Single Cell RNA Sequencing Reveals Heterogeneity of Human MSC Chondrogenesis: Lasso Regularized Logistic Regression to Identify Gene and Regulatory Signatures

Abstract

AbstractBone marrow-derived mesenchymal stem cells (MSCs) exhibit the potential to undergo chondrogenesis in vitro, forming de novo tissues with a cartilage-like extracellular matrix that is rich in glycosaminoglycan and collagen type II. However, it is now apparent that MSCs comprise an inhomogeneous population of cells, and the fate of individual subpopulations during this differentiation process is not well understood. We analyzed the trajectory of MSC differentiation during chondrogenesis using single cell RNA sequencing (scRNA-seq). Using a machine learning technique – lasso regularized logistic regression – we showed that multiple subpopulations of cells existed at all stages during MSC chondrogenesis and were better-defined by transcription factor activity rather than gene expression. Trajectory analysis indicated that subpopulations of MSCs were not intrinsically specified or restricted, but instead remained multipotent and could differentiate into three main cell types: cartilage, hypertrophic cartilage, and bone. Lasso regularized logistic regression showed several advances in scRNA-seq analysis, namely identification of a small number of highly influential genes or transcription factors for downstream validation, and cell type classification with high accuracy. Additionally, we showed that MSC differentiation trajectory may exhibit donor to donor variation, although key influential pathways were comparable between donors. Our data provide an important resource to study gene expression and to deconstruct gene regulatory networks in MSC differentiation.