Molar furcation and root elongation regions show distinct cellular dynamics

The furcation is the structure at the bottom of the dental pulp chamber that separates the roots of multi-rooted teeth (Fig 1A). Thus, studying furcation development is crucial for understanding tooth root patterning, as the furcation is intimately connected to the determination of the root number. To investigate the furcation development process, we systematically analyzed coronal sections (as defined in Fig 1B) of the mandibular first molars of control mice from PN0.5 to PN14.5, during which the furcation forms. We divided the developing tooth root into two parts: the furcation development region (FDR) in the middle and the non-furcation development region (NFDR) where root elongation occurs on either side (Fig 1A). Within each of these two parts, we further divided the apical portion into the lateral apical area (LAA) and middle apical area (MAA) (Fig 1C) for downstream analysis.

Fig 1

Postnatal furcation development of the mandibular first molar.

(A) MicroCT scanning of mandibular first molars from control mice at PN14.5 showing the furcation morphology. Arrow indicates tooth root furcation. The pink line indicates furcation development region (FDR); blue lines indicate non-furcation development region (NFDR). (B) Schematic drawing of the coronal and sagittal planes of mouse molars. (C) Schematic drawing of the lateral apical area (LAA) and middle apical area (MAA) in coronal sections of the developing tooth root. (D) H&E staining (a-e), immunofluorescence staining of epithelial marker Krt14 (f-j) and cell proliferation marker Ki67 (k-o), and in situ RNAscope of odontoblast differentiation marker Dspp (p-t) in the FDR of mandibular first molars in the coronal view from control mice at indicated stages. (E) H&E staining (a-e), immunofluorescence staining of Krt14 (f-j) and Ki67(k-o), and RNAscope in situ hybridization of Dspp (p-t) of the NFDR of mandibular first molars in the coronal view from control mice at indicated stages. Arrows in b-e of D, E indicate the ends of dentin protrusions elongating from the buccal and lingual sides; in g-i of D, E indicate the tips of the dental epithelial processes; in l-m of D, E indicate regions with actively mesenchymal cell proliferation activity; and in q-t of D and E indicate the most apically located cells expressing Dspp. Asterisks in E indicate the differences present in the NFDR in comparison to the FDR. Scale bars: 100μm in A; 200μm in D and E. Epi, epithelium; De, dentin.

We compared cellular events between the FDR (Fig 1D) and NFDR (Fig 1E), including epithelial elongation, mesenchymal cell proliferation and differentiation. We found these two regions started to show differences after PN4.5. In the FDR, HERS elongated horizontally, formed a bridge across the LAA and MAA at PN8.5 (Fig 1D/f-i/), then lost its bilayered structure completely (Fig 1D/j/). In the NFDR, HERS did not fuse in the MAA, and the tip of HERS maintained its bilayered structure for a longer period (Fig 1E/f-j/). The proliferating dental mesenchymal cells in the FDR expanded from the LAA to the MAA (Fig 1D/k-m/), and the mesenchymal cell proliferation diminished and eventually disappeared in both the LAA and the MAA after PN7.5 (Fig 1D/n-o/). However, there was a difference in the dynamics of cell proliferation in the FDR and NFDR, as mesenchymal cells in the NFDR kept actively proliferating after PN7.5 (Fig 1E/k-o/). In addition, mesenchymal cells in the LAA and MAA differentiated into odontoblasts sequentially (lateral to medial) in the FDR (Fig 1D/p-t/), enabling the dentin to elongate apically and form the hard tissue bridge in the middle. In contrast, in the NFDR, only the mesenchymal cells in the LAA underwent odontoblast differentiation (Fig 1E/p-t/). Therefore, the dentin in the NFDR kept elongating apically without fusion in the MAA. To summarize our results, the cellular dynamics of FDR and NFDR are differently patterned, which may contribute to the morphological difference between furcation development and root elongation. The distinctly patterned cellular dynamics in the FDR and NFDR of the developing tooth root suggest that there may be a specific group of genes/molecules which control the location-specific fate of dental cells and participate in furcation development and root patterning.

Lhx6+ cells are a subpopulation of Gli1+ progenitor cells closely associated with furcation development

Lhx6 is a key patterning gene during early craniofacial development [25,26]. Since genes that are important for embryonic development usually have a role in the postnatal growth of the same tissue, it is likely that Lhx6 may be involved with postnatal root patterning. To specifically investigate the role of Lhx6 in tooth development, we first analyzed its expression pattern from E13.5 to PN7.5. We found that Lhx6 was specifically expressed in the dental mesenchymal cells at both prenatal and postnatal stages of tooth development (Figs 2A–2C and S1). Consistent with previous findings [22,27], at the bud and cap stages, Lhx6 was ubiquitously expressed in the dental mesenchyme (S1A and S1B Fig). However, from E16.5 onwards, Lhx6 expression was enriched in the apical region (S1C Fig). This enrichment of Lhx6 transcripts in the apical dental mesenchyme persisted after birth, although its overall expression level decreased gradually (Fig 2A–2C). This apical enrichment of Lhx6 expression prior to tooth root development suggested that it might play a crucial role in tooth root patterning and morphogenesis.

Fig 2

Lhx6 labels a subpopulation of Gli1+ progenitor cells closely associated with furcation development.

(A-C) RNAscope assays at indicated stages showing expression pattern of Lhx6 during postnatal root morphogenesis. Sagittal sections of mandibular first molars were analyzed. Dotted lines indicate border between dental epithelium and mesenchyme. Epi, epithelium; Mes, mesenchyme. Arrows indicate non-furcation development region (NFDR), and arrowheads point to furcation development region (FDR). (D-F) Co-localization of tdTomato (as indicated by red signal) with β-gal (as indicated by green signal) within mandibular first molars of PN4.5 Lhx6-Cre^ER;tdT;Gli1-LacZ mice after tamoxifen injection at PN2.5. Sagittal sections of mandibular first molars were analyzed. Yellow signal indicates cells expressing both Lhx6 and Gli1. Red box in D is shown magnified in E, and blue box in D is shown magnified in F. (G-I) Lineage tracing of Lhx6+ cells in mandibular first molars of PN4.5, PN12.5 and PN21.5 Lhx6-Cre^ER;tdT mice after tamoxifen injection at PN2.5. Sagittal sections of mandibular first molars were analyzed. Arrowheads indicate Lhx6+ cells and their progeny. (J-K) Progeny of Lhx6+ cells (as indicated by red signal) in the furcation express odontoblast differentiation marker Dspp (as indicated by green signal) in mandibular first molars of PN14.5 Lhx6-Cre^ER;tdT mice after tamoxifen injection at PN2.5. Sagittal sections were analyzed. Box in J is shown magnified in K. (L) Lineage tracing of Gli1+ cells in mandibular first molars of PN21.5 Gli1-Cre^ER;tdT mice after tamoxifen injection at PN2.5. Sagittal sections of mandibular first molars were analyzed. The schematic at the bottom of each set of panels indicates the induction protocol. Scale bars: 200μm in A, B, C, D, G, H, I, J, L; 50μm in K; 20μm in E, F. TMX: tamoxifen.

Gli1+ cells supporting mouse molar root growth also reside in the apical region of dental mesenchyme [7,11]. The finding that Lhx6+ cells populate the same apical region as the Gli1+ cells suggested that these two cell populations are closely related. To elucidate this relationship, we generated Lhx6-Cre^ER;tdT;Gli1-LacZ reporter mice, in which the cells expressing tdTomato and beta-galactose represent Lhx6+ and Gli1+ cells, respectively. Tamoxifen was injected at PN2.5, and the mandibles were collected two days after induction. We noticed that Tdtomato signal could be detected from both the odontoblasts of the crown and apical dental mesenchymal cells (Fig 2D), which is consistent with the in situ RNAscope results (Fig 2A–2C). As shown in Fig 2D–2F, Lhx6+ cells co-localized with some Gli1+ cells in the apical dental mesenchyme, suggesting that these apical Lhx6+ cells are a subpopulation of Gli1+ cells.

Considering that Gli1+ cells are a heterogeneous population containing tooth root progenitor cells that contribute to the entire dental mesenchyme [7,11], we further tested the contribution of Lhx6+ cells to tooth root development. We found that progeny of labeled Lhx6+ cells were closely associated with furcation development during root development, as strong tdTomato signal was detected predominantly in the furcation region compared to other parts of the tooth root mesenchyme (Fig 2G–2I). We confirmed that some cells derived from the Lhx6+ lineage (tdTomato+ cells) overlapped with expression of odontoblast marker Dspp, indicating Lhx6+ cells can differentiate into odontoblasts in the furcation (Fig 2J–2K). In comparison to Gli1+ progeny, the Lhx6+ cells only contributed to a subset of the root dental mesenchyme (Fig 2I and 2L), which supported our observation that Lhx6 labels a subpopulation of Gli1+ cell population during root development. To summarize our findings, Lhx6+ cells are a subpopulation of Gli1+ root progenitor cells, and these cells are closely associated with furcation formation during tooth root development.

Loss of Lhx6 leads to furcation and root patterning defects in mouse molars

Combining the expression pattern of Lhx6 with lineage tracing results led us to hypothesize that Lhx6 may serve as a key regulator for tooth furcation development and root patterning. To test this hypothesis, we generated Lhx6 knockout mice using the Lhx6-Cre^ER mouse line. The Lhx6-Cre^ER knock-in allele in these mice abolishes Lhx6 gene function; therefore, homozygous mice (referred to here as Lhx6^-/- unless otherwise specified) have a knockout phenotype similar to other null mutations of this gene [28]. These Lhx6^-/- mice were born alive and appeared grossly normal but later showed an obvious reduction in body size compared to controls (S2A and S2B Fig). The Lhx6^-/- mice ultimately died three to four weeks after birth; however, we were still able to observe their molar development because by PN18.5 tooth root development is complete and the molars have erupted into the oral cavity [5,11].

We first confirmed the Lhx6 protein was ablated in the molar mesenchyme of Lhx6^-/- mice at the newborn stage (S2C and S2D Fig). We analyzed the phenotypes of both maxillary and mandibular first molar roots from PN4.5 to PN21.5. We found that the tooth crowns were normal at PN4.5, when crown formation completes (S2E and S2F Fig), and continued appearing normal through PN21.5 (Fig 3). The lengths of the tooth roots in Lhx6^-/- mice were mildly affected (Fig 3A–3H). Strikingly, both the maxillary and mandibular first molars failed to form a furcation, leading to a change in their number of roots. In control mice there were three roots in the maxillary first molar (Fig 3A and 3B) and two in the mandibular first molar (Fig 3C and 3D). However, both the maxillary (Fig 3E and 3F) and mandibular first molars (Fig 3G and 3H) were single-rooted in Lhx6^-/- mice. To investigate the mechanisms underlying the tooth root phenotype, we used the mandibular first molar as a model for further studies. Histological analysis further confirmed microCT scanning results (Fig 3I–3N). In Lhx6^-/- mice, dentin could not be detected in the furcation region at PN14.5 (Fig 3L), and the furcation failed to form even at PN21.5 (Fig 3M). Furthermore, periodontal tissues in the furcation region of the Lhx6^-/- mice were also abnormal: the alveolar bone was underdeveloped, and periodontal ligament development was also impaired (Fig 3K and 3N).

Fig 3

Loss of Lhx6 leads to furcation defects in mouse molars.

(A-B) Image and microCT scanning of maxillary first molars of PN21.5 control mice. (C-D) Image and microCT scanning of mandibular first molars of PN21.5 control mice. (E-F) Image and microCT scanning of maxillary first molars of PN21.5 Lhx6^-/- mice. (G-H) Image and microCT scanning of mandibular first molars of PN21.5 Lhx6^-/- mice. (I-N) H&E staining of mandibular first molars of control (I-K) and Lhx6^-/- mice (L-N) at indicated stages in sagittal view. K and N are magnified images of the boxes in J and M, respectively. Asterisks in F, H, L and M indicate absence of furcation. Dashed lines in K and N indicate the border between tooth hard tissues and periodontal ligament, while dotted lines in K and N indicate the border between periodontal ligament and alveolar bone. Scale bars: 500μm in I, J, L, M; 100μm in D and H; 90μm in B and F. De, dentin; PDL, periodontal ligament; Ab, alveolar bone.

We sought to investigate why loss of Lhx6 leads to tooth root malformation without disturbing crown development, although it is expressed both during crown and root development. Previous studies have shown that there is functional redundancy between Lhx6 and Lhx8 in regulating early tooth development, as single mutation of either Lhx6 or Lhx8 does not affect crown development, while double mutations result in an absence of tooth germ formation [27,29,30]. We found that the expression patterns of Lhx6 and Lhx8 largely overlap during crown development, suggesting that one of these genes could compensate for the loss of the other during this process (S3A–S3D Fig). However, at E16.5 when crown patterning has almost completed but root patterning has not yet started, the expression patterns of Lhx6 and Lhx8 started to differ (S3E–S3H Fig). During the postnatal root development stage, Lhx8 was highly expressed in pre-odontoblasts but not in the apical region where Lhx6 transcripts were highly enriched (S3I–S3L Fig). This dynamic Lhx6/8 co-localization pattern may partly explain why loss of Lhx6 produced an abnormal tooth root phenotype without affecting the crown.

Lhx6 regulates mesenchymal proliferation and epithelial elongation during tooth root patterning and morphogenesis

Cell proliferation, apoptosis, and epithelium-mesenchyme interaction are all fundamental to patterning and morphogenesis throughout the body [31]. To investigate how Lhx6 regulates tooth root patterning and morphogenesis, we assessed its involvement in these cellular events. We found that loss of Lhx6 adversely affected mesenchymal cell proliferation within the apical region of the dental mesenchyme (Fig 4A–4C) without affecting mesenchymal apoptosis (S4A and S4B Fig). In control mice, apical dental mesenchymal cells in both the LAA and MAA showed active proliferation at PN7.5 (Fig 4A), whereas in Lhx6^-/- mice, actively proliferating Ki67+ mesenchymal cells were mainly located in the LAA (Fig 4B) and mesenchymal cell proliferation was severely reduced in the MAA (Fig 4B). In addition, we found that the distribution pattern of proliferating mesenchymal cells in the FDR of Lhx6^-/- mice at PN7.5 was more similar to the pattern seen in the NDFR, rather than the FDR, of control mice at the same stage (Fig 1E/m/).

Fig 4

Loss of Lhx6 compromises dental mesenchymal proliferation in the furcation development region.

(A-B) Ki67 staining of FDR from control and Lhx6^-/- mice at PN7.5. Coronal sections of mandibular first molars were analyzed. Yellow dotted lines indicate the border of the actively proliferating apical mesenchymal region. Arrows in B indicate the region where proliferating cells reside in the Lhx6^-/- mice, and asterisk in B indicates significantly decreased cell proliferation activity in the MAA of Lhx6^-/- mice. (C) Comparison of Ki67+ mesenchymal cells in the apical regions of FDR between control and Lhx6^-/- mice at PN7.5. The data represented as mean ± SD. The asterisk indicates p < 0.05. (D-I) Distribution pattern of Cdkn1c in control mice at PN4.5. Coronal sections of mandibular first molars were analyzed. E and H are magnified images of the LAA in D and G, respectively, while F and I are magnified images of the MAA in D and G, respectively. Arrows indicate the region with low expression levels of Cdkn1c. Asterisks indicate the high expression level of Cdkn1c in the MAA region. Yellow dotted lines indicate the border between apical mesenchymal regions with low and high Cdkn1c expression levels. Black dotted lines indicate the border between the dental epithelium and dental mesenchyme. (J) Comparison of the Cdkn1c expression levels in the FDR of control and Lhx6^-/- mice at PN4.5 by RT-qPCR. Data are represented as mean ± SD. Asterisk indicates p < 0.05. (K-M) Distribution pattern of Cdkn1c in Lhx6^-/- mice at PN4.5. Coronal sections of mandibular first molars were analyzed. L shows magnified image of the LAA in K while M shows magnified image of the MAA in K. Arrows indicate region with low expression levels of Cdkn1c. Asterisks indicate the high expression level of Cdkn1c in the MAA region. Yellow dotted lines indicate the border between apical mesenchymal regions with low and high Cdkn1c expression levels. Black dotted lines indicate the border between the dental epithelium and dental mesenchyme. Scale bar:100μm in A, B, D, G, K; 20μm in E, F, H, I, L and M. MAA, middle apical area; LAA, lateral apical area; FDR, furcation development region; NFDR, non-furcation development region.

Cell proliferation is controlled by the cell cycle, which is negatively regulated by cyclin-dependent kinase inhibitors (CKIs) [32]. A previous study has shown that the CKI Ckdn1c is negatively regulated by Lhx6 during palatogenesis [33]. However, the expression pattern of Ckdn1c and whether its expression is regulated by Lhx6 during tooth root development are not known. To answer these questions, we analyzed the distribution pattern of Ckdn1c transcripts in control mice and compared the expression of Cdkn1c between control and Lhx6^-/- mice. We noticed that the expression patterns of Cdkn1c in the FDR and NFDR were different in control mice (Fig 4D–4I). Specifically, the expression levels of Cdkn1c in both the LAA and MAA of the FDR were almost undetectable in control mice (Fig 4D and 4F). However, in the NFDR of control mice, Cdkn1c mRNA expanded into the MAA but not into the LAA (Fig 4G–4I). We also found that the expression level of Cdkn1c was increased in the FDR of Lhx6^-/- mice compared to wild type mice (Fig 4J), suggesting that Lhx6 is also a negative regulator of Cdkn1c in the FDR during tooth root development. We further found that the distribution pattern of Cdkn1c transcripts was altered in the FDR of Lhx6^-/- mice (Fig 4K–4M), and was similar to the pattern seen in the NFDR of control mice (Fig 4G–4I).

We also learned that the expression pattern of Cdkn1c was complementary and mutually exclusive to the distribution pattern of proliferating dental mesenchymal cells in the FDR of both Lhx6^-/- and control mice (Fig 4A and 4B, 4D and 4F and 4K–4M), as the region with low Cdkn1c expression level showed high mesenchymal cell proliferation activity and vice versa. These distribution patterns of Cdkn1c transcripts and proliferating dental mesenchymal cells supported the notion that Cdkn1c is a negative regulator of dental mesenchymal proliferation, and that Lhx6 may regulate mesenchymal proliferation through inhibiting the expression of Cdkn1c.

Although Lhx6 is exclusively expressed in the dental mesenchyme, HERS failed to fuse and the tips of HERS did not dissociate in the Lhx6^-/- mice at PN8.5 (S4C–S4F Fig), suggesting a secondary defect in the dental epithelium as a result of loss of mesenchymal Lhx6.

Lhx6 coordinates dental mesenchymal progenitor cell differentiation via regulating canonical Wnt signaling pathway

Dentin bridge formation by odontoblasts in the MAA is critical for furcation development. We therefore evaluated the odontoblast differentiation of dental mesenchymal cells in the FDR of Lhx6^-/- mice compared to controls. We confirmed that loss of Lhx6 disturbed the dentin bridge formation (Fig 5A and 5B) and odontoblast differentiation of dental mesenchymal cells specifically in the MAA of the FDR using Dspp as an odontoblast differentiation marker (Fig 5C and 5D). We also detected compromised periodontal ligament (PDL) differentiation in the same area using periostin as a PDL differentiation marker (Fig 5E and 5F). In parallel, we found that Gli1+ cells were undetectable in the FDR of control mice at PN9.5 (Fig 5G) while they persisted in the FDR of Lhx6^-/- mice (Fig 5H). Combined with the finding that Gli1+ cells residing in the FDR disappeared after PN9.5 while those located in the NFDR persisted (S5 Fig), these data suggested that the apical dental mesenchyme cells in the presumptive FDR of Lhx6^-/- mice behave as in the NFDR.

Fig 5

Lhx6 coordinates dental mesenchymal differentiation.

(A-B) H&E staining, (C-D) in situ RNAscope assay of Dspp, (E-F) immunofluorescence staining of periostin, and (G-H) X-gal staining of FDR from control and Lhx6^-/- mice at indicated stages. Coronal sections of mandibular first molars were analyzed. (I-L) Co-localization of Dspp (green) and tdTomato (red) in mandibular first molars of Lhx6-Cre^ER/+;tdT (control mice with tdTomato reporter) and Lhx6-Cre^ER/ER;tdT (Lhx6^-/- mice with tdTomato reporter) mice at PN14.5 after tamoxifen injection at PN2.5. Sagittal sections were analyzed. K and L are magnified images of the boxes in I and J, respectively. Asterisks in B, D, F, H, and L indicate where the most significant differences were observed between control and Lhx6^-/- mice. Scale bars: 200μm in I, J; 100μm in A, B, C, D, E, F; 50μm in G, H, K, and L. The schematic at the bottom indicates the induction protocol. TMX: tamoxifen.

Based on our data showing that Lhx6+ cells contribute to the odontoblasts at the root furcation, we asked whether the odontoblast differentiation defect caused by loss of Lhx6 was cell-autonomous or not. We compared the co-localization patterns of tdTomato and Dspp in Lhx6-Cre^ER/ER;tdT (Lhx6^-/- mice with tdTomato reporter) and Lhx6-Cre^ER/+;tdT mice (Lhx6^+/- mice with tdTomato reporter, which didn’t show any abnormal furcation phenotype). In the Lhx6-Cre^ER/+;tdT mice, Lhx6+ cells contributed to the Dspp+ cells in the furcation. However, the progeny of labelled Lhx6+ cells did not express Dspp and did not differentiate into odontoblasts in the furcation of Lhx6-Cre^ER/ER;tdT mice (Fig 5I–5L), suggesting that loss of Lhx6 changed the odontogenic fate of the progenitor cells it labels in a cell-autonomous manner.

To understand how Lhx6 regulates odontoblast differentiation, we used laser-capture microdissection to collect the apical part of the FDR of mandibular first molars from both Lhx6^-/- and control mice for RNAseq. We harvested samples at PN4.5 based on our previous finding that crown development finishes and root development initiates at this stage [11], as well as our observation that the cellular dynamics in the FDR and NFDR diverged after this stage. Hierarchical clustering showed that gene expression profiles of Lhx6^-/- and control mice were well separated (S6A Fig). We identified 284 differentially expressed genes (DEGs, ≥ 1.5-fold, p < 0.05), of which 120 were upregulated and 164 were downregulated (S6A Fig). Pathway analysis using PANTHER classification system revealed that these DEGs belonged to several signaling pathways including Wnt, Cadherin, and Integrin (S6B Fig). Wnt signaling pathway was the top pathway affected by Lhx6 loss (S6B Fig). Canonical Wnt signaling pathway plays important roles in regulating root formation [5]. We therefore narrowed our focus to this pathway. Using Axin2 as a readout [34], we found that the activity of canonical Wnt pathway was down-regulated in the FDR of the Lhx6^-/- mice, especially in the MAA (Fig 6A–6E).

Fig 6

Altered spatial expression patterns and levels of putative downstream targets of Lhx6.

In situ RNAscope assays of Axin2 (A-D), Sfrp2 (F-I, K-N), Smoc2 (P-S) and Frzb (U-X) in the FDR of control and Lhx6^-/- mice at indicated stages. Coronal sections of mandibular first molars were analyzed. Quantification of the expression levels of Axin2 (E) and Sfrp2 (O) between control and Lhx6^-/- mice at PN7.5, and comparison of Sfrp2 (J), Smoc2 (T) and Frzb (Y) at PN4.5 using RT-qPCR. Data are represented as mean ± SD. Boxes in A, C, F, H, K, M, P, R, U and W are shown magnified in B, D, G, I, L, N, Q, S, V and X, respectively. The arrow in B indicates the high Axin2 expression level in the MAA of control mice, and arrows in G, L, Q, and V indicate the low gene expression levels in the MAA of control mice. Asterisks in D, I, N, S, and X indicate the region where the most significant differences were observed between control and Lhx6^-/- mice. Asterisks in E, J, O, T, and Y indicate p < 0.05. Scale bars: 100μm in A, C, F, H, K, M, P, R, U and W; 20μm in B, D, G, I, L, N, Q, S, V and X.

To investigate the possible mechanisms by which Lhx6 regulates the Wnt signaling pathway, we next analyzed the expression patterns of the top 15 DEGs (S1 Table). In control mice, three DEGs, including Smoc2, Sfrp2 and Frzb (Sfrp3), showed spatially specific expression patterns that differed between the FDR and NFDR (S6C–S6N Fig). These three genes had low expression levels in both the MAA and LAA of the FDR (S6C, S6E, S6G, S6I, S6K and S6M Fig). However, the expression levels of Smoc2 and Sfrp2 were high in the MAA of the NFDR (S6D, S6F, S6H and S6J Fig) while Frzb transcripts were mainly enriched in the LAA of the NFDR (S6L and S6N Fig). We further validated the expression changes of Smoc2, Sfrp2 and Frzb in Lhx6^-/- mice as compared to control mice by in situ RNAscope and RT-qPCR. We found that all three of these Lhx6 putative target genes showed increased expression levels in the FDR of Lhx6^-/- mice (Fig 6F–6Y), which is highly consistent with the RNAseq results. We also found that loss of Lhx6 disrupted the expression patterns of Smoc2, Sfrp2 and Frzb in the FDR, and the expression patterns of these three genes in the FDR of Lhx6^-/- mice (Fig 6F–6I, 6K–6N, 6P–6S and 6U–6X) were similar to their expression patterns in the NFDR of control mice (S6C–S6N Fig).

Previous studies have shown that Sfrp2 is a Wnt antagonist [35,36]. In the Lhx6^-/- mice, Sfrp2 transcripts were enriched in the MAA (Fig 6H, 6I, 6M and 6N) where the decreased expression of Axin2 was mainly detected (Fig 6C and 6D). Therefore, we further narrowed our focus to Sfrp2 as a potential regulator through which Lhx6 regulates the activity of Wnt signaling pathway. To test this hypothesis, we first cultured mouse dental mesenchymal cells as described in our previous study [7] and then treated them with mouse recombinant Sfrp2 protein. We found that Sfrp2 down-regulated the activity of canonical Wnt signaling pathway, as revealed by the expression of Axin2 (Fig 7A). Wnt signaling pathway regulates odontoblast differentiation, such that inactivating canonical Wnt signaling pathway blocks odontoblast differentiation and results in complete ablation of root dentin formation [37,38]. In cell culture, we also found that supplementation of Sfrp2 inhibited odontoblast differentiation, as revealed by decreased Dspp expression (Fig 7B). Briefly, these data suggested that Sfrp2 is among the key regulators through which Lhx6 controls the activity of Wnt signaling pathway as well as odontoblast differentiation during furcation development. Because Sfrp2 can mediate Wnt signaling through binding to Wnt ligands [35,36], we sought to determine which Wnt ligands were affected by Sfrp2. We found that four canonical Wnt ligands were expressed during postnatal root development, namely Wnt10a, Wnt6, Wnt4 and Wnt3a, although the expression levels of these ligands did not significantly change between control and Lhx6^-/- mice (S7 Fig). Previous studies showed that Sfrp2 is able to bind to Wnt6, Wnt4 and Wnt3a, and mediates their induction of Wnt signaling [35,36,39]. It remains unclear whether Sfrp2 also interacts with Wnt10a. Loss of Wnt10a results in a similar furcation phenotype to that observed in Lhx6^-/- mice [17,18], and Wnt10a can induce the expression of Dspp [40]. We therefore investigated the potential interaction between Sfrp2 and Wnt10a. Using co-immunoprecipitation, we found Wnt10a can physically bind to Sfrp2 (Fig 7C). In addition, treating cells with mouse recombinant Wnt10a protein was able to rescue the compromised odontoblast differentiation caused by Sfrp2 (Fig 7D), suggesting that the increased Sfrp2 in Lhx6^-/- mice may antagonize the function of Wnt10a and contribute to the furcation phenotype.

Fig 7

Lhx6 regulates dental mesenchymal cell differentiation through regulating canonical Wnt signaling pathway by inhibiting Sfrp2.

(A-B) RT-qPCR assay for Axin2 (A) and Dspp (B) in cultured dental pulp cells from PN2.5–3.5 control mice with or without recombinant Sfrp2 treatment. Data are represented as mean ± SD, and asterisks indicate p < 0.05. (C) Co-immunoprecipitation using Flag-tagged Sfrp2 and GFP-tagged Wnt10a expressed in 293T cells. Sfrp2 was immunoprecipitated (IP) and Wnt10a was immunoblotted (IB) for the physical interaction analysis. Comb.: simultaneous transfection with Flag-tagged Sfrp2 and GFP-tagged Wnt10a vectors. (D) RT-qPCR for Dspp in cultured dental pulp cells from PN2.5–3.5 control mice treated with recombinant Sfrp2, Wnt10a or their combination. Data are represented as mean ± SD. Different lowercase letters above bars indicate statistically significant intergroup differences (p < 0.05), and same lowercase letters above bars indicate no statistically significant intergroup differences (p > 0.05). (E-G) microCT images of the maxillary first molars of PN21.5 control (E), Lhx6^-/- (F) and Lhx6^-/- rescue (Lhx6-Cre^ER/ER;Ctnnb1^floxE3/+; G) mice after tamoxifen injection at PN4.5. (H-M) microCT images of the mandibular first molars of PN21.5 control (H, K), Lhx6^-/- (I, L) and Lhx6^-/- rescue (Lhx6-Cre^ER/ER;Ctnnb1^floxE3/+; J, M) mice after tamoxifen injection at PN4.5. Arrowhead in M indicates the furcation of rescue mice. (N-S) H&E staining of mandibular first molars of PN21.5 control (N, Q), Lhx6^-/- (O, R) and Lhx6^-/- rescue (Lhx6-Cre^ER/ER;Ctnnb1^floxE3/+; P, S) mice after tamoxifen injection at PN4.5. Sagittal sections of mandibular first molars were analyzed. Q, R and S are magnified images of the boxes in N, O and P, respectively. Dashed lines in Q, R and S indicate the border between tooth hard tissues and periodontal ligament, while dotted lines in Q, R and S indicate the border between periodontal ligament and alveolar bone. The schematic at the bottom indicates the induction protocol. De, dentin; PDL, periodontal ligament; Ab, alveolar bone. Scale bars: 100μm in E-J and Q-S; 500μm in K-P. TMX: tamoxifen.

To further validate that the downregulation of canonical Wnt activity in the Lhx6+ lineage is involved in the odontoblast differentiation and furcation defects of Lhx6^-/- mice, we generated a Lhx6-Cre^ER/ER;Ctnnb1^floxE3/+ compound mutant mouse model in which, in addition to loss of Lhx6 expression, Wnt signaling activity is constitutively activated in the Lhx6+ cell derivatives after tamoxifen induction at PN4.5. We first analyzed the roots of double heterozygous compound mutants (i.e., Lhx6-Cre^ER/+;Ctnnb1^floxE3/+), and found there was no abnormal furcation phenotype (S8 Fig). We found that furcation development was partially rescued in the compound mutant mice (Fig 7E–7M). Histological analysis detected dentin and organized odontoblasts at the furcation, although it was situated more apically than in wild type controls (Fig 7N–7S). PDL development, though not alveolar bone development, was also rescued (Fig 7N–7S). These data supported our hypothesis that down-regulated Wnt activity in the Lhx6+ lineage is among the key factors underlying odontoblast differentiation and furcation defects in Lhx6^-/- mice.

Ethics statement

All animal studies were approved by the Institutional Animal Care and Use Committee of the University of Southern California. The animals were handled according to approved IACUC protocol #11765 at the University of Southern California.

Animals

C57BL/6J and transgenic mouse lines were used, including Lhx6-Cre^ER (JAX#010776, The Jackson Laboratory) [28], ROSA26^{loxP-STOP-loxP-tdTomato} (tdTomato conditional reporter, JAX#007905, The Jackson Laboratory) [55], Gli1-LacZ (JAX#008211, The Jackson Laboratory) [56], Gli1-Cre^ER (JAX#007913, The Jackson Laboratory) [57] and Ctnnb1^floxE3 [58]. All mice were housed in pathogen-free conditions, and newborn pups were documented as postnatal stage PN0.5. Both male and female mice at indicated stages were collected, genotyped and analyzed.

Tamoxifen administration

For Cre^ER activation, tamoxifen (Sigma, T5648) dissolved in corn oil (Sigma, C8267) with a final concentration of 20mg/ml was injected intraperitoneally (10μL/1g body weight, single injection unless otherwise specified) at PN2.5 for lineage tracing. To activate the expression of Ctnnb1 in the Lhx6+ cell lineage, the tamoxifen dosage was adjusted to 8μL/1g body weight, and the tamoxifen was injected at PN4.5.

microCT analysis

Both mandibles and maxillae were dissected under a stereomicroscope (Leica L2) and fixed in 4% paraformaldehyde (PFA) overnight at room temperature. microCT scanning was carried out with Skyscan 1174v1.2 (Bruker Corporation, USA) at the Center for Craniofacial Molecular Biology, University of Southern California. Images at a resolution of 16.7μm were acquired with the X-ray source at 50kVp and 800μA. The reconstruction of 3D images was completed using Avizo/Amira 9.5.0 (FEI Visualization Sciences Group, France).

Histology assay

Mouse mandibles were dissected and fixed in 10% neutralized buffered formalin (NBF, Sigma) overnight at room temperature, then decalcified with 10% EDTA in 1xPBS for 1–14 days based on mouse age. Decalcified mandibles were dehydrated and embedded in paraffin. Tissue blocks were sectioned at 5μm using a microtome (Leica) and mounted on SuperFrost Plus slides (Fisher). Hematoxylin and Eosin (H&E) staining was completed following standard protocols.

Immunofluorescence assay

Immunofluorescence assays were performed on paraffin-embedded sections prepared as described above. Sections were dried for 1 hour at 55°C, deparaffinized and rehydrated. After antigen retrieval (Vector, H-3300), sections were blocked for 1 hour at room temperature in blocking solution (PerkinElmer, FP1020), and then incubated with primary antibodies diluted in blocking solution at 4°C overnight. After washing three times with PBST (0.1% Tween20 in 1xPBS), sections were incubated with secondary antibodies at room temperature for 1 hour. For periostin and Lhx6 protein detection, fluorescent signal amplification Tyramide SuperBoost Kits, including goat anti-rabbit Alexa Fluor-647 (B40926, Thermo Fisher Scientific) and goat anti-mouse Alexa Fluro-488 (B40912, Thermo Fisher Scientific) were used according to the supplier’s protocols. DAPI (Sigma, D9542) was used for nuclear staining. Detailed information on antibodies used in this study is available in S2 Table.

TUNEL assay

Specimens were harvested, fixed, decalcified, and sectioned as described for immunofluorescence assays. Apoptotic cells were detected using Click-it Plus TUNEL Assay for In Situ Apoptosis Detection kit (Thermo Fisher, C10617) following the manufacturer’s recommended protocol.

In situ RNAscope assay

Tissues were fixed in 10% NBF overnight at room temperature, decalcified with DEPC-treated 10% EDTA in 1×PBS at 4°C, dehydrated sequentially with 15% sucrose in 1xPBS (4°C, overnight) and 30% sucrose in 1xPBS/OCT (4°C, overnight), and embedded in OCT (Sakura, Tissue-Tek, 4583) on liquid nitrogen immediately. Frozen tissue blocks were sectioned at 8μm on a cryostat (Leica CM3050S) and mounted on SuperFrost Plus slides. In situ RNA expression detection was performed using RNAscope 2.5 HD Assay-RED kit (Advanced Cell Diagnostics, 322350) and RNAscope 2.5 HD Multiplex Fluorescent v2 (Advanced Cell Diagnostics, 323110) according to the manufacturer’s instructions. Probes used for Lhx6, Lhx8, Dspp, Axin2, Sfrp2, Frzb (Sfrp3), Wnt3a, Wnt4, Wnt6 and Wnt10a were designed and synthesized by Advanced Cell Diagnostics. Detailed information on probes is provided in S3 Table.

X-gal staining

The mandibles were fixed in 0.2% glutaraldehyde overnight in 4°C, decalcified in 10% EDTA supplemented with 2mM MgCl₂. The samples were dehydrated through a sucrose series and embedded in OCT. To detect β-galactosidase (β-gal) activity in tissue sections, cryosections were stained in X-gal staining solution for 3–4 h at 37°C in the dark according to standard procedure [7].

Laser capture microdissection and RNA-sequencing

First mandibular molars were collected from both control and littermate mutant mice at PN4.5. After dissection, the mandibles were washed once in pre-chilled DEPC-treated 1xPBS and transferred into a mold filled with pre-chilled OCT and frozen using liquid nitrogen. Next, samples were sectioned into slices of 12μm thickness on a cryostat and mounted on Polyethylenenaphthalate (PEN) membrane slides (Zeiss). The sections were fixed and stained with H&E staining strictly following the manufacturer’s protocol. We performed laser capture microdissection of the apical tissues at the furcation development region of both control and mutant mice using the Zeiss PALM Laser Capture Microdissection System. To get sufficient total RNA, tissues from eight mandibular first molars were combined together to prepare one RNA library. The total RNA was isolated using an RNeasy Plus Micro Kit (Qiagen, 74034), and the quality of RNA was determined using an Agilent 2100 Bioanalyzer. The cDNA library preparation and sequencing were performed at the UCLA Technology Center for Genomics & Bioinformatics. Single-end reads with 75 cycles were performed on the Illumina Nextseq 500 platform for 3 pairs of samples. Raw reads were trimmed, aligned to the mm10 genome using STAR (version 2.6.1d), and normalized using RPKM. Differential expression was calculated by selecting transcripts with fold change ≤ -1.5 or ≥ 1.5 and a significance level of p < 0.05. The pathway analysis was done using PANTHER classification system [59].

Cell culture, recombinant protein treatment

Dental apical mesenchymal cells of the mandibular first molars of PN2.5 to PN3.5 mice were dissected, cut into small pieces, and seeded into 24-well plate culture dishes (Corning) with α-MEM supplemented with 20% FBS, 100U/ml penicillin and 100μg/ml streptomycin (Life Science Technologies). For each well, at least 6 first mandibular molar tooth germs were required to get sufficient cells. The cells were cultured without disturbance for 24h to allow initial cell attachment, then fresh medium was added and the cells were cultured until 80% confluent. Mouse recombinant Sfrp2 protein (R&D) and Wnt10a (LsBio) with a final concentration of 2.5μg/ml were dissolved in odontoblast differentiation medium (1% FBS, 5mM β-glycerophosphate (β-GP, Sigma, G9422), 50μg/ml ascorbic acid (Sigma, A4403), and 10nM dexamethasone (DEX, Sigma, D4902), and added into each well. The cells were harvested 7 days after odontogenic induction for RT-PCR assays.

Tissue dissection, RNA extraction and RT-qPCR

The apical third of the first mandibular molar was dissected out first, and then the middle third of the apical tissue was collected for RNA extraction and RT-qPCR validation. Briefly, the total RNA was isolated using an RNeasy Plus Micro Kit (Qiagen, 74034), and the cDNA was synthesized using an iScript cDNA Synthesis kit (Bio-Rad). The quantitation reaction was done with SsoFast EvaGreen Supermix (Bio-Rad) on a Bio-Rad CFX96 Real-Time System. The primer sequences were obtained from PrimerBank [60] and are listed in S4 Table.

Co-IP assays

Flag-Sfrp2 (MR204070, OriGene) and GFP-Wnt10a (MG206595, OriGene) vectors were transfected into 293T cells. Cells were harvested 48h after transfection and lysed in lysis buffer (50mM Tris-HCl (pH7.5), 150mM NaCl, 2mM EDTA, 0.1% NP-40, 10% glycerol and protease inhibitor cocktail). Lysates were immunoprecipitated with anti-Flag antibody and protein G-Sepharose 4 fast flow (GE Healthcare). Immune complexes were washed three times with IP buffer and immunoblotted with anti-Flag or anti-GFP antibodies (S2 Table).

Image acquisition

The Immunofluorescence, TUNEL and RNAscope images were captured using an All-in-one Fluorescence Microscope (Keyence, BZ-X710) at the Center for Craniofacial Molecular Biology, University of Southern California.

ImageJ image analysis

ImageJ was used to determine the percentage of immunostained area out of the total area. To calculate the percentage of positive immunostaining, selected regions were first converted to 8-bit binary images, and then positive signals were measured with the “Analyze Particles” function. The derived area was then divided by the total area.

Sample size

We analyzed n≥3 samples for each experimental group for all experiments unless otherwise stated.

Statistical analysis

Statistical analysis was completed with GraphPad Prism. Significance was assessed by independent two-tailed Student’s t-test or analysis of variance, and the chosen level of significance for all statistical tests in this study was p <0.05.