Finding new ways of targeting cancer cell metabolism by combining high-throughput metabolic profiling of chemical and genetic perturbations
Embargoed until 2026-10-11
Author
Date
2023Type
- Doctoral Thesis
ETH Bibliography
yes
Altmetrics
Abstract
Cancer treatment has made significant progresses over the last decades, improving outcomes of medication for many patients. However, the repertoire of drugs that can be used to target selected cancer cell vulnerabilities is often small, non-selective and patients respond only for a limited time to the same drug treatment. Advances in gene editing techniques, such as CRISPR have greatly facilitated target discovery and expanded the space of novel promising gene targets for cancer therapy, which revealed an even more prominent role of cellular/cancer metabolism in mediating processes like drug tolerance, quiescence or metastasis formation. However, the ability to discover novel drugs that target vulnerabilities is lagging behind, progressively moving the bottleneck from target identification to effective drug inhibitors. Hence there is a need to find innovative experimental designs and novel approaches in drug discovery to expand the drug space with which known and newly identified cancer cell vulnerabilities can be targeted. New computational and experimental approaches have emerged to characterize broader functional impacts of small molecules that go beyond quantifying their growth inhibitory activity. High-throughput and multi-dimensional profiling technologies, such as molecular profiling or high-content imaging measure features that enable a deeper functional characterization of compound actions. However, despite technological advances in molecular profiling techniques, like transcriptomic or proteomics an approach to directly probe drug-induced metabolic changes at large scale is missing to date. Even though metabolism has proven to be an important regulator for cancer progression and homeostasis and hence a promising target for cancer therapy (i.e. more than 50% of newly identified cancer targets are metabolic). Moreover, what type of information can be extrapolated from the systematic analysis of drug-induced metabolic changes and how metabolic drug signatures relate to other drug molecular signatures (e.g. gene expression signatures) remains an open question. Can one use drug metabolic signatures to improve and speed up the functional characterization of small molecules and identify how such compounds can be used to systematically interfere and modulate cellular function?
In this thesis, we generated an unprecedented resource of metabolic responses of cancer cells to largely diverse chemical and genetic perturbations and developed new tools for the normalization and analysis of large compendia of non-targeted metabolomics profiles. With this, we aimed at comparing chemical versus chemical and chemical versus genetic perturbations to perform drug mode of action annotation and identify drugs interfering with promising gene targets for therapy.
In chapter two, we aimed at developing and testing a new framework for drug mode of action prediction. For this, we systematically mapped the metabolic effects of a large space of diverse chemical perturbations in lung cancer cells. We used a high-throughput metabolomics framework to monitor alterations in abundance of 2’296 putatively annotated metabolites caused by 1’520 drugs in A549 lung cancer cells. Surprisingly, although only 27% of the drugs induced measurable growth inhibition, a large majority (87%) resulted in intracellular metabolic changes. With over 3.4 million drug-metabolite dependencies, we produced a functional reference table of drug interference with metabolism, which enabled us to conduct high-throughput characterization of compound libraries across diverse drug-therapeutic classes in a single-pass screen. The metabolic changes uncovered previously undetected drug effects, broadening drug functional annotations and therapeutic applications. Here we predicted and validated one novel DHODH inhibitor and four glucocorticoid receptor agonists. Overall, we were able to validate six of seven predictions on drug MoA, generated by comparing drug-metabolic fingerprints. Additionally, we demonstrated the complementary nature of metabolome profiling to various phenotypic and molecular profiling approaches, opening up new opportunities to combine molecular profiling technologies with phenotypic screens and enhance the efficiency, scope, and precision of preclinical drug discovery.
In chapter three, we aimed at developing and testing a new framework to identify commercially available drugs that interfere with newly identified gene targets in cancer. To this end, by utilizing high-throughput non-targeted metabolomics, we analyzed changes in the abundance of 1’985 putatively annotated metabolites induced by CRISPR Cas-9 mediated knockout of 216 genes, of which more than 50% were identified to be lethal targets in lung adenocarcinoma cells. Notably, although only 36% of genetic perturbations resulted in measurable growth inhibition, the majority caused intracellular metabolic changes within 48- (71%) and 72-hours (76%) post-transfection. We observed significant alterations in the local metabolic networks of 49% of identified enzymatic perturbations, as early as two steps from the target, and could identify specific metabolic changes associated with several knocked out key regulators of cellular function, such as TYMS, GSS, XDH, p53, KRAS, and EGFR. Hence, validating that our profiling method accurately captures metabolic changes relevant to gene perturbation. By comparing metabolic profiles of genetic perturbations with profiles of cells undergoing treatment with 1’520 diverse small compounds collected and discussed in chapter two, we generate a map of drug-gene similarities. Here, we showed that drug-gene pairs with known target correspondence are significantly more similar than drug-gene pairs without target correspondence. Furthermore, we highlighted how one can use this map of drug-gene similarities to mine for chemical and genetic perturbations that elicit similar metabolic responses, which are indicative of similar intracellular effects, ultimately generating testable hypothesis on drugs that interfere with newly identified gene targets in cancer. Additionally, we propose three promising drug-gene associations suggesting for new drugs to interfere with the mitochondrial import inner membrane translocase subunit Tim9, dynactin subunit 2, a gene involved in dynein activation and glutathione synthase.
In summary, through our work, we showed that metabolic profiling of treatment effects unraveled a large space of before unknown and unappreciated intracellular drug effects, in particular for drugs that don’t induce growth inhibition. Such drugs, even if not inhibiting processes essential for rapid cancer cell growth, could be potentially attractive means to interfere with non-growing or metastatic cancer cells. While a metabolic readout recapitulated local changes of metabolism targeting drugs it also allows to recover molecular patterns of drugs of other therapeutic applications. By metabolic profiling of drug treatment effects, we generated and validated functional annotation of compound effects and highlighted its complementarity with other multi-dimensional profiling techniques opening the door for integration with multiplexed molecular profiling technologies to enhance the efficiency and precision of preclinical drug discovery. Furthermore, we showed that the comparison of chemical and genetic induced metabolic changes allows to systematically mine for drugs that elicits similar intracellular changes as gene knockouts, allowing to generate hypothesis and predictions on drugs that interfere with newly identified gene targets in cancer.
Overall, we established an effective workflow for metabolic profiling of chemical and genetic perturbations to functionally characterize the MoA of small chemical compounds and we highlight its potential to generate testable hypothesis on drugs interfering with newly identified gene targets in cancer. Hence, our approach opens new opportunities in drug screening to enhance and speed up the discovery of new and potent drugs as well as of drugs that interfere with newly identified gene targets to improve cancer therapy. Show more
Permanent link
https://doi.org/10.3929/ethz-b-000635899Publication status
publishedExternal links
Search print copy at ETH Library
Contributors
Examiner: Sauer, Uwe
Examiner: Zampieri, Mattia
Examiner: Picotti, Paola
Examiner: Beltrao, Pedro
Publisher
ETH ZurichOrganisational unit
03713 - Sauer, Uwe / Sauer, Uwe
More
Show all metadata
ETH Bibliography
yes
Altmetrics