Bone marrow is the main part of lifelong blood generation and bone regeneration. In the bone marrow microenvironment (BM niches), different mesenchymal cells, vascular system and differentiated hematopoietic cells interact to regulate the maintenance and differentiation of hematopoietic and mesenchymal stem cells (HSC and MSC) respectively. However, its cellular and spatial organization is still controversial. Recently, scientists from European Molecular Biology Laboratory (EMBL) and other institutions have combined the single cell and space transcriptome system to draw the molecular, cellular and spatial composition of different hematopoietic microenvironment in mice. The major bone marrow resident cell types have been analyzed, their spatial location and the source of hematopoietic stimulating factors have been elucidated. Finally, a RNA-Magnet has been developed. The algorithm can accurately deduce the three-dimensional organization structure of cells from single cell transcriptome data.
Sequencing sample information:
From 8 to 12 weeks, femur, tibia, hip and spine of C56Bl/6J female mice
10x Genomics system, Illumina NextSeq 500
Cell sorting platform:
FACS was used for sorting: CD45-CD71+ cells (erythroid progenitor cells).
CD45-CD71- (non hematopoietic cells)
One adopt ScRNA-Seq Identification and description BM Resident cell type
The researchers first performed scRNA-Seq on the bone marrow of mice. In order to capture high abundance and rare BM resident cells at the same time, after the total BM of mice was scRNA-Seq, the rare cell types were enriched and scRNA-Seq was carried out by gradually consuming abundant cell types or bone tissue that was never digested by BM or enzyme digestion.
The 7497 cells were analyzed by t-SNE. Separate out Thirty-two Subgroups They correspond to different cell types or stages of differentiation. The types of hematopoietic cells were identified by marker gene expression and GO analysis, including dendritic cells, neutrophils, monocytes, T cells and B cells at different developmental stages, and erythroid progenitors with low expression of CD45 and expression of erythroid marker (e.g. CD71).
Fig. 1 Identification of BM resident cell types by scRNAseq
The remaining 2% cells were non hematopoietic cells, and did not express CD45 and CD71. They include Mog (Mag), smooth muscle cells (Tagln, Acta2), myofibroblasts, 9 different Pdgfra positive mesenchymal cells and two endothelial cells (EC) populations (Cdh5, Pecam). Endothelial cells include endothelial cells expressing Sca1 (Ly6a) and sinusoidal endothelial cells expressing Emcn. Mesenchymal cells contain chondrocytes (Acan, Sox9) and osteoblasts (Osteocalcin / Bglap, Col1a1) and several other cell types, including three different fibroblasts cluster, and the other two cell populations showing high transcriptome similarity with SCF- green fluorescent protein (GFP) positive and Cxcl12 GFP positive Cxcl12-abundant reticular (CAR) cells reported previously. It is noteworthy that these two subgroups expressed the genes of adipocytes and osteoblasts, so they were named Adipo-CAR and Osteo-CAR cells respectively. Adipo-CAR cells expressed high level of leptin receptor (Lepr) and showed high transcriptional similarity with LepR - Cre cells. On the contrary, Osteo-CAR cells expressed higher levels of osteoblast specific transcription factor osterix (Sp7) and lower level Lepr. In addition, the researchers identified a group of mesenchymal cells expressing Ng2 and Nestin, which are similar to the Ng2 + Nestin + MSC described previously in the transcriptome and named Ng2 + MSC.
Finally, the researchers used pseudo temporal analysis to construct the differentiation locus of all mesenchymal cells. Ng2 + MSCs was the initial differentiation cell, and osteoblasts, CAR cells, chondrocytes and fibroblasts were downstream. No osteoclasts, neurons and mature megakaryocytes were found in the study. This may be due to the limitation of the size of the captured cells.
Fig. 2 Identification of BM resident cell types
Two Spatial localization of BM resident cells by combining single cell and spatial transcriptome
The researchers developed an improved laser capture microdissection combined with sequencing (LCM-Seq) to collect 76 microanatomical regions from the BM tissue according to the presence of sinus and small artery vessels or the distance from the intima of the talus. LCM-Seq was used to describe the composition of the non vascular microenvironment of the intima, subperiosteum, sinus and arteriole.
Figure 3. Spatial distribution of BM resident cells by combining single cell and spatial transcriptome analysis.
In order to evaluate the newly developed method of spatial transcriptome, we compared the expression of marker gene in different microenvironments in scRNA-Seq and corresponding spatial transcriptome data. The osteoblast gene is selectively enriched in endothelial cells, and the endothelial cells are enriched in the sinusoidal endothelial rich region, while the arterial endothelial gene is enriched in the arterioles. The marker gene sets of hematopoietic, Schwann cells and myofibroblasts were not significantly correlated with any microenvironment, indicating that the distribution of these cell types in the microenvironment was relatively uniform, or that the data coverage in LCM-Seq was insufficient. On the contrary, the expression of marker gene in the other 12 cell populations was significantly different from each other. In order to systematically assess the preferential location of these BM cell types for candidate microenvironments, the researchers used CIBERSORT algorithm to estimate the frequency of cell clusters defined by scRNA-Seq in spatial transcriptome data. The results indicate that the integration of spatial transcriptome and single cell transcriptome data can locate most of the identified and unknown BM cell populations in different endospores, sinus and arterioles, and non vascular microenvironments.
Three Verification of cell type localization
In order to confirm the spatial relationship of BM cell types identified by LCM-Seq, the researchers used immunofluorescence staining on bone sections using marker gene combinations targeting specific cell populations.
Fig. 4 localization of CAR cell subtypes by immunofluorescence
Gene expression analysis showed that the expression of alkaline phosphatase (Alpl) and Cxcl12 could distinguish Osteo-CAR (Cxcl12 + Alpl +), Adipo-CAR cells (Cxcl12 + Alpl-) and Cxcl12-Alpl + cell types, such as osteoblasts, Ng2 + MSC and arterial Cxcl12-Alpl. The common staining with marker gene Emcn showed that Cxcl12 + Alpl- Adipo-CAR cells in central BM were mainly wrapped in arterioles and were consistent with LCM-seq results. On the contrary, Cxcl12 + Alpl + Osteo-CAR cells in central BM usually show non arterial localization and highly reticular form. The important thing is in Sca1 + GFP Dim Cxcl12 + Alpl + Sca1 is also observed near the arteriole. Low Osteo-CAR cells, of which GFP High The Osteo-CAR protuberance densely covers the arterioles. These results demonstrate the non sinusoidal localization of Osteo-CAR cells, and qualitatively confirm that these cells are located in the arterioles and non vascular regions, and are consistent with those predicted by LCM-Seq.
Fig. 5 localization of other mesenchymal cells by immunofluorescence.
Next, the researchers used CD31, SM22, Pdpn, Pdgfr, Col1a1 and Sca1 as marker to specifically identify and locate smooth muscle cells (SM22 + Pdpn-), fibroblasts (Pdpn + Pdfgr +), osteoblasts (+ +) and arterial arteries. The results demonstrate that LCM-Seq can identify unknown cell types, such as Adipo and Osteo-CAR cells, and have the ability to locate in BM.
Four utilize ScRNA-Seq Accurate data prediction BM Spatial relationship of resident cell types
The researchers compiled a complete annotated list of cell adhesion receptors and their homologous plasmalemma or extracellular matrix binding ligands, and developed the RNA-Magnet algorithm to predict cell type specific localization of BM through differential expression of cell adhesion molecules. In order to study whether the RNA-Magnet algorithm can determine the spatial location relationship between BM cell types, the researchers introduced four microenvironment groups: osteoblasts, endothelium cells, endothelial cells and smooth muscle cells in the arteriolar microenvironment. The simulation results showed that the adhesion of BM cells to different microenvironments was closely related to their spatial transcriptional differences. It proved that the RNA-Magnet algorithm could predict cell localization by single cell gene expression data.
Figure 6. Inference of cell interaction from single cell gene expression data using RNA-Magnet.
Five Cellular and spatial sources of cytokines and growth factors in bone marrow
The key biological processes occurring in BM are thought to be mediated by the synergy of multiple cytokines and growth factors. However, little is known about the cytokine producing cells and their spatial organization and function in the BM microenvironment. In order to confirm the Cxcl12 expression of Adipo- and Osteo-CAR cells at the protein level, the researchers developed FACS strategy to differentiate these cell types and confirmed it based on FACS index-scRNAseq. Comparative analysis showed that Adipo-CAR cells were CD45-CD71-Ter119-CD41-CD51 +
VCAM1 + CD200 Mid CD61 Low Osteo-CAR and NG2 + MSC overexpressed CD200 and CD61. In all BM cell types, CAR cells produce the largest number of cytokines and growth factors, indicating that CAR cells are "professional cytokine producing cells". Based on these observations, the researchers put forward a model in which the differential localization of cells producing specialized cytokines led to the establishment of a specific microenvironment.
Fig. 7 cellular and spatial sources of key cytokines in BM
6. BM Intercellular signal interaction analysis of cell types
Finally, the researchers applied the RNA-Magnet algorithm to soluble signal transduction mediators (such as cytokines, growth factors, etc.) and their receptors, and obtained a systematic overview of potential intercellular signal transduction interactions. The network obtained by RNA-Magnet analysis forms two disconnected signal clusters, which are made up of mature or non hematopoietic cells. This indicates that immune cells and non immune cells are preferentially communicating within their respective groups, indicating that different BM biological processes are mediated by specific combination signals from different local microenvironments.
Fig. 8 system level analysis of signal transduction potential in BM
In conclusion, the new method developed by researchers can help to reveal the composition of BM cells, namely the three-dimensional organization structure and how cells interact. Researchers can use this method to study the molecular mechanism of hematologic diseases, such as leukemia, and develop new therapies. In addition, the new method can be applied in many tissues, to determine the spatial structure of cells in tissues, and to study the complex pathological mechanism of human diseases.
Single cell transcriptome sequencing