Mesenchymal stem cells (MSCs) gain an increasing focus in the field of regenerative medicine due to their differentiation abilities into chondrocytes, adipocytes, and osteoblastic cells. However, it is apparent that the transformation processes are extremely complex and cause cellular heterogeneity. The study aimed to characterize differences between MSCs and cells after adipogenic (AD) or osteoblastic (OB) differentiation at the proteome level. Comparative proteomic profiling was performed using tandem mass spectrometry in data-independent acquisition mode. Proteins were quantified by deep neural networks in library-free mode and correlated to the Molecular Signature Database (MSigDB) hallmark gene set collections for functional annotation. We analyzed 4108 proteins across all samples, which revealed a distinct clustering between MSCs and cell differentiation states. Protein expression profiling identified activation of the Peroxisome proliferator-activated receptors (PPARs) signaling pathway after AD. In addition, two distinct protein marker panels could be defined for osteoblastic and adipocytic cell lineages. Hereby, overexpression of AEBP1 and MCM4 for OB as well as of FABP4 for AD was detected as the most promising molecular markers. Combination of deep neural network and machine-learning algorithms with data-independent mass spectrometry distinguish MSCs and cell lineages after adipogenic or osteoblastic differentiation. We identified specific proteins as the molecular basis for bone formation, which could be used for regenerative medicine in the future.
Keywords: SWATH; data-independent acquisition mass spectrometry; differentiation markers; human stromal/mesenchymal stem cells; machine learning; protein profiling.