Systematic detection of functional proteoform groups from bottom-up proteomic datasets

The cellular proteome, the ensemble of proteins derived from a genome, catalyzes and controls thousands of biochemical functions that are the basis of living cells. Whereas the protein coding regions of the genome of the human and many other species are well known, the complexity and composition of proteomes largely remains to be explored. This task is challenging because mechanisms including alternative splicing and post-translational modifications generally give rise to multiple distinct, but related proteins – proteoforms – per coding gene that expand the functional capacity of a cell.Bottom-up proteomics is a mass spectrometric method that infers the identity and quantity of proteins from the measurement of peptides derived from these proteins by proteolytic digestion. Whereas bottom-up proteomics has become the method of choice for the detection of translation products from essentially any gene, the inherent missing link between measured peptides and their parental proteins has so far precluded the systematic assessment of proteoforms and thus limited the resolution of proteome maps. Here we present a novel, data-driven strategy to assign peptides to unique functional proteoform groups based on peptide correlation patterns across large bottom-up proteomic datasets. Our strategy does not fully characterize specific proteoforms, as is achievable in top-down approaches. Rather, it clusters peptides into functional proteoform groups that are directly linked to the biological context of the study. This allows the detection of tens to hundreds of proteoform groups in an untargeted fashion from bottom-up proteomics experiments.We applied the strategy to two types of bottom-up proteomic datasets. The first is a protein complex co-fractionation dataset where native complexes across two different cell cycle stages were resolved and analyzed. Here, our approach enabled the systematic detection and evaluation of assembly specific proteoforms at an unprecedented scale. The second is a protein abundance vs. sample data matrix typical for bottom-up cohort studies consisting of tissue samples from the mouse BXD genetic reference panel. In either data type the method detected state-specific proteoform groups that could be linked to distinct molecular mechanisms including proteolytic cleavage, alternative splicing and phosphorylation. We envision that the presented approach lays the foundation for a systematic assessment of proteoforms and their functional implications directly from bottom-up proteomic datasets.