Metabolomics using dissolution DNP-NMR

Dissolution Dynamic Nuclear Polarization (dDNP), a hyperpolarization technique, offers a way to dramatically enhance the signals in Nuclear Magnetic Resonance (NMR) experiments, by transferring spin polarization from electrons to nuclei like carbon-13 (¹³C). With a more than 10.000 fold increase in the signal-to-noise ratio (SNR), on a very short time scale, dDNP has opened up the possibilities of directly monitoring metabolism in an organism. Such real-time metabolism is measured by injecting a hyperpolarized ¹³C-labelled tracer into the subject and measuring the biochemical transformations through changes in hyperpolarized NMR signals as a function of time. The first real-time experiment performed clinically was done in 2013, studying tumours in prostate cancer patients. Since then, hundreds of real-time studies have been performed, elucidating metabolic flux in different types of organs and organisms.
Besides such real-time studies, with narrow focus on the immediate metabolic products of a tracer, the field of metabolomics studied through dDNP NMR remains relatively unexplored. Newer studies using hyperpolarized NMR on natural abundance ¹³C samples from tomato extracts, show that the method is highly repeatable and yields easily interpretable spectra with the information necessary for biological classification.
The focus of this thesis is metabolomics performed by combining metabolic fingerprints obtained with dDNP NMR with the tracer based method of Stable Isotope Resolved Metabolomics (SIRM). In this combined methodology a biological system is injected with a ¹³C labelled tracer and, after an  incubation period, the resulting metabolic products are harvested to be analysed in a single dDNP enhanced NMR spectrum (dDNP SIRM fingerprint). This approach has been shown to have potential for insight into metabolic pathways in the cancer cell, useful for classification of cancer and for identification of the biomarkers of cancer.
Specifically, this thesis explores the potential for dDNP SIRM fingerprints to contain the information needed to classify between prostate cancer samples of different aggressivity. Two studies were conducted as part of this thesis, one on prostate cancer cell lines and one on tissue samples from Transgenic Adenocarcinoma of Mouse Prostate (TRAMP) mice. The process from production of the metabolite samples, to the acquisition of the dDNP NMR spectra and subsequent data analysis is described. Statistical analysis was performed with a combination of Random Forest (RF) and Support Vector Machines (SVM), to obtain both feature importance and classification. The goal in the statistical analysis was both to use the dDNP SIRM fingerprints to classify between samples of different cancer types, and to use the highly resolved data to pinpoint which metabolites support such a classification. The latter has the potential to enable selection and evaluation of potential biomarkers of aggressivity in prostate cancer.