Mapping the spatial atlas of the human bone tissue integrating spatial and single-cell transcriptomics

Abstract

Bone is a multifaceted tissue requiring orchestrated interplays of diverse cells within specialized microenvironments. Although significant progress has been made in understanding cellular and molecular mechanisms of component cells of bone, revealing their spatial organization and interactions in native bone tissue microenvironment is crucial for advancing precision medicine, as they govern fundamental signaling pathways and functional dependencies among various bone cells. In this study, we present the first integrative high-resolution map of human bone and bone marrow, using spatial and single-cell transcriptomics profiling from femoral tissue. This multi-modal approach discovered a novel bone formation-specialized niche enriched with osteoblastic lineage cells and fibroblasts and unveiled critical cell-cell communications and co-localization patterns between osteoblastic lineage cells and other cells. Furthermore, we discovered a novel spatial gradient of cellular composition, gene expression and signaling pathway activities radiating from the trabecular bone. This comprehensive atlas delineates the intricate bone cellular architecture and illuminates key molecular processes and dependencies among cells that coordinate bone metabolism. In sum, our study provides an essential reference for the field of bone biology and lays the foundation for advanced mechanistic studies and precision medicine approaches in bone-related disorders.

Graphical Abstract

Figure 1.

Spatial multi-modal profiling of human bone. (A) Sampling region (see ‘Materials and methods’ section). (B) Data modalities. ST, spatial transcriptomics. (C) ST feature plots. nFeature, the number of unique genes per spot; nCount, the number of unique transcripts per spot. (D) Uniform manifold approximation and projection (UMAP) of scRNA-seq data from four samples (n = 26 574) and average marker gene expression after z-score transformation. Colors along the bottom correspond to the cell types. MP, mononuclear phagocyte; pDC, plasmacytoid dendritic cell. (E–G) Characterization of spatial transcriptomics data: gene expression (E), cell enrichment (F) and pathway activity (G). Prob, probability. Lo, low. Hi, high.

Figure 2.

Characterization of tissue organization utilizing spatial transcriptomics data. (A) Diagram of molecular niche definition. Each spot may contain multiple cells or various types of cells. Niches are identified based on the gene expression matrix within each spot. Spots that are grouped into the same cluster through clustering analysis are considered to have similar cellular compositions, defining a molecular niche. The cells depicted in spot 1 are the OstLin cells and fibroblasts identified in our study. Fibro, Fibroblast. (B) UMAP of spots based on spatial transcriptomics data (left) and spatial distribution of niches (right). (C) Volcano plot of differential gene expression among niches. (D) Pathway enrichment analysis of differentially expressed genes. (E) Cell enrichment analysis of niches. (F) Visualization and cell enrichment score comparison of MesLin cells in the bone tissue. (G) PROGENy pathway activities across different cell types. (H) PROGENy pathway activities across different niches.

Figure 3.

Dissecting cell–cell communication across spatial tissue. (A) Cell–cell communication network. Edge richness represents the number of ligand–receptor interactions. Macro, Macrophage. Granu, granulocyte. (B) Sankey plots of top 50 ligand–receptor interactions between fibroblasts and OstLin. (C and E) Visualization of selected ligand–receptor interactions for spatial transcriptomics data. Left is the relative expression of the ligand and the receptor; right is the interaction strength. (D and F) LR score comparison of selected ligand–receptor interactions in molecular niches.

Figure 4.

Characterization of tissue organization inferred from cell abundance estimated from cell2location. (A) Diagram of cell composition niche definition. Each spot may contain multiple cells of one type or multiple types. Spots that are grouped into the same cluster through clustering analysis based on the estimated cell proportion matrix are considered to have similar cellular compositions, defining a cell-type niche. The cells depicted in spot 1 are the OstLin cells and fibroblasts identified in our study. Fibro, Fibroblast. (B) UMAP of spots based on spatial transcriptomics data (left) and spatial distribution of niches (right). (C) Cell enrichment analysis of niches. (D) The degree of intercellular dependence within a spot. (E) Immunofluorescence staining of fibroblasts and OstLin cells with VIM and RUNX2 antibodies. (F) Co-localization of cell types and signaling pathways within a spot. (G) Co-localization of signaling pathways within a spot.

Figure 5.

Characterization of spatial signaling gradient domain. (A) Gradient heatmap of distance from bone trabeculae. (B) Variation of average RUNX2 expression with distance gradient. (C) Immunofluorescence staining of RUNX2. (D and E) Variation of cell-type proportion with distance gradient. (F and G) Variation of pathway activities with distance gradient. (H) Immunohistochemical staining of TGFβ activation (p-SMAD3) and comparison of p-SMAD3 intensity.