Roland Geyer (Presenter)
Bern University Hospital
Authorship: Roland Geyer (1), Alexander B. Leichtle (1), Michael Hayoz (1), Christos Nakas (1,2), Cédric Bovet (1), Anke Tönjes (3) and Martin G. Fiedler (1)
(1) Bern University Hospital, Switzerland, (2) Laboratory of Biometry of the University of Thessaly (Greece), (3) Department for Internal Medicine, University Leipzig, Leipzig, (Germany)
| Short Abstract Clinical Metabolomics usually has to deal with highly diverse sample sets in its search for distinguishing features with a potential as biomarker. Additional to a specific disease or treatment of interest a myriad of lifestyle and genetic factors influence the metabolic profile of each individual sample. Preanalytical and analytical variation in larger sample cohorts adds further variability. A semi-targeted screening of samples based on mass transitions covering known metabolic pathways can stratify (group) sample sets and identify outliers. The global/non-targeted screening could afterwards evaluate known or so far unknown metabolites with potential as biomarkers using smaller sample sets. |
Long Abstract
Introduction
Clinical Metabolomics usually has to deal with highly diverse sample sets in its search for distinguishing features with a potential as biomarker. Additional to a specific disease or treatment of interest a myriad of lifestyle and genetic factors influence the metabolic profile of each individual sample. Preanalytical and analytical variation in larger sample cohorts adds further variability.
Samples from oral glucose tolerance test (oGTT) were used to evaluate changes in the metabolite profile related to time as well as related to a potential development or early stage of diabetes. The objective was to compare the ability of non-targeted and semi-targeted screening methods to separate sample groups and to elucidate the metabolites distinguishing these groups (potential biomarker). Semi-targeted means here the use of defined mass transitions whereby each can cover a range of isobaric compounds. It allows only a relative quantification. We assumed the non-targeted screening would more clearly separate e.g. the two time points due to the higher number of metabolites and that shifts in metabolite profiles will be seen with both methods.
Methods
Blood samples of a sorbian population were obtained 30 and 120min after intervention by an 75g oral glucose tolerance test (oGTT) at the University Hospital of Leipzig (1). The serum samples were prepared with a commercially available kit to extract acylcarnitines, glycerophospho- and sphingolipids as well as amino acids. Eluates were analyzed by triple quadrupole mass spectrometry (QqQ; Waters Xevo TQ-S) with or without chromatographic separation and by scanning with high resolutions mass spectrometry (HRMS; Waters Synapt G2-S HDMS) coupled to the same reversed phase chromatography (column Cortecs 100x2.1, 1.6µm; Waters Acquity UPLC I-Class). Peak picking and alignment was done for non-targeted UPLC-MS data with the Progenesis QI software (Waters). MRM-DIFF, a data processing tool for multiple reactions monitoring (MRM)-based differential analysis (3) was used for extracting peaks from the UPLC-MS/MS data including many isobaric phospholipids. The different sets of raw data were transposed and imported in SIMCA (Umetrics) to enable a comparison by multivariate statistical analysis methods (e.g., PCA, OPLSD-DA). Scripts from the r-project were used for a paired test design (multilevel sPLS-DA) (4).
Results
Based on a 12min chromatographic separation of 111 mass transitions covering glycerophosphatidylcholine- and sphingolipids in positive electrospray mode more than 1800 peaks were extracted by peak picking and alignment software, while scanning with quadrupole time-of-flight (QToF) HRMS revealed >4400 features. Peaks from these semi-targeted or non-targeted analysis were annotated by their retention time and mass or mass transition. After exclusion of peaks with insignificant change or unsuitable peak widths QToF data had about 3 times more features (~800 vs 2300, respectively). However, neither prinicipal component analysis (PCA) nor OPLS-DA algorithm could separate samples sets of the two time points. Inter-individual variability was higher than the metabolic effects of concentrated glucose intake. Using a paired test design with multilevel sPLS-DA algorithm well separated the samples of the two time points based on both, the data from targeted and from non-targeted measurements.
Non of the analytical features from targeted lipid analysis with highest influence on distinguishing 30 and 120min samples (n=15 with highest loadings) was among the important chromatographic and/or mass resolved peaks from the non-targeted screening. With solely QToF based screening the respective (lipid) metabolites would have been probably missed as targets for subsequent quantitative measurements.
As loading values for the triple quadrupole peaks were higher but data amounts and analysis time much smaller compared to full scan HRMS peaks, the use of mass transitions may be suitable for an initial grouping and identifying outliers in sample sets.
Even flow injection analysis (without chromatographic separation) can be sufficient as an initial measurement, which will be shown for e.g. mass transitions related to glycerophosphatidylcholines and acylcarnitines.
The examples do indicate that non-targeted and (semi)targeted mass spectrometry rather can be used complementary instead of executing it sequentially. This would mean a paradigm shift in the screening for potential biomarker in complex sample sets for clinical metabolomics.
Conclusion
A (semi)targeted screening of samples based on mass transitions covering known metabolic pathways can stratify (or group) sample sets and identify outlier. The global/non-targeted screening with its broader mass and peak coverage can afterwards evaluate known or so far unknown metabolites with potential as biomarkers using smaller samples sets.
References & Acknowledgements:
References
[1] Mai M, Tönjes A, Kovacs P, Stumvoll M, Fiedler GM, et al. (2013) Serum Levels of Acylcarnitines Are Altered in Prediabetic Conditions. PLoS ONE 8(12): e82459. doi:10.1371/journal.pone.0082459
[2] http://www.biocrates.com/products/research-products/absoluteidq-p180-kit
[3] Tsugawa et al. 2015, Front. Genet. 5:471. doi: 10.3389/fgene. 2014.00471
[4] Kim-Anh Le Cao, Ignacio Gonzalez, Sebastien Dejean with key contributors Florian Rohart, Benoit Gautier, contributions from Pierre Monget, Jeff Coquery, FangZou Yao and Benoit Liquet. (2015). mixOmics: Omics Data Integration Project. R package version 5.1.2. http://CRAN.R-project.org/package=mixOmics
| Description | Y/N | Source |
| Grants | no | |
| Salary | no | |
| Board Member | no | |
| Stock | no | |
| Expenses | no |
IP Royalty: no
| Planning to mention or discuss specific products or technology of the company(ies) listed above: | no |