Specialized post-arterial capillaries facilitate adult bone remodelling

Abstract

The vasculature of the skeletal system is crucial for bone formation, homoeostasis and fracture repair, yet the diversity and specialization of bone-associated vessels remain poorly understood. Here we identify a specialized type of post-arterial capillary, termed type R, involved in bone remodelling. Type R capillaries emerge during adolescence around trabecular bone, possess a distinct morphology and molecular profile, and are associated with osteoprogenitors and bone-resorbing osteoclasts. Endothelial cell-specific overexpression of the transcription factor DACH1 in postnatal mice induces a strong increase in arteries and type R capillaries, leading to local metabolic changes and enabling trabecular bone formation in normally highly hypoxic areas of the diaphysis. Indicating potential clinical relevance of type R capillaries, these vessels respond to anti-osteoporosis treatments and emerge during ageing inside porous structures that are known to weaken compact bone. Our work outlines fundamental principles of vessel specialization in the developing, adult and ageing skeletal system.

Fig. 1. scRNA-seq analysis of bone endothelial cells.

a, Schematic overview of scRNA-seq work flow. b, Uniform Manifold Approximation and Projection (UMAP) visualization of total bone ECs, with colour-coded subclusters of bmECs (n = 12,002), mpECs (n = 6,307), aECs (n = 2,809), rECs (n = 1,768) and pECs (n = 329) with a total n = 23,215 cells. c, UMAP plot of total ECs separated by age group (juvenile = 7,973, adult = 10,721 and aged = 4,521). d, Bar plots showing the proportion of cells in each subcluster per age group. e, Feature plots show markers for subclusters of bone endothelial cells: bmECs/type L ECs expressing Rgs4, Gpr126, Vcam1, Adamts5 and Stab2; mpECs/type H ECs expressing Apln, Piezo2, Ramp3, Aplnr and Lamb1; aECs expressing Sema3g, Gja4, Fbln5, Vegfc and Bmx; rECs/type R expressing C1qtnf9, Cav1, Fmo2 and Aqp7; and proliferating pECs expressing Top2a, Neil3, Hist1h2ae and Mki67. f, Heatmap showing the expression of the top four markers for each cluster. g, Heatmap depicting expression of Cdh5 and the absence of BM stromal cell markers (Pdgfra and Pdgfrb) as well as absence of lymphatic EC markers (Prox1 and Lyve1) in bone EC subclusters. h, Heatmap showing selected markers distinguishing EC subclusters. Black box highlights combination of markers allowing the identification of rECs (Flt4⁻Emcn⁺Cav1⁺). Colour bars illustrate the expression level (e, g) and expression level (log₂ fold change) (f, h). Data in b,d–h are derived from an integrated scRNA-seq dataset, whereas data in c show individual samples from different age groups.

Fig. 2. Identification of remodelling endothelial cells.

a, UMAP plots showing colour-coded subclusters in juvenile and adult bone ECs. Dashed black line encompasses rEC clusters. Red arrowhead indicates expansion of type R ECs in adult. b, Bar plots showing colour-coded subclusters in juvenile and adult bone ECs. Relative representation (in %) of rECs and aECs is indicated. c, High-magnification images of femurs immunostained for EMCN (red), VEGFR3 (blue) and CAV1 (green) showing the emergence of EMCN⁺VEGFR3⁻CAV1⁺ (yellow) rECs (white arrowheads) around trabecular bone (TB). d, Two-photon microscopic image of immunostained EMCN⁺VEGFR3⁻ rECs (white arrowheads) around TB visualized by second-harmonic generation. e, Representative confocal images of 3-, 6-, 8- and 12-week-old femurs immunostained for EMCN (red) and VEGFR3 (green). White arrowheads indicate increasing age-dependent abundance of EMCN⁺VEGFR3⁻ rECs around TB. f, Quantitation of rECs (EMCN⁺VEGFR3⁻ area) showing the age-dependent increase in the 3-, 6-, 8- and 12-week-old TB relative to samples from 3 weeks. n = 3 mice per group. Mean ± s.e.m. P values, one-way analysis of variance (ANOVA). g, Schematic representation of type R capillary expansion in postnatal and adult long bone. Data in a show individual samples from different age groups, whereas b is based on integrated scRNA-seq data. Source data

Fig. 3. Properties and origin of type R capillaries.

a, Perfusion of EMCN⁺VEGFR3⁻ type R capillaries (arrowheads) near TB demonstrated by injected 2,000 kDa TRITC dextran (yellow) in 12-week-old wild-type femur. b, Representative confocal images of 12-week-old Efnb2-H2B-GFP (green) femur section co-stained for EMCN (red) and VEGFR3 (blue). Efnb2⁺EMCN⁺ rECs (white arrowheads) are connected to Efnb2⁺EMCN⁻ arterioles and arteries (green arrowheads). c,d, Type III collagen (COL3A1) (c) and type IV collagen (COL4A1) (d) are tightly associated with EMCN⁺VEGFR3⁻ type R capillaries (arrowheads) in 12-week-old wild-type femur, whereas the surrounding sinusoidal vessels show a loose reticular fibre network. e, High-magnification images showing filopodia (arrowheads) extending from EMCN⁺VEGFR3⁻ rECs around 12-week-old TB. f, Proliferating rECs (white arrowheads) near 6-week-old TB. Cdh5-mTnG reporter (nGFP, red) shows EC nuclei co-stained with KI-67 (green). g, Scheme of genetic fate-mapping strategy. 4-OHT administration (1 mg per mouse) is indicated by black arrow and red arrows mark time points of analysis. h,i, Tile-scan confocal images (h) and higher magnification of insets (i) showing fate-tracked Flt4-CreERT2 R26-mTmG (GFP, green)-labelled ECs in femur at 8 weeks (48 h after Cre induction) and 12 weeks (4 weeks after Cre induction). White arrowheads indicate EMCN⁺CAV1⁺ rECs and yellow arrowheads mark GFP⁺ traced rECs. j, Quantitative analysis of GFP⁺ rECs (EMCN⁺CAV1⁺FLT4-GFP⁺) at 48 h and 4 weeks post-induction, respectively. n = 3 mice per group. Mean ± s.e.m. P values were obtained using an unpaired two-tailed t-test. k, Schematic illustration of genetic fate mapping of rECs in Flt4-CreERT2 R26-mTmG femur. Source data

Fig. 4. Fate mapping of Aplnr⁺ ECs.

a, UMAP plot (derived from integrated scRNA-seq data of all age groups) illustrating Aplnr expression in rECs (arrow) but not aECs (arrowhead). Scheme of genetic fate-mapping strategy with Aplnr-CreERT2 mice. 4-OHT administration (1 mg per mouse) is indicated by black arrows and red arrows mark time points of analysis. b,c, High-resolution confocal images and tile-scan overview images of fate-tracked Aplnr-CreERT2 R26-mTmG (GFP, green) ECs in femur at 3 weeks (72 h after 4-OHT administration and Cre activation), 6 weeks (3 weeks after Cre activation) (b) and at 12 weeks (9 weeks after Cre activation) (c). Small panels show higher magnifications of TB and diaphyseal area. White arrowheads mark GFP⁺EMCN⁺ bmECs, yellow arrowheads GFP⁺CAV1⁺EMCN⁺ rECs, green arrowheads GFP⁻CAV1⁺ aECs and red arrowheads fate-tracked GFP⁺CAV1⁺EMCN⁻ ECs inside arteries. White dashed lines in b and c indicate arteries and arterioles. d, Bar plots showing the proportion of area covered by CAV1⁺GFP⁺ rECs and aECs, CAV1⁺GFP⁻ aECs and CAV1⁻GFP⁺ ECs at P21, 6 weeks and 12 weeks post-induction. e, Schematic illustration of genetic fate mapping of rECs in Aplnr-CreERT2 R26-mTmG femur. Source data

Fig. 5. Coupling of rECs and remodelling bone.

a,b, High-resolution confocal images showing EMCN⁺VEGFR3⁻ rECs (arrowheads) around TB in relation to RUNX2⁺ osteoprogenitors (green) (a) and ATP6V1B1/B2⁺ osteoclasts (green) (b). c,d, Maximum-intensity projection showing Sp7-mCherry⁺ osteoblasts (yellow) and ATP6V1B1/B2⁺ osteoclasts (green) in in relation to EMCN⁺VEGFR3⁻ rECs around 6-week-old femoral TB (c). Single xy plane (d). Green arrowheads mark rECs near osteoclasts and yellow arrowheads near osteoblasts. e,f, Distribution of EMCN⁺VEGFR3⁻ type R vessels in relation to Sp7-mCherry⁺ osteoblasts (yellow arrowheads) and ATP6V1B1/B2⁺ osteoclasts (green arrowheads) around 6-week-old and 12-week-old TB (e). Isolated xy plane (f). g, Confocal images showing VEGFR3⁻VEGFR2⁺ ECs (white arrowheads) around TB in young and aged patient samples. VEGFR3 (yellow), VEGFR2 (red) and nuclei (DAPI, blue). DAPI, 4,6-diamidino-2-phenylindole.

Fig. 6. Loss of Dach1 reduces rEC number and trabecular bone.

a, UMAP plot (integrated scRNA-seq dataset of all age groups) showing Dach1 expression in rECs (arrow). Colour bar illustrates the expression level. b, Representative confocal images of DACH1 immunostaining (green) in 6-week-old Cdh5-mTnG (red) reporter femur. DACH1⁺ rECs near TB (green) are marked by white arrowheads. c, Scheme of tamoxifen-induced EC-specific Dach1 inactivation. d, Tile-scan confocal images of EMCN⁺VEGFR3⁻CAV1⁺ rECs (white arrowheads) and EMCN⁻VEGFR3⁻CAV1⁺ aECs (yellow arrowheads) in 12-week-old Dach1^iΔEC loss-of-function and littermate control femur. Growth plate (GP) is indicated. e, Quantitation of EMCN⁺ vessel density, EMCN⁺VEGFR3⁻ vessel density, and number of CAV1⁺ arteries in 12-week-old Dach1^iΔEC and control femur (n = 3–4 female mice per group). Mean ± s.e.m. P values, unpaired two-tailed t-test. Emcn⁺ vessel density plotted with Welch’s correction. f, High-magnification images of metaphysis near GP (left) and of EMCN⁺VEGFR3⁻ vessels (white arrowheads; right) around TB in 12-week-old male Dach1^iΔEC and control femurs. White dashed lines in d and f indicate type H area. g, Representative 3D reconstruction of µCT measurements of 12-week-old Dach1^iΔEC and control femoral bone. Dashed yellow lines indicate area analysed. h, Quantitation shows relative bone volume, represented as bone volume/tissue volume (BV/TV) and trabecular thickness (per mm) (n = 3–4 female mice per group and n = 3 male mice per group). Mean ± s.e.m. P values were plotted using an unpaired two-tailed t-test. Source data

Fig. 7. EC-specific Dach1 overexpression enhances trabecular bone formation.

a, Schematic representation of tamoxifen-inducible Dach1 overexpression in Aplnr⁺ ECs. b, Tile-scan confocal images showing expansion of CAV1 (green, white arrowheads) immunostained arteries and type R capillaries in the Dach1 gain-of-function (Dach1^OE) femur relative to littermate control at postnatal day 30 (P30). c, High-magnification confocal images of EMCN (red) and VEGFR3 (green) immunostained femurs showing the expansion of type R capillaries near TB in Dach1^OE mutants relative to littermate control. White arrowheads indicate type R capillaries. d, Quantitation of arteries/arterioles and CAV1⁺ area in control and Dach1^OE mutants. n = 3 mice per group. Mean ± s.e.m. P values were obtained by unpaired two-tailed t-test. e, Representative three-dimensional (3D) reconstruction of µCT measurements of P30 Dach1^OE and control femoral TB. f, Quantitation of parameters: bone volume/tissue volume (BV/TV), trabecular number represented by number of trabeculae per millimetre, connectivity density (connectivity density per cubic millimetre) and cortical thickness (mm). n = 3 mice per group. Mean ± s.e.m. P values were obtained by an unpaired two-tailed t-test. g, Schematic overview of scRNA-seq workflow for non-haematopoietic cells from Dach1^OE and littermate control long bone. h,i, UMAP plots of BMSCs (n = 10,315) with colour-coded subclusters, namely osteoblasts (OBs), osteocytes (OCYs), septoclasts (SCs), proliferating BMSCs (pBMSCs), mpMSCs and dpMSCs (dpMSCs1 and dpMSCs2) (h). UMAP distribution of Dach1^OE and control BMSCs, as indicated by colour (i). j, Bar plots showing proportion of cells in Dach1^OE and control BMSC subclusters. k, Heatmap showing the top three marker genes for each BMSC subcluster. l, Heatmap illustrating differentially expressed genes in Dach1^OE and control BMSCs related to hypoxia, growth factors, adipogenic and osteogenic transcription factors, and regulators of osteogenesis. Texts in red are the areas of interest. Colour bars in k and l illustrate the expression level (log₂ fold change). Data in h–l are derived from an integrated scRNA-seq dataset of two conditions. Source data

Fig. 8. Age-associated changes in bone remodelling and regeneration.

a, UMAP visualization of bone ECs from 75-week-old bone with colour-coded subclusters. b, Bar plot showing the proportion of cells from each EC subcluster in adult and aged bone. Percentage (%) differences are indicated for rECs and aECs. Data are representative of an individual sample (Methods) (a) and integrated scRNA-seq data from all age groups (b). c,d, High-magnification confocal images of TB (c) and compact bone (CB) (d) immunostained for EMCN (red), VEGFR3 (blue) RUNX2 (yellow arrowheads, marking osteoprogenitors) and ATP6V1B1B2 (green arrowheads, marking osteoclasts). e, High-magnification two-photon microscopy images of CB immunostained for EMCN (red) and VEGFR3 (blue) together with second-harmonic generation (white) showing the changes during ageing. f, Confocal tile-scan images showing EMCN⁺VEGFR3⁻ rECs in untreated 12-week-old and 75-week-old mice femur and in response to treatment with alendronate (ALN) or PTH, as indicated. Note expansion of EMCN⁺ VEGFR3⁻ vessels in CB in response to ageing or treatments (arrowheads). g, Representative confocal images showing EMCN⁺VEGFR3⁻ rECs and OSTERIX (green) immunostaining in 75-week-old control, ALN-treated or PTH-treated CB (white arrowheads, marking type R capillaries during ageing). White dashed lines in d–g indicate compact bone (CB) area. h,i, Quantitation of EMCN⁺CAV1⁺ vessel density across age groups (h), EMCN⁺VEGFR3⁻ vessel density in 75-week-old control, ALN-treated or PTH-treated groups, and number of OSTERIX⁺ cells in CB after ALN and PTH treatment compared with respective controls (i). n = 3 mice per group. Mean ± s.e.m. P values were obtained using ordinary ANOVA. Source data

Extended Data Fig. 1. Molecular heterogeneity of endothelial cells in long bone.

a. Representative FACS plots showing gating strategies for the isolation of ECs from Cdh5-mTnG reporter bone. b. Tile scan confocal images showing sections of 6, 22 or 55-week-old Cdh5-mTnG reporter bone co-stained for EMCN (red) and VEGFR3 (blue). The avascular growth plate (GP) and regions containing bmECs, mpECs and rECs are indicated. Small panels show higher magnifications of these areas. Arrowheads mark type R capillaries. Scale bars, 100μm (overview images) and 500μm (small panels). c. Heatmap showing the expression of selected marker genes for each EC subcluster (based on integrated scRNA-seq data for all age groups). d. Representative confocal images showing immunostaining for EMCN (red) and SOX11 (green, white arrowheads), marking pECs in the young but not adult metaphysis. e. CXADR (green) staining of bmECs in young, adult, and aging femur. Growth plate (GP) and epiphysis are indicated. f. MAdCAM1 (green, white arrowheads) expression in diaphyseal bmECs in young and adult femur. Scale bars, 50μm (d), 500μm (e), 100μm (f).

Extended Data Fig. 2. Expansion of rECs in adolescent femur.

a, b, Freshly isolated wild-type femurs at the age of 3, 6, 8 or 12 weeks (W), as indicated. Tile scan confocal images of EMCN⁺ VEGFR3⁻ ECs (arrowheads) in sectioned femur at the indicated ages (b). Epiphysis and growth plate are labelled. Scale bars, 1000μm (a), 500μm (b). c. The UMAP plot (integrated scRNA-seq data of all age groups) showing Bmx expression in arterial endothelial cells (aECs), marked by arrowhead. Scheme of Bmx-CreERT2 genetic fate mapping. 4-hydroxy tamoxifen (4-OHT) administration (1mg/mouse) is indicated by black arrow and red arrows mark time points of analysis. d, e, High-resolution confocal images and tile scan overview images of fate-tracked Bmx-CreERT2 R26-mTmG (GFP, green) in femur at the indicated time points after 4-OHT administration. Insets in (e) show metaphysis (i) and diaphysis (ii), respectively. White arrowheads mark CAV1⁺ EMCN⁻ aECs and yellow arrowheads GFP⁺ CAV1⁺ EMCN⁻ aECs. Green arrowheads indicate EMCN⁺ CAV1⁺ rECs, which are devoid of GFP signal. Scale bars, 100μm (d) and 500μm or 100μm (e).

Extended Data Fig. 3. Molecular features of bone endothelial cells.

a-d, Monocle trajectory analysis of bone EC subpopulations with coloured cell clusters (a), representation of individual clusters in trajectory (b), and cell type-specific relative gene expression shown in pseudo-time (c, d). Results are derived from integrated scRNA-seq data of all age groups. e, Schematic representation of major bone EC subpopulations along with designated markers.

Extended Data Fig. 4. Type R vessels are associated with remodelling bone.

a, High-magnification two-photon microscopy images of compact bone immunostained for EMCN (red) and CAV1 (blue) together with second-harmonic generation (SHG, white). Type R capillaries (arrowheads) and trabecular bone (TB) are indicated. Scale bars, 50μm. b, c, Maximum-intensity projections of sections from 3, 6, 8 or 12-week-old femur immunostained for EMCN (red) and OSTERIX (green) (b) or EMCN (red), VEGFR3 (blue) and ATP6V1B1/B2 (green) (c). Arrowheads mark type R capillaries near trabecular bone (TB). Scale bars, 50μm. Diagram on the right depicts association of rECs with osteoblast-lineage cells and osteoclasts. d. Injection scheme and time points of imaging after labelling of wild-type mice with Calcein Green (CG) and Alizarin Red (AZ). Tile scan imaging of sectioned femur from 8-week-old wild-type after double labelling. Insets depict active resorption in the metaphysis with CG⁻ AR^high labelling (top row), active osteogenesis (CG^high AR^high) in compact bone (middle row), and active remodelling (CG^low AR^high) of trabecular bone (bottom row). n=3 mice. Scale bars, 500μm (overview images) and 50μm (insets).

Extended Data Fig. 5. Features of type R vessels.

a, b. High-resolution confocal images showing VEGFR3 (yellow) and VEGFR2 (red) immunostaining together with DAPI (blue). VEGFR3⁻ VEGFR2⁺ capillaries (white arrowheads) around trabecular bone (TB) and nearby VEGFR3⁺ VEGFR2⁺ vessels (yellow arrowheads) in samples from young (a) and aged patients (b) are indicated. Scale bars, 200μm. c. Tile scan confocal images of EMCN⁺ VEGFR3⁻ ECs in 12-week-old female and male murine femur. Trabecular bone (TB) is indicated. Scale bars, 500μm. d. Quantitation of EMCN⁺ VEGFR3⁻ vessel density in 12-week-old female and male femur (n =3 mice per group). Mean ± SEM. P values, unpaired two-tailed t-test. e. EMCN (red) and VEGFR3 (blue) immunostaining showing the presence of VEGFR3⁻ EMCN⁺ vessels (white arrowheads) around trabecular bone (TB) in female and male femur. Scale bars, 100μm. Source data

Extended Data Fig. 6. Endothelial-specific inactivation of Dach1.

a, t-SNE visualization of Dach1 gene expression in ECs of different organs. b, Confocal tile scan overview images of 12-week-old Dach1^iΔEC and control retina stained for CD31 (red). Scale bars, 300μm. c, Quantitation of CD31⁺ vessel density in 12-week-old Dach1^iΔEC retinas relative to control (n =4 mice per group). Mean ± SEM. P values, unpaired two-tailed t-test. d, High-resolution confocal images of Dach1^iΔEC and control intestinal villi stained for DACH1 (green), VEGFR2 (red), and DAPI (blue). Scale bars, 50μm and insets 15μm. e, Quantitation of average VEGFR2⁺ vessel density, villus size, and average vascular area per villus in 12-week-old Dach1^iΔEC and littermate control (n =4 mice per group). Mean ± SEM. P values, unpaired two-tailed t-test. f, Confocal tile scan overview images of 12-week-old female Dach1^iΔEC and control femur stained for EMCN (red) and OSTERIX (green, OSX). White dashed lines and yellow arrowheads indicate reduction of region containing trabecular bone (TB) in Dach1^iΔEC mutants. Scale bars, 500μm. g. High-magnification images showing EMCN (red) and ATP6V1B1/B2 (green) immunostained female Dach1^iΔEC and control femur. Growth plate (GP) is indicated. Scale bars, 100μm. h. Quantitation of OSX⁺ cells and ATPV6B1B2 area in control and Dach1^iΔEC mutant (n=3 and 4 mice per group). Mean ± SEM. P values, unpaired two-tailed t-test. i. Quantitation shows connectivity density and cortical thickness (per mm), (n = 3-4 female mice per group and n=3 male mice per group). Mean ± SEM. P values, plotted using unpaired two-tailed t-test. j. DACH1 (green) co-stained with VE-CADHERIN (red) in siCONTROL and siDACH1 cultured HUVECs. k. Downregulation of arterial marker genes in siDACH1 HUVECs relative to siCONTROL. n=6 (3 independent experiments) for each group. Mean ± SEM. P values, unpaired two-tailed t-test with Welch’s correction. Source data

Extended Data Fig. 7. Effect of DACH1 overexpression in endothelial cells.

a, b. Confocal images of femur sections immunostained for EMCN (red) together with OSTERIX (OSX, green) (a) or ATP6V1B1/B2 (green) (b). Scale bars, 100μm. c. Quantitation of OSX⁺ osteoblast-lineage cells and ATP6V1B1/B2⁺ osteoclasts in Dach1^OE and control femur (n = 3 in each group). Mean ± SEM. P values, unpaired two-tailed test with Welch’s correction. d-f. Injection scheme and time points of imaging after labelling of Dach1^OE and control with Calcein Green (CG) and Alizarin Red (AZ) (d) and tile scan imaging of sectioned femur from P30 Dach1^OE and control after double labelling (e). Scale bars, 500μm. Bottom panels in (e) show stained trabecular bone. Scale bars, 20μm. f. Quantitation showing extent of bone surface actively mineralizing as mineralizing surface/bone surface (MS/BS) in percentage, mineral apposition rate (MAR) as per micrometre/day, and bone formation rate (BFR) (n=3 mice per group). Mean ± SEM. P values, unpaired two-tailed t-test with Welch’s correction for MS/BS, MAR-trabecular bone and BFR. g. Maximum-intensity projections showing prominent HIF1α⁺ immunostaining (yellow arrowheads) in 6-week-old Sp7-mCherry⁺ trabecular osteoblasts, whereas HIF1α⁺ is strongly decreased after emergence of EMCN⁺ VEGFR3⁻ type R vessels at 12 weeks. Scale bars, 50 μm. h. Maximum-intensity projections of 12-week-old femurs showing immunostaining for hypoxia-related markers in relation to EMCN⁺ VEGFR3⁻ type R vessels. Shown are immunostaining for Glucose-6-Phosphate Isomerase (GPI), 5′-nucleotidase/CD73 and Haem Oxygenase-1 (HMOX1)-expressing macrophages (yellow arrowheads), all of which are elevated near juvenile trabecular bone relative to the equivalent region in adult. White arrowheads mark EMCN⁺ VEGFR3⁻ type R vessels. i, j. UMAP plots of bone ECs (n=18383) with colour-coded subclusters (f), namely sinusoidal bmECs, metaphyseal mpECs, arterial aECs, remodelling rECs, and proliferating pECs. UMAP distribution of Dach1^OE and control ECs, as indicated by colour (g). k. Bar plots showing proportion of cells in Dach1^OE and control samples. l. UMAP plots comparing Dach1 expression in Dach1^OE (bottom) and control (top) EC subclusters. m. Heatmap of differentially expressed genes between EC subclusters. n. Heatmap of differentially expressed genes related to hypoxia and, tissue oxygenation in Dach1^OE and control bone ECs. o. Confocal images showing reduced HIF1α immunostaining (white arrowheads) in Dach1^OE metaphysis and around trabecular bone relative to littermate control. Scale bars, 50μm. Results in panels i, j, m, and n show the integrated mutant and control scRNA-seq data, whereas panels k and l show separated samples derived from the integrated data. Source data

Extended Data Fig. 8. Molecular signalling and metabolic regulation of rECs.

a. Violin plots showing selected genes related to reduction of bmEC markers and increased arterialization in Dach1^OE and control. b. Heatmap of differentially expressed genes related to Notch signalling in Dach1^OE and control bone ECs. c. Representative confocal images showing increased Delta-like 4 (DLL4) (green, white arrowheads) expression in vessels around Dach1^OE trabecular bone (TB). Scale bars, 100μm. d. Scheme of tamoxifen-induced (TMX) EC-specific Dll4 inactivation with Cdh5-CreERT2 line. e, f, Tile scan confocal images of EMCN⁺ VEGFR3⁻ ECs (white arrowheads) (e) and CAV1⁺ aECs (green arrowheads) (f) in 12-week-old Dll4^iΔEC loss-of-function and control femur. Trabecular bone (TB) is indicated. Scale bars, 500μm. g, Quantitation of EMCN⁺ VEGFR3⁻ vessel density (left), and number of CAV1⁺ arteries (right) in 12-week-old Dll4^iΔEC and control femur (n =3 female mice per group). Mean ± SEM. P values, unpaired two-tailed t-test. h. Heatmap showing differentially expressed transcripts for secreted factors in bone EC subpopulations. i, j, Representative images (i) and quantitation (j) of calcified nodules in human mesenchymal stem cells (HMSC) cultured in osteogenic differentiation medium (ODM) supplemented with 100ng/ml Complement C1q and Tumour Necrosis Factor-Related Protein 9 (CTRP9), Neurotrophin 3 (NTF3), Platelet-Derived Growth Factor D (PDGFD), or Semaphorin 7A (SEMA7A) (n=3 independent experiments for each group). Scale bars, 100μm. Graph shows quantitation of absorbance at 405nm. Mean ± SEM. P values, ordinary one-way ANOVA and plotted with Dunnett’s multiple comparisons test. k, l, Representative images (k) and quantitation (m) of multinucleated TRAP+ osteoclasts generated from bone marrow-derived monocytes/macrophages treated with RANKL in osteoclast medium (OCM) supplemented with 100ng/ml CTRP9, NTF3, PDGFD, or SEMA7A n=8 (4 independent experiments) for each group. Scale bars, 500μm. Graph shows quantitation of TRAP⁺ multinucleated cells between OCM and treated conditions. Mean ± SEM. P values, ordinary one-way ANOVA and plotted with Dunnett’s multiple comparisons test. Panels a and b show separated conditions derived from the integrated mutant and control data, whereas panel h is based on the integrated scRNA-seq data. Source data

Extended Data Fig. 9. Modulation of type R vessels by anti-osteoporosis treatments.

a, Scheme for the treatment of adult mice with parathyroid hormone (PTH) or Alendronate. Black arrows indicate onset of treatment, red arrows time points of imaging. b, Confocal images of OSTEOPONTIN (blue) immunostaining of the metaphysis near the growth plate (GP) together with EMCN (red), CAV1 (green) and DAPI (white) to show the response to treatments. Scale bars, 100μm. c, UMAP visualization of sorted bone ECs from control and treatment groups with colour-coded subclusters. d, Bar plots showing the proportions of EC subpopulations in 12-week-old mice. Differences in percentage of rECs and aECs compared to control are indicated. Panels c and d show the individual samples from an integrated scRNA-seq dataset of three conditions. e, f, Confocal images showing EMCN⁺ VEGFR3⁻ CAV1⁺ rECs (white arrowheads) around trabecular bone (TB) (e) and association of ATP6V1B1/B2⁺ (green) osteoclasts with EMCN⁺ VEGFR3⁻ type R capillaries (f) in 12-week-old control and indicated treatment conditions. Scale bars, 50μm. g, Effect of treatments on EMCN⁺ VEGFR3⁻ rECs (white arrowheads) and RUNX2⁺ (green) osteoprogenitors around trabecular bone (TB). Scale bars, 100μm.

Extended Data Fig. 10. Validation of anti-osteoporosis treatments.

a, b, Experimental design, 3D reconstruction of µCT measurements and quantitation of key parameters for the treatment of adult mice with Alendronate (a) and Parathyroid hormone (PTH) (b). PTH results have been previously reported. Quantitation shows relative bone volume represented as bone volume/tissue volume (BV/TV), connectivity density as the connectivity density per cubic millimetre, trabecular thickness (per mm), and trabecular number represented as number of trabeculae per millimetre (n = 5 mice per group). Mean ± SEM. P values, unpaired two-tailed t-test with Welch’s correction for BV/TV and connectivity density for both treatments. Source data