Chemical proteomics for drug target deconvolution and to study biological systems

Abstract

Mass spectrometry-based proteomics has become an irreplaceable method for the unbiased and system-wide study of protein chemistry. As modern proteomics methods can not only be applied to quantify protein expression levels, but also monitor changes in post-translational modifications, subcellular localization, turnover rate, and lastly protein structures, it has become a method of choice for drug discovery and dissecting cellular processes. The work presented in this thesis is the combined effort of developing modification-free chemical proteomics methods and their interchangeable application for both drug target deconvolution and cell transition elucidation. In addition, as proteomics becomes more accessible and multidimensional, a focus was set on the development of user-friendly databases and analytical workflows, facilitating the exploration of generated data and the adoption of our methods.

In paper I, we compiled a database containing the proteome responses of cancer cells to 56 anti-cancer drugs and used Orthogonal Projections to Latent Structures Discriminant Analysis (OPLS-DA) modeling to extract specific regulated proteins. We show that our approach determines both proteins for target deconvolution and mechanisms of action elucidation. Additionally, after empirically determining the minimal number of contrasting drugs, we performed three experiments in different cancer cell lines at an increased analytical depth, showing the importance of cell-line dependencies in drug discovery. Lastly, we created an intuitive graphical user interface for easy exploration of all acquired data and allowing straightforward analysis of custom-generated data sets.

In paper II, we extended Thermal Proteome Profiling (TPP) to identify enzyme-substrate associations. We leveraged the fact that the thermal stability of the substrates might change upon the addition of PTMs, terming the new method System-wide identification and prioritization of enzyme substrates by thermal analysis (SIESTA). We showed the generality of our method with three enzyme systems, TXNRD1, AKT1, and PARP10, all with their distinct PTMs. The putative substrates identified by SIESTA show a good overlap with known ones and allow for prioritized validation based on the magnitude of the thermal shift. Additionally, we also determine the interactome of the three different cosubstrates as well as protein-protein interactions with the added enzymes.

In paper III, we studied protein stability alterations and expression changes of cells transitioning from pluripotency to differentiated cells. We reprogrammed human foreskin fibroblasts into induced pluripotent stem cells, which we differentiated into embryoid bodies. Developing PISA-Express allowed us to simultaneously quantify protein stability alterations and expression changes along these transitions. Merging these changes into a single analysis by Sankey diagrams, we identified key differences between pluripotent and terminally differentiated cells. This included ribosomal proteins, which show lower thermal stability in pluripotent cells compared to differentiated ones. Further experiments showed that this is caused by a lower pool of functionally assembled ribosomes, controlled by SBDS. Collectively, we developed a new method for the simultaneous analysis of protein thermal stability and expression during cell transitions and created a matching innovative analysis approach. Ultimately, a web interface was created allowing for the exploration of our data and the user-friendly analysis of custom-generated data based on our Sankey diagram approach.

False negatives lead to blind spots for every drug target deconvolution method. Therefore, in paper IV we develop a novel method leveraging the kosmotropic effect of ions belonging to the Hofmeister series for TPP and PISA-style experiments, providing much-needed orthogonality. Intending to efficiently pool multiple concentrations samples, as in a PISA-style experiment, we develop a straightforward quenching approach for the kosmotropic ions. We benchmarked this novel method against several drugs in complex cellular lysate, detecting known direct targets, as well downstream events when treating intact cells. We compared the ion-based protein precipitation with the temperature-based one in a full-curve experiment as well as in a PISA experiment using a clinically approved kinase inhibitor, showing the partial orthogonality between them. Ultimately, we performed a PISA experiment with minute amounts of samples on microscopy over slides, potentially opening chemical proteomics methods for paucicellular samples.

Finally, in paper V we develop a novel method for the system-wide characterization of drug residence time in cellular lysate and intact cells based on PISA. To validate our approach, we studied two kinase inhibitors as well as a covalent small molecule, allowing us to prioritize their target landscape. Interestingly, we detected no correlation of drug residence time with either the magnitude of the thermal shift or the binding affinity, underlying the importance of determining this parameter. Additionally, we created a higher throughput version of our method, allowing the characterization of the target landscape, drug residence time, and binding affinity in a single mass spectrometer injection. Ultimately, we extended our method for use in intact cells, following the approach developed in paper III, providing a more holistic picture compared to the lysate, as they can actively import, metabolize, and export small molecules. For custom-generated data we also implemented an easy-to-use graphical user interface, allowing for the analysis of various PISA design experiments.