Towards a semi-automated analysis of fish plasma by 1H NMR metabolomics - applications to aquaculture

Aquafeed is one of the main costs in aquaculture and major efforts are being made to find sustainable ingredients that could improve production sustainability and efficiency. Our research focused on the inclusion of glycerol in the feed of European seabass (Dicentrarchus labrax), as it is a cheap carbon substrate expected to enter metabolism through the glycolytic pathway. The Nuclear Magnetic Resonance (NMR) spectroscopy provides accurate and high-throughput metabolic profiles, allowing quantification and evaluation of the effects of experimental diets in the context of fish nutrition. The aim of the present study was to assess the applicability of the automated online profiling routine Bayesil to fish blood plasma and compare with the manual targeted metabolite profiling approach. After a 60-day feeding trial (0%, 2.5%, 5.0% glycerol supplemented diets), fish were sacrificed at 6 h and 24 h after last meal and blood was sampled. Filtered plasma samples were analysed by NMR, followed by two metabolite-profiling approaches: i) an automated spectral profiling provided by the Bayesil platform, and ii) a manual metabolite profiling. Spectra were analysed with the automated Bayesil routine, identifying and quantifying 51 metabolites, in a reproducible, fast and user-independent method. On the other hand, the manual approach only managed to quantify 20 metabolites. However, under supraphysiological concentrations of glycerol, namely in the 5.0% diet, the Bayesil algorithm was not able to properly integrate the glycerol and glycine signals. The Bland-Altman and PLS-DA analysis revealed differences; nonetheless, with the exception of acetic acid, creatinine, formate, glutamine and threonine, all other metabolites presented a low bias and/or a strong correlation coefficient. Overall, it is expected a reasonable tradeoff between the accuracy in metabolite quantification and the increase of variables. The automated Bayesil database may be applied as a useful tool for systematic monitoring of fish plasma profile in aquaculture research and industrial quality control analysis, to generate robust variance-based multivariate statistical models. To our knowledge, this is the first time a NMR-based high-throughput automated routine is evaluated for aquaculture research.