch activation. By mimicking native ligand dynamics in a controlled environment, this system provides unparalleled flexibility to dissect the nuances of Notch signaling. Importantly, it enabled investigation of how spatially and temporally regulated Notch signaling orchestrates the complex processes underlying ameloblast differentiation and maturation.

Potential crosstalk between ameloblasts and odontoblasts has been proposed previously. Although an intact basement membrane typically separates the two cell types, this barrier disintegrates at the onset of pre-dentin secretion, providing a narrow time frame for direct cell–cell interactions.⁴² We have now revealed a plausible mechanism for this interaction that involve Notch ligands (e.g., Jagged or Delta), which may become accessible to ameloblast processes upon basement membrane breakdown. Further investigation of this Notch signaling pathway, particularly through advanced imaging techniques, could yield critical insights into how odontoblasts communicate with ameloblasts to coordinate dentin and enamel formation.

Our human iAM organoid platform, combined with our engineered Notch activation scaffold, enabled us to dissect DLX3’s role independently in ameloblasts and odontoblasts. We demonstrate that DLX3 is required cell-autonomously in human ameloblasts during terminal maturation, where it drives the expression of key enamel matrix genes, including ENAM, MMP20, and WDR72. This discrepancy between mouse^24,25 and human findings may reflect species-specific regulatory differences, timing of DLX3 activity, or compensatory effects by other Dlx family members in the mouse. By bypassing the requirement for odontoblast co-culture, our model complements previous in vivo studies and establishes a uniquely human-specific framework to define DLX3’s mechanisms in enamel formation, motivating future work to define its direct targets in AM.

In vivo, polarized Notch signaling plays an essential role in coordinating epithelial organization and enamel secretion, as demonstrated by ADAM10-dependent regulation of ameloblast polarity and stratum intermedium integrity.⁴⁸ The ability of our soluble C3-DLL4 scaffold to activate Notch signaling in a controlled and tunable manner provides a foundation to further explore how spatially localized Notch activation contributes to enamel patterning and polarity. Future iterations of this system could be adapted for matrix-tethered or spatially patterned DLL4 presentation, enabling mechanistic studies of polarized Notch signaling and advancing its translational potential for enamel regeneration.

Our innovative protein design approach to activate Notch has yielded critical insights into the molecular mechanisms of Notch signaling during enamel organ development and facilitated the generation of a more advanced platform. This work deepens our understanding of Amelogenesis Imperfecta and lays the groundwork for future research aimed at developing therapies for one of the world’s most prevalent diseases—dental caries (tooth decay)—the most common chronic condition in children.

Materials and methods

iPSC derived ameloblast differentiation

Briefly, hiPSCs (WTC-11 human induced pluripotent stem cells) (Coriell, #GM25256) were seeded on 12-well plates coated with growth factor-reduced Matrigel (Corning, #356231) and cultured in mTeSR medium (StemCell Technologies, #85850) until cells reach confluency with medium changes daily. On the first day of differentiation (Day 0), stem cell media is replaced with base iAM media consisting of EpiCult-C media (StemCell Technologies, #05630) supplemented with 1× EpiCult-C Proliferation Supplement (STEMCELL Technologies), 0.1 μmol/L β-mercaptoethanol (BME) (Sigma, #M7522), 1× Hydrocortisone Stock Solution (96 μg/mL), 0.05X GlutaMAX (Gibco), 0.4× NEAA (Gibco) and 1× Penicillin-Streptomycin (Gibco) and 400 nmol/L smoothened agonist (SAG) (Selleckchem, # S7779). On day 3 of differentiation 150 pmol/L of monomeric bone morphogenic protein-4 (BMP4) (rndsystems, #314-BP-010) is added daily until day 7. At day 8, the base media is supplemented with 1 μmol/L of BMP-I inhibitor (LDN-193189) (Tocris, #6053), 5 μmol/L of GSK3-Inhibitor (CHIR99021) (Selleckchem, #4423), 500 pmol/L epidermal growth factor (EGF) (rndsystems, #236-EG) and 70 pmol/L of Neurotrophin-4 (NT4) (rndsystems, #268-N4). No media refresh is added on day 9. On day 10, repeat with the same small molecules used on day 8. From day 12 onwards the cultures were then extended until day 16 by adding 1× N2 Supplement (Gibco #17502-048), 400 nmol/L SAG, 300 pmol/L BMP4, 5 μmol/L of GSK3-Inhibitor (CHIR99021), 500 pmol/L of EGF 2 nmol/L transforming growth factor beta 1(TGFβ1) (rndsystems, #7754-BH) 70 ppmol/L of Neurotrophin-4 (NT4) (rndsystems, #268-N4) and 300 pmol/L of Activin (Peprotech, #120-14 P) and for the early ameloblast stage at day 16. The media containing the small molecules were changed every other day from day 12 until day 16.

Production and characterization of the C3-DLL4 notch activator

The C3-DLL4 Notch activator was produced using a previously established computationally designed scaffold with defined valency and geometry.²³ The C3 scaffold, a homotrimeric helical bundle, was conjugated to DLL4 using SpyLigation, which forms a stable isopeptide bond between SpyCatcher (SC) and SpyTag (ST). The C3-SpyCatcher fusion was expressed in Escherichia coli, while DLL4-SpyTag was produced in mammalian Expi293F cells, ensuring proper folding and post-translational modifications. Both components were purified using affinity chromatography, and successful conjugation was confirmed by SDS-PAGE (sodium dodecyl sulfate–polyacrylamide gel electrophoresis) and Coomassie staining, showing a higher molecular weight band corresponding to the assembled C3-DLL4 complex. For cell treatment, a 10 µM stock solution of C3-DLL4 was prepared by conjugating DLL4-ST with C3-SpyCatcher, followed by serial dilutions to achieve precise ligand presentation and receptor engagement.

Notch activation assay in reporter cell line

The Notch activation assay was performed as described before²³ using U2OS-N1-Gal4/UAS-H2B-mCitrine cells^49,50 to evaluate the effect of C3-DLL4 on Notch signaling. Cells were seeded in a 96-well tissue culture-treated plate at a density of 15 000 cells per well in DMEM supplemented with 10% FBS, 1× Penicillin-Streptomycin, 0.5X NEAA (Gibco), 1× GlutaMax, and 1× Sodium Pyruvate. Doxycycline (Dox, 1 µg/mL) was added at seeding to induce Notch1-Gal4 (N1-Gal4) expression, which is essential for the assay. On Day 1, cells were treated with C3-DLL4. Through serial dilutions of C3-DLL4 the final concentrations of 200, 100, 50, 10 and 1 nmol/L (DLL4 molar equivalent) were added to respective wells in replicates. Untreated wells served as negative controls. After treatment, the cells were gently mixed and incubated under standard culture conditions at 37 °C with 5% CO₂. On Day 2, the media was replaced with fresh Dox-containing media to maintain Notch induction. On Day 4, the Notch activation was assessed by measuring H2B-mCitrine fluorescence. Fluorescence intensity was quantified to determine the dose-dependent activation of Notch signaling. Control wells without C3-DLL4 were analyzed to establish baseline fluorescence, ensuring accurate interpretation of Notch activation levels.

iPSC derived odontoblast differentiation

We modified the odontoblast differentiation protocol previously described by our lab^8,26 to avoid serum usage and developed serum-free protocol for the OB differentiation. This differentiation protocol begins with hiPSCs, treated with dual SMAD inhibitors, SB431542 and LDN-193189 (Tocris, #6053), to inhibit TGF-β and BMP signaling pathways, respectively. This promotes ectodermal lineage commitment, guiding the hiPSCs toward an odontogenic trajectory. From Day 0 to Day 11, the WNT signaling pathway is activated using CHIR99021, a GSK-3β inhibitor, to direct mesenchymal lineage specification and to facilitate the generation of neural crest cells (iNCs), a precursor population essential for odontoblast development. On Day 12, BMP4 is introduced to drive the differentiation of iNCs into odontoblast precursors in serum-free conditions.

Specifically, p75 + iNC cells were cultured in Serum-free (SFM) Odontogenic Medium, consisting of DMEM + Glutamax (Gibco 10566016), 100 nmol/L dexamethasone (Sigma-Aldrich D4902), 15% KnockOut Serum Replacement (KOSR), 0.5X NEAA (Gibco), 0.005 mM ITS-A (Gibco 51300-044), 5 mmol/L β-glycerophosphate (Sigma-Aldrich G9422), and 50 μg/mL L-ascorbic acid (Sigma-Aldrich #A4544) for 14 days (OB). Odontogenic Medium was supplemented with 50 ng/mL BMP4 (Stemcell Technologies #78211) for Day 11 to Day 18, followed by 25 ng/mL BMP4 (Stemcell Technologies #78211) and 400 nmol/L SAG (Stemcell Technologies #73412) from Day 18 until Day 25 supplemented with 100 ng/mL C6^26,29 for 14 days (iOB C6); followed by treatment with or without Notch activator (50 nmol/L C3-DLL4)²⁶ (iOB C6 N) on Day 18 for 24 h followed by change of media on Day 19 followed until Day 25. All cultures were performed on Matrigel-coated plates at a 1:30 dilution and incubated at 37 °C with 5% CO₂. Each differentiation was performed in triplicate, with undifferentiated hiPSC as the negative control.

Establishing a functional co-culture model for ameloblast and odontoblast

hiPSCs were differentiated into ameloblast and odontoblast lineages to establish a functional co-culture model for ameloblasts and odontoblast lineages under defined serum-free conditions. For ameloblast differentiation, hiPSCs were cultured in a 2D monolayer to generate early ameloblasts (ieAM) from Day 0 to Day 16. On Day 16, differentiated iAM cells were trypsinized using TrypLE Express (Thermo Scientific or Gibco, #12604013). The cells were gently resuspended and then transferred into 24-well ultra-low attachment plates for three-dimensional (3D) suspension culture in Epicult Plus medium (Stemcell Technologies #06070), allowing them to self-organize into ieAM organoids. The ieAM organoid cultures were incubated at 37 °C with 5% CO₂, and the media were changed every other day. These organoids were cultured until the next 7 days, stabilizing their maturation. Concurrently, odontoblast cells were made similarly in a separate plate by culturing iOBs differentiated from iNCs (neural crest) that were differentiated from hiPSCs in a serum-free odontogenic differentiation medium. The differentiated iOB cells were plated as monolayer mixed in 25% (v/v) of Matrigel (Corning, #356231) diluted in odontogenic media in a glass-bottomed 24-well plate (Corning, #3603). The next day, ieAM organoids suspended in the ameloblast base medium and 10 μM ROCKi (Y-27632, Selleckchem, #S1049) were added on top of the iOB monolayer and then incubated for 24 h at 37 °C in 5% CO₂. The co-culture was supplemented with fresh media (1:1 mixture of ameloblast and odontogenic media) every three consecutive days. The co-culture was collected on Day 37 for further analysis.

DAPT inhibition assay

To assess the effect of Notch inhibition on ameloblast-odontoblast interactions, a DAPT inhibition assay was performed after 4 days of the co-culture (Day 29). DAPT (10 μmol/L) (SantaCruz Biotech #sc-201315) was added to the co-culture medium to inhibit Notch signaling, and the treated co-culture was incubated for 24 h at 37 °C with 5% CO₂. After 24 h, the media was replaced with a fresh co-culture medium to continue differentiation. The co-culture was maintained under standard conditions and collected on Day 37 for further analysis of ameloblast maturation and enamelin secretion.

Notch activation in the co-culture assay using C3-DLL4

Notch activation was induced in the co-culture system by treating ieAM organoids with C3-DLL4 (50 nmol/L) after 4 days of co-culture on Day 29. The treatment was maintained for 24 h, after which the media was replaced on Day 30. Co-cultures were maintained with media changes every 3 days until Day 37, when they were collected for further analysis.

Pathway analysis (TopPath)

To reanalyze the pathway involved in ameloblast maturation, we used TopPath pipeline, as described previously with minor changes. In brief, the talklr R package²⁷ was used to pinpoint ligand-receptor interactions specific to each cell type during the transition stage of ameloblast maturation from early ameloblast to secretory ameloblast. The DEsingle⁵¹ and scMLnet⁵² tools were applied to analyze downstream signaling by creating multilayer networks that connect ligands to receptors and transcription factors to their corresponding differentially expressed target genes. Pathway activity scores were then calculated, reflecting the percentage (0–100%) of total activity across all pathways evaluated in the analysis. The pathways considered in the analysis were: TGFβ, BMP, GDF, GDNF, NODAL, ACTIVIN, WNT, ncWNT, EGF, NRG, FGF, PDGF, VEGF, IGF, INSULIN, HH, EDA, NGF, NT, FLT3, HGF, NRXN, OCLN, NOTCH.

Generation of DLX3 KO iPSC

One million WTC11 iPSC were electroporated with Cas9 (0.3 μmol/L, Sigma) and gRNA targeting DLX3 (1.5 μmol/L, Synthego) as RNP complex using Amaxa nucleofector (Human Stem Cell kit 2) in the presence of ROCK inhibitor. Individual colonies were hand-picked and plated into 96-well plates. DNA was extracted using Quick Extract DNA extraction solution (Epicentre #QE09050), and nested polymerase chain reaction (PCR) was performed using Phusion Flash polymerase (ThermoFisher, #F631S). The PCR product was purified using ExoSap-IT (Thermofisher) and sent for Sanger sequencing analysis (Genewiz, Azenta Life Sciences) to identify potential KO clones. gRNA sequence: TAGCTGGAGTAGATCGTACG. PCR primer sequences: F: GAAGGCGTCGTGAGCGAAG, R: TAGCCTGGAGGGAAAACACG. KO was verified at the protein level after 14 days of odontoblast differentiation and 16 days of ameloblast differentiation.

Confirming knockout at the protein level

Cells were lysed directly on the plate using a lysis buffer containing 20 mmol/L Tris-HCl (pH 7.5), 150 mmol/L NaCl, 15% glycerol, 1% Triton X-100, 1 mol/L β-glycerophosphate, 0.5 mol/L NaF, 0.1 mol/L sodium pyrophosphate, orthovanadate, PMSF, and 2% SDS. Following lysis, 25 U of Benzonase Nuclease (EMD Chemicals, Gibbstown, NJ) and a 100× phosphatase inhibitor cocktail were added to the lysate. To prepare the samples for analysis, 4× Laemmli sample buffer (Bio-Rad, #1610747), consisting of 950 μL of sample buffer and 50 μL β-mercaptoethanol (Sigma, #M7522), was added. The mixture was then heated at 95 °C for 10 min. Subsequently, 15 μL of the protein sample was loaded onto an SDS-PAGE gel using a Protean TGX precast gradient gel (4%−20%) (Bio-Rad, #17000546) and transferred onto a nitrocellulose membrane (Bio-Rad, #1620115) using a semi-dry transfer system (Bio-Rad). The membranes were blocked for 1 h with 5% BSA, followed by incubation overnight at 4 °C with primary antibodies on a rocker. The primary antibodies used were AMBN (Santa Cruz #sc-271012, 1:500), SP6 (Atlas #HPA024516, 1:1 000), DLX3 (Abnova #H00001747-M09, 1:500), DSPP (Santa Cruz #7363-2, 1:500), H3 (Abcam #Ab1791, 1:1 000), and β-Actin (Cell Signaling #13E5, 1:10 000) all prepared in 5% BSA. The following day, the membranes were washed three times with 1× TBST at 10-min intervals. Afterward, they were incubated for 1 h at room temperature with an anti-rabbit IgG HRP-conjugated secondary antibody (Bio-Rad, #1721019) (1:10 000) and an anti-mouse IgG HRP-conjugated secondary antibody prepared in 5% milk. After incubation, the membranes were rewashed with 1× TBST (three times, 10-min intervals) and developed using the Immobilon Luminol reagent assay (EMP Millipore). Finally, protein bands were visualized using a Bio-Rad ChemiDoc Imager.

RNA extraction and QRT-PCR analysis

RNA was isolated from the cells utilizing Trizol reagent (Life Technologies) in accordance with the manufacturer’s guidelines. RNA samples underwent treatment with Turbo DNase (Thermo Fisher Scientific) to eliminate genomic DNA contamination and were subsequently quantified utilizing the Nanodrop ND-1000 spectrophotometer (Thermo Fisher Scientific). For cDNA synthesis, 1 µg of RNA was reverse transcribed using the iScript™ cDNA Synthesis Kit (Bio-Rad) or the Applied Biosystems™ High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific). Quantitative real-time PCR (qPCR) was performed on 10 ng of cDNA per reaction using SYBR Green (Applied Biosystems) and the 7300 Real-Time PCR System (Applied Biosystems). The conditions for the PCR were established as follows: an initial incubation at 50 °C for 2 min, followed by a denaturation step at 95 °C for 10 min. This was succeeded by 40 amplification cycles, each consisting of a denaturation phase at 95 °C for 15 s and an annealing/extension phase at 60 °C for 1 min. All quantitative PCR (qPCR) reactions were conducted in triplicate, utilizing β-actin (Forward: TCCCTGGAGAAGAGCTACG, Reverse: GTAGTTTCGTGGATGCCACA) as the endogenous control. The comparative ΔCt (threshold cycle) method was employed to evaluate relative gene expression, and the primer sequences used for specific markers of odontoblast differentiation included NESTIN (Forward: GAAACAGCCATAGAGGGCAA, Reverse: TGGTTTTCCAGAGTCTTCAGTGA) and DSPP (Forward: TGACAGCAATGATGAGAGTG, Reverse: CACTGGTTGAGTGGTTACTG).

RNA sequencing library preparation and analysis

Total RNA was extracted from six iAM organoid samples (two biological replicates per condition). RNA quality and concentration were assessed using the Agilent High Sensitivity D5000 ScreenTape® system (cat# 5067-5592, VWR cat# 76645-084) and the Qubit RNA HS Assay Kit (Thermo Fisher Scientific, cat# Q32852), following the manufacturer’s instructions. High-quality RNA was used for library construction with the Illumina Stranded mRNA Prep, Ligation kit (16 samples, cat# 20040532), according to the manufacturer’s protocol. Library size distribution was evaluated with the D5000 ScreenTape, indexed with Illumina adapters, and purified using AMPure XP beads (Beckman Coulter, cat# A63880). Sequencing was performed on the Illumina NextSeq 2000 platform using a P2, 200-cycle sequencing kit (cat# 20100986), generating paired-end 100 bp reads. Reads were aligned to the human reference genome (Ensembl GRCh38) with HISAT2 (v2.2.1), and gene-level counts were generated using FeatureCounts with GENCODE GRCh38 annotations. Principal component analysis was performed with DESeq in R, while differential expression analysis was carried out using DESeq2. Gene Ontology enrichment was assessed using the TopGO R package and the DAVID 6.8 online tool. RNA-seq data have been deposited in the NCBI GEO under accession number (GSE307437).

Mouse single-cell datasets

To cross-validate findings from the human dataset, we analyzed two independently published mouse scRNA-seq datasets.^28,53 The first dataset, cited in Figs. S1 and S5, is a curated integration of mouse dental tissues from multiple studies, available as a loom file from Mendeley Data⁵⁴ (https://data.mendeley.com/datasets/2kskdknngb/1). Loom files were converted into AnnData objects and used to generate heatmaps. The second dataset comprises whole mouse embryos from embryonic day 8 (E8) to birth (postnatal day 0, P0) and is accessible via CELLxGENE (https://cellxgene.cziscience.com/collections/45d5d2c3-bc28-4814-aed6-0bb6f0e11c82). For this dataset, the dental epithelium cluster was extracted from the whole epithelial dataset and analyzed using the standard Scanpy clustering workflow. Clusters were annotated based on known marker genes.

Pseudotime trajectory visualization

Pseudotime expression trajectories for ameloblast markers were generated using diffusion pseudotime values from the dental epithelium subset. For each selected marker gene (Dlx3, Pitx2, Vwde, Dspp, Ambn, Amelx, Mmp20, Enam, Amtn, Klk4, Tuft1, Wdr72, Odam), expression values were ordered by pseudotime and smoothed using a Gaussian kernel (σ = 35) via scipy.ndimage.gaussian_filter1d. Cluster intervals corresponding to known ameloblast states (OE, DE, pAM, eAM, sAM, mAM) were overlaid as color-coded bars along the x-axis. Plots were created in Matplotlib (v3.8.0) with standardized styling (Fig. S3d). A further-smoothed version of these trajectories, generated by downsampling and applying a light Savitzky–Golay filter, including those for Dspp, Enam, and Wdr72, was incorporated into the schematic summary (Fig. 7) to illustrate transcriptional dynamics across ameloblast maturation.

Bulk-to-single-cell correlation analysis

To identify single cells most transcriptionally similar to our organoid bulk RNA-seq profiles, we computed Pearson correlation coefficients between each cell in the mouse dataset and reference bulk expression vectors for each group (C3-DLL4 Treated, Control) using scipy.stats.pearsonr (Fig. S3c). Analyzes were restricted to a shared set of genes between the bulk and single-cell datasets. For each group, the top 15% of cells (≥85th percentile correlation) were defined as the highest-matching subset. These cells were visualized on precomputed UMAP embeddings, with all other cells shown in gray to highlight group-specific similarity patterns. Figures were generated in Matplotlib with consistent sizing, color coding, and scaling. This correlation-based mapping enabled the identification of single-cell populations closely matching bulk transcriptional signatures, as presented alongside cluster-level annotations.

Notch activation and inhibition in ameloblast organoid differentiation

hiPSCs were differentiated into 2D induced ameloblasts (iAMs) from D0 to Day 16. At Day 16, iAMs were transferred to low-attachment plates to form 3D ieAM organoids in Epicult Plus medium (Stemcell Technologies #06070). Notch activation was induced at Day 24 using the C3-DLL4 Notch activator (50 nmol/L), a trimeric scaffold conjugated to DLL4, or it was incubated with Notch inhibitor, DAPT (10 μmol/L) (SantaCruz Biotech #sc-201315) (Fig. S2a), for 24 h. On Day 25, the media was replaced, and the ieAM organoids continued maturing into isAMs by Day 31 (Fig. S2a). Likewise, DLX3 knockout clones (Clone 10 and Clone 13) were differentiated following the same protocol.

Immunostaining and confocal imaging of iAM organoids

For immunostaining, the organoids were fixed in 4% paraformaldehyde (PFA), then immersed in 1× PBS for 3 × 5-min washes and then immersed in 0.5% TritonX 100 at RT for 10 min to facilitate permeabilization. Later, Organoids were blocked in solution (1× PBS containing 10% BSA, 5% normal goat serum, and 0.1% Triton-X) for 2 h. All incubations were done in tubes on a nutator. They were then suspended in antibody dilution buffer (1× PBS containing 10% BSA, 5% normal goat serum, and 0.2% Triton-X) and primary antibodies diluted at concentrations recommended by the manufacturer and incubated overnight at 4 °C. On the second day, cells were washed 3× for 6 min with 1× PBS, resuspended in antibody dilution buffer containing secondary antibodies at 1:200 and DAPI (1:50), and then incubated overnight at 4 °C. On the third day, organoids were washed 3× for 6 min in 1× PBS and mounted in Vectashield Antifade Mounting Media on a glass concavity microscope slide, one to three organoids per well. Organoids were then imaged using a Leica (DMi8) SP8 LIGHTNING confocal microscope (25× and 40× objectives) equipped with HyD and PMT spectral detectors and Leica LASX acquisition software [version 3.5.5IR], or a High Resolution Widefield Nikon ECLIPSE Ti Fluorescent microscope and Nikon Ti2 confocal microscope with Nikon NIS-Elements software.

Quantification of immunofluorescence signals

Immunofluorescence signals for the proteins of interest (ENAM, AMELX, and MMP20) were quantified as follows. First, raw images were processed using Fiji (ImageJ v2.3.091/92) to split each image stack into individual TIFF files corresponding to DAPI (4′,6-diamidino-2-phenylindole) and the target protein. To normalize quantification across each organoid, nuclei were segmented using the StarDist3D pipeline on ZEISS arivis Cloud (Apeer) and then counted with ImageJ’s 3D Object Counter, ensuring that threshold parameters remained consistent across images acquired under similar conditions.

The fluorescent signals for the proteins of interest were quantified in three dimensions by segmenting puncta with the ImageJ 3D Object Counter. The ImageJ measure function was then used to determine both the mean fluorescence intensity and the integrated intensity for each punctum. The overall mean signal for each organoid was normalized to the nuclei count. Finally, all results were compiled into a spreadsheet and analyzed with GraphPad Prism to assess the statistical significance between groups and conditions.

Kidney capsule transplantation

All animal procedures were performed in compliance with relevant ethical regulations and approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Washington, Seattle. Adult male NOD-SCID mice (RRID: IMSR_JAX:001303), 8 weeks of age, were used for all experiments. Mice were housed under specific pathogen-free conditions with ad libitum access to standard chow and water. ismAM organoids were differentiated for 25 days following the protocol described above, with C3-DLL4 treatment throughout. On the day of transplantation, organoids in suspension were transferred into beveled, kinked PE50 tubing (BD Intramedic, #427517). Adult male NOD-SCID mice (n = 6) were anesthetized in an induction chamber with isoflurane in oxygen, then maintained under 2% isoflurane via a nose cone. Body temperature was maintained at 37 °C using a heating pad for the duration of the procedure. The dorsal flank was shaved and sterilized with betadine, followed by alcohol wipes, before making a 1–2 cm incision. The kidney was gently externalized with a cotton swab, and the capsule was nicked near the caudal end using a 22-gauge needle. The beveled PE50 tubing was inserted beneath the capsule, and the organoid suspension was delivered under the control of a Hamilton syringe. The capsule opening was sealed with a cotton swab to promote clotting and retain the graft. The kidney was then returned to the abdominal cavity, the peritoneum was closed with absorbable sutures, and the skin was closed with surgical staples. Following surgery, mice were placed in a heated recovery cage until fully awake, then returned to their home cages and transported back to the vivarium. All procedures were performed in compliance with ethical regulations under IACUC protocol #4152-01. At 21 days post transplantation, mice were euthanized via CO₂ inhalation. Kidneys with intact grafts were harvested, fixed in 4% paraformaldehyde at 4 °C for 1 h, transferred to 15% sucrose for 2 h, and then to 30% sucrose overnight at 4 °C. Kidneys were bisected and embedded in embedding cryo-molds (Sakura, #25608-916) with Tissue-Tek O.C.T. compound (Sakura, #4583). The embedded tissue was snap-frozen by immersion in 2-methylbutane (EMD, #MX0760-1) chilled in a liquid nitrogen slurry to −80 °C, then stored at −80 °C until sectioning. Cryosections (15 µm) were cut on a pre-chilled cryostat (−20 °C) onto Superfrost Plus microscope slides (Fisherbrand, #12-550-15) and stored at −80 °C until further analysis.

Micro-computed tomography (microCT)

As a proof of concept, we used high-resolution microCT imaging to identify calcified tissue within a sample organoid implanted kidney. We adapted methods from a previous study³³ in order to enable full 3D reconstruction of the kidney without manual edge detection while not desiccating the specimen. Briefly, we 3D printed an imaging platform that press fit within a 15 ml Falcon tube, filled the distal 13 mL of the tube with 30% sucrose-soaked gauze, inserted the platform into the proximal end of the tube, placed the kidney on the platform, and sealed the tube. High-resolution microCT images were then obtained for the entire specimen (Scanco vivaCT 40; 10.5 μm voxel size, 45 kVp, 177 μA). The raw image data were preprocessed using a Gaussian Filter algorithm to remove image noise (Sigma = 1.2, Support = 2.0), followed by segmentation of calcified material within the specimen using standard image thresholding techniques and, finally, manual comparison of the grayscale and binary images to confirm that threshold values maintain morphologic fidelity (201 mg HA/cm³).^55,56 For each 2D image, the surface of the kidney was identified through an automated secondary thresholding procedure to delineate soft tissue. The resulting 3D image depicts the kidney (rose color) and calcified tissue(blue).

Histological analysis

Hematoxylin and eosin (H&E) staining

For hematoxylin and eosin (H&E) staining, the frozen fixed tissue sections were air-dried at room temperature for 10 min and subsequently fixed in 4% paraformaldehyde (PFA) for 10 min. Slides were rinsed in PBS and incubated with Mayer’s hematoxylin for 5 min, followed by a 5 min wash in running tap water and bluing in 0.2% ammonia water for 1 min. Slides were then counterstained with eosin Y for 2 min, dehydrated through a graded ethanol series (70%, 95%, 100%) and cleared in xylene. Coverslips were mounted using a resinous mounting medium (Permount, Fisher Scientific), and images were acquired using Nikon NSPARC brightfield microscope.

Trichrome staining

For modified Masson’s Trichrome staining, the frozen fixed tissue sections were air-dried and incubated in pre-warmed Bouin’s solution at 56 °C for 1 h. Slides were cooled to room temperature and washed in running tap water for 10 min until the yellow color had cleared completely. Sections were stained with Weigert’s iron hematoxylin for 10 min, rinsed in tap water, and incubated in Biebrich scarlet-acid fuchsin solution for 10 min. Differentiation was performed in phosphomolybdic/phosphotungstic acid solution for 15 min without rinsing. Slides were then transferred directly to aniline blue solution for 10 min, rinsed in 1% acetic acid for 1 min, and dehydrated through ethanol and xylene. Coverslips were mounted using either resinous medium for permanent brightfield imaging or aqueous mounting medium (e.g., VECTASHIELD) for subsequent immunofluorescence analysis. Representative images were captured using Nikon NSPARC brightfield microscope.

Von Kossa and Alizarin Red S staining

For Von Kossa staining, frozen sections fixed in 4% paraformaldehyde (EMS, #15710) were incubated in 5% silver nitrate solution (Sigma-Aldrich, #209139) under ultraviolet light for 1 h 30 min. Sections were rinsed several times in deionized distilled water (5 min each rinse), then incubated in 5% sodium thiosulfate solution (Sigma-Aldrich, #217263) for 5 min to remove unreacted silver.

For calcium detection, sections were stained in 1% Alizarin Red S solution (pH4.2) (Sigma-Aldrich, #A5533) for 30 min in the dark, followed by thorough rinsing in deionized distilled water (5 min each rinse). Sections were counterstained with either nuclear fast red (EMS, #26078-05) or 0.2% aqueous Fast Green FCF solution (EMS, #26053-02) for 30 s, then briefly rinsed in deionized distilled water (5 min each rinse). Slides were mounted with Permount Mounting (#SP15100-EA) and stored at room temperature until imaging.

Graphics and illustrations

The illustrations in the graphical abstract and in Figures were created with BioRender.com.

Statistical analysis

All quantifications show the mean, and error bars are ± SEM. Ordinary one-way ANOVA was used for multiple comparisons. A two-tailed, unpaired t-test was used for comparing groups of two using GraphPad Prism. P-values < 0.05, 0.01, 0.001, 0.000 1 are indicated with *, **, *** and ****, respectively. Software used and methods for analysis and quantification of each data in this manuscript are described in the method section.

Author contributions

Conceptualization: A.P.P., J.M., and H.R.-B.; A.P.P. and H.R.-B.; conceived and analyzed the designed proteins in ameloblast and odontoblast differentiation; Methodology: A.P.P., A.A., R. Mout, R. Moore, S. N., B. J. A., P. H., R.K., M. R., T.G., J.M., and H.R.-B.; investigation: A.P.P., A.A.; performed organoid assays; funding acquisition: A.P.P., H.R.-B., and J.M.; resources: R. Mout, G.Q.D., D.B. and H.R.-B.; supervision: R. Mout, H.R.-B., and J.M.; visualization: A.P.P., A.A. and H.R.-B.; A.P.P., R. Mout, A.A., and R. Moore prepared the figures; formal analysis: A.P.P., A.A.; data curation: A.P.P.; project administration: H.R.-B.; writing – original draft, A.P.P., A.A., H.R.-B, and J.M.; writing – review and editing, A.P.P., R. Mout, R. Moore, A.A., H.R.-B. and J.M., G.Q.D., D.B.

Data availability

The lead contact will share all data analyzed and reported in this paper upon request. This paper does not report the original code. Any additional information required to reanalyze the data reported in this work is available from the lead contact upon request. The datasets analyzed for this study were submitted to the GEO repository (GSE307437). The published human datasets analyzed for this study were retrieved from the GEO repository (GSE184749).

Code availability

The lead contact will share all data analyzed and reported in this paper upon request. This paper does not report the original code. Any additional information required to reanalyze the data reported in this work is available from the lead contact upon request. The datasets analyzed for this study were submitted to the GEO repository (GSE307437). The published human datasets analyzed for this study were retrieved from the GEO repository (GSE184749).

Materials availability

This study did not generate new unique reagents.

Competing interests

A.P.P., A.A., J.M., and H.R.-B. are co-inventors on a patent application entitled “A Method to Direct the Differentiation of Human Induced Pluripotent Stem Cells into Early Ameloblasts” (PCT/US2022/053517 filed 12/20/2022), and a patent application entitled “System and Method to Direct the Differentiation of Human Induced Pluripotent Stem Cells Derived Odontoblasts” (PCT/US2023/072209 filed 08/15/2023).

Supplementary information

The online version contains supplementary material available at 10.1038/s41368-026-00429-4.