Circulating fatty acids from high-throughput metabolomics platforms as potential biomarkers of dietary fatty acids

Abstract

Background: Some fatty acids, i.e. n-3 and n-6 polyunsaturated fatty acids (PUFA), from metabolomics platforms based on nuclear magnetic resonance imaging (NMR) or liquid chromatography mass-spectrometry (LC-MS) are suggested to reflect dietary exposure. NMR and LC-MS are both relatively fast and cheap, however few studies have investigated their validity. Linoleic acid (LA) and docosahexaenoic acid (DHA), measured using gas chromatography (GC), are established biomarkers of dietary n-6 and n-3 PUFA intake, respectively.

Objective: To examine if circulating fatty acids derived from two commonly applied metabolomics platforms (using NMR and LC-MS) provide similar information compared to GC in two pooled population-based cohorts, one patient cohort, and in a randomized controlled trial (RCT).

Methods: Spearman rank correlations were conducted between LA and DHA in cholesteryl esters (CE) from GC and whole serum/plasma LA and DHA from the metabolomics platforms in a pooled population-based cohort of men and women (n ˜ 1100) (primary analysis). Secondary correlation analyses included fatty acid classes such as n-3 PUFA, n-6 PUFA, saturated fatty acids (SFA), monounsaturated fatty acids (MUFA) and total PUFA. Additionally, correlations were investigated for LA, DHA and the five fatty acid classes in phospholipids (PL), triacylglycerols (TAG) and non-esterified fatty acids (NEFA) in a RCT of n = 60 as well as in a population with biopsy-verified non-alcoholic fatty liver disease (NAFLD) (n = 59). Misclassification was examined using cross-tabulation and visualized using alluvial plots.

Results: Moderate to strong correlations (r = 0.51-0.81) were observed for LA and DHA in multiple lipid fractions in all cohorts using the NMR platform. For the pooled cohort, LA (r = 0.67, P < 0.0001) and DHA (r = 0.68, P < 0.0001) assessed in CE were strongly correlated with LA and DHA derived using NMR. Nearly half (49%) were correctly classified into their respective quartiles. Using LC-MS, only DHA (r = 0.44, P < 0.0001) demonstrated moderate correlations with DHA from GC.

Conclusions: Unless fatty acid data from GC analysis is available or feasible, NMR-based technology might be a better option than a LC-MS-based platform, at least for certain PUFA. This should be taken into account in future studies aiming to use circulating fatty acids as dietary biomarkers for the investigation of diet-disease relationships.

Keywords: Epidemiology; Fatty acids; Metabolomics; Nutrition; Validation.