Application Of Single-cell Technology In Studying Lepr~+ Bone Marrow Stromal Cells And Development Of A Single-cell ATAC-seq Analysis Tool

With the rapid development of single-cell sequencing technologies,analysis and integration of single-cell omics data have raised great computational challenges.Although an increasing number of bioinformatics methods have been developed for analyzing and interpreting single-cell RNA-seq(sc RNA-seq)data,how to integrate multiple datasets and correct batch effects to fully understand sc RNA-seq data still remains challenges.Besides,analysis of single-cell epigenetic sequencing data,such as single-cell ATAC-seq(scATAC-seq),is even more challenging due to the sparse signals and high levels of noise in the data.Therefore,this study includes two parts:1)how to integrate sc RNA-seq methods to dissect the heterogeneity of Lepr~+bone marrow stromal cells,and 2)the development of a bioinformatics tool for scATAC-seq data analysis.Bone marrow stromal cells(BMSCs)are mesenchymal cell populations with potential to differentiate into bone,fat,and cartilage in vitro.Leptin receptor(Lepr)is expressed by a wide array of BMSCs.Lepr~+cells are key components of the bone marrow hematopoietic microenvironment,and highly enrich skeletal stem and progenitor cells(SSPCs)that maintain homeostasis of the adult skeleton.However,the heterogeneity and lineage hierarchy within this population has been elusive.In the first part of the study,we took advantage of sc RNA-seq technology to understand the heterogeneioty of Lepr-Cre-traced bone marrow stromal cells(Lepr~+BMSC)under homeostatic and stress conditions including aging,rosiglitazone feeding,irradiation and bone fracture.In order to fully interpret sc RNA-seq datasets,we utilized sc RNA-seq analysis strategies including batch effect correction,cell type identification,trajectory analysis,differentially expression analysis,gene set enrichment analysis,single cell gene regulatory network analysis,and RNA velocity analysis.Firstly,we built a transcriptional atlas of Lepr~+BMSC and identified three cell lineages including adipogenic,osteogenic and chondrogenic lineage cells,revealing the heterogeneity and differentiation potential of adipogenic lineage clusters in response to different stress conditions.Next,we constructed a maturation trajectory of osteogenic lineage cells and predicted two key transcription factors(i.e.,Hoxb2 and Npdc1)in the fracture repair process by gene regulatory network analysis.Finally,we reveraled the heterogeneity of periosteum cells and found a novel Lepr-Cre~+Sca-1~+subset in the long bone periosteum.Gene expression is a multifaceted process controlled by the combinatorial activity of regulatory elements.Expression heterogeneity could arise from cell-to-cell differences in gene activation and repression regulatory mechanisms such as the action of TFs,chromatin modifiers,and other regulatory factors.The scATAC-seq method is a powerful tool to reveal the regulatory logic of gene expression programs in single cells.However,the high signal sparsity and high levels of noise in scATAC-seq data raised new computational challenges for data analysis.Thus,in the second part of this study,we developed a bioinformatics tool for scATAC-seq data analysis,which is termed Single-cell Chromatin Accessibility-based cluster Recognition and Trajectory reconstruction(scART).In order to reduce noise in the sparse data,scART combined KNN imputation and TF-IDF weighting scheme and infered cell-to-cell similarities by cosine distance.Compared with published methods,scART exhibited its superiority in the accuracy and sensitivity of cell clustering,especially when used for datasets with low sequencing depth.Besides,scART was capable to construct developmental trajectories from scATAC-seq data.By reanalysis of a published dataset,scART revealed the dynamic changes of cortical layer neurogenesis during mouse embryo forebrain development.Furthermore,scART can be used for motif enrichment analysis and gene accessibility analysis when processing scATAC-seq data.