Proteomics and machine learning-based approach to decipher subcellular proteome of mouse heart.

Abstract

Protein compartmentalisation to distinctive subcellular niches is critical for cardiac function and homeostasis. Here, we employed a rapid and robust workflow based on differential centrifugal-based fractionation with mass spectrometry (MS)-based proteomics and bioinformatic analyses for systemic mapping of the subcellular proteome of mouse heart. Using supervised machine learning (ML) of 450 hallmark protein markers from 16 subcellular niches, we further refined the subcellular information of 2083 proteins with high confidence. Our data validation focused on specific subcellular niches such as mitochondria, cell surface, cardiac dyad, myofibril and nuclear, unfolding dominant subcellular localisation of proteins in their native environment of mouse heart. We further provide targeted nuclear enrichment and co-immunoprecipitation-based proteomic validation from the heart of nuclear-localising protein networks. This study provides novel insights into the molecular landscape of different subcellular niches of the heart and serves as a draft map for heart subcellular proteome.