Illuminating the human phosphoproteome by dose-resolved proteomics of clinical kinase inhibitors

Introduction

Pathological kinase signaling is among the key causes of human diseases, as it shifts the phosphoproteome out of its homeostasis and leads to uncontrolled cell growth or inflammation. Rationally, kinase inhibitors have become a popular treatment strategy to restore the phosphoproteome. Although many phosphorylation sites (p-sites) can now be analyzed by mass spectrometry, the biological contexts and functions of these p-sites remain hidden in the dark for most cases. Consequently, the complete sets of mechanisms of action(s) of kinase inhibitors in cells remain similarly vague. The known polypharmacology of most inhibitors must be carefully considered in any perturbation experiment and can only be dissected by exploiting affinity differences from dose-resolved experiments. Only by systematically screening inhibitors in various cell line models in a dose-dependent manner, it becomes now possible to understand the consequences of kinase inhibition on a systems level.

Methods

We developed a scalable "decryptM" screening platform in a 96-well format and combined it with a mostly automated phosphoproteomic workflow. MS raw files were searched with MaxQuant and boosted with SIMSI and Percolator. Relevant p-site-specific dose-response curves were identified using the novel tool "CurveCurator". Kinase inhibition estimates were obtained from curated seed p-sites. New kinase substrate relationships were obtained by characteristic potency signatures across the data set.

Results

We performed systematic and short dose-resolved perturbations of a total of 134 approved and phase 3 clinical kinase inhibitors across 5 different cell lines. This results in a complete data matrix of 715 decryptM experiments that are based on 7,865 individual perturbations at a depth of around 20-25"000 p-sites per drug and cell line. All 28 million dose-response curves were fitted and statistically evaluated using our new analysis tool, CurveCurator, which combines statistical significance with biological effect size into a single relevance score that outperforms traditional Benjamini-Hochberg procedures in terms of specificity and sensitivity. Overall, the global degree of perturbation is between 1-10%, constraining the response always close to the affected pathways. Harnessing the potency dimension of decryptM experiments, which hints at the initially inhibited kinase(s), we identified many known and unknown p-sites that follow their direct upstream kinase inhibition/activation signature across the entire data matrix. Additionally, we also collected clear evidence for numerous questionable kinase-substrate relationships in public databases. On a higher level, we also obtained an activity-based view of the dependencies of different kinases and cellular backgrounds.

Conclusion

This is the largest decryptM study existing which allowed a systematic grouping of signaling relevant p-sites into a kinase regulative context by exploiting the potency dimension of drugs.