Coupling of angiogenesis and odontogenesis orchestrates tooth mineralization in mice

In vertebrates, the vascular network develops throughout the body to meet tissue demands for oxygen and nutrients and to secrete organotypic paracrine molecules, known as angiocrine factors, that drive cell differentiation and tissue morphogenesis (Potente et al., 2011; Augustin and Koh, 2017; Rafii et al., 2016). Recent advances in histological technology have significantly enhanced our understanding of the spatiotemporal association between angiogenesis and osteogenesis during embryonic and postnatal bone development, particularly in the context of endochondral ossification (Wang et al., 2007; Maes et al., 2010; Kusumbe et al., 2014). The interrelationship between vascular growth and bone formation is known as angiogenic–osteogenic coupling (Maes and Clemens, 2014).

Like bones, teeth are part of the skeletal system. They are essential for daily functions such as mastication and speech. The dental pulp is the anatomically central part of teeth and consists primarily of connective tissues, neurovascular bundles, and odontoblasts, columnar cells lining the surface of the dental pulp (Li et al., 2017; Rombouts et al., 2017; Tucker and Sharpe, 2004). Odontoblasts secrete dentin, a calcified tissue inside the enamel, in the process of dentinogenesis, which is similar to osteoid secretion of osteoblasts during bone development (Thesleff and Sharpe, 1997). Because of the similarity of these processes, the cellular interaction between blood vessels and odontoblasts would reasonably be expected in dentinogenesis, similar to the angiogenic–osteogenic coupling known to occur in bones. However, the tooth is the hardest tissue in the mammalian body, even harder than bones (Simmer et al., 2010). This caveat has hampered precise histological examinations and genetic approaches recently applied to characterize bone angiogenic–osteogenic coupling (Kusumbe et al., 2014, 2015). Therefore, the mechanism and function of angiogenesis in tooth development have thus far remained uncharacterized.

Rodent incisors, which are widely used as an experimental model to study the biology and function of teeth, grow continuously throughout the lifespan, driven by self-renewal of adult stem cells (Kaukua et al., 2014; Biehs et al., 2013). This process does not occur in human teeth. In the present study, to optimize the relevance of mouse studies to the biology of human teeth, we used mouse molars, specifically the first mandibular molar, which lacks adult stem cells and the capacity for renewal after development, similar to human teeth (Tummers and Thesleff, 2003; Li et al., 2015).

Here, to overcome the technical limitations of tooth histology, we first optimized the decalcification solution and processing length before tissue sectioning, and succeeded in visualizing the 3D structure of the tooth vasculature at high resolution. Using this technique in combination with single-cell transcriptomics, we identified that reciprocal feedback between a specialized endothelial population and odontoblasts orchestrates tooth formation.

Initially, we thoroughly optimized the immunohistochemistry protocol, including decalcification time and equipment, ultimately visualizing the 3D structure of the tooth vasculature at high resolution (Fig. 1, A–I). Blood vessels running through the middle of the dental pulp are known to branch from alveolar vessels (Rombouts et al., 2017) and extend small vessels toward the outer surface of the dental pulp, forming a dense capillary plexus (Fig. 1 F). Vascular endothelial growth factor (VEGF) secretion and multiple VEGF receptors are vital for angiogenic processes during development (Potente et al., 2011). Therefore, we examined expression of the VEGF receptors Vegfr1 and Vegfr2 using Vegfr2-BAC-GFP⁺Vegfr1-BAC-DsRed⁺ mice (Ishitobi et al., 2010; Okabe et al., 2014). In teeth, Vegfr1,2 expression was predominantly localized to endothelial cells (ECs; Fig. 1, E–I). Interestingly, many of the ECs exhibited robust filopodia extensions and concomitantly expressed high levels of Vegfr1 and Vegfr2 (Fig. 1, G–I; and Fig. S1, A–E), in contrast to the side-by-side expression pattern of Vegfr2 and Vegfr1 in endothelial tip and stalk cells of other tissues (Gerhardt et al., 2003; Potente et al., 2011). These capillaries perforated the basal layer of odontoblasts, which was marked by Osterix-positive (Osx⁺) nuclei, thin cytoplasm, and apical processes consisting of an intermediate filament protein, Nestin (Lendahl et al., 1990), whereas in bone tissue, osteoblasts and the vasculature are mutually exclusive (Fig. 1, J–M; and Video 1). These ECs around odontoblasts lacked smooth muscle cells, confirming they indeed formed capillaries (Fig. S1 F). The emergence and expansion of Vegfr1^highVegfr2^high vasculature occurred at postnatal day 12 (P12), just before the start of mastication and active dentinogenesis (Fig. S1, G–V). Considering these spatial and temporal characteristics, we termed the sprouting Vegfr1^highVegfr2^high capillaries around odontoblasts POTCs. CD31 was not highly expressed in POTCs (Fig. 1, N–P), unlike type H vessels in developing bones (Kusumbe et al., 2014). Next, we examined the cellular composition of dental pulp and alveolar bone marrow (Fig. 2 A). We detected hematopoietic cells (HCs) and osteoclasts in bone marrow but not in dental pulp (Fig. 2, B–F), suggesting that hematopoiesis and osteoclastogenesis do not occur in the dental pulp.

To evaluate the role of ECs in tooth formation, we targeted their abundant expression of Vegfr1 and Vegfr2 by crossing a tamoxifen-inducible endothelial-specific Cre line, Cdh5-BAC-Cre^ERT2 mice (Okabe et al., 2014), with Vegfr2-flox or Vegfr1-flox mice, generating endothelial-specific Vegfr2 and Vegfr1 knockout mice (Cdh5–BAC–Cre^ERT2+Vegfr2^flox/flox and Cdh5–BAC–Cre^ERT2+Vegfr1^flox/flox, hereafter referred to as Vegfr2^iΔEC and Vegfr1^iΔEC). Stiffness and molar extrusion were reduced in the molars of Vegfr2^iΔEC mice at P12 (Fig. 3, A–C). Consistent with the high dependence of retinal endothelial tip cells on Vegfr2 (Gerhardt et al., 2003), Vegfr2^iΔEC mice exhibited ablated POTCs and endothelial apoptosis, together with decreased odontoblast and dentin widths relative to control mice (Fig. 3, D–S). Conversely, Vegfr1^iΔEC did not manifest detectable defects in gross molar phenotype (Fig. 4, A and B) but did show increased filopodia and vessel areas (Fig. 4, C–J). Mice with combined knockout of Vegfr1 and Vegfr2 (Cdh5–BAC–Cre^ERT2+Vegfr1^flox/floxVegfr2^flox/flox) exhibited ablation of periodontal capillaries and impaired molar formation, similar to Vegfr2^iΔEC mice (Fig. 4, K–N), suggesting that the Vegfr2 deletion–mediated vascular damage is dominant over hypervascularity caused by Vegfr1 deletion. The thick vessels in the center of the dental pulp were far less affected in the above mutant mice, suggesting that POTCs were particularly angiogenic and plastic in the tooth vasculature. Vegfr2^iΔEC mice were slightly smaller than control mice at P12, likely reflecting the systemic effect. To mitigate this effect, we conducted a delayed deletion protocol in which 4OHT injection was started at P10 (Fig. S2 A). In this protocol, body length was not affected, but POTC numbers were significantly decreased, and odontoblast maturation and tooth strength were impaired (Fig. S2, B–J). The phenotype in this delayed protocol is likely ascribed to the direct effects of Vegfr2 deletion on POTCs, although some indirect effects caused by impairment in other vascular beds cannot be excluded definitively.

In bone development, several cell types, including osteoblasts, hypertrophic chondrocytes, and ECs, contribute to angiogenesis as Vegfa-producing cells (Sivaraj and Adams, 2016). Therefore, we examined Vegfa expression in teeth using in situ hybridization, revealing that Vegfa was predominantly expressed in odontoblasts (Fig. 5, A–D). Osx-Cre^ERT2 mice, which are widely used to delete genes in osteogenic lineage cells, express Cre in odontoblasts, as demonstrated by crossing with indicator flox-LSL-GFP mice (Fig. 5 E). Thus, we generated Osx-Cre^ERT2–specific Vegfa knockout mice (Osx–Cre^ERT2+Vegfa^flox/flox). VEGF immunohistochemistry of Osx-Cre^ERT2+Vegfa^fl/fl mice confirmed that expression of VEGF in osteoblasts was largely diminished (Fig. 5 F). Like Vegfr2^iΔEC mice, Osx–Cre^ERT2+Vegfa^flox/flox mice exhibited impaired molar extrusion (Fig. 5 G). Histologically, Osx–Cre^ERT2+Vegfa^flox/flox mice exhibited ablated POTCs and decreased odontoblast and dentin widths (Fig. 5, H–S). These data indicated that odontoblast-derived VEGF is crucial for the growth of POTCs, although a low level of VEGF from other cells expressing Osx–Cre^ERT2 might contribute to the phenotype of Osx–Cre^ERT2+Vegfa^flox/flox mice. In contrast to osteoblast Vegfr2 expression and function (Hu and Olsen, 2016; Sivaraj and Adams, 2016), odontoblast Vegfr2 was dispensable for tooth formation (Fig. S3, A–I).

Next, we conducted metabolomic analysis of mice to determine how teeth responded metabolically to endothelial Vegfr2 deletion. Principal component analysis revealed global metabolome changes in Vegfr2^iΔEC mice relative to controls (Fig. 6 A). Volcano plot analysis identified that metabolites related to ketone body metabolism were significantly increased in Vegfr2^iΔEC mice, while the ATP/ADP ratio and creatine phosphate level were decreased in Vegfr2^iΔEC mice, suggesting a decreased energy pool and alternatively elevated ketone body metabolism (Fig. 6, B and C; and Table S1). Moreover, tracing intraperitoneally injected stable isotope phosphate (H₃P[¹⁸O]₄) revealed decreased tissue levels of H₃P[¹⁸O]₄ (Fig. 6 D), suggesting impaired phosphate delivery in Vegfr2^iΔEC mice. Because phosphate, as well as calcium, are necessary for odontoblasts to mineralize dentin, lack of phosphate transport from ECs could potentially be responsible for impaired tooth formation in Vegfr2^iΔEC mice. Indeed, imaging mass spectrometry (IMS) revealed decreased levels of metabolites associated with the phosphate pathway in Vegfr2^iΔEC mice (Fig. 6 E). Systemic alteration of these metabolites in the circulation, in addition to localized vascular changes, could also contribute to the observed effects in teeth of Vegfr2^iΔEC mice.

To comprehensively understand the changes in cellular composition and gene expression profile in Vegfr2^iΔEC mice, we conducted single-cell RNA sequencing (scRNA-seq) on cells collected from dental pulps of control and Vegfr2^iΔEC mice (Fig. 7 A and Fig. S4, A and B). Using Uniform Manifold Approximation and Projection (UMAP), we defined 12 cell clusters representing six cell types, including HCs, odontoblasts, myelinated Schwann cells, nonmyelinated Schwann cells, ECs, and mesenchymal stem cells based on known markers (Pagella et al., 2021; Tai-Nagara et al., 2020; Fig. 7, B and C; Fig. S4, C and D; and Table S2). HCs were identified as extravasated macrophages, as immunohistochemistry did not detect rounded myeloid cells (Fig. 2 C). Odontoblasts were divided into one fibroblastic cluster (OB1) and two mature clusters (OB2 and OB3), based on expression of Fap (fibroblastic) and Runx2 (mature) odontoblast markers. Intriguingly, the number of mature odontoblasts was decreased in Vegfr2^iΔEC mice, although the number of fibroblastic cells was increased (Fig. 7 D). A number of genes were affected by Vegfr2 deletion (Table S3). Gene Ontology (GO) enrichment analysis identified that the biological processes most prominently inhibited in odontoblasts from Vegfr2^iΔEC mice included osteoblast differentiation and ossification, a mineralization process similar to dentinogenesis. Conversely, those enhanced in Vegfr2^iΔEC mice included glycolytic process, blood vessel development, and cellular response to hypoxia (Fig. 7 E and Table S4). A heatmap created by extraction of dentinogenesis genes revealed a broad decrease in Vegfr2^iΔEC mice, especially in OB2, the mature odontoblast cluster (Fig. 7 F). Conversely, Hif-1 target genes, including Vegfa, were increased in Vegfr2^iΔEC mice, especially in OB1, the fibroblastic odontoblast cluster (Fig. 7 F). Expression of representative genes for dentinogenesis, including Dmp1, Dspp, and Runx2, was confirmed by quantitative PCR (qPCR) analysis (Fig. S4 E). Reclustering ECs (EC1 and EC2) generated five subclusters, including arterial, venous, capillary endothelial (Capil1 and 2), and POTCs (Fig. 8 A) based on known markers (Vanlandewijck et al., 2018; Kalucka et al., 2020) and expression of Kdr (Vegfr2) and Flt1 (Vegfr1; Fig. 8 B). Consistent with histological data (Fig. 3, D–G), scRNA-seq analysis confirmed that POTCs were predominantly ablated in Vegfr2^iΔEC mice (Fig. 8 C). Genes known to be enriched in endothelial tip cells (del Toro et al., 2010) were broadly downregulated in ECs of Vegfr2^iΔEC mice (Fig. 8 D and Table S5). Indeed, extraction of differentially expressed genes in POTCs revealed that tip cell genes (Nid2, Esm1, and Ramp3) and Vegf receptors (Kdr and Flt1) were most strikingly upregulated (Fig. 8 E).

The above findings prompted us to determine how ECs support odontoblast maturation and dentinogenesis. As oxygen and nutrients are generally the major effectors provided by blood vessels, we performed organ culture of isolated dental pulps (Fig. 9 A). All of the dentinogenesis markers, including Dmp1, Dspp, and Runx2, were significantly downregulated by hypoxic conditions or serum deprivation (Fig. 9 B). On the other hand, scRNA-seq data identified that paracrine molecules known to promote odontoblast maturation or dentinogenesis, such as Tgfb1, Ptn, and Jag2 (Melin et al., 2000; Jin et al., 2021), but not Vegfa or Noggin, were abundantly expressed in POTCs, which were ablated in Vegfr2^iΔEC mice (Fig. 9 C and Fig. S4 F). To examine the significance of these angiocrine factors, we used the organ culture of dental pulps from Vegfr2^iΔEC mice. Interestingly, supplementation of recombinant TGFβ and pleiotrophine (PTN) proteins partially, but statistically significantly, restored decreased expression of dentinogenesis markers in dental pulps from Vegfr2^iΔEC mice (Fig. 9 D). To determine which cell types interacted with POTCs, we performed CellPhoneDB analysis (Efremova et al., 2020), a bioinformatics tool that builds cell–cell communication networks using receptor–ligand information. We found that POTCs most abundantly interacted with OB1 and OB2 (Fig. 9 E). Affected ligand–receptor interactions included interactions related to vascular growth and maturation or mineralization of odontoblasts (Fig. 9 F). These data support the idea that not just oxygen and nutrients, but also multiple angiocrine factors, contribute to the regulation of odontoblasts as the effectors of POTCs.

In the present study, we developed a new imaging technique that allowed unprecedented examination of the tooth vasculature. Using multiomics analyses, we identified a reciprocal feedback loop between odontoblasts and a specialized vessel subtype, POTCs, during dental development. In the feedback loop, odontoblasts secrete VEGF to promote angiogenic POTC growth, and POTCs in turn provide odontoblasts with oxygen, phosphate, and odontogenic factors necessary to support their maturation and dentinogenesis. We propose that this interdependent coupling of angiogenesis and odontogenesis could be referred to as “angiogenic–odontogenic coupling,” and is a comparable process to angiogenic–osteogenic coupling in bones, although the mechanisms of these two couplings are distinct (Fig. 9 G).

Interaction between blood vessels and parenchymal cells is essential to ensure the diversity of organotypic vasculature. Therefore, mechanisms of these interactions likely differ depending on the organ. The cell-to-cell interaction observed in teeth is similar to that of the developing retina. Retinal astrocytes in hypoxic regions of the retina secrete VEGF to promote angiogenesis. In turn, newly formed blood vessels provide the oxygen and leukemia inhibitory factor to relieve hypoxia and promote astrocyte maturation (West et al., 2005; Kubota et al., 2008). Future studies will examine the role of hypoxia-inducible factors in odontoblasts in tooth vascular growth.

Together, the teeth and bones comprise the skeletal system, which is mineralized and hardened during development. However, angiogenic–osteogenic coupling in bones differs from angiogenic–odontogenic coupling in teeth. First, in bones, osteoblasts are not the primary source of VEGF, which is instead predominantly expressed by hypertrophic chondrocytes (Gerber et al., 1999; Duan et al., 2015), suggesting that the direct contribution of osteoblasts to angiogenesis is minor compared with the indispensable role of odontoblast-secreted VEGF in tooth angiogenesis and development. It is also noteworthy that VEGF inhibition increased the layer of hypertrophic chondrocytes in bones (Gerber et al., 1999). The characteristics of the specialized vessel subtypes in bones and teeth also differ. There are two primary vessel populations in bone marrow: type S (CD31^lowEmcn^low) and type H (CD31^highEmcn^high). CD31^highEmcn^high vessels are the primary contributors to osteogenesis (Kusumbe et al., 2014). However, there is no such distinction in teeth. POTCs express CD31 and Emcn at levels similar to other vessels located in the inner portion of the dental pulp. Moreover, the Vegfr2^highVegfr1^high feature of POTCs is not present in bone type H vessels in bones (not depicted).

In summary, our findings uncovered a fundamental process of intercellular communication required for tooth development. The concept of angiogenic–odontogenic coupling could contribute to the development of effective regenerative therapies for tooth loss. Moreover, the mechanisms of how blood vessels orchestrate tooth mineralization extend the knowledge base regarding skeletal development and the diversity of organotypic vasculature and angiocrine signaling.

Animal care was approved by the Institutional Animal Care and Use Committee of Keio University and performed in accordance with the Guidelines of Keio University for Animal and Recombinant DNA Experiments. CAG-LSL-GFP (Kawamoto et al., 2000), Vegfr1-BAC-DsRed, and Vegfr2-BAC-GFP (Ishitobi et al., 2010; Okabe et al., 2014) mice have been described previously. Cdh-BAC-Cre^ERT2mice, which were previously described (Okabe et al., 2014), were crossed with Vegfr2-flox mice (Hooper et al., 2009) and Vegfr1-flox mice (Ho et al., 2012). Cdh5-BAC-Cre^ERT2Vegfr2^flox/+ and Cdh5-BAC-Cre^ERT2Vegfr1^flox/+ mice were used as littermate controls, respectively. Osx–Cre^ERT2 mice (Maes et al., 2010; Ono et al., 2016) were mated with Vegfa-flox mice (provided by Genentech; Gerber et al., 1999). The mutant mice were crossed with C57BL/6J mice >10 times. For tamoxifen-inducible animals, 4-hydroxytamoxifen (40 μg) dissolved in oil/EtOH (4:1) at 2 mg/ml was injected s.c. at the indicated time points. For analysis, mice of both sexes were used without exact determination.

Surgically dissected mandibles were fixed overnight in 4% paraformaldehyde in PBS. For decalcification, samples were immersed in 0.5 M EDTA for 7 d at P6, 14 d at P12, and 28 d at P18 and P24. Samples were carefully embedded in optimum cutting temperature compound under a microscope to ensure that sectional planes were parallel and were then cryosectioned at 40-μm thickness (HM550; Thermo Fisher Scientific) using tungsten carbide knives (Thermo Fisher Scientific). All sections were stained as described below.

Immunohistochemistry of whole-mount samples or tissue sections was performed as described previously (Kubota et al., 2011). The primary monoclonal antibodies used were hamster anti-CD31 (MAB1398Z; 1:1,000; Chemicon), Nestin (Rat401; 1:200; BD Pharmingen), VEGFR2 (Avas12a1; 1:200; BD Pharmingen), and CD45 (30-F11; 1:500; BD Pharmingen). The primary polyclonal antibodies used were Alexa Fluor 488–conjugated anti-GFP (A21311; 1:500; Molecular Probes), Endomucin (sc-65495; 1:500; Santa Cruz), Osterix (ab22552; 1:500; Abcam), Cathepsin K (ab19027; 1:500; Abcam), and cleaved caspase 3 (9694; 1:200; Cell Signaling). Secondary antibodies used were Alexa Fluor 488–conjugated IgGs (A11034, A11006, A11055; 1:500; Molecular Probes) or Cy3/Cy5 DyLight549/DyeLight649-conjugated IgGs (711-165-152, 112-165-167, 127-165-160, 711-605-152, 112-605-167, 127-605-160; 1:500; Jackson ImmunoResearch). For nuclear staining, specimens were treated with DAPI (D-1306; Molecular Probes). For in situ hybridization, mandibular sections generated in RNAse-free conditions were hybridized with digoxigenin-labeled antisense RNA probes, as described previously (Kubota et al., 2011).

Fluorescent images were obtained using a confocal laser scanning microscope (FV3000; Olympus). Imaris software (Bitplane) was used to obtain orthogonal views and create a 3D video from multislice Z-stack images. Quantification of dentin thickness and odontoblast height was measured at the maximal points in the tooth sections. Measurement of vessel area, filopodia number, and Osx⁺ cells was conducted in 300 × 300-µm fields of view in each sample using the FV10-ASW 3.0 Viewer (Olympus) and ImageJ software (National Institutes of Health).

Dental pulps were isolated en bloc from first molars removed from the mandibles and then thoroughly washed with PBS. Thereafter, the isolated whole-mount dental pulps were transferred into 1 ml DMEM (Sigma-Aldrich), with or without 10% FBS, and cultured in 96-well culture plates (Corning) at 37°C in 20% (normoxic) or 1% (hypoxic) O₂ for 24 h. In some experiments, recombinant human PTN (1 μg/ml; R&D Systems) or 20 ng/ml human TGFβ (Abcam) proteins were added to culture medium.

Dental pulps were incubated for 30 min at 37°C in DMEM containing 1% collagenase D (from Clostridium histolyticum; Sigma-Aldrich), 1 U/ml dispase (Thermo Fisher Scientific), and 1 U/ml DNase (Invitrogen) before cells were dissociated by gentle trituration. Cells were isolated using Dynabeads (Veritas), according to the manufacturer’s instructions. To isolate ECs, cells were incubated with an anti-CD31 antibody (BD Biosciences) preconjugated to Dynabeads (M-450) anti-Rat IgG, and positively selected as CD31⁺ cells.

Total RNA was prepared from fresh or cultured dental pulps, and reverse transcription was performed using Superscript III (Invitrogen). qPCR assays were performed with the ABI 7500 Fast Real-Time PCR System using TaqMan Fast Universal PCR Master Mix (Applied Biosystems) and TaqMan Gene Expression Assay Mix with Dmp1 (Mm00803833_g1), Dspp (Mm00515666_m1), Runx2 (Mm00501578_m1), Col1a1 (Mm01302043_g1), Vegfr1 (Mm00438780_m1), and Vegfr2 (Mm00440099_m1). A mouse β-actin (Mm00607939_s1) assay mix served as an endogenous control. Data were analyzed using 7500 Fast System SDS Software 1.3.1.

To analyze metabolic pathways in dental pulp cells, H₃PO₄[¹⁸O]₄ (1 mg/body weight [g]; Taiyo Nissan) was intraperitoneally injected 30 min before sacrificing pups. Dissected dental pulps were immediately frozen and stored at −80°C until use. Tissues were used for both metabolomic analysis and IMS. Metabolite extraction for metabolomics was performed as described previously (Sugiura et al., 2014; Yamamoto et al., 2014; Morita et al., 2018). Briefly, frozen tissue blocks, together with internal control compounds (2-morpholinoethanesulfonic acid), were homogenized in ice-cold methanol (500 μl) using a manual homogenizer (Finger Masher [AM79330]; Sarstedt), followed by addition of an equal volume of chloroform and 0.4 times the volume of ultrapure water (LC/MS grade; Wako). The suspension was then centrifuged at 15,000 g for 15 min at 4°C. After centrifugation, the aqueous phase was filtered using an ultrafiltration tube (Ultrafree-MC; UFC3 LCC NB; Human Metabolome Technologies). The filtrate was concentrated with a vacuum concentrator (SpeedVac; Thermo Fisher Scientific). The concentrated filtrate was dissolved in 50 μl ultrapure water and used for metabolomic analysis. H₃PO₄ and H₃P[¹⁸O]₄ levels were measured as described below.

For metabolomic analysis of central glucose metabolic pathways, including glycolysis, the TCA cycle, and the pentose phosphate pathway, anion metabolites were measured using an orbitrap-type mass spectrometer (MS; Q-Exactive Focus; Thermo Fisher Scientific) connected to a high-performance ion chromatography (IC) system (ICS-5000+; Thermo Fisher Scientific), enabling us to perform highly selective and sensitive metabolite quantification owing to the IC separation and Fourier transfer MS principle (Hu et al., 2015). We note that IC can measure not only organic compounds, but also inorganic phosphoric acids. The IC device was equipped with an anion electrolytic suppressor (Dionex AERS 500; Thermo Fisher Scientific) to convert the potassium hydroxide gradient into pure water before the sample entered the MS. The separation was performed using a Dionex IonPac AS11-HC, 4-μm-particle-size column (Thermo Fisher Scientific). The IC flow rate was 0.25 ml/min, supplemented postcolumn with a 0.18-ml/min methanol makeup flow. The potassium hydroxide gradient conditions for IC separation were 1–100 mM for 0–40 min, 100 mM for 40–50 min, and 1 mM for 50.1–60 min at a column temperature of 30°C.

MALDI imaging analyses were performed as described previously (Sugiura et al., 2016). Briefly, thin sections (8 µm) were prepared using a cryomicrotome (CM3050; Leica Microsystems). Sections were attached to indium tin oxide–coated glass slides (Bruker Daltonics) and coated with 9-aminoacridine as a matrix (10 mg/ml, dissolved in 80% ethanol) by manually spraying with an artist brush (Procon Boy FWA Platinum; Mr. Hobby). MALDI imaging was performed using an orbitrap MS (QExactive Focus; Thermo Fisher Scientific) coupled with an atmospheric pressure scanning microprobe MALDI ion source (AP-SMALDI10; TransMIT). For the orbitrap MS, signals within a mass range 50–400 were acquired with a mass resolving power of 70,000 at m/z 200. Thereafter, spectral data were transformed to image data and analyzed using ImageQuest v1.0.1 (Thermo Fisher Scientific).

Molars were isolated immediately after animals were euthanized. A compression test was conducted using a mechanical testing machine (Model TK-252C; Muromachi Kikai Co.). A load was placed at the center of the molar at a rate of 2 mm/min until the molar broke, and maximum load (N) was measured as tooth strength.

Dental pulps were collected and incubated for 30 min at 37°C in DMEM containing 1% collagenase D (from C. histolyticum; C5138-1G; Sigma-Aldrich), 1 U/ml dispase (17105-041; Thermo Fisher Scientific), and 1 U/ml DNase (Invitrogen) before cells were dissociated by trituration. The cells were dissociated to generate a single-cell suspension by filtering thorough a 40-μm nylon mesh. Cells were stained with propidium iodide (421301; 1:500; BioLegend). Propidium iodide–negative cells were sorted with a Cell Sorter SH800Z (Sony). Single cells were captured using a Chromium instrument (10x Genomics). Libraries were constructed using a commercial microdroplet-based platform, Chromium Single Cell 3′ v3.1 (10x Genomics) according to the manufacturer’s instructions. A DNBSEQ-G400 platform (MGI) was used to sequence 28 and 91 base paired-end reads. Sequenced reads were aligned to the reference genome (mm10) using Cell Ranger v6.0.1 software (10x Genomics). The data were processed and analyzed using Seurat package v4.0.1. UMAP was used for the dimensionality reduction that captured both the local and global structure in scRNA-seq data. Clusters identified by k-means clustering analysis were visualized in 2 dimensions as feature or violin plots and annotated based on expression of canonical cell-type markers. Differentially expressed genes were extracted using the Wilcoxon rank sum test in the Seurat package. Gene expression was averaged in each cluster, and the correlation matrix was calculated by Pearson correlations. GO analysis of upregulated genes in control cells (log fold-change [FC] > 0.25) and Vegfr2^iΔEC cells (log FC > 0.3) was conducted using the Database for Annotation, Visualisation, and Integrated Discovery (DAVID). Heatmaps showed the z-scores of average UMI counts in each cluster, visualized by the seaborn v0.11.0 package. Single-cell gene expression profiles from each cluster identified by scRNA-seq were further analyzed using CellPhoneDB (Efremova et al., 2020).

Results are expressed as means ± SD. Comparisons between two groups were evaluated using a two-tailed Student’s t test. Comparisons between multiple groups were evaluated using two-way ANOVA followed by Bonferroni’s multiple comparison test. P values <0.05 were considered statistically significant.

Fig. S1 shows immunohistological and qPCR data for the temporal transition of Vegf receptor expression in teeth. Fig. S2 shows the phenotype of Vegfr2^iΔEC mice in the delayed deletion protocol. Fig. S3 shows the phenotype of Osx–Cre^ERT2+Vegfr2^flox/flox mice characterized by no apparent blood vessel or odontoblast defects. Fig. S4 shows the data for quality control and additional information for cell type annotation in scRNA-seq. Video 1 shows a 3D reconstructed video of immunohistochemistry on the first molar at P12 showing that POTCs perforate the basal layer of odontoblasts. Table S1 shows raw data for metabolome analysis. Table S2 lists the top 100 enriched genes for each cluster in scRNA-seq analysis. Table S3 lists differentially expressed genes in control and Vegfr2^iΔEC mice. Table S4 lists differentially expressed genes in odontoblast clusters (OB1, OB2, OB3) of control and Vegfr2^iΔEC mice. Table S5 lists differentially expressed genes in endothelial clusters (EC1, EC2) of control and Vegfr2^iΔEC mice.

Raw scRNA-seq data is available in GEO (GSE181175). There is no restriction on data availability.

We thank Prof. Henry M. Kronenberg (Harvard University, Cambridge, MA) for kindly providing Osx-Cre^ERT2 mice.

This work was supported by Grants-in-Aid for Specially Promoted Research from the Ministry of Education, Culture, Sports, Science, and Technology of Japan (18H05042, 18K19553, and 19H03397); by Japan Agency for Medical Research and Development–PRIME (AMED-PRIME; JP20gm6210017h0002 and 21gm6210017h0003); by Japan Science and Technology Agency (MoonshotR&D; JPMJMS 2024); and by research grants from the following: Inamori Foundation, Kao Foundation for Arts and Culture, Takeda Science Foundation, Mochida Memorial Foundation, Mitsubishi Foundation, Cell Science Research Foundation, SENSHIN Medical Research Foundation, Sumitomo Foundation, Daiichi Sankyo Foundation of Life Science, Naito Foundation, Uehara Memorial Foundation, and Toray Science Foundation.

Author contributions: Designed experiments: Y. Kubota. Performed experiments: T. Matsubara, T. Iga, Y. Sugiura, T. Sanosaka, and I. Tai-Nagara. Analyzed data: T. Matsubara, Y. Sugiura, D. Kusumoto, T. Sanosaka, J. Kohyama, H. Okano, and M. Suematsu. Provided experimental materials: N. Takeda, G.-H. Fong, K. Ito, and M. Ema. Edited the manuscript: D. Kusumoto, T. Sanosaka, H. Okano, and M. Suematsu. Wrote the paper: Y. Kubota.