Mass spectrometry-based proteomics and metabolomics in biomedicine
Mass spectrometry has a long and illustrious history as an analytical science, even before the development of electrospray and MALDI made MS easily applicable to the study of proteins. Developments over the last few years have steadily increased the scope of MS-based studies in molecular biology but important challenges remain, chief among them the lack of comprehensiveness compared to oligonucleotide based systems. However, this limitation is now falling away, as I will show with deep proteomic analysis of human cancer cell lines. More than 10,000 different proteins can now be identified in such systems, in a relatively short time, shedding new light on similarities and differences to each other and to in vivo cells. Developments in sample preparation make these capabilities available in clinically relevant material as well, such as formalin-fixed, paraffin embedded cancer samples.
Enabled by these developments, mass spectrometry-based proteomics is now being employed in a wide variety of applications spanning the entire breadth of molecular biology (Aebersold and Mann, Nature, 2016). Arguably, cell signaling is one of the areas in which this approach has made the greatest and most unique contributions. Here, I will summarize the current status of the methodology and applications of the ‘EasyPhos’ method that we have developed recently (Humphrey et al. Nat. Biotech, 2015). This technology now enables studying complex signaling events in vivo and we have used it to uncover the long sought substrates of the Parkinson’s kinase LRRK2 (Steger et al. eLife 2016, 2017). In the circadian rhythm it has revealed that a large percentage of the phospho-proteomics is coordinately regulated during the day and night cycle, and that many of the target sites appear to fine tune the metabolic machinery (Robles et al. Cell Metabolism, 2017). Recently, we have used EasyPhos to unravel signaling events downstream of opioid receptors in the brain in the context of analgesia and addiction. Importantly, this knowledge can be used to selectively ablate the undesired signaling effects (Liu et al. Science, 2018). Finally, we describe a multi-layered proteomics approach to discover a new biomarker in ovarian cancer that correlates with long term survival after chemotherapy. We use proteomics together with cell biological follow-up – especially DNA damage assays – to reveal the mechanism of action of this biomarker (Coscia et al, Cell 2018). We also discuss application of the same basic MS-technology to metabolomics and lipidomics.
We have also extended this concept to the analysis of cellular interactomes (Hein et al. Cell 2015) and transcription factor complexes. Body fluids have long been of great interest to researchers in proteomics because of their potential to directly ‘phenotype’ individuals with minimally invasive procedures. However, the high dynamic range – along with other challenges – have long stymied this approach. We have recently revisited this area using the latest technological advances. The ‘protein correlation profiling’ approach allows us to study the plasma proteome rapidly in a wide range of conditions (Geyer et al. Cell Systems 2016). We have now increased the protein coverage several-fold using a novel scan mode termed BoxCar and applied our workflow to a number of clinical studies (Geyer et al. MSB 2016). These will be described in the talk together with a perspective of how plasma proteome profiling could be implemented in the clinic.
These biological and clinical applications all require sophisticated and robust bioinformatic analyses. Over the years we have developed several search engines and workflows for MS-based proteomics, but here I will focus on the ‘Clinical Knowledge Graph’ in which we used a number of ontologies to import a very large set of data points and their relationships into a graph based structure. This graph database currently contains more than 10 million edges and 100 million relationships. We demonstrate how the Clinical Knowledge Graph can be used to interpret proteomics results, integrate them with other omics data types and even help to inform treatment decisions (Alberto Santos, Ana Rita Colcao, … Matthias Mann, in preparation).