Potential Diagnostic Biomarker Detection for Prostate Cancer Using Untargeted and Targeted Metabolomic Profiling

Abstract

Prostate cancer (PCa) remains one of the leading causes of cancer mortality in men worldwide, currently lacking specific, early detection and staging biomarkers. In this regard, modern research focuses efforts on the discovery of novel molecules that could represent potential future non-invasive biomarkers for the diagnosis of PCa, as well as therapeutic targets. Mounting evidence shows that cancer cells express an altered metabolism in their early stages, making metabolomics a promising tool for the discovery of altered pathways and potential biomarker molecules. In this study, we first performed untargeted metabolomic profiling on 48 PCa plasma samples and 23 healthy controls using ultra-high-performance liquid chromatography coupled with electrospray ionization quadrupole time-of-flight mass spectrometry (UHPLC-QTOF-[ESI+]-MS) for the discovery of metabolites with altered profiles. Secondly, we selected five molecules (L-proline, L-tryptophan, acetylcarnitine, lysophosphatidylcholine C18:2 and spermine) for the downstream targeted metabolomics and found out that all the molecules, regardless of the PCa stage, were decreased in the PCa plasma samples when compared to the controls, making them potential biomarkers for PCa detection. Moreover, spermine, acetylcarnitine and L-tryptophan had very high diagnostic accuracy, with AUC values of 0.992, 0.923 and 0.981, respectively. Consistent with other literature findings, these altered metabolites could represent future specific and non-invasive candidate biomarkers for PCa detection, which opens novel horizons in the field of metabolomics.

Keywords: biomarkers; diagnosis; metabolomics; prostate cancer.

Figure 1

Volcano plot: red-marked m/z values are decreased in the patients’ group compared to the control group; blue-marked m/z values are increased in the patients’ group compared to the control group.

Figure 2

(a, left) PLS-DA plot with a covariance of 39.3% and (b, right) sPLS-DA plot with a covariance of 35.2% showing the discrimination between patient and control groups.

Figure 3

PLS-DA score plots showing the differences between controls and patients (subgroups PI-PIV).

Figure 4

Ranking of top 15 molecules with higher VIP scores, according to PLS-DA analysis (their identification can be made by using Table S1).

Figure 5

Metabolite concentrations (μM) between all patients and controls for the five selected molecules in the targeted analysis.

Figure 6

ROC curves for the five molecules selected in the targeted analysis.