Revisiting multi-omics-based predictors of the plasma triglyceride response to an omega-3 fatty acid supplementation

Abstract

Background: The aim of the present study was to identify the metabolomic signature of responders and non-responders to an omega-3 fatty acid (n-3 FA) supplementation, and to test the ability of a multi-omics classifier combining genomic, lipidomic, and metabolomic features to discriminate plasma triglyceride (TG) response phenotypes.

Methods: A total of 208 participants of the Fatty Acid Sensor (FAS). Study took 5 g per day of fish oil, providing 1.9-2.2 g eicosapentaenoic acid (EPA) and 1.1 g docosahexaenoic (DHA) daily over a 6-week period, and were further divided into two subgroups: responders and non-responders, according to the change in plasma TG levels after the supplementation. Changes in plasma levels of 6 short-chain fatty acids (SCFA) and 25 bile acids (BA) during the intervention were compared between subgroups using a linear mixed model, and the impact of SCFAs and BAs on the TG response was tested in a mediation analysis. Genotyping was conducted using the Illumina Human Omni-5 Quad BeadChip. Mass spectrometry was used to quantify plasma TG and cholesterol esters levels, as well as plasma SCFA and BA levels. A classifier was developed and tested within the DIABLO framework, which implements a partial least squares-discriminant analysis to multi-omics analysis. Different classifiers were developed by combining data from genomics, lipidomics, and metabolomics.

Results: Plasma levels of none of the SCFAs or BAs measured before and after the n-3 FA supplementation were significantly different between responders and non-responders. SCFAs but not BAs were marginally relevant in the classification of plasma TG responses. A classifier built by adding plasma SCFAs and lipidomic layers to genomic data was able to even the accuracy of 85% shown by the genomic predictor alone.

Conclusion: These results inform on the marginal relevance of SCFA and BA plasma levels as surrogate measures of gut microbiome in the assessment of the interindividual variability observed in the plasma TG response to an n-3 FA supplementation. Genomic data still represent the best predictor of plasma TG response, and the inclusion of metabolomic data added little to the ability to discriminate the plasma TG response phenotypes.

Keywords: bile acids; gut microbiota; metabolic health; metabolomics; precision nutrition; short-chain fatty acids.

Figure 1

Contrast analysis between short-chain fatty acid levels and clinical parameters. (A, B) show levels of HDL-C and vitamin B-12, respectively, for responders and non-responders before and after the intervention. (C) shows the change in HDL-C levels related to the change in isovaleric acid levels for responders and non-responders. (D) shows the change of vitamin B-12 linked to the change in isobutyric acid levels for both groups. HDL-C, High-density lipoprotein cholesterol.

Figure 2

Predictive performance of metabolomic, lipidomic, and genomic features. The predictive performance of each feature is represented by the area under the ROC-curve (AUC) and the balance accuracy (ACC) metrics. AUC, area under the curve; ACC, accuracy; BA, bile acids; SCFA, short-chain fatty acids; TG + CE, triglycerides and cholesterol esters; SNP, single nucleotide polymorphism.

Figure 3

Predictive performance of the response to an n-3 FAs supplementation. SNP, single nucleotide polymorphism; SCFA, short-chain fatty acid; TG&CE, triglycerides and cholesterol esters; BA, bile acids; n-3 FA, omega-3 fatty acid. The balance accuracy for each feature or combination of features is shown the top of the bars.