A spatially organized Cd24a+/ Pax9+ stem cell core governs postnatal tooth establishment

Abstract

Mineralized tissues are fundamental in mechanical support and protection in vertebrates. Their formation by progenitor cells during development remains elusive. Here, we show that the postnatal establishment of the molar teeth was governed by a spatially organized core of Cd24a⁺/Pax9⁺ progenitors that persisted into adulthood. Cd24a⁺ cells gave rise to the dentin-pulp complex, while Pax9⁺ ones mainly generated periodontal tissues. During development, guided by alveolar bone-derived PDGFB (platelet-derived growth factor, B polypeptide), Cd24a⁺/Pax9⁺ cells gradually concentrated on the apical region during the crown-to-root transition, collectively migrated and formed dental root. Cell ablation and conditional Wnt knockout notably compromised tooth establishment. Single-cell RNA sequencing, CUT&Tag, and spatial mapping further revealed distinct features of Cd24a⁺/Pax9⁺ cells and their cellular organization. Last, the CD24⁺/PAX9⁺ core was also present in human teeth, suggesting it as a conserved developmental program. Together, our work underscores the role of spatially organized dental stem cells in the postnatal establishment of a model mineralized organ in mammals.

Fig. 1.. Cd24a and Pax9 exhibited a spatially organized and dynamic expression pattern during postnatal tooth development.

(A) RNA velocity analysis indicated two potential mesenchymal progenitor hubs (green circles) that gave rise to other cell compositions. These hubs were positive for Cd24a and Pax9 expression. (B) Summary for scRNA-seq analysis of PN7 dental cells. RNA velocity revealed two potential mesenchymal progenitor hubs, with one mainly differentiated into DP cells (route 1) and the other exhibiting a tendency to differentiate into dental follicle and perivascular cells (route 2). (C) Heatmap of differentially expressed genes between Cd24a⁺/Pax9⁺ and Cd24a⁻Pax9⁻ cells. (D) RNAscope revealed Cd24a (red) and Pax9 (green) gene expressions at different developmental stages. The mandibular first molars (M1) at postnatal stages PN1 (a), PN2 (b), PN3 (c), PN7 (d), and PN14 (e) were analyzed. White dotted lines outlined the dental pulp. White arrows pointed to the position of the Cd24a⁺/Pax9⁺ interface. Epi: epithelium. Scale bars, 200 μm (whole sections) and 50 μm (zoom-in). (E) Cd24a- and Pax9-positive cells formed a unique spatially organized expression pattern during postnatal tooth development. The fluorescent signal intensity for Cd24a (red) and Pax9 (green) was quantitatively and spatially analyzed by line scan analysis in ImageJ. The x axis represents the vertical position within the tissue section, while the y axis indicates the relative value of the fluorescent signal. (F) RNAscope staining of Gli1, Cd24a, and Pax9 in the mouse molar at PN3 and PN7. Scale bars, 100 μm (whole) and 50 μm (zoom-in).

Fig. 2.. Lineage tracing revealed distinct but integrated contribution of Cd24a⁺ and Pax9⁺ cells to tooth establishment.

(A) Experimental design for lineage tracing using Cd24a-CreERT2 and Pax9-CreERT2 mice. 3′UTR, 3′ untranslated region. (B to D) Lineage tracing of Cd24a⁺ (red) and Pax9⁺ (green) cells using the Cd24a-CreERT2;tdT and Pax9-CreERT2;tdT mice. TM was intraperitoneally injected at PN1 (B), PN7 (C), and PN14 (D), with mandibles harvested for analysis at 48 hours postinduction and at PN21. The lower right panel included a schematic drawing of the M1 tooth harvested at PN21. Red and green dots outlined the tdTomato signals of Cd24a and Pax9 tracing, respectively. Scale bars, 200 μm (whole sections) and 50 μm (zoom-in). (E) Dental marker analysis of lineage-traced Cd24a⁺ and Pax9⁺ cells. Lineage tracing was performed with TM induction at PN7. Tissues were analyzed at PN21. Red, tdTomato⁺ cells (Cd24a or Pax9 tracing); green, dental markers, including NESTIN and Dspp (odontoblasts), αSMA (perivascular cells), and POSTN (PDL). Arrowheads: colocalization signals. Scale bars, 200 μm (whole sections), 50 μm (zoom-in), and 20 μm (for αSMA in Pax9-CreERT2;tdT zoom-in). (F) Quantitative analysis of dental tissue contributions by lineage-traced Cd24a⁺ and Pax9⁺ cells, respectively. Lineage tracing was performed with TM induction at PN7. Tissues were analyzed at PN21. Error bars (means ± SD) represented data from three independent samples with two sections per sample.

Fig. 3.. Cell-autonomous generation of dental tissues by Cd24a⁺/Pax9⁺ cells in vivo under the kidney capsule of mTmG mice.

(A) Schematic experimental procedures. The PN7 tooth germs of Cd24a-or Pax9-CreERT2 transgenic mice were extracted, and EGFP⁺ and EGFP⁻ cells were sorted and embedded in collagen I before renal capsule transplantation. Grafts were analyzed at 4 weeks posttransplantation. IHC, immunohistochemistry. (B) Cd24a-EGFP⁺ cells could still differentiate into odontoblasts in vivo at an ectopic site. H&E staining (top) indicated that Cd24a-EGFP⁺ cells developed odontoblast-like structures. Immunostaining for NESTIN (middle, red) and RNAscope for DSPP (bottom, green) confirmed odontoblast differentiation. The tdTomato signal (red) marked mTmG host-derived tissues. The mouse normal tooth stained with the same markers serves as the positive control (right). Scale bars, 100 μm for whole sections of EGFP⁺ and normal tooth and 50 μm for zoom-in views. (C) Pax9-EGFP⁺ cells could develop into odontoblasts and PDL in vivo at an ectopic site. H&E staining (top) indicated that Pax9-EGFP⁺ cells developed odontoblast-like structures and PDL-like structures. Immunostaining for POSTN (middle, green) and NESTIN (bottom, red) confirmed PDL and odontoblast differentiation, respectively. The tdTomato signal (red) marked mTmG host-derived tissues. The mouse normal tooth stained with the same markers serves as the positive control (right). Scale bars, 100 μm for whole sections of EGFP⁺ and the normal tooth and 50 μm for zoom-in views.

Fig. 4.. Cd24a⁺/Pax9⁺ cell ablation compromised tooth development.

(A) Experimental design for TM-induced cell ablation. (B) Cell ablation of Cd24a⁺ cells. TM was injected at PN3, PN5, and PN7. H&E staining (left) and RNAscope analysis of Cd24a and Dspp (middle and right) on M1 molar sections at PN9. Scale bars, 100 μm (left), 200 μm (middle), and 50 μm (zoom-in). (C) Ablation of Cd24a⁺ cells significantly compromised tooth root development. Tooth height and the sum of the mesial and distal root lengths were quantified. Error bars (means ± SD) represented data from three independent samples with three sections per sample. Statistics: Student’s t test by SPSS version 26. ***P < 0.001. (D) Cell ablation of Pax9⁺ cells. TM was injected at PN7, PN9, and PN11. H&E staining (left) and RNAscope analysis of Pax9 and Dspp (middle) and immunostaining of POSTN (middle and right) on M1 molar sections at PN13. Scale bars, 200 μm (left), 200 μm (middle), and 50 μm (zoom-in). Black dotted lines outlined the PDL of the furcation region in the H&E staining panel. (E) Pax9⁺ cell ablation significantly compromised tooth development. Root lengths and PDL area were quantified. Error bars (means ± SD) represented data from three independent samples with three sections per sample. Statistics: Student’s t test by SPSS version 26. ***P < 0.001. (F) Schematic of subrenal capsule transplantation assay using Cd24a/Pax9-DTA donor in wild-type mice. (G to J) Representative grafts from Cd24a-CreERT2;DTA mice (G) or Pax9-CreERT2;DTA mice (I) following corn oil (left) or TM (right) treatment. Scale bars, 500 μm. The root length of M1 was significantly decreased upon ablation of Cd24a⁺ cells (H) or Pax9⁺ cells (J). Statistics: Student’s t test by SPSS version 26. **P < 0.01. Error bars (means ± SD) represented data from 8 to 10 independent samples in (H) and 5 independent samples in (J).

Fig. 5.. Characterizing lineage-specific features of Cd24a⁺/Pax9⁺ dental stem cells.

(A) Schematic experimental procedure. (B) UMAP visualization of cell clusters from the combined Cd24a⁺/Pax9⁺ cells in the M1 molar at PN9. Cd24a⁻Pax9⁻ cells exhibited three distinct cell compositions based on marker gene expression. Cd24a⁺ and Pax9⁺ cells were very similar to each other compared with Cd24a⁻Pax9⁻ cells. (C) GSEA analysis for Cd24a⁺Pax9⁺, Cd24a⁺Pax9⁻, and Cd24a⁻Pax9⁺ cell subsets. The lineage-committed Cd24a⁻Pax9⁻ cells were used for comparison. (D) GSEA revealed significant enrichment of Wnt signaling in progenitor cells (Cd24a⁺Pax9⁺, Cd24a⁺Pax9⁻, and Cd24a⁻Pax9⁺). The normalized enrichment score (NES), P value, and false discovery rate (FDR) were shown. (E) Scatter plots depicting the expression levels [log₂(FPKM + 1)] of WNT pathway–related genes Fzd2 and Ctnnb1 across Cd24a⁺/Pax9⁺, Cd24a⁺/Pax9⁻, Cd24a⁻/Pax9⁺, and Cd24a⁻/Pax9⁻ cell populations. Data represented the means ± SD. Statistics: one-way ANOVA with Dunn’s test by SPSS version 26. ns, not significant; *P < 0.05, **P < 0.01, and ***P < 0.001. (F) Four distinct gene clusters were identified from Cd24a⁺Pax9⁺, Cd24a⁺Pax9⁻, and Cd24a⁻Pax9⁻ cells. Each cluster in the heatmap showed the top 50 genes, and the panel on the right showed gene expression patterns in the Mfuzz trend graph format. (G) Two gene clusters with distinct patterns were identified from comparing Cd24a⁻Pax9⁺ to Cd24a⁻Pax9⁻ cells. Each cluster in the heatmap showed the top 50 genes. (H) Cd24a-CreERT2;Ctnnb1^fl/fl (CKO) and Ctnnb1^fl/fl (Ctrl) mice were induced with TM for two consecutive days at postnatal stages PN1 and euthanized at PN18. Scale bars, 100 and 200 μm for whole sections of H&E staining and immunostaining for NESTN and CTNNB1, respectively, and 50 μm for zoom-in views. White arrows point to the positive signals of CTNNB1.

Fig. 6.. PDGFB from alveolar bones guided the cell migration of the stem cells core.

(A) Workflow of migration analysis. (B) Flow cytometry scatter plots showing the sorted cell populations from Pax9-CreERT2 mice. WT, wild type. (C) RT-qPCR confirmation of Cd24a and Pax9 expression in the sorted cells. Error bars (means ± SD) represented data from three independent experiments with duplicates. (D and E) Migration patterns and velocities were tracked for sorted cells with different Pax9 and Cd24a profiles by time-lapse microscopy. Top: migration paths of individual cells over time. Bottom: corresponding images with tracks superimposed. Right: quantitative analysis of migration velocities. n, number of cells. Statistics: one-way ANOVA with Dunn’s test by SPSS version 26. ***P < 0.001 and **P < 0.01. (F) Workflow for laser capture microdissection. (G) Dot plot visualization of ligand-receptor interactions. (H) Transwell migration assay indicated that the collective migration of Pax9⁺ cells relied on PDGFB derived from alveolar bones. Error bars (means ± SD) represented data from three independent experiments in duplicates (three recorded fields per sample). Statistics: one-way ANOVA test by SPSS version 26. *P < 0.05 and **P < 0.01. (I) Immunostaining confirmed PDGFB (a), PDGFRA (b), and PDGFRB (c) expression in lineage-traced Pax9⁺ cells. White arrowheads: colocalization signals. Scale bars, 100 μm (whole sections) and 20 μm (zoom-in). (J) Coimmunostaining of PDGFB (green) and EMCN (red) in PN3 roots. Red arrowheads: colocalized PDGFB and EMCN signals; white arrowheads: non-endothelial PDGFB signals. Scale bars, 100 μm (whole sections) and 50 μm (zoom-in). (K) Schematic of bead-induced subrenal capsule assay. (L) Representative images of grafted teeth with beads presoaked in PDGFB, Combo, or NC. Beads (blue dots) were placed on one side of the root. Scale bars, 500 μm. (M) Quantification of root length asymmetry between bead-facing and opposite sides. Statistics: one-way ANOVA with post hoc t test using SPSS version 26. *P < 0.05. Error bars (means ± SD) represented data from five independent samples.

Fig. 7.. scRNA-seq and spatial mapping revealed the organization of the Cd24a⁺/Pax9⁺ zone.

(A) Schematic experimental procedures. EGFP⁺/tdTomato⁺ dental cells were sorted from TM-induced Cd24a-CreERT2;tdT and Pax9-CreERT2;tdT mice at PN9 for scRNA-seq. (B) Cd24a⁺ clusters were identified from scRNA-seq. The UMAP plot delineated five main clusters. (C) Pax9⁺ clusters were identified from scRNA-seq. The UMAP plot delineated five main clusters. (D) Distinct marker genes were characterized for Cd24a⁺ clusters. The plot illustrated the relative expression levels and the proportion of cells expressing the marker genes within each cluster. The markers used for validation were circled out. (E) Characterization and spatial mapping of Cd24a⁺ clusters. Markers for clusters 0 to 3, including S100A1, RUNX2, RGS5, and HMGB2, were costained with tdTomato to identify the corresponding cells and their spatial locations. Zoom-in views provided detailed views of regions with costaining signals, marked by white arrowheads. Scale bars, 100 μm (whole) and 20 μm (zoom-in). (F) Representative model image illustrating the spatial organization of Cd24a⁺ cell clusters. (G) Distinct marker genes were characterized for Pax9⁺ clusters. The plot illustrated the relative expression levels and the proportion of cells expressing key genes within each cluster. The markers used for validation were circled out. (H) Characterization and spatial mapping of Pax9⁺ clusters. Markers for clusters 0 to 4, including POSTN, KCNIP4, LDHB, HMGB2, and TAGLN, were costained with tdTomato to identify the corresponding cells and their spatial location. Zoom-in views provided detailed views of regions with costaining signals, marked by white arrowheads. Scale bars, 200 μm (whole) and 20 μm (zoom-in). (I) Representative model image illustrating the spatial organization of Pax9⁺ cell clusters.

Fig. 8.. The CD24⁺/PAX9⁺ stem cell core was also present in human teeth.

(A) Illustrated progression of human tooth development from stages 1 to 6, showing key morphological changes. (B) Collected tooth specimens from patients with healthy impacted third molars and premolars for orthodontic need, corresponding to stages 1 and 2, 3 and 4, and 5 and 6. Scale bars, 5 mm. (C) RNAscope and immunostaining of human tooth germ sections at different stages indicated that the presence of the CD24⁺/PAX9⁺ stem cell core. Top: RNAscope of CD24 (green). Middle: immunofluorescence of PAX9 protein (red). Scale bars, 500 μm (whole) and 100 μm (zoom-in). (D) Quantitative analysis of spatial fluorescence for CD24 (green) and PAX9 (red) at different developmental stages. Line graphs depicted the fluorescent signal’s position within the tissue sections. The x axis represented the vertical position within the tissue section, while the y axis indicated the relative value of the fluorescent signal.

Fig. 9.. Model for the Cd24a⁺/Pax9⁺ stem cell core–mediated postnatal tooth development.

Model image for the Cd24a⁺/Pax9⁺ stem cell core and its involvement in postnatal tooth development. Specifically, Cd24a⁺ cells formed, most exclusively, the dentin-pulp complex, while Pax9⁺ cells mainly generated periodontal tissues, including PDL, cementum, and apical follicle tissues, with minor contribution to the pulp. The migration of the stem cell core was guided by PDGFB secreted from the alveolar bones, which determined the direction of root development.