You have accessJournal of UrologyProstate Cancer: Localized: Surgical Therapy V1 Apr 2018PD38-07 MULTIPLATFORM METABOLOMICS REVEALS PREDICTIVE PROSTATE CANCER RECURRENCE PHENOTYPES FOLLOWING RADICAL PROSTATECTOMY Callan Brownfield, Chaevien Clendinen, David Gaul, Rebecca Arnold, Arthur Edison, Facundo Fernandez, and John Petros Callan BrownfieldCallan Brownfield More articles by this author , Chaevien ClendinenChaevien Clendinen More articles by this author , David GaulDavid Gaul More articles by this author , Rebecca ArnoldRebecca Arnold More articles by this author , Arthur EdisonArthur Edison More articles by this author , Facundo FernandezFacundo Fernandez More articles by this author , and John PetrosJohn Petros More articles by this author View All Author Informationhttps://doi.org/10.1016/j.juro.2018.02.1754AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookTwitterLinked InEmail INTRODUCTION AND OBJECTIVES By far the most commonly used technologies in metabolomics are Nuclear Magnetic Resonance (NMR) and Liquid Chromatography-Mass Spectrometry (LC-MS). The structural diversity of metabolites, however, dictates that maximum metabolome coverage can only be achieved by applying multiple complementary technologies simultaneously. Here, we use fused 1H NMR and LC-MS (HILIC, RP, +ve, -ve) datasets to determine the alterations in metabolic pathways indicative of prostate cancer (PCa) recurrence following prostatectomy. A priori prediction of the likelihood of PCa recurrence prior to prostatectomy would allow avoiding unnecessary surgery and associated complications in patients not likely to be cured by prostatectomy. We provide strong evidence that it is feasible to predict recurrence after radical prostatectomy based on the patient metabolic phenotype prior to surgery. METHODS De-identified serum obtained prior to planned prostatectomy was aliquoted and specifically prepared for either NMR, HILIC-MS, or RP-MS. MS experiments were performed using an Ultimate3000 LC system coupled to a Q-Exactive HF equipped with a HESI source in both +ve and -ve mode. NMR experiments were performed on a Bruker Avance III HD 600 MHz system equipped with a triple resonance 5 mm cold probe. All MS data were processed using Progenesis QI. Database searching NMR peaks and accurate mass and MS2 matching were used to produce a list of tentative IDs. Lipid IDs were obtained via Lipidsearch using accurate MS and MS2. Multivariate data analysis was done using MATLAB and the PLS Toolbox. All patients had long term post-prostatectomy follow-up and were categorized as recurrent or non-recurrent based on standard criteria, including biochemical, clinical and/or radiologic recurrence. RESULTS Five datasets resulted from our multiplatform approach. HILIC +ve and -ve mode and NMR datasets were used to provide coverage for predominately polar metabolites, while reverse phase (RP) +ve and -ve mode datasets involved predominantly non-polar metabolites. HILIC +ve and -ve datasets provided over 190 uniquely identified metabolites, with 50 unique additional metabolites obtained from NMR. The biggest changes were observed in purine and TCA cycle metabolites. Though some metabolites showed no significant differences, recurrence was shown to have increased metabolism purine metabolism. Patients without post-operative cancer recurrence had significantly higher serum glucose prior to surgery while those that recurred following surgery had significantly higher pre-operative lactate. In most cases, information between these HILIC-MS and NMR data were consistent and correlations between like metabolites across different platforms were significant. Most small molecules identified in the HILIC and NMR datasets were not covered by the RP datasets. RP +ve and -ve mode provided information on lipidome alterations. Over 450 lipids were identified from both positive and negative mode LC-MS experiments. The most common altered lipids being triglycerides (TG), about half of which increased in recurrence. Across all chain lengths and saturations, recurrent patients showed more significantly changed lipids. Though some lipids were seen in the HILIC and NMR datasets, lipid coverage in the RP datasets were much greater and inclusive.Feature selection from the fused dataset employing a genetic algorithm yielded 29 features that distinguished between patients that had recurrence or went into remission following surgery with approximately 99 percent accuracy and 98 percent specificity. Lipids, lactate, and phenylalanine were among the features that underlie the separation between the two groups. The use of all platforms was vital in characterizing the metabolic phenotype underlying PCa recurrence. CONCLUSIONS A multiplatform analytical approach provides in-depth views of PCa metabolome and exceptionally accurate prediction of recurrence from a single pre-operative blood sample. © 2018FiguresReferencesRelatedDetails Volume 199Issue 4SApril 2018Page: e738-e739 Advertisement Copyright & Permissions© 2018MetricsAuthor Information Callan Brownfield More articles by this author Chaevien Clendinen More articles by this author David Gaul More articles by this author Rebecca Arnold More articles by this author Arthur Edison More articles by this author Facundo Fernandez More articles by this author John Petros More articles by this author Expand All Advertisement Advertisement PDF downloadLoading ...
Search from 250 M+ research papersSearch from 250 M+ research papers
Open Access
Nov 5, 2018
Jung-Hoon Kim + 4
