Unraveling bone marrow architecture

Different types of stromal cells in the bone marrow associate to form niches that support differentiating blood cells and ensure lifelong production of all major blood lineages. A study now combines single-cell and spatial transcriptomics with imaging to infer the cellular composition and spatial architecture of specific niches.

Haematopoietic stem and progenitor cells differentiate to generate all major blood cell lineages and are supported by the bone marrow stroma. These stromal cells organise into multicellular niches that produce cytokines, chemokines, adhesion molecules and other factors to regulate the self-renewal and differentiation of blood cells1. Defining the cellular composition of these niches, their location within the tissue and their secretion of signalling factors is indispensable for understanding how the bone marrow ensures a balanced blood cell production. Insights into the cellular composition and architecture of the bone marrow microenvironment derive largely from studies that focused on the niches for haematopoietic stem cells (HSC)2,3,4,5. In these studies, fluorescent reporters were used to visualise HSC localisation and their association with candidate niches cells, as well as to carry out conditional ablation of specific stromal cells5 or cell-specific deletion of regulatory cytokines to identify key contributors2,3,4. On the basis of these and other studies, it is now recognised that most HSCs reside adjacent to sinusoids and megakaryocytes, with a small fraction localising near arterioles and a smaller fraction localising near the endosteum1. Distinct subtypes of endothelial cells, perivascular cells and other stromal and haematopoietic cells have been characterised as components of these HSC niches and represent key contributors to niche function1. However, controversies remain with regard to the cellular composition of sinusoidal, arteriolar and endosteal niches and the cellular sources of important regulatory cytokines, such as CXCL12 and SCF2,3,4,5,6,7. Although the primary focus of the field has been on the identification of HSC niches, evidence suggests that specific niches for lineage-committed progenitors exist as well1, but these remain poorly characterized. In this issue of Nature Cell Biology, Baccin, Al-Sabah, Velten et al.8 reveal surprising cellular and functional heterogeneity between arteriolar, sinusoidal, non-vascular, endosteal, and subendosteal areas of the bone marrow and clarify the main sources of regulatory cytokines.

Baccin, Al-Sabah and Velten et al.8 first characterised the cellular composition of the bone marrow using droplet-based, single-cell RNA-sequencing, with a focus on the major haematopoietic and stromal cell lineages (Fig. 1). These initial analyses revealed a high degree of heterogeneity within bone marrow stromal cells. The authors then microdissected sinusoidal, arteriolar, endosteal, subendosteal and non-vascular areas of the bone marrow using laser capture microdissection (LCM). This technique was integrated with RNA sequencing (RNA-seq) in an optimised LCM-sequencing approach9 to determine which stromal populations localised to each region (Fig. 1). Notably, the authors validated the location of various stromal cells in their respective bone marrow regions with immunofluorescence imaging of bone sections. Finally, they predicted the association of specific cell types with the five bone marrow regions using a bioinformatics approach evaluating the expression of cellular adhesion receptors and their ligands. In addition, the authors used this method to define differential cytokine and growth factor expression patterns within distinct bone marrow areas.

Fig. 1: Scheme showing the experimental workflow used by Baccin, Al-Sabah, Velten et al. to dissect the cellular architecture of the bone marrow (BM) microenvironment.

ScRNA-seq, single-cell RNA sequencing.

Previous work established that perivascular stromal cells form a three-dimensional network that ensheathes and supports bone marrow vessels1. Cells in this population have been identified as CAR (CXCL12-abundant reticular) cells using Cxcl12-gfp reporter mice, Nestin-GFPdim cells using Nestin-gfp reporter mice and LepR+ cells using LepR-cre mice1. These perivascular stromal cells are key components of the HSC niche and produce regulatory cytokines, such as CXCL12 and SCF, which maintain HSC function2,3,4. Functional and cellular heterogeneity clearly exists within these perivascular stromal cell populations10,11,12. Baccin, Al-Sabah, Velten et al.8 now add further insights into the complexity of these niche components with the discovery of two distinct subpopulations within the broad CAR population. They termed these populations Adipo-CAR and Osteo-CAR cells, based on differential expression of adipocyte and osteolineage genes, respectively. Notably, Adipo-CAR cells showed a gene expression pattern similar to that of LepR+ cells, suggesting that they are the same population. These observations are particularly interesting, as LepR-cre lineage tracing experiments done by members of the Morrison laboratory established LepR+ cells as the main source of bone cells13 and revealed adipocyte potential in a subset of LepR+ cells (detected using Adipoq-CreER mice)10. Additional in vivo studies will be required to clarify the overlap between and the in vivo differentiation potential of LepR+, Adipo-CAR and Osteo-CAR cells.

The authors further demonstrated that Adipo-CAR cells are enriched in the sinusoids, whereas the spatial transcriptomics data indicated that Osteo-CAR cells preferentially locate to arteriolar and non-vascular niches. Validating this finding, the authors confirmed that Osteo-CAR cells express alkaline phosphatase (Alpl) and showed that ALPL+Cxcl12+ Osteo-CAR cells are less tightly associated with sinusoids, and more closely associated with arterioles, compared to ALPL-Cxcl12+ Adipo-CAR cells. Using a similar strategy, Baccin, Al-Sabah, Velten et al.8 also verified the presence of smooth muscle cells and Podoplanin (PDPN) positive fibroblasts in arterioles, whereas PDPN+COL1Alow fibroblasts localised to bone-facing endosteal areas. The authors were further able to distinguish individual cell populations from their LCM data and their associated abundance by adopting a previously developed computational tool, CIBERSORT14, and experimentally validated these predictions using flow cytometry and bulk RNA-seq.

Having identified these previously unappreciated marrow components and their localisation, Baccin, Al-Sabah and Velten et al.8 revealed that each bone marrow region has a unique cytokine signature. Namely, Osteo-CAR and Adipo-CAR cells were the main source for haematopoietic cytokines in the bone marrow, suggesting that haematopoietic progenitors in different bone marrow regions are exposed to distinct cytokine milieus that might skew differentiation towards specific lineages. In further support of this exciting hypothesis, the authors developed a computational approach, termed RNA-Magnet, to infer interactions between stromal cells and haematopoietic cells. RNA-Magnet maps physical interactions between single cells and ‘attractor’ cell populations based on the imputed gene expression of cellular adhesion receptors and their associated ligands in ‘sending’ and ‘receiving’ cells. In doing so, the authors were able to assign weights to attractor cell populations and individual cells, describing their relative adhesiveness and the direction of their interaction. Impressively, this algorithm confirmed the niche localisation of the authors’ proposed Adipo-CAR and Osteo-CAR populations, as well as that of previously demonstrated cell populations.

In summary, the study by Baccin, Al-Sabah and Velten et al.8 complements previous literature using single-cell RNA-seq to investigate the mouse bone marrow microenvironment in normal and perturbed haematopoiesis11,12,15. In combination with the growing number of reference atlases of human and mouse haematopoiesis, the workflow provided by the authors holds tremendous promise for the identification of additional cellular and molecular interactions in the niche and an improved understanding of their role in native haematopoiesis and disease. Taken together, the transcriptome of each stromal population, its location in the bone marrow, differential and regional bone marrow cytokine enrichment and candidate interacting cells provide an invaluable blueprint for future studies aimed at deciphering the role of specific vascular and endosteal structures that support haematopoiesis.


  1. 1.

    Wei, Q. & Frenette, P. S. Immunity 48, 632–648 (2018).

    CAS  Article  Google Scholar 

  2. 2.

    Ding, L. & Morrison, S. J. Nature 495, 231–235 (2013).

    CAS  Article  Google Scholar 

  3. 3.

    Ding, L., Saunders, T. L., Enikolopov, G. & Morrison, S. J. Nature 481, 457–462 (2012).

    CAS  Article  Google Scholar 

  4. 4.

    Greenbaum, A. et al. Nature 495, 227–230 (2013).

    CAS  Article  Google Scholar 

  5. 5.

    Kunisaki, Y. et al. Nature 502, 637–643 (2013).

    CAS  Article  Google Scholar 

  6. 6.

    Asada, N. et al. Nat. Cell Biol. 19, 214–223 (2017).

    CAS  Article  Google Scholar 

  7. 7.

    Xu, C. et al. Nat. Commun. 9, 2449 (2018).

    Article  Google Scholar 

  8. 8.

    Baccin, C. et al. Nat Cell Biol. (2019).

  9. 9.

    Nichterwitz, S. et al. Nat. Commun. 7, 12139 (2016).

    CAS  Article  Google Scholar 

  10. 10.

    Zhou, B. O. et al. Nat. Cell Biol. 19, 891–903 (2017).

    CAS  Article  Google Scholar 

  11. 11.

    Baryawno, N. et al. Cell 177, 1915–1932.e16 (2019).

    CAS  Article  Google Scholar 

  12. 12.

    Tikhonova, A. N. et al. Nature 569, 222–228 (2019).

    CAS  Article  Google Scholar 

  13. 13.

    Zhou, B. O., Yue, R., Murphy, M. M., Peyer, J. G. & Morrison, S. J. Cell Stem Cell 15, 154–168 (2014).

    CAS  Article  Google Scholar 

  14. 14.

    Newman, A. M. et al. Nat. Methods 12, 453–457 (2015).

    CAS  Article  Google Scholar 

  15. 15.

    Wolock, S. L. et al. Cell Rep. 28, 302–311.e5 (2019).

    CAS  Article  Google Scholar 

Download references

Author information



Corresponding author

Correspondence to H. Leighton Grimes.

Ethics declarations

Competing interests

The authors declare no competing interests.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Lucas, D., Salomonis, N. & Grimes, H.L. Unraveling bone marrow architecture. Nat Cell Biol 22, 5–6 (2020).

Download citation

Further reading


Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing