David Ochoa, Mindaugas Jonikas, Robert T Lawrence, Bachir El Debs, Joel Selkrig, Athanasios Typas, Judit Villén, Silvia DM Santos, Pedro Beltrao
Cells need to constantly adapt to internal and external conditions in order to maintain homoeostasis. During cellular decision-making, signal-transduction networks dynamically change in response to cues, triggering cellular state-defining responses. Multiple mechanisms exist to transfer this information from sensors to the corresponding molecular responses, one of the fastest being the reversible post-translational modification of proteins (PTMs). Through these targeted modifications, such as phosphorylation, the cell can quickly alter enzymatic activities, protein interactions or sub-cellular localization in order to produce a coordinated response to a given stimulus (Pawson 2004). Protein phospho-regulation constitutes a highly conserved regulatory mechanism relevant for a broad set of biological functions and diseases (Beltrao 2012).
Over the past decades, our view of cellular signaling has advanced from an idea of isolated and linear cascades to highly complex and cooperative regulatory networks (Jordan 2000, Gibson 2009). Different perturbations in cellular conditions often activate different sets of interconnected kinases, ultimately triggering appropriate cellular responses. The complete understanding of such cell-fate decisions would require the systematic measurement of changes in kinase activities under multiple perturbations, but the small number of quantified regulatory events (i.e. tens) that were possible to date has limited our knowledge of cellular decision making and its molecular consequences (Kim 2011, Bendall 2011, Niepel 2013, Garmaroudi 2010).
Advances in mass-spectrometry and enrichment methods now allow measuring changes in thousands of phosphorylated peptides at a very high temporal resolution (Olsen 2013, Kanshin 2015, Humphrey 2015). Recent studies on human quantitative phosphorylation include responses at different cell-cycle stages (Dephoure 2008, Olsen 2010), after DNA damage (Beli 2012), EGF stimulation (Olsen 2006, Mertins 2012), prostaglandin stimulation (de Graaf 2014) and different kinase inhibitions (Weber 2012, Oppermann 2013, Kettenbach 2011, Hsu 2011) among many others. More recently, improvements in experimental and computational methods have fostered the study of differential regulation of phosphosites and kinases in different cancer types (Casado 2013), the modeling of drug resistance (Wilkes 2015) and inference of more precise pathway models (Terfve 2015). We suggest that the integrated analysis of phosphoproteomic responses after a wide panel of heterogeneous perturbations can expedite our understanding of cell decision-making processes.
In this study, we have compiled condition-dependent changes in human protein phosphorylation derived from 2,940,379 phosphopeptide quantifications in 435 heterogeneous perturbations. After quality control and normalization, we infer and benchmark the changes in 215 kinase activities in 399 conditions. We show that the similarity of kinase regulatory profiles can be used as a fingerprint to compare conditions in order to, for example, identify perturbations that modulate the kinase activity changes of a condition of interest. The large number of conditions analyzed allowed us to identify the kinases that are broad regulators (i.e. generalist kinases), found to be central kinases of the signaling network. Individual kinase profiles across conditions were also informative to recapitulate known kinase-kinase interactions and to identify novel co-regulated complexes and phosphosites acting as potential effectors.
To extensively study the heterogeneity and specificity of the human signaling response, we compiled and standardized 41 quantitative studies reporting the relative changes in phosphopeptide abundance after perturbation (see Materials and Methods). From the detected peptides, we collected identifications for 119,710 phosphosites in 12,505 proteins, 63% of which already reported in phosphosite databases (Fig EV1). For these sites, we normalized a total of 2,940,379 quantitative changes in phosphopeptide abundance in a panel of 435 biological conditions covering a broad spectrum of perturbations including targeted kinase inhibition, induced hESC differentiation or cell cycle progression, among many others (Appendix Fig S1, Table EV1). Only 1% of all phosphorylated sites were reported in more than 60% of the studies, whereas 52% of the sites were