Plurifaceted proteomics in studying cellular dynamics and action mechanisms of anticancer drugs

Abstract

Mass spectrometry (MS)-based proteomics has developed tremendously in recent years and was the leading technology for many novel methods to study protein chemistry. Contrary to classical approaches based on Western blot, MS-based approaches are mostly unbiased. In addition to protein expression levels, today several protein chemical properties can be examined by MS-based proteomics making it unique in comparison to transcriptomics approaches. These properties include post-translational modifications (PTMs), protein localization, synthesis/degradation, and lastly protein thermal stability, adding novel dimensions or facets to characterize the proteome. However, these facets are difficult to combine as they are mainly orthogonal and are therefore often analyzed separately. This thesis presents a simplified and higher throughput version of current protein stability analyses and showcases the advantages of combined/merged analysis of proteomics facets including our new method, as well as expression and redox proteomics to study anticancer treatments and cellular dynamics.

In paper I, we studied the dynamics of cancer cells in vitro with and without anticancer treatment over the course of 48 h by monitoring protein expression every 6 h. We discovered that naturally occurring proteome variations are on par with anticancer treatment killing 50% of the cells after 48 h. Then, we acquired a deep proteomics dataset of untreated HCT116 and A375 cell lines. Surprisingly, we observed downregulation of proteins involved in cell division and upregulation of proteins involved in metabolism as early as 12 h after treatment, suggesting that growth inhibition happens earlier than usually assumed and even at low cell confluence. In paper II we developed the proteome integral solubility alteration (PISA) assay that increases the throughput of pre-existing drug target deconvolution methods based on protein stability/solubility measurements and reduces the analysis time and cost as well as sample requirements. We provided theoretical calculations showing that the integral of the curve correlates well with melting temperature estimations in Thermal Proteome Profiling (TPP) and tested our assumptions with publicly available TPP datasets. Then we performed a proof of principle experiment using the well-studied methotrexate (MTX) and 5-fluorouracile (5-FU) as test drugs highlighting the targets as outliers with our method. Furthermore, we demonstrated that PISA assay can also be used for concentration series analysis as in 2D-TPP. Finally, we showcased the higher throughput of PISA compared to TPP by simultaneously analyzing nine drugs in one multiplexed analysis. In paper III we used a combination of chemical proteomics approaches to study the target and mechanism of action (MOA) of Auranofin (AF) (Ridaura®). Functional Identification of Target by Expression Proteomics (FITExP), TPP, and redox proteomics combined highlighted that thioredoxin reductase 1 (TXNRD1) is indeed a top target hit of AF, and that the main MOA of the drug is through disruption of the redox balance in the cell. Finally, we showed that protein thermal shifts can be associated with altered cysteine oxidation levels in proteins, suggesting that TPP is suitable to study disulfide bond formation/reduction and map some cysteines to the active sites of sulfiredoxin 1 (SRXN1) and peroxiredoxin 5 (PRDX5) as examples. Overall, our study demonstrates that using only one of the proteomics methods is not sufficient to accurately pinpoint drug target and MOA, but a combination of multiple complementary methods should be used instead.

Following the success of our strategy in paper III, we decided to utilize the same strategy in paper IV using PISA developed in paper II instead of TPP, to study two novel inhibitors of TXNRD1, namely TRi-1 and TRi-2. For these studies, we used the mouse cancer cell lines B16-F10 and Lewis lung carcinoma (LLC) to provide a comprehensive analysis of the therapeutic effects of the inhibitors. Since we showed that Txnrd1 (the mouse version of the human TXNRD1 protein) is a main target of AF, we decided to use it as our reference point to evaluate the specificity of the two new compounds. AF had a broader effect on the proteome than TRi-1 and TRi-2 and had more target hits in PISA analyses. This suggested that AF has lower specificity than TRi-1 and TRi-2. TRi-1 was the most specific Txnrd1 inhibitor with a better target ranking in FITExP and the least target hits in PISA followed by TRi-2 and AF. Thus, we showed that TRi-1 and TRi-2 are indeed more specific inhibitors of Txnrd1. In addition to these findings, we also highlighted that only AF triggers a high nuclear factor erythroid 2-related factor 2 (Nrf2) antioxidant response, suggesting that this response is not necessarily Txnrd1-dependent. Finally, we detected selenocysteine-specific elongation factor (Eefsec), mini-chromosome maintenance complex-binding protein (Mcmbp), glycogen synthase kinase-3 (Gsk3) a and b, as target hits of AF, which would explain at least partially three different effects of AF treatment. Collectively, our data represents a resource for redox biologists interested in Txnrd1 inhibition. Our approach provides a framework for target deconvolution using proteomics approaches.

Finally, in paper V we studied the dynamics of the proteome in transition between various cell types. We reprogrammed human foreskin fibroblasts (hFF) into induced pluripotent stem cells (iPSCs), which we differentiated through embryoid bodies (EBs) formation. We examined protein expression and stability after each cell type transition using PISA-Express, a new version of the PISA assay developed in paper II. We merged the readout from protein expression and thermal stability in one analysis using Sankey diagrams to detect changes in protein properties during proteome transitions resulting in the ProteoTracker web interface (http://www.proteotracker.genexplain.com/). Using this innovative analysis, we discovered that ribosomes are less stable in pluripotent stem cells (PSCs) compared to differentiated cells and that this difference stems from the deficiency of one ribosome maturation factor, Shwachman-Bodian-Diamond syndrome protein (SBDS). Knock-down (KD) of SBDS slowed down translation and increased expression of the master pluripotency markers homeobox protein NANOG (NANOG), and octamer-binding transcription factor 4 (OCT4). Collectively, we developed a new method for simultaneous analysis of protein thermal stability and expression, a new analysis and visualization tool, and provided evidence that control of translation through ribosome biogenesis is a natural mechanism used by PSCs to maintain the pluripotency state.