Metabolic fingerprinting of high-fat plasma samples processed by centrifugation- and filtration-based protein precipitation delineates significant differences in metabolite information coverage.

Metabolomics and metabolic fingerprinting are being extensively employed for improved understanding of biological changes induced by endogenous or exogenous factors. Blood serum or plasma samples are often employed for metabolomics studies. Plasma protein precipitation (PPP) is currently performed in most laboratories before LC-MS analysis. However, the impact of fat content in plasma samples on metabolite coverage has not previously been investigated. Here, we have studied whether PPP procedures influence coverage of plasma metabolites from high-fat plasma samples. An optimized UPLC-QTOF/MS metabolic fingerprinting approach and multivariate modeling (PCA and OPLS-DA) were utilized for finding characteristic metabolite changes induced by two PPP procedures; centrifugation and filtration. We used 12-h fasting samples and postprandial samples collected at 2h after a standardized high-fat protein-rich meal in obese non-diabetic subjects recruited in a dietary intervention. The two PPP procedures as well as external and internal standards (ISs) were used to track errors in response normalization and quantification. Remarkably and sometimes uniquely, the fPPP, but not the cPPP approach, recovered not only high molecular weight (HMW) lipophilic metabolites, but also small molecular weight (SMW) relatively polar metabolites. Characteristic SMW markers of postprandial samples were aromatic and branched-chain amino acids that were elevated (p<0.001) as a consequence of the protein challenge. In contrast, some HMW lipophilic species, e.g. acylcarnitines, were moderately lower (p<0.001) in postprandial samples. LysoPCs were largely unaffected. In conclusion, the fPPP procedure is recommended for processing high-fat plasma samples in metabolomics studies. While method improvements presented here were clear, use of several ISs revealed substantial challenges to untargeted metabolomics due to large and variable matrix effects.