A Wnt-mediated transformation of the bone marrow stromal cell identity orchestrates skeletal regeneration

Abstract

Bone marrow stromal cells (BMSCs) are versatile mesenchymal cell populations underpinning the major functions of the skeleton, a majority of which adjoin sinusoidal blood vessels and express C-X-C motif chemokine ligand 12 (CXCL12). However, how these cells are activated during regeneration and facilitate osteogenesis remains largely unknown. Cell-lineage analysis using Cxcl12-creER mice reveals that quiescent Cxcl12-creER⁺ perisinusoidal BMSCs differentiate into cortical bone osteoblasts solely during regeneration. A combined single cell RNA-seq analysis demonstrate that these cells convert their identity into a skeletal stem cell-like state in response to injury, associated with upregulation of osteoblast-signature genes and activation of canonical Wnt signaling components along the single-cell trajectory. β-catenin deficiency in these cells indeed causes insufficiency in cortical bone regeneration. Therefore, quiescent Cxcl12-creER⁺ BMSCs transform into osteoblast precursor cells in a manner mediated by canonical Wnt signaling, highlighting a unique mechanism by which dormant stromal cells are enlisted for skeletal regeneration.

Fig. 1. Cxcl12-creER marks a quiescent subset of CXCL12⁺LepR⁺ BMSCs.

a Structure of Cxcl12-iCreER-bGHpA bacterial artificial chromosome (BAC) transgene. b–o Short-chase analysis of Cxcl12-creER⁺ cells. b Cxcl12^GFP/+; Cxcl12-creER; R26R^tdTomato (pulsed at P21) distal femurs with growth plates on top. Gray: DIC. Scale bar: 500 µm. n = 3 mice. c, d Immunostaining for endomucin (Emcn, c) and CD31 (d). Sn: sinusoids, At: arterioles. Arrows: sinusoid-attaching GFP^hightdTomato⁺ cells. Gray: DAPI. Scale bar: 10 µm. n = 5 mice. e Percentage of cells attaching Emcn⁺ sinusoids among GFP^hightdTomato^neg (green) or GFP^hightdTomato⁺ (red) cells. n = 5 mice. f Flow cytometry analysis of CD45/Ter119/CD31^neg bone marrow cells. Blue lines: control cells. n = 5 mice. g Leptin receptor (LepR) expression based on flow cytometry analysis. Blue line: isotype control. n = 3 mice. h LepR staining. Gray: DAPI (Center), DIC (metaphysis and endosteal zone). Scale bar: 20 µm. i Quantification based on histology. Percentage of BMSCs expressing Cxcl12-GFP (green dots), Cxcl12^CE-tdTomato (red dots) and LepR (purple dots). n = 7 fields from four mice (central marrow and endosteal surface), n = 5 fields from 4 mice (metaphysis). j Percentage of CD45/Ter119/CD31^neg subpopulations in S/G2/M phase. n = 5 mice. k Intracellular CXCL12 protein levels of CD45/Ter119/CD31^neg subpopulations. MFI: median fluorescence intensity. n = 4 mice. l EdU serial pulse assay of Cxcl12-creER⁺ cells. Cxcl12-creER; R26R^tdTomato mice were pulsed with tamoxifen at P21, then with EdU every 8 h for 3 days prior to analysis. Percentage of EdU⁺ cells among GFP^hightdTomato^neg (green) or GFP^hightdTomato⁺ (red) cells. n = 5 mice. m Intracellular SCF protein levels of CD45/Ter119/CD31^neg subpopulations. MFI median fluorescence intensity. n = 5 mice. n, o Col1(2.3 kb)-GFP; Cxcl12-creER; R26R^tdTomato. n Left panel: trabecular bone. Right panel: endosteal space. Gray: DIC. Scale bar: 20 µm. n = 4 mice. o Flow cytometry analysis of CD45/Ter119/CD31^neg cells harvested from Col1(2.3 kb)-GFP; Cxcl12-creER; R26R^tdTomato at P28 (pulsed at P21) bone marrow. Blue lines: control cells. n = 4 mice. ****p < 0.0001, ***p < 0.001, **p < 0.01, *p < 0.05, two-tailed, Mann–Whitney’s U test (e, l). Two-tailed, one-way ANOVA followed by Tukey’s post hoc test (i–k, m). All data are presented as mean ± s.d. Source data are provided as a Source Data file.

Fig. 2. Single-cell characterization of Cxcl12-creER⁺ BMSCs.

a Single-cell RNA-seq analysis, UMAP-based visualization of major classes of Cxcl12-GFP⁺ cells (Clusters 0–8). Red dotted contour: reticular cells (Cluster 0). Center panels: tdTomato expression, feature plot (top), violin plot (Clusters 0–2) (bottom). Right panels: feature plots. Blue: high expression. n = 5858 cells merged from two biological replicates (Mouse #1: 2028 cells, Mouse #2: 3830 cells). b Split-dot-based visualization of representative gene expression in Cluster 0 (C-0, reticular cells) and Cluster-1 (C-1, pre-osteoblasts). Upper: Mouse #1, Lower: Mouse #2. Circle size: percentage of cells expressing a given gene in a given cluster (0–100%), color density: expression level of a given gene. c–f CFU-F assay. c Cxcl12^CE-tdTomato⁺ (left) and Ubc^CE-tdTomato⁺ (right) bone marrow cells at P28 (pulsed at P21). Scale bar: 5 mm. d percentage of tdTomato⁺ colonies among total CFU-Fs. n = 4 (Cxcl12^CE), n = 7 (Ubc^CE) mice. **p < 0.01, two-tailed, Mann–Whitney’s U test. Data are presented as mean ± s.d. e Survival curve of individual tdTomato⁺ clones over serial passages. n = 26 (Cxcl12^CE), n = 48 (Ubc^CE) clones. Gehan–Breslow–Wilcoxon test. f In vitro trilineage differentiation assay of Cxcl12^CE-tdTomato⁺ clones (Passage 2–7). Upper panels: Adipogenic conditions. LipidTOX staining (left). Green: LipidTOX-Alexa488, red: tdTomato. Oil red O staining (Right). Center panels: Osteogenic conditions. Green: Col1(2.3 kb)-GFP, red: tdTomato. Alizarin Red staining (right). Lower panel: Chondrogenic conditions, Alcian Blue staining. Insets: differentiation medium negative controls. Scale bar: 200 µm. n = 12 clones. Source data are provided as a Source Data file.

Fig. 3. Cell-fate analysis reveals dormancy of Cxcl12-creER⁺ BMSCs.

a–l Long-chase analysis of Cxcl12-creER⁺ cells. a–c Col1(2.3 kb)-GFP; Cxcl12-creER; R26R^tdTomato distal femurs with growth plates on top, analyzed at 1Y. Pulsed at P3 (a), P21 (b), and 8 W (c). Yellow double arrows: tdTomato^neg area beneath growth plates, white brackets: tdTomato⁺ domain in marrow space. Scale bar: 500 µm. Gray: DIC. d Quantification of tdTomato⁺ reticular cells in marrow space at 1Y, based on distance from growth plate. Pulsed at P3 (blue line), P21 (green line) or 8 W (red line). n = 6 (P3, P21-pulsed), n = 5 (8W-pulsed) mice. e Cxcl12^GFP/+; Cxcl12-creER; R26R^tdTomato bone marrow at 1 y (pulsed at P21), central marrow space. Gray: DAPI. Scale bar: 20 µm. n = 3 mice. f Quantification of tdTomato⁺ reticular cells in marrow space during extended chase, pulsed at 8 W. n = 4 (9 W), n = 5 (3 M), n = 3 (6 M, 1 y, and 1 y 6 M) mice. g Col1(2.3 kb)-GFP; Cxcl12-creER; R26R^tdTomato bone marrow at 1 y (pulsed at P21), endosteal space. Scale bar: 20 µm. n = 4 mice. h Percentage of Col1(2.3 kb)-GFP⁺tdTomato⁺ cells per total Col1(2.3 kb)-GFP⁺ cortical bone osteoblasts. (Left): flow cytometry. n = 4 (+1 W), n = 3 (+8 W) mice. (Right): labeled at P3, P21, 8 W, and 5 M, analyzed at 1 y 6 M. OB osteoblasts, OCY osteocytes. n = 3 per group. i Col1(2.3 kb)-GFP; Cxcl12-creER; R26R^tdTomato distal femurs at 9 W (pulsed at 8 W). 1 (endosteal area): boxed area on left panel. Gray: DIC. Scale bar: 500 µm (left), 20 µm (right). n = 3 mice. j Cxcl12-creER; R26R^tdTomato distal femur bone marrow at 1Y (pulsed at 8 W). Lower left panel: central marrow space with Cxcl12^GFP/+, right panels: endosteal space and trabecular bone with Col1(2.3 kb)-GFP. Gray: DIC. Scale bar: 20 µm. n = 3 mice. k, l Percentage of Col1(2.3 kb)-GFP⁺tdTomato⁺ cells per total Col1(2.3 kb)-GFP⁺ cells by using flow cytometry (k). n = 3 (+2 days), n = 4 (+1 W), n = 7 (+4 W) and n = 6 (+8 W) mice. l Labeled at 8 W, analyzed at 9 W, 3 M, 6 M, 1 y, and 1 y 6 M. n = 3 (9 W, 3 M, 1 y, and 1 y 6 M), n = 4 (6 M) mice. ****p < 0.0001, ***p < 0.001, two tailed, one-way ANOVA followed by Tukey’s post hoc test (f, h, k, l). Two-tailed, Mann–Whitney’s U test (h). All data are presented as mean ± s.d. Source data are provided as a Source Data file.

Fig. 4. Cortical bone regenerative responses of Cxcl12-creER⁺ BMSCs.

(a–n) Fates of Cxcl12-creER⁺ cells or Osx-creER⁺ cells during cortical bone regeneration after drill-hole injury. a Experimental schemes for drill-hole cortical injury model. b Timeline for tamoxifen injection, drill-hole injury and analysis. Cxcl12^GFP/+; Cxcl12-creER; R26R^tdTomato (c, f) or Col1(2.3 kb)-GFP; Cxcl12-creER; R26R^tdTomato (g–k) femur diaphyseal cortical bones (pulsed at 7 days before surgery) with periosteum on top. CB cortical bone, BM bone marrow. Gray: DIC. n = 5 mice per each group. c, d Injured area at 2 days after injury, femurs with EdU administration shortly before analysis. Scale bar: 200 µm (c), 50 µm (d). e Quantification of EdU⁺tdTomato⁺ cells. n = 6 (Intact), n = 7 (Injured). EdU was administered twice (6 and 3 h) prior to analysis. **p < 0.01, two-tailed Mann–Whitney’s U test. f, g Injured area at 7 days after injury, femurs with EdU administration shortly before analysis. Scale bar: 200 µm (f), 20 µm (g). h, i Injured area at 14 days after injury, femurs with EdU administration shortly before analysis. Scale bar: 200 µm (h), 20 µm (i). j–l Injured cortical bone at 56 days after injury. Scale bar: 200 µm (h, l), 20 µm (i). m Number of tdTomato⁺ cells within intact and injured are of Cxcl12-creER; R26R^tdTomato cortical bone (blue and red dots, respectively) or injured area of Osx-creER; R26R^tdTomato cortical bone (orange dots). n = 5 mice per group. **p < 0.01, two-tailed, Mann–Whitney’s U test. n Injured cortical bone at 56 days after injury, Ocn-GFP; Osx-creER; R26R^tdTomato femurs. All data are presented as mean ± s.d. Source data are provided as a Source Data file.

Fig. 5. Cxcl12-creER⁺ BMSCs respond to bone marrow ablation-induced osteogenesis.

a–c Experimental schemes and procedures of bone marrow ablation model. a Right femurs were operated, while left femurs were unoperated and used as internal control. (b, i–vi): Step-by-step demonstration of surgical procedures. c Timeline for tamoxifen injection, ablation surgery and analysis. d–j Col1(2.3 kb)-GFP; Cxcl12-creER; R26R^tdTomato distal femur bone marrow (pulsed at 7 days before surgery) with growth plates on top. Three days (d), 7 days (e), and 14 days after surgery (f, h). h Ablated area, boxed area in (f). Scale bar: 50 µm. g contralateral control side. Gray: DIC. Scale bar: 500 µm. n = 4 (Day 3), n = 3 (Day 7), n = 5 (Day 14), n = 3 (Conrol) mice. i Flow cytometry analysis of CD45/Ter119/CD31^neg bone marrow cells. Lower panels: fold change of total tdTomato⁺ cells (left) or Col1-GFP⁺tdTomato⁺cells (right) in ablated side over contralateral control side. Scale bar: 20 µm. n = 6 mice. **p < 0.01, two-tailed, Mann–Whitney’s U test. Data are presented as mean ± s.d. j Regenerated cartilaginous tissue adjacent to growth plate. Gray: DIC. Source data are provided as a Source Data file.

Fig. 6. Injury-induced identity conversion of Cxcl12-creER⁺ BMSCs.

a–d Single-cell RNA-seq analyses of lineage-traced Cxcl12-creER⁺ cells. a Cxcl12^CE-tdTomato⁺ cells were isolated from contralateral (CONT) and ablated (ABL) femurs separately at 7 days after marrow ablation. Two datasets were integrated by Seurat/CCA. b, c UMAP-based visualization of major classes of mesenchymal Cxcl12^CE-tdTomato⁺ cells (Cluster 0–11). Left panel, red dotted contour: reticular (Clusters 0–2). Lower panels: feature plots. Blue: high expression. Right panel: cells colored by conditions (CONT, ABL). Biological replicate 1, n = 2939 cells (CONT: 2066 cells, ABL: 873 cells), pooled from n = 7 mice. d Split-dot-based visualization of representative gene expression. Clusters 0 and 1 (reticular) and Cluster 6 (pre-osteoblast) shown for reticular-signature genes (Cxcl12, Adipoq) and osteoblast-signature genes (Col1a1, Bglap, Spp1, Alpl, Postn, Tnc, Pth1r, Sox9, Runx2, Sp7, Atf4). *p < 0.0001, Wilcoxon rank sum test. Circle size: percentage of cells expressing a given gene in a given cluster (0–100%), Color density: expression level of a given gene. e, f CFU-F assay of Cxcl12^CE-tdTomato⁺ bone marrow cells, harvested from contralateral (CONT, left) and ablated (ABL, right) femurs at Day 7 (pulsed at Day-7). Scale bars: 5 mm. g Percentage of tdTomato⁺ colonies among total CFU-Fs. n = 5 mice per group. *p < 0.05, two-tailed, Mann–Whitney’s U test. Data are presented as mean ± s.d. Source data are provided as a Source Data file.

Fig. 7. Cxcl12-creER⁺ BMSCs transform into a skeletal stem cell-like state.

a Inducible partial cell ablation of Cxcl12-creER⁺ stromal cells using an inducible diphtheria toxin fragment A (DTA) allele. Flow cytometry-based verification of iDTA-mediated cell ablation of Cxcl12-creER⁺ stromal cells. n = 3 (tdT), n = 6 (DTA) mice. *p < 0.05, two-tailed, Mann–Whitney’s U test. b, c DTA-mediated ablation of Cxcl12-creER⁺ cells during cortical bone regeneration. Rosa26^{lsl-tdTomato/iDTA} (Control) and Cxcl12-creER; Rosa26^{lsl-tdTomato/iDTA} (DTA) femur diaphyseal cortical bones, 14 days after cortical drill-hole surgery (pulsed at 7 days before surgery). b Cortical bone injured areas. Upper panels: alkaline phosphatase (ALP) staining. Gray: DIC. Scale bar: 200 µm. Lower panels: 3D-µCT. Scale bar: 1 mm. c Bone volume/tissue volume (BV/TV, left) and bone mineral density (BMD, right) of the injured area. n = 6 (Control), n = 4 (DTA) mice. **p < 0.01, two-tailed, Mann–Whitney’s U test. d–i Single-cell RNA-seq analyses of lineage-traced Cxcl12-creER⁺ cells. Cxcl12^CE-tdTomato⁺ cells were isolated from ablated (ABL) femurs at 14 days after marrow ablation. e UMAP-based visualization of major classes of mesenchymal Cxcl12^CE-tdTomato⁺ cells (Clusters 0–9). Red dotted contour: reticular cells, green dotted contour: preosteoblasts. n = 1825 cells pooled from n = 7 mice. f Feature plots of skeletal stem cell marker genes. g Heatmap of representative pseudotime-dependent genes. h Pseudotime and single cell trajectory analysis by Monocle. i Most significant Gene Ontology (GO) terms (Biological Process) based on DAVID functional annotation of differentially expressed genes (DEGs). Black: general, red: Wnt signaling-related, blue: bone-related terms. All data are presented as mean ± s.d. Source data are provided as a Source Data file.

Fig. 8. Wnt-mediated transformation orchestrates bone regeneration.

a, b Functional analysis of Wnt/β-Catenin signaling in Cxcl12-creER⁺ cells during cortical bone regeneration. Cxcl12-creER; Ctnnb^+/+; R26R^tdTomato (Control, left), Cxcl12-creER; Ctnnb^fl/fl; R26R^tdTomato (βcat cKO, right) femur diaphyseal cortical bones, 14 days after cortical drill-hole surgery (pulsed at 7 days before surgery). a Cortical bone injured areas. Upper panels: alkaline phosphatase (ALP) staining. Gray: DIC. Scale bar: 200 µm. Lower panels: 3D-µCT. Scale bar: 1 mm. b Bone volume/tissue volume (BV/TV, left) and bone mineral density (BMD, right) of the injured area. n = 4 (Control), n = 8 (Cxcl12-creER; Ctnnb^fl/+; R26R^tdTomato, βCat cHet) mice, n = 8 (βCat cKO) mice. c Functional analysis of Sox9 and Runx2 in Cxcl12-creER⁺ cells during cortical bone regeneration. Cxcl12-creER; Sox9^+/+ or Runx2^+/+;R26R^tdTomato (Control, left), Cxcl12-creER; Sox9^fl/+ or Runx2^fl/+;R26R^tdTomato (Sox9 cHet or Runx2 cHet, center) and Cxcl12-creER; Sox9^fl/fl or Runx2^fl/fl; R26R^tdTomato (Sox9 cKO or Runx2 cKO, right) femur diaphyseal cortical bones, 14 days after cortical drill-hole surgery (pulsed at 7 days before surgery). BV/TV (left) and BMD (right) of the injured area. n = 7 (Sox9 Control), n = 4 (Sox9 cHet, cKO) mice. n = 6 (Runx2 Control, cKO), n = 8 (Runx2 cHet) mice. d Functional analysis of Apc in Cxcl12-creER⁺ cells during cortical bone regeneration. Cxcl12-creER; Apc^+/+; R26R^tdTomato (Control, left), Cxcl12-creER; Apc^fl/+; R26R^tdTomato (Apc cHet, right) femur diaphyseal cortical bones, 7 days after cortical drill-hole surgery (pulsed at 7 days before surgery). BV/TV (left) and BMD (right) of the injured area. n = 6 (Control), n = 4 (Apc cHet) mice. e, f Functional analysis of Wnt/β-Catenin signaling in Dlx5-creER⁺ cells during cortical bone regeneration. Ctnnb^fl/fl; R26R^tdTomato (Control, left), Dlx5-creER; Ctnnb^fl/fl; R26R^tdTomato (Dlx5-βcat cKO, right) femur diaphyseal cortical bones, 14 days after cortical drill-hole surgery (pulsed at 7 days before surgery). e 3D-µCT. Lower panels: BV/TV (left) and BMD (right) of the injured area. Scale bar: 1 mm. n = 6 mice per group. ****p < 0.0001, ***p < 0.001, *p < 0.05, two-tailed, one-way ANOVA followed by Tukey’s post hoc test (b, c). Two-tailed, Mann–Whitney’s U test (d, f). All data are presented as mean ± s.d. Source data are provided as a Source Data file.

Fig. 9. A Wnt-mediated conversion of stromal cell identity orchestrates skeletal regeneration.

a Quiescent CXCL12⁺ bone marrow stromal cells (BMSCs) in perisinusoidal space are marked by Cxcl12-creER upon tamoxifen injection. These cells are preferentially localized in a non-endosteal location of the marrow space, associated with a reticular morphology, abundant cytokine expression and little colony-forming activities. These cells do not become cortical bone osteoblasts in homeostasis, but can be readily converted to marrow adipocytes. b A proposed diagram of the bone marrow stromal cell lineage. In homeostasis, CXCL12^high reticular cells and COL1A1^high osteoblasts represent two opposite differentiated states of the bone marrow stromal cell lineage. The transition between these two cell types is continuous, wherein intermediate-state cells represent a precursor cell population. c Wnt-mediated cellular plasticity of CXCL12⁺ reticular cells during injury responses. In homeostasis, Cxcl12-creER⁺ reticular cells are in a Wnt-inhibitory state by expressing potent Wnt inhibitors, such as Sfrp1, Sfrp2, and Sfrp4. During injury responses, Cxcl12-creER ⁺ reticular cells are brought into a reticular-osteoblast hybrid state by injury-induced cues, associated with upregulation of osteoblast-signatures and a critical canonical Wnt component, β-catenin. Subsequently, Wnt/β-catenin signaling induces cell identity conversion of Cxcl12-creER⁺ reticular cells into pre-osteoblasts, which are featured by a skeletal stem cell-like state with partial commitment to the osteoblast lineage. These cells can be recruited to cortical bone defect and functionally contribute to skeletal regeneration.