MSACL 2017 EU Abstract

Optimisation and Validation of a High-Throughput Semi-Targeted Method by GC-MS with Metabolite Libraries for Large Scale Molecular Epidemiological Research

Antonis Myridakis (Presenter)
Imperial College London

Bio: I am a postdoctoral research associate at the Division of Computation and Systems Medicine. My research at the Imperial College is focused on development of mass spectrometric metabolomic approaches. Furthermore, I employ metabolomics in order to investigate the link between gut microbiome metabolites and the pathogenesis of cardiometabolic and neurodevelopmental diseases. In the past, my main research activities were focused on the exposure and risk assessment of endocrine disrupting and neurotoxic chemicals in pregnant women and their children. I studied at the University of Crete, Greece. I was awarded a PhD in “Environmental and Analytical Chemistry” (2015), a MSc in “Environmental Science and Engineering” (2011) and a BSc in Chemistry (2009).

Authorship: Antonis Myridakis, Julien Chilloux, Ana Luisa Neves, Dheeraj Narendra, Isabel Garcia Perez and Marc-Emmanuel Dumas
Division of Computational and Systems Medicine, Department of Surgery and Cancer, Faculty of Medicine, Imperial College London, Exhibition Road, South Kensington, London SW7 2AZ, UK

Short Abstract

Metabolomics is one of the most fast growing fields in systems biology. Gas chromatography−mass spectrometry (GC-MS) approaches combine the coverage of the untargeted pipelines, with the library-facilitated, rapid metabolite identification. We developed a new holistic and high-throughput GC-MS profiling protocol samples with the use of metabolite libraries and validated for human urine and plasma samples. We optimized sample preparation, peak integration, quality control and data preprocessing steps. We compared several sample clean-up conditions and software packages for the data analysis. We present a detailed pipeline from sample aliquoting to the assigned metabolites table. The method was validated by analysing 356 human urine and plasma samples. 84 and 56 assigned metabolites passed the quality criteria, respectively for plasma and urine.

Long Abstract


The relationship between metabolism and disease risk is well-characterized [1-3] and metabolomics, the systematic study of the unique metabolic profile of a cell, tissue, organ or organism, play a key-role in the biomedical research [4]. The most widely-used techniques for metabolic profiling, are mass spectrometry (MS) [5,6] and nuclear magnetic resonance (NMR) spectroscopy [7]. Generally, NMR produces more reproducible and robust metabolomic data while requires minimal sample preparation. On the other hand, MS is more sensitive so it is able to detect metabolites at much lower levels. Furthermore, MS resolution is greatly improved by its coupling with either liquid (LC) or gas (GC) chromatography [8]. LC-MS approaches do not require derivatisation of the polar metabolites and give a large number of metabolic features, however, the structural assignment is extremely time-consuming or even practically impossible for some features. On the contrary, GC-MS approaches offer the great advantage of the availability of retention time/mass spectrum libraries, opening perspectives for a semi-targeted approach [9]. The use of these libraries enable the rapid and nearly automatic assignment of the metabolic features. Although several GC-MS metabolomic based methods and studies have been published [10], to the best of our knowledge, a complete, detailed, high-throughput and step-by-step methodology from sample preparation to data processing and analytical batch stitching for large-scale untargeted metabolomic analysis of human plasma and urine samples is not available.

Here we present a detailed analytical and data processing pipeline, from sample aliquoting to the generation of the final assigned metabolites/intensities table which includes the optimisation of key steps as, I) sample clean-up, II) peak integration, III) expansion of the commercially available metabolite libraries, IV) quality control and IV) batch stitching. Furthermore, we validated this method by analysing 192 urine and 166 plasma samples from two human cohorts.


Samples were spiked with isotopically labeled internal standards. Urine samples were treated with urease (Canavalia ensiformis) in order to deplete urea. Eight different clean-up protocols were compared: four protein/lipid precipitation and four liquid-liquid extraction ones. Metabolites were derivatised with MOX (methoxyamine) followed by MSTFA (N-methyl-trimethylsilyl-trifluoroacetamide) before their injection to an Agilent 7890B-5977C single quadrupole GC-MS system. Several commercial and in-house software packages were compared for the data extraction/processing while the metabolic features were annotated using the Fiehn library.


Protein/lipid precipitation with cold methanol was highlighted as the most effective clean-up technique and demonstrated the broadest metabolome coverage, offering in parallel a reasonable timeframe for its evaporation to dryness. Peak picking with Gavin [11] showed substantially more accurate peak picking than AMDIS/MPP software packages by I) zeroing not integrated peaks, II) backfilling non-detects by integrating noise and III) allowing the visual inspection of each metabolite peak and finally IV) by enabling the use of less sensitive ions for the saturated peaks. Additionally, we are expanding the Fiehn library (>800 metabolites with their corresponding spectra and retention times) with endogenous metabolites (>500 available standards). Furthermore, reproducibility, blank contamination, instrument drift, run order and batch effects were evaluated and depending on the dataset, were minimised with the use of in-house scripts [12]. Moreover, a strategy was designed the instrument maintenance and the day-to-day routine analysis of large numbers of samples. Finally, the optimised method was applied to 358 human urine and plasma samples. 84 and 56 assigned metabolites passed the quality criteria, respectively for plasma and urine.

Conclusions & Discussion

The key goals of this project are the development and validation of an analytical and data extraction/processing pipeline for high throughput untargeted profiling with GC-MS for large-scale molecular epidemiological research. A detailed protocol is presented from sample aliquoting to the generation of the assigned metabolite table. Furthermore, our method was successfully applied to two human cohorts and was validated for both human urine and plasma samples.

References & Acknowledgements:


1. Holmes, E., Li, J. V., Marchesi, J. R., & Nicholson, J. K. (2012). Gut microbiota composition and activity in relation to host metabolic phenotype and disease risk. Cell Metab, 16(5), 559-564. doi: 10.1016/j.cmet.2012.10.007

2. Le Chatelier, E., Nielsen, T., Qin, J., Prifti, E., Hildebrand, F., Falony, G., . . . Pedersen, O. (2013). Richness of human gut microbiome correlates with metabolic markers. Nature, 500(7464), 541-546. doi: 10.1038/nature12506

3. Pedersen, H. K., Gudmundsdottir, V., Nielsen, H. B., Hyotylainen, T., Nielsen, T., Jensen, B. A., . . . Pedersen, O. (2016). Human gut microbes impact host serum metabolome and insulin sensitivity. Nature, 535(7612), 376-381. doi: 10.1038/nature18646

4. Johnson, C. H., Ivanisevic, J., & Siuzdak, G. (2016). Metabolomics: beyond biomarkers and towards mechanisms. Nat Rev Mol Cell Biol, 17(7), 451-459. doi: 10.1038/nrm.2016.25

5. Zhou, B., Xiao, J. F., Tuli, L., Ressom, H. W. (2012) . LC-MS-based metabolomics. Mol Biosyst, 8 (2), 470-481. doi: 10.1039/c1mb05350g

6. Garcia, A. and Barbas, C. (2011) Gas chromatography-mass spectrometry (GC-MS)-based metabolomics. Methods Mol Biol. 708, 191-204. doi: 10.1007/978-1-61737-985-7_11.

7. Larive, C. K., Barding, G. A., Jr., & Dinges, M. M. (2015). NMR spectroscopy for metabolomics and metabolic profiling. Anal Chem, 87(1), 133-146. doi: 10.1021/ac504075g

8. Dumas, M. E., Kinross, J., & Nicholson, J. K. (2014). Metabolic phenotyping and systems biology approaches to understanding metabolic syndrome and fatty liver disease. Gastroenterology, 146(1), 46-62. doi: 10.1053/j.gastro.2013.11.001

9. Fiehn, O. (2016). Metabolomics by Gas Chromatography-Mass Spectrometry: Combined Targeted and Untargeted Profiling. Curr Protoc Mol Biol, 114, 30 34 31-30 34 32. doi: 10.1002/0471142727.mb3004s114

10. Mastrangelo, A., Ferrarini, A., Rey-Stolle, F., Garcia, A., & Barbas, C. (2015). From sample treatment to biomarker discovery: A tutorial for untargeted metabolomics based on GC-(EI)-Q-MS. Anal Chim Acta, 900, 21-35. doi: 10.1016/j.aca.2015.10.001

11. Behrends V, Tredwell GD & Bundy JG. (2011). A software complement to AMDIS for processing GC-MS metabolomic data. Anal Biochem, 415, 206-208. doi: 10.1016/j.ab.2011.04.009

12. Muhammad, A. K., Ebbels, T. M. D., Spagou, K., Masson, P., Want, E. J. (2012). Optimizing the Use of Quality Control Samples for Signal Drift Correction in Large-Scale Urine Metabolic Profiling Studies. Anal Chem, 84(6), 2670–2677. doi: 10.1021/ac202733q


The present study is initiated and funded by the European Community’s Seventh Framework Program “Metagenomics in Cardiometabolic Diseases, METACARDIS”, grant agreement HEALTH-F4-2012-305312.

Financial Disclosure

Board Memberno

IP Royalty: no

Planning to mention or discuss specific products or technology of the company(ies) listed above: