CXCL12-abundant reticular cell lineage is the key source of endosteal osteoblasts and adipocytes in adult bone marrow

Abstract

Introduction

CXCL12-abundant reticular (CAR) cells, characterized by high expression of CXCL12, are perisinusoidal bone marrow stromal cells that establish the hematopoietic stem cell niche. Although CAR cells are postulated to become precursor cells of both osteoblasts and adipocytes, the spatiotemporal dynamics of CAR cells—particularly their contribution to osteoblasts—remain poorly understood. Here, we defined the contribution of the endogenous Cxcl12-expressing (Cxcl12⁺) lineage to the skeleton from development to adulthood.

Methods

A novel Cxcl12-Cre knock-in mouse model was generated that allows permanent lineage tracing of CAR cells and their progeny.

Results

Lineage tracing revealed that the Cxcl12⁺ lineage arose perinatally and was initially localized around sinusoids in the bone marrow. During aging, these cells expanded and progressively populated non-sinusoidal regions distal to the sinusoids. While the contribution of the lineage to osteogenesis was negligible during postnatal growth, the lineage accounted for over 60% of endosteal osteoblasts in both cortical and trabecular bones in skeletally mature adult mice. Additionally, the lineage gave rise to the majority of bone marrow adipocytes. This lineage did not contribute to periosteal osteoblasts or extra-bone adipocytes. Furthermore, intermittent administration of the anabolic agent parathyroid hormone significantly promoted osteoblast differentiation from this lineage. Single-cell RNA sequencing confirmed the presence of both osteogenic and adipogenic precursors within the traced population.

Conclusions

The Cxcl12-expression-based lineage tracing approach revealed the CAR cell lineage as the key source of endosteal osteoblasts and adipocytes in adult bone marrow, suggesting their important role in skeletal homeostasis. This lineage may represent a promising therapeutic target for age-related bone diseases.

Highlights

• •Cxcl12-cre marks Cxcl12⁺ CAR cells in marrow and endosteal compartments.
• •Cxcl12⁺ lineage shifts from perisinusoidal to non-perisinusoidal with age.
• •Cxcl12⁺ lineage supplies endosteal osteoblasts and marrow adipocytes.
• •Cxcl12⁺ lineage forms osteoblasts during adulthood but not during growth.
• •Anabolic PTH promotes osteoblast differentiation from the Cxcl12⁺ lineage.

1. Introduction

Bone is a multifunctional organ essential for providing skeletal support, hosting hematopoiesis, and maintaining mineral homeostasis [1]. The bone marrow, the primary site of hematopoiesis, is involved in bone growth, maintenance, and regeneration. The bone marrow harbors both hematopoietic stem cells (HSCs), the source of all blood lineages, and bone marrow stromal cells (BMSCs), which support HSCs and serve as a cellular source for bone formation [[2], [3], [4], [5], [6]]. Under defined in vitro conditions, BMSCs are capable of differentiating into osteoblasts, chondrocytes, and adipocytes [2,3]. BMSCs constitute a heterogeneous population that includes stromal cells with multipotent capabilities; consequently, these cells have been extensively investigated as a prospective source for regenerative medicine [2,3]. Advances in single-cell RNA sequencing (scRNA-seq), lineage tracing using genetically modified mice, and fluorescence-activated cell sorting (FACS) have led to the progressive identification of functional subsets within the heterogeneous stromal cell population, defined by the expression of specific molecular markers [[6], [7], [8], [9], [10], [11], [12], [13]]. However, the exact cellular composition, tissue distribution, and specific functions of BMSCs require further investigation.

CXCL12-abundant reticular (CAR) cells, characterized by high expression of C–X–C motif chemokine 12 (CXCL12), have been identified as a critical subset of BMSCs [10,[13], [14], [15], [16]]. CAR cells are perisinusoidal stromal cells that establish a reticular network throughout the bone marrow, thereby supporting HSCs [14,17,18]. CXCL12, produced by CAR cells, is a cytokine essential for the retention and maintenance of CXCR4-expressing HSCs, positioning CAR cells as a central component of the HSC niche [4,14,17,19,20]. Furthermore, historical data suggest that CAR cells serve as a source for precursor cells of osteoblasts and adipocytes involved in bone formation [10,[13], [14], [15],21]. Omatsu et al. [14] reported that CAR cells (Cxcl12-expressing cells; abbreviated as Cxcl12⁺ cells) isolated from Cxcl12-GFP mice were capable of differentiating into both osteoblasts and adipocytes in vitro and expressed key osteogenic (Runx2) and adipogenic (Cebpα, Pparg) transcription factors. Moreover, three-dimensional analysis using Cxcl12-GFP mice revealed that CAR cells, initially considered restricted to the perisinusoidal space, consist of at least two subpopulations: “sinusoidal” CAR cells adjacent to sinusoids and “non-sinusoidal” CAR cells distal to the sinusoids [10,18]. Based on their localization and gene expression patterns, the former is hypothesized to be involved in adipogenesis and the latter in osteogenesis [10,13,15].

Lineage tracing using a tamoxifen-inducible Cxcl12-CreER mouse model preferentially marked a quiescent, sinusoidal subset of CAR cells localized to the central bone marrow. Under physiological conditions, although a portion of these labeled cells differentiated into adipocytes, no corresponding differentiation into osteoblasts was detected in the femoral diaphysis [15]. However, in a fracture healing model and under in vitro conditions, the Cxcl12-CreER model demonstrated that CAR cells are multipotent and capable of differentiating into osteoblasts, chondrocytes, and adipocytes [15]. This finding highlights a discrepancy between the intrinsic properties of CAR cells—specifically their differentiation potential [10,13,14,21] and localization [10,18]—and their lineage contribution under physiological conditions [15]. A potential explanation is that the Cxcl12-CreER model utilizes a Bacterial Artificial Chromosome transgene; consequently, Cre expression may not fully recapitulate the expression of the endogenous Cxcl12 gene. Furthermore, this approach only labels cells actively expressing Cxcl12 at the time of tamoxifen induction, which precludes tracing the lineage from ontogeny. Although osteoblasts have diverse origins and distinct stem cell populations contribute to bone formation during growth versus homeostasis [6,9,22,23], the spatiotemporal dynamics of CAR cells—particularly their contribution to osteoblasts throughout development and adulthood—remain poorly understood in vivo.

In this study, we generated a Cxcl12-Cre mouse model by targeting Cre recombinase to the endogenous Cxcl12 locus. By crossing these mice with a reporter line, we permanently labeled CAR cells and their progeny to trace their lineage in vivo. Using this model, we aimed to elucidate the contribution of CAR cells to the skeleton from development through adulthood.

2. Methods

All animal experiments were conducted under the approval of the University of Osaka Graduate School of Dentistry Animal Ethics Committee (approval no.: Animal Dentistry R-05-020-0) and the University of Osaka Safety Committee for Genetic Recombination Experiments (approval no.: 04930).

2.1. Experimental animals

2.1.1. Generation of Cxcl12-Cre mice

Cxcl12-Cre mice, which express Cre recombinase in Cxcl12⁺ cells, were generated using the CRISPR-Cas9 system. To specifically insert the Cre cDNA into exon 2 of the endogenous Cxcl12 locus [14,17], the 5′ homology arm (HA) was amplified from murine genomic DNA through polymerase chain reaction (PCR) using the primers 5′-HA: 5′-GGGCCCGGG-CCTGGGTCCTTTTGTCCTTTAGA-3' (SmaI-attached) and 5′-CAAATTTTGGTGTACGGTCAGTAAATTGGACATGACTGGTTTACCTAGAGGAGGTG-3'. The 5′-HA reverse primer contained the 5′ sequence of the Cre cDNA for subsequent overlapping extension PCR. The Cre-polyA cDNA was amplified through PCR using the primers 5′-ATGTCCAATTTACTGACCGTACA-3′ and 5′-CGAAGAACCGGCAGGGCATCGGTAGCTCAGGCTAACTTGTTTATTGCAGCTTATAA-3'. The reverse primer for Cre cDNA contained the 3′-HA sequence for subsequent overlapping extension PCR. First, the 5′-HA and Cre-polyA cDNA were amplified and fused via overlap extension PCR (5′-HA-Cre-polyA). The 3′-HA was amplified with primers 5′-AGCCTGAGCTACCGATGCCCTGC-3′ and 5′-GGGACTAGTAACAGCTTGGAAAGCAAGAGTGA-3' (SpeI-attached). The amplified product was fused to the 5′-HA-Cre-polyA construct to generate the 5′-HA-Cre-polyA-3′-HA construct, which was inserted into the pLSODN-2D vector. The insert was sequenced to verify sequence fidelity. The purified single-stranded oligodeoxynucleotide was electroporated into zygotes derived from C57BL/6 J mice using the CRISPR-Cas9 system (gRNA; CCAgtcagcctgagctaccgatg). The zygotes were cultured to the two-cell stage and transferred into the oviducts of pseudopregnant ICR female mice. Littermates were genotyped using the following 5′-Cxcl12-Cre or 3′-Cxcl12-Cre primers: 5′-Cxcl12-Cre: 5′-TTGTGATTTCCAGGGCTTAGGAT-3′ and 5′-GCGCGCCTGAAGATATAGAAGAT-3’; 3′-Cxcl12-Cre: 5′-TGGAGTTTCAATACCGGAGATCA-3′ and 5′-TGACAAGACAGACCCAACTGTCA-3’. Amplified products were gel-purified and confirmed with sequencing. Mice confirmed to be knock-in mice were crossed with C57BL/6 J mice for further breeding. Founder mice were crossed at least five times to segregate chromosomes for potential off-target effects of the gRNA.

2.1.2. Mating and maintenance of experimental animals

To visualize the Cxcl12⁺ lineage, Cxcl12-Cre mice were crossed with a ZsGreen (ZsG) reporter mouse line (Rosa26-Flox-Stop-Flox-ZsGreen) that expresses ZsG in a Cre-dependent manner, generating Cxcl12-Cre; ZsG mice. The ZsG mice (#007906, B6.Cg-Gt (ROSA)26Sortm6(CAG-ZsGreen1)Hze/J) were purchased from The Jackson Laboratory (Bar Harbor, ME, USA). Cxcl12-Cre and ZsG mice were each maintained in a heterozygous state by crossing them with C57BL/6 J mice (CLEA Japan, Inc., Osaka, Japan). Mouse genotypes were determined through PCR using the primers listed in Table 1.

Table 1.

Mouse genotyping primers.

Species	Primer name	Target	Sequence, 5′-3′
Cxcl12-Cre	Cre gene	Cre gene	F: TGGAGTTTCAATACCGGAGATCA
			R: TGACAAGACAGACCCAACTGTCA
ZsGreen	Rosa26-ZsGreen	ZsGreen-induced Rosa26 allele	F: AACCAGAAGTGGCACCTGAC
			R: GGCATTAAAGCAGCGTATCC
ZsGreen	Rosa26-WT	WT Rosa26 allele	F: AAGGGAGCTGCAGTGGAGTA
			R: CCGAAAATCTGTGGGAAGTC

WT, wild type.

2.1.3. Sex as a biological variable

To account for the effects of sex hormones on skeletal formation, male mice were primarily used for analysis. However, owing to the difficulty in determining the sex of pups younger than 1 week, data from these earlier time points include both males and females.

2.2. Histological analysis

2.2.1. Tissue sample preparation

Histological analysis and tissue sample preparation were performed as previously described [24]. Mice were euthanized, and tissue samples were collected. Tissues were fixed with 4% paraformaldehyde (4% PFA) overnight at 4°C. Hard tissues were subsequently decalcified in a 10% solution of ethylenediaminetetraacetic acid disodium salt (Dojindo Laboratories, Kumamoto, Japan) at 4°C. Tissues were processed for paraffin embedding according to standard procedures, and 4-μm-thick paraffin sections were prepared. These sections were used for hematoxylin and eosin staining and immunohistochemistry. Imaging was performed using an Eclipse 600 microscope (Nikon, Tokyo, Japan), an All-in-One Fluorescence Microscope BZ-X800 (Keyence, Osaka, Japan), and a SP8 LIGHTNING confocal microscope (Leica Microsystems, Wetzlar, Germany). Image analysis was conducted using BZ-X800 software or ImageJ software (https://imagej.net/ij/download.html).

2.2.2. Immunohistochemistry

Paraffin sections were deparaffinized and rehydrated, followed by antigen retrieval as necessary. Sections were blocked with Tris-buffered saline (TBS) containing 1% bovine serum albumin and 0.1% Tween 20 (both from Sigma-Aldrich, St. Louis, MO, USA) for 30 min at room temperature. Primary antibodies were applied and incubated overnight at 4°C.

The primary antibodies used were: rat anti-endomucin monoclonal antibody (1:300, clone V·7C7, Cat# 65,495, Santa Cruz Biotechnology, Dallas, TX, USA), rabbit anti-Forkhead box C1 protein (FOXC1) monoclonal antibody (1:300, clone EPR20685, Cat# ab227977, Abcam, Cambridge, UK), rabbit anti-Mouse Osteocalcin polyclonal antibody (1:100, Cat#M173, Takara Bio Inc., Tokyo, Japan), rabbit anti-COL1A1 polyclonal antibody (1:300, Cat# ab254113, Abcam), rabbit anti-Perilipin-1 monoclonal antibody (1:300, clone D1D8, Cat# 9349, Cell Signaling Technology, Danvers, MA, USA), rabbit anti-SOX9 monoclonal antibody (1:500, clone EPR14335, Cat# ab185230, Abcam), rabbit anti-tartrate-resistant acid phosphatase (TRAP) polyclonal antibody (1:300, Cat# M183, Takara Bio Inc.), APC-conjugated rat anti-mouse TER119/erythroid cells monoclonal antibody (1:300, clone TER-119, Cat# 116,211, BioLegend, San Diego, CA, USA), APC-conjugated rat anti-mouse CD45 monoclonal antibody (1:300, clone 30-F11, Cat# 103,111, BioLegend), and goat anti-mouse leptin receptor (LepR) polyclonal antibody (1:200, Cat# AF497, R&D Systems, Minneapolis, MN, USA).

After washing with TBS, fluorescently labeled secondary antibodies were applied and incubated for 60 min at room temperature, as needed. The secondary antibodies were: Alexa Fluor 594-conjugated donkey anti-goat IgG (1:200, Invitrogen, Waltham, MA, USA), Alexa Fluor 647-conjugated goat anti-rat IgG (1:200, Invitrogen), and Alexa Fluor 647-conjugated goat anti-rabbit IgG (1:200, Invitrogen). After washing with TBS, nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI) (Sigma-Aldrich). The sections were washed with TBS and mounted with VECTASHIELD mounting medium (Vector Laboratories, Inc., Burlingame, CA, USA).

2.2.3. Bone histomorphometry

At 12 weeks of age, Cxcl12-Cre mice and wild-type control mice were subcutaneously injected with tetracycline hydrochloride (Sigma-Aldrich) at a dose of 0.01 mg/g body weight, followed 72 h later by a subcutaneous injection of calcein (Dojindo Laboratories) at 0.02 mg/g body weight. Mice were sacrificed 48 h after the second injection, and femurs were dissected and fixed in 70% ethanol. Non-decalcified blocks were prepared after Villanueva bone staining. Thin sections from the frontal plane of the distal femur and ground sections from the transverse plane of the femoral diaphysis were prepared and used for bone histomorphometry.

2.2.4. Temporal osteoblast/osteocyte quantification analysis

Femurs from 1 to 48-week-old Cxcl12-Cre; ZsG mice were used to analyze the temporal contribution of the Cxcl12⁺ lineage to osteoblasts and osteocytes populations in the diaphyseal endosteum and metaphyseal trabecular bone. Paraffin sections were prepared in the longitudinal plane along the femoral long axis, and then deparaffinized and rehydrated. After washing with TBS, nuclei were counterstained with DAPI (Sigma-Aldrich). The sections were washed with TBS and mounted with VECTASHIELD mounting medium (Vector Laboratories). At each time point, 6–9 mice were analyzed, and 3–4 sections per mouse were used for quantification. Osteoblasts were defined as cells lining the endosteal surfaces of cortical and trabecular bones, and osteocytes as cells embedded within cortical bone. The percentage of ZsG⁺ cells in each population was quantified. To ensure accurate cell counting, differential interference contrast images were acquired and used for analysis as needed.

2.3. X-ray analysis

Mice were sedated with a triple anesthetic cocktail (medetomidine hydrochloride: 0.3 mg/kg, midazolam: 4 mg/kg, and butorphanol tartrate: 5 mg/kg), and whole-body skeletal imaging was performed using an X-ray apparatus (SOFTEX M − 60, Softex Co., Ltd., Tokyo, Japan).

2.4. Parathyroid hormone (PTH) administration experiment

Human PTH (1-34) was purchased from FUJIFILM Wako Pure Chemical Corporation (Osaka, Japan). Twelve-week-old Cxcl12-Cre; ZsG mice were intraperitoneally administered PTH (80 μg/kg/injection) three times a week for 3 weeks. Mice were sacrificed 48 h after the final PTH injection for histological analysis.

2.5. Bulk gene expression analysis

Cxcl12 mRNA expression levels were measured in primary cultured BMSCs derived from 4-week-old Cxcl12-Cre mice and wild-type control mice. The proximal and distal metaphyses of the femurs and tibias were excised, and bone marrow cells were harvested from the diaphysis. The collected cells were suspended in serum-free α-Minimum Essential Medium (αMEM), pipetted, filtered through a 70 μm cell strainer, and plated. Cells were cultured for 2 weeks in αMEM supplemented with fetal bovine serum (FUJIFILM Wako Pure Chemical Corporation). Total RNA was extracted from the cultured cells using the RNeasy Mini Kit (Qiagen, Tokyo, Japan). cDNA was synthesized from the obtained RNA using ReverTra Ace® qPCR Master Mix (TOYOBO, Osaka, Japan). Cxcl12 mRNA expression levels were measured using real-time PCR (StepOne Real-Time PCR Systems, Thermo Fisher Scientific). Expression levels were normalized to the endogenous control, glyceraldehyde-3-phosphate dehydrogenase (Gapdh) mRNA. The following primers were used: Cxcl12 (Forward: CCAGAGCCAACGTCAAGCAT; Reverse: CATCCGTGCAACAATCTGAA) and Gapdh (Forward: GGTGTGAACCACGAGAAA; Reverse: AGTCGCAGGAGACAA).

2.6. Single-cell RNA sequencing

Femurs and tibias were dissected from 24-week-old Cxcl12-Cre; ZsG mice (n = 3), and surrounding soft tissues were carefully removed. The proximal and distal metaphyses of the femurs and tibias were excised, and the diaphyseal bone tissues were immersed in Hank's balanced salt solution (HBSS) (FUJIFILM Wako Pure Chemical Corporation). The collected tissues were enzymatically dissociated through incubation at 37°C for 60 min in HBSS containing 0.2% collagenase (FUJIFILM Wako Pure Chemical Corporation) and 0.2% dispase (Godo Shusei Co., Ltd., Tokyo, Japan). The isolated cell suspension was filtered through a 70 μm cell strainer, followed by red blood cell lysis using RBC Lysis Buffer (pluriSelect, Leipzig, Germany). After washing with HBSS, the cell suspension was adjusted to 1.0 × 10⁷ cells/ml in HBSS. ZsG-positive cells were isolated using a BD FACSAria™ III cell sorter (Becton-Dickinson, Piscataway, NJ, USA), and the sorted cells were resuspended in HBSS to a final concentration of 1.0 × 10⁷ cells/ml.

The sorted cell suspension was processed using the 10x Chromium X system with Chromium GEM-X Single Cell 5′ v3 reagents (10x Genomics, Pleasanton, CA, USA), according to the manufacturer's instructions. Subsequently, reverse transcription for first-strand cDNA synthesis was performed within the GEMs to generate a single-cell library with individually barcoded cDNAs for each cell. The single-cell library was sequenced using NovaSeq X Plus (Illumina, CA, USA).

The resulting sequence data underwent primary analysis with Cell Ranger 9.0.0 and secondary analysis with Seurat 5.0.0 for cell clustering and extraction of differentially expressed genes specific to each cell cluster. First, the cellranger count pipeline was used for filtering, mapping, and unique molecular identifier counting to extract the gene expression matrix. Next, using Seurat 5.0.0, cells were filtered from the sequence data based on the following criteria: Percent.mt < 20% (percentage of mitochondrial gene counts less than 20%) and nFeature_RNA > 500 (number of detected genes per cell greater than 500). After filtering, principal component analysis was performed on 2434 cells, followed by dimensionality reduction and two-dimensional plotting using uniform manifold approximation. Genes with characteristic differential expression in each cluster were extracted. Cell type annotation was performed by referencing the CellMarker database (http://xteam.xbio.top/CellMarker/) and BMSC datasets [10,13,15,25].

2.7. Statistical analysis

All experimental data are presented as means ± standard deviations (SD). Statistical analyses and graph generation were performed using Microsoft Excel (Microsoft, Redmond, WA, USA) or GraphPad Prism 10 (La Jolla, CA, USA). Two-group comparisons were performed using Student's t-test, and multiple-group comparisons were performed using Tukey's multiple comparison test. In this study, a P-value of <0.05 was considered statistically significant.

3. Results

3.1. Establishment of Cxcl12-Cre mice

We generated Cxcl12-Cre mice by targeting the Cre recombinase gene to one allele of the endogenous Cxcl12 gene (Fig. 1a). Evaluation of Cxcl12-Cre; ZsG mice, generated by crossing Cxcl12-Cre mice with a Cre-dependent ZsG reporter line, demonstrated that at 2 weeks of age, ZsG-positive (ZsG⁺) cells were confined to the bone marrow and uniformly distributed throughout the marrow cavity (Fig. 1b, left panel). ZsG⁺ cells were hardly observed among endosteal osteoblasts, osteocytes, or within the periosteum (Fig. 1b, right panel). Within the bone marrow, ZsG⁺ cells were distributed adjacent to the outer wall of endomucin⁺ sinusoids (Fig. 1c, left panel) and were positive for FOXC1, a known marker of CAR cells [26] (Fig. 1c, right panel). These findings indicated that the Cxcl12-Cre mouse line selectively labeled CAR cells with ZsG.

Fig. 1.

Generation and characterization of Cxcl12-Cre mice. a. Schematic of the Cxcl12-Cre mouse generation strategy. The Cre cDNA was inserted into exon 2 of one endogenous Cxcl12 allele. b, c. Validation of Cxcl12 expression localization. (b) Representative fluorescence images of femurs from 2-week-old Cxcl12-Cre; ZsGreen (Cxcl12-Cre; ZsG) mice. The right panel displays a high-magnification view of the diaphyseal cortical bone area. Arrowheads indicate osteoblasts. (c) Immunofluorescence staining with anti-endomucin (EMCN) and anti-FOXC1 antibodies of femurs from 2-week-old Cxcl12-Cre; ZsG mice. Arrowheads in the right panel indicate FOXC1⁺ZsG⁺ cells. GP, growth plate; TB, trabecular bone; CB, cortical bone; BM, bone marrow; Ob, osteoblast; Peri, periosteum; ∗, sinusoidal lumen. d.Cxcl12 mRNA expression levels in bone marrow stromal cells. Data represent means ± SD; n = 3. Student's t-test. e, f. Assessment of the whole body. (e) Representative gross appearance at 12 weeks of age (left) and body weight quantification (right). n = 3–20. Student's t-test. (f) Representative radiographic images of whole skeleton at 12 weeks of age (left) and femoral length quantification (right). n = 4–8. Student's t-test. No overt phenotypic differences are observed between Cxcl12-Cre and control mice. g–j. Femoral bone histomorphometry. Representative Villanueva-stained sections of the metaphysis (g) and diaphysis (i) in 12-week-old control and Cxcl12-Cre mice. (h) Histomorphometric analysis of metaphyseal trabecular bone and (j) diaphyseal cortical bone at 12 weeks of age. n = 4. Student's t-test. ∗∗p < 0.01, ∗∗∗p < 0.001. BV/TV, bone volume/tissue volume; Tb·Th, trabecular thickness; BFR/BS, bone formation rate/bone surface; N.Ob/BS, number of osteoblasts/bone surface; N.Oc/BS, number of osteoclasts/bone surface; Ct.Ar, cortical area; Ct.Wi, cortical width; Es.BFR/BS, endosteal bone formation rate/bone surface; BFR/Ct.Ar, bone formation rate/cortical area; N.Ob/Es.BS, number of osteoblasts/endosteal bone surface.

In Cxcl12-Cre mice, similar in design to the Cxcl12-GFP mice widely used in CAR cell research [10,[14], [15], [16], [17], [18],21], one allele of the endogenous Cxcl12 gene is functionally inactivated by the knock-in of the Cre gene (Fig. 1a). Therefore, we measured Cxcl12 mRNA expression levels in primary cultured BMSCs. The results indicated a downward trend in Cxcl12 mRNA expression in BMSCs from Cxcl12-Cre mice compared with those from control mice (Fig. 1d). Next, we assessed the phenotypic effects of this genetic modification. Cxcl12-Cre mice exhibited no gross abnormalities in appearance or body weight compared with the control group (Fig. 1e). Furthermore, no overt alterations were observed in the whole-body skeleton or femur length (Fig. 1f). Since CXCL12 is expressed in various organs beyond the bone marrow [27], we examined the systemic effects of endogenous Cxcl12 gene modification. No significant differences were detected in the weights of the heart, liver, spleen, or kidney between Cxcl12-Cre and control mice (Supplemental Fig. 1). Histological examination similarly revealed no obvious abnormalities in these organs (data not shown).

Detailed bone histomorphometric analysis of the femur revealed that, compared with controls, 12-week-old Cxcl12-Cre mice exhibited a significant increase in trabecular bone volume, trabecular thickness, and bone formation rate (BFR), as well as an increased osteoblast number and a significant decrease in osteoclast number in the metaphyseal trabecular bone (Fig. 1g and h). This increase in trabecular bone volume was likely attributable to the increased osteoblast number and decreased osteoclast number. In contrast, no significant differences were observed in cortical area, cortical width, BFR, or the number of endosteal osteoblasts in the diaphyseal cortical bone (Fig. 1i and j). Histologically, there were no apparent alterations in the overall bone tissue structure or cell morphology of Cxcl12-Cre mice compared with that of controls (Fig. 1g and i). Despite a localized increase in femoral trabecular bone volume, the absence of systemic skeletal abnormalities or alterations in overall tissue architecture supported the utility of this mouse model for in vivo lineage tracing of CAR cells and their progeny.

3.2. CAR cells and progeny reside within the bone marrow cavity

Lineage tracing of CAR cells was performed by labeling Cxcl12⁺ cells with ZsG using Cxcl12-Cre; ZsG mice. At embryonic day 15.5 (E15.5), the femur existed as a cartilaginous primordium lacking a marrow cavity; consequently, no ZsG⁺ cells were observed (data not shown). By E17.5, ZsG⁺ cells appeared in the periosteum and at the periphery of the nascent bone marrow (Fig. 2a). At postnatal day 1 (P1), ZsG⁺ cells were located in the bone marrow cavity, with a few remaining in the periosteum (Fig. 2b). At 1 week of age, ZsG⁺ cells were uniformly distributed throughout the bone marrow cavity, with the majority localizing adjacent to the sinusoidal outer walls (perisinusoidal ZsG⁺ cells) (Fig. 2c). Subsequently, at 2, 4, and 12 weeks of age, a distinct population of ZsG⁺ cells distal to the sinusoids (non-perisinusoidal ZsG⁺ cells) expanded progressively (Fig. 2c). The total number of ZsG⁺ cells in the diaphyseal bone marrow increased during growth (Fig. 2c and d). This increase was attributed to a rise in both perisinusoidal ZsG⁺ cells (Fig. 2e) and, more markedly, non-perisinusoidal ZsG⁺ cells (Fig. 2f). Throughout the 1- to 24-week period, these ZsG⁺ cells remained confined to the bone marrow and endosteal compartments.

Fig. 2.

Lineage tracing of CAR cells. a, b. Perinatal distribution of ZsGreen (ZsG)⁺ cells. Representative H&E staining (left panels) and fluorescence/immunofluorescence images (right panels) in femurs of Cxcl12-Cre; ZsG mice at embryonic day 17.5 (E17.5) (a) and postnatal day 1 (P1) (b). Green arrowheads indicate ZsG⁺ cells; red signals indicate endomucin (EMCN)⁺ sinusoids. DIC, differential interference contrast; GP, growth plate. c–f. Spatiotemporal dynamics of ZsG ⁺ cell localization during skeletal growth. (c) Immunofluorescence staining with anti-EMCN antibody in the metaphyseal trabecular bone area (upper row) and diaphyseal bone marrow (lower row) of femurs from Cxcl12-Cre; ZsG mice. Arrows denote perisinusoidal ZsG⁺ cells; arrowheads mark non-perisinusoidal ZsG⁺ cells. (d) Frequency of ZsG⁺ cells relative to total bone marrow cells in the diaphysis (n = 4–9). (e) Number of ZsG⁺ cells per sinusoidal perimeter (n = 4–5) and sinusoidal perimeter per unit area (n = 4–5). (f) Proportions of perisinusoidal and non-perisinusoidal ZsG⁺ cells (n = 4–9). Tukey's multiple comparison test. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. GP, growth plate; BM, bone marrow.

3.3. CAR cells differentiate into osteoblasts and adipocytes

Next, we characterized the ZsG⁺ cells of the Cxcl12⁺ lineage in the bone marrow and endosteal compartments. Immunohistochemistry revealed that Cxcl12⁺ lineage differentiated into both osteoblasts (Fig. 3a) and adipocytes (Fig. 3b). In the femurs of 12-week-old Cxcl12-Cre; ZsG mice, 19.9 ± 6.6% of the osteocalcin (Ocn)⁺ osteoblasts on the endosteal cortical bone surface were ZsG-positive (mean ± SD) (Fig. 3a). In addition, 36.8 ± 5.8% of the collagen type I alpha 1 chain (Col1a1)⁺ osteolineage cells, including osteoblasts and pre-osteoblasts on the endosteal compartments were ZsG-positive (Supplemental Fig. 2a). In contrast, Col1a1⁺ osteolineage cells in the periosteum were ZsG-negative (ZsG⁻) (Supplemental Fig. 2a). Regarding adipocytes, 91.0 ± 2.9% of the Perilipin⁺ adipocytes within the bone marrow were ZsG-positive, whereas ZsG⁺ adipocytes were hardly observed in extra-skeletal adipose tissues (Fig. 3b, Supplemental Fig. 2b). Furthermore, ZsG⁺ cells did not colocalize with SOX9⁺ chondrocytes, TRAP⁺ osteoclasts, TER119⁺ erythroid lineage cells, or CD45⁺ hematopoietic cells (data not shown). Given that mature osteoblasts and adipocytes do not express Cxcl12 [10,[13], [14], [15],21,26], these lineage tracing results suggested that the Cxcl12⁺ lineage may differentiate into endosteal osteoblasts and bone marrow adipocytes under physiological conditions in vivo.

Fig. 3.

Characterization of the CAR cell lineage. a, b. Osteogenic and adipogenic differentiation of ZsGreen (ZsG)⁺ cells. Immunofluorescence images showing expression of osteocalcin (Ocn) (a) or Perilipin (b) in femurs of 12-week-old Cxcl12-Cre; ZsG mice. (a) Arrowheads indicate Ocn⁺ZsG⁺ osteoblasts on the endosteal surface of the diaphyseal cortical bone. (b) Arrowheads indicate Perilipin⁺ZsG⁺ adipocytes residing in the bone marrow. BM, bone marrow; CB, cortical bone. c–e. Single-cell RNA sequencing (scRNA-seq) transcriptomic profiling. (c) Experimental workflow schematic. ZsG⁺ cells were isolated via fluorescence-activated cell sorting (FACS) from the long bone diaphysis of 24-week-old Cxcl12-Cre; ZsG mice for analysis. (d) Uniform Manifold Approximation and Projection (UMAP) plot depicting cell clustering. n = 2434 cells. (e) UMAP-based visualization of mRNA expression patterns for ZsGreen and representative marker genes of major clusters. Dotted lines demarcate CAR cells/osteolineage cells (clusters 8, 11) or adipolineage cells (cluster 6). Prog, progenitor; Cxcl12, C-X-C motif chemokine ligand 12; Lepr, leptin receptor; Cebpa, CCAAT/enhancer binding protein alpha; Pparg, peroxisome proliferator-activated receptor gamma; Lpl, lipoprotein lipase; Runx2, Runt-related transcription factor 2; Sp7, Sp7 transcription factor; Alpl, alkaline phosphatase; Col1a1, collagen type I alpha 1 chain; Bglap, bone gamma-carboxyglutamate protein; Pth1r, parathyroid hormone 1 receptor; S100a8, S100 calcium binding protein A8; Cxcr2, C-X-C motif chemokine receptor 2.

The ZsG⁺ population was not exclusively composed of Ocn⁺ osteoblasts and Perilipin⁺ adipocytes, indicating cellular heterogeneity, suggesting that the population encompasses both terminally differentiated cells and their precursors. To delineate the cellular composition at the molecular level, we sorted ZsG⁺ cells from the long bone diaphysis and performed single-cell RNA sequencing analysis (Fig. 3c). Based on marker gene expression patterns, we identified a CAR cell cluster (Cluster 11), a cluster containing osteolineage cells (Cluster 8), and an adipolineage cell cluster (Cluster 6) (Fig. 3d and e, Supplemental Fig. 3). These clusters exhibited Zsgreen mRNA expression, suggesting the Cxcl12⁺ lineage (Fig. 3e). A small number of Zsgreen-expressing cells were detected in clusters containing monocytes and neutrophils (Clusters 1, 7, 9). The CAR cell/osteolineage cell cluster (Cluster 8) expressed Cxcl12 and the key osteoblast transcription factors Runx2 and Sp7 (Osterix), as well as low levels of the early osteoblast markers Alpl and Col1a1, but lacked the mature osteoblast marker Bglap (Ocn) (Fig. 3e). Given that mature osteoblasts do not express Cxcl12, and that cells adjacent to osteoblasts in endosteal compartments do express Cxcl12 [10,13,14,21], this cluster may include a population of osteoblast precursors. The adipolineage cell cluster (Cluster 6) expressed the adipocyte transcription factors Cebpa and Pparg, alongside the early adipocyte marker Lpl (Fig. 3e), suggesting that these cells are adipogenic precursors. Thus, our scRNA-seq data suggested that the Cxcl12⁺ lineage in the bone marrow included precursor cells of both osteoblasts and adipocytes.

3.4. CAR cells constitute the source of osteoblasts in adulthood

We next assessed the temporal contribution of the Cxcl12⁺ lineage to the osteoblasts in the femoral diaphyseal endosteum and metaphyseal trabecular bone. For this analysis in Cxcl12-Cre; ZsG mice, we defined osteoblasts as cells lining the endosteal surfaces of cortical and trabecular bones, and osteocytes as cells within cortical bones. We then quantified the percentage of ZsG⁺ cells in each population. In the diaphyseal cortical bone, ZsG⁺ osteoblasts were virtually absent during the juvenile period from 1 to 4 weeks of age (Fig. 4a and b). However, ZsG⁺ osteoblasts emerged in adulthood, with their proportion among total osteoblasts progressively increasing to 23% at 12 weeks (corresponding to 20 human years [28]), 50% at 24 weeks (30 human years), and 66% at 48 weeks of age (42.5 human years) (Fig. 4b). Furthermore, from 12 weeks onward, ZsG⁺ osteocytes became apparent within the cortical bone, and their numbers increased with time (Fig. 4b). In contrast, ZsG⁺ osteoblasts were hardly observed in the cortical bone periosteum at any time point between 1 and 24 weeks of age (Fig. 4a and c). Similar results were observed at 48 weeks of age, although small clusters of ZsG⁺ cells were occasionally detected within the periosteum (Supplemental Fig. 4a). In the metaphyseal trabecular bone, few ZsG⁺ osteoblasts were observed at 1–2 weeks of age. Subsequently, the proportion of ZsG⁺ osteoblasts among all trabecular osteoblasts increased to 20% at 4 weeks, 42% at 12 weeks, and 74% at 24 weeks of age (Fig. 4d and e). At 48 weeks of age, the metaphyseal trabecular bone surfaces were predominantly lined by ZsG⁺ osteoblasts (Supplemental Fig. 4b and 4c). These findings suggested that osteoblasts derived from the Cxcl12⁺ lineage showed an earlier increase in trabecular bone than in cortical bone.

Fig. 4.

Contribution of the CAR cell lineage to osteoblasts. a–c. Longitudinal analysis of CAR cell-derived osteoblasts in diaphyseal cortical bone. (a) Representative fluorescence images of the femoral endosteal surface from 1- to 24-week-old Cxcl12-Cre; ZsGreen (ZsG) mice. The black dashed box indicates a high-magnification view of the 12-week-old mouse sample. Arrowheads indicate ZsG⁺ osteoblasts. (b) Quantification of ZsG⁺ osteoblast frequency among total osteoblasts on the endosteal surface and ZsG⁺ osteocyte frequency among total osteocytes in the cortical bone of the femoral diaphysis at each age (n = 6–9). (c) Representative transverse fluorescence image of the tibia from 24-week-old Cxcl12-Cre; ZsG mice. Tukey's multiple comparison test. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. BM, bone marrow; CB, cortical bone; Peri, periosteum; DIC, differential interference contrast. d, e. Longitudinal analysis of CAR cell-derived osteoblasts in metaphyseal trabecular bone. (d) Representative fluorescence images of the femoral metaphyseal trabecular bone from 1- to 24-week-old Cxcl12-Cre; ZsG mice. Arrowheads indicate ZsG⁺ osteoblasts. (e) Quantification of ZsG⁺ osteoblast frequency in the metaphyseal trabecular bone. n = 6–9. Tukey's multiple comparison test. ∗∗∗∗p < 0.0001. GP, growth plate; TB, trabecular bone. f, g. Effect of intermittent parathyroid hormone (PTH) administration in 12-week-old Cxcl12-Cre; ZsG mice. (f) Representative fluorescence images of the diaphyseal cortical bone, and quantification of ZsG⁺ osteoblast frequency in control and PTH-treated groups. (g) Representative immunofluorescence images of the central diaphyseal marrow stained with anti-endomucin (EMCN) antibody, and quantification of perisinusoidal ZsG⁺ cells per sinusoidal perimeter. n = 7. Student's t-test. ∗p < 0.05.

Collectively, these results suggested that Cxcl12⁺ cell lineage constituted the key source of osteoblasts for bone marrow-facing, endosteum-lined surfaces (i.e., endosteal cortical bone and trabecular bone). Notably, this contribution may become prominent during skeletally mature adulthood rather than during the active skeletal growth phase.

Given that the Cxcl12⁺ cell lineage became the source of osteoblasts during skeletal maturation, we hypothesized that the anabolic effects of PTH—a therapeutic agent for osteoporosis—are mediated by the lineage. Our scRNA-seq analysis confirmed that the CAR cell/osteolineage cell cluster (Cluster 8) expressed the PTH receptor Pth1r (Fig. 3e). Intermittent PTH administration exerts an anabolic effect by promoting the proliferation and differentiation of osteoblast progenitors and stimulating bone formation [29,30]. Accordingly, we intermittently administered PTH to Cxcl12-Cre; ZsG mice to trace the resulting dynamics of the Cxcl12⁺ lineage ZsG⁺ cells. PTH treatment led to a significant increase in the proportion of ZsG⁺ osteoblasts on the endosteal cortical surface, coupled with an expansion of ZsG⁺ cells in the adjacent bone marrow (Fig. 4f). In contrast, the number of perisinusoidal ZsG⁺ cells remained unchanged (Fig. 4g). These results suggest that the anabolic action of PTH promotes the differentiation of the Cxcl12⁺ lineage located near the bone surface into osteoblasts.

3.5. Spatial distribution of CAR cells and LepR ⁺ cells during development and growth

In BMSCs of adult mice, CAR cells largely overlap with cells positive for the LepR [10,31]. Therefore, we histologically examined whether CAR cells and LepR⁺ cells represent overlapping populations at each stage of development and growth. At E17.5, the ZsG⁺ cells in our model showed negligible overlap with LepR⁺ cells (Fig. 5a). LepR⁺ cells were distributed at the periosteum and primary spongiosa (Fig. 5a), consistent with previous findings [9,11]. However, during postnatal growth, the degree of colocalization, particularly in the perisinusoidal regions, progressively increased, culminating in extensive colocalization in adulthood (12 weeks of age) (Fig. 5b and c). These results suggest that CAR cells and LepR⁺ cells may initially represent distinct cell populations with different origins and characteristics; however, they appear to gradually converge in the bone marrow during growth.

Fig. 5.

Spatial distribution of CAR cells relative to LepR + cells. a. Developmental distribution of CAR cells and LepR ⁺ cells. Representative immunofluorescence images of femurs from Cxcl12-Cre; ZsGreen (ZsG) mice at embryonic day 17.5 (E17.5), stained with anti-Leptin receptor (LepR) antibody. Green arrowheads indicate ZsG⁺ cells; white arrowheads and white arrows indicate LepR⁺ cells in the periosteum and primary spongiosa, respectively. b, c. Age-dependent colocalization dynamics of CAR cells and LepR⁺ cells. (b) Representative immunofluorescence images of anti-LepR antibody staining at 1, 2, and 12 weeks of age. (c) Quantitative analysis of LepR⁺ cell frequency within the perisinusoidal ZsG⁺ population at each time point. Green arrowheads indicate LepR⁻ZsG⁺ cells; yellow arrowheads indicate LepR⁺ZsG⁺ cells. n = 6–8. Tukey's multiple comparison test. ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001.

4. Discussion

Using a Cxcl12-expression-based lineage tracing approach, we demonstrated that Cxcl12⁺ cells, considered as CAR cells [10,14,18,21,26], contributed to both osteoblasts and adipocytes under physiological conditions. Our study further revealed a spatiotemporal division of labor in bone formation: while the CAR cell lineage plays a limited role in osteogenesis during postnatal skeletal growth, the lineage serves as the source of endosteal osteoblasts during adulthood. These findings are likely important for understanding the mechanisms that maintain bone mass during adulthood. Given that CAR cells have also been identified in adult human bone marrow and are implicated in human hematopoiesis and osteogenesis [32], the fundamental principles elucidated in our mouse model are possibly conserved. These results suggest that the CAR cell lineage may represent a promising therapeutic target for skeletal disorders such as osteoporosis.

This study showed that the CAR cell lineage serves as the key source of osteoblasts in adulthood, a role spatially confined to the endosteal surfaces of cortical and trabecular bone. This contribution was demonstrated in our Cxcl12-Cre; ZsG model, in which ZsG⁺ endosteal osteoblasts were virtually absent from the cortical endosteum at 4 weeks of age but substantially increased thereafter (Fig. 4a and b). This spatiotemporal contribution is consistent with the findings of Zhou et al., who utilized Lepr-Cre mice to show that LepR⁺ cells are the major source of osteoblasts in adulthood [8]. Given the extensive overlap between CAR cells and LepR⁺ cells in adult bone marrow [10,31], these findings suggested that this stromal cell population constitutes the primary source for osteoblasts on endosteal surfaces. However, LepR⁺ progenitor cells are present in both the bone marrow and periosteum [8,9,11,22,33], contributing to bone formation by supplying osteoblasts from both endosteal and periosteal directions [8,9,11,22,33]. In contrast, the CAR cell lineage in our Cxcl12-Cre mouse model was largely confined to the bone marrow and endosteal compartments. This lineage was not detected in the growth plate cartilage and, throughout most of adulthood, was absent from the periosteum (Fig. 4a and c, Supplemental Fig. 2a and 2b), although we observed discrete clusters of ZsG⁺ cells emerging in the periosteum at 48 weeks of age (Supplemental Fig. 4a). This specific localization of the CAR cell lineage indicated that endosteal osteoblasts may be supplied by CAR cells residing within the bone marrow. This distinction can establish the Cxcl12-Cre mouse as a valuable tool for dissecting the dynamics of intramedullary BMSCs, independent of contributions from the periosteum. Therefore, this mouse model may be useful for analyzing the pathogenesis of bone diseases associated with abnormal osteoblastic differentiation of BMSCs, such as fibrous dysplasia and osteosarcoma [34,35].

Our Cxcl12-expression-based lineage tracing enabled the visualization of the CAR cell fate in vivo, providing evidence supporting the hypothesis that distinct CAR cell subpopulations have specialized functions [10,13,15]. This hypothesis posits that a “non-sinusoidal” subpopulation of CAR cells, located distal to the sinusoids and particularly in the endosteal compartment, functions as the progenitor source for osteoblasts [10,[13], [14], [15],21]. Our findings provided several lines of evidence to support this hypothesis. First, while ZsG⁺ cells were localized to the outer wall of sinusoids at 1 week of age, the proportion of ZsG⁺ cells distal to the sinusoids increased with aging, accompanied by the appearance of ZsG⁺ osteoblasts (Fig. 2c, f, 4a, 4b). Second, intermittent PTH administration induced the expansion of non-sinusoidal ZsG⁺ cells near the endosteum and ZsG⁺ osteoblasts (Fig. 4f and g). Furthermore, our scRNA-seq analysis confirmed that the PTH receptor gene, Pth1r, was expressed in the CAR cell/osteolineage cluster (Fig. 3e). Taken together, these results suggested a stepwise differentiation process for CAR cells, wherein perisinusoidal CAR cells (or their progeny) may migrate from the sinusoidal niche and subsequently differentiate into osteolineage cells and osteoblasts in endosteal compartments, thereby contributing to cortical bone formation. This cellular dynamic is consistent with previous findings demonstrating that osteoblasts during physiological bone remodeling are sourced from nearby precursor cells [7].

In addition to osteogenesis, this study showed that the CAR cell lineage serves as the source of adipocytes in the bone marrow (Fig. 3b). This provides in vivo evidence for the adipogenic potential of CAR cells under physiological conditions, a concept previously demonstrated by in vitro studies and scRNA-seq analyses [[14], [15], [16],36]. Previous studies demonstrated that the deletion of the transcription factor early B-cell factor 3 drives CAR cells to differentiate into osteoblast precursor cells [21], while the deletion of FOXC1 results in their differentiation into adipogenic precursor cells [25]. Further research is needed to understand the regulatory mechanisms of CAR cell differentiation into precursors of osteoblasts and adipocytes, as their elucidation could contribute to understanding the pathogenesis of age-related bone diseases, such as osteoporosis and marrow adiposity [36,37].

The interpretation of our results warrants consideration of several technical limitations of the Cxcl12-Cre mice used in this study. First, the recombination efficiency of the Cre-loxP system is rarely 100%; therefore, we cannot exclude the possibility that some CAR cells were not traced. Second, and more critically, the relationship between perisinusoidal and non-sinusoidal CAR cells requires further investigation. In this study, we used Cxcl12-Cre; ZsG mice, in which ZsG is expressed upon activation of the Cxcl12 locus, allowing us to trace the fate of these cells. We interpreted the early appearance of perisinusoidal ZsG⁺ cells and subsequent increase of non-sinusoidal cells as indicative of dynamic progression. However, because both perisinusoidal and non-sinusoidal CAR cells express Cxcl12 mRNA [10,18], our model cannot formally distinguish between two possibilities: (1) a direct lineage progression, where perisinusoidal cells give rise to non-sinusoidal cells, or (2) the existence of two separate and independent lineages. A previous report on cell lineage tracing using Cxcl12-CreER knock-in mice, which harbor an exogenous CreER construct under the Cxcl12 promoter, preferentially labeled the perisinusoidal subset and found only limited osteogenic contribution under homeostasis [15]. This suggested the possibility that perisinusoidal and non-sinusoidal CAR cells may represent distinct, independent cell populations.

5. Conclusions

Through lineage tracing using Cxcl12-Cre mice, this study revealed that the CAR cell lineage functions as the key source of endosteal osteoblasts and marrow adipocytes in skeletally mature mice. CAR cells may act as key regulators of adult bone homeostasis, particularly regarding the balance between bone formation and marrow adiposity. Consequently, these findings may have significant implications for both understanding the pathophysiology of age-related skeletal diseases and developing therapies for conditions such as osteoporosis.

CRediT authorship contribution statement

Takumi Shibahara: Validation, Investigation, Funding Acquisition. Katsutoshi Hirose: Conceptualization, Methodology, Validation, Investigation, Resources, Data Curation, Writing – Original Draft, Writing – Review & Editing, Visualization, Supervision, Project Administration, Funding Acquisition. Makoto Abe: Conceptualization, Methodology, Investigation, Resources, Writing – Original Draft, Writing – Review & Editing, Funding Acquisition. Yu Usami: Investigation, Resources, Supervision. Miho Hyodo: Investigation. Yoshiaki Hayashi: Methodology, Investigation, Visualization. Daisuke Motooka: Methodology, Investigation, Visualization, Supervision. Kanta Wakamori: Investigation. Narikazu Uzawa: Supervision. Shinsuke Ohba: Conceptualization, Supervision, Writing – Original Draft, Writing – Review & Editing. Satoru Toyosawa: Conceptualization, Validation, Investigation, Writing – Original Draft, Writing – Review & Editing, Visualization, Supervision, Project Administration, Funding Acquisition.

Ethics approval statement

This experiment was conducted under the approval of the University of Osaka Graduate School of Dentistry Animal Ethics Committee (Approval No.: Animal Dentistry R-05-020-0), and the University of Osaka Safety Committee for Genetic Recombination Experiments (Approval No.: 04930).

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request. The single-cell RNA-seq data presented herein have been deposited in the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO), and are available under accession number GSE334241.

Funding statement

This work was supported by JSPS KAKENHI grants (24K12884 [obtained by K·H.], 22H04922 (AdAMS) [obtained by M.A.], 21H03110 [obtained by S.T.], 24K02613 [obtained by S.T.]), JST SPRING JPMJSP2138 [obtained by T.S.], and Daiichi-Sankyo “Habataku” Support Program for the Next Generation of Researchers [obtained by K·H.].

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article: