Quality Assurance and Quality Control in LC-MS-based Targeted Lipidomics of Human Blood Plasma Bo Burla (Presenter) National University of Singapore Presenter Bio: At SLING in Singapore for more than two years, I am working at the interface between the analytics in the lab and our data team to explore and implement new data processing workflows and pipelines for our lipidomics platform.

Authors: Bo Burla (1), Jeremy John Selva (2), Shanshan Ji (1), , Peter Benke (2), Anne K. Bendt (1), Federico Torta (2), Markus R. Wenk (1,2)
(1) Singapore Lipidomics Incubator (SLING), Life Sciences Institute and (2) Department of Biochemistry, YLL School of Medicine, National University of Singapore, Singapore

Ensuring quality and reproducibility of large-scale, large-panel LC-MS-based lipidomics analyses is critical but still challenging. Experimental design, sample processing, matrix effects, instrumental variations and data processing can all introduce variations and inter- and intra-assay batch effects, which can bias or cofound experimental readouts. We here explore frequently used and novel quality assurance (QA) and quality control (QC) strategies based on internal standards, different QC sample types, reference materials and different corresponding data processing methods and software tools. We demonstrate these strategies with analyses of large-scale studies and specifically designed tests.

Introduction

Large-scale lipidomics analyses of human plasma samples are increasingly used in clinical research to elucidate physiological and pathophysiological mechanisms, and to identify prognostic and diagnostic disease markers [1-3]. Critical but still challenging for such applications are the aspects of analytical quality and reproducibility, especially when analyzing hundreds or thousands of clinical samples. The experimental design, sample processing variabilities, sample matrix effects, instrumental fluctuations and data processing can all introduce variabilities in the measurements and cause inter- and intra-assay batch effects, which may bias or cofound experimental readouts and consequently affect conclusions.

Also in lipidomics, systematic QA/QC measures are key elements to ensure the analytical quality. A lot of work has been done on QA/QC in untargeted metabolomics [4]. Many of these strategies are also applicable for large-scale, large-panel lipidomics. However, QA/QC methods tailored for the characteristics of targeted lipidomic assays have not been explored in detail yet, but may improve the analytical quality and utility of such assays.

Methods

For the targeted lipidomics analyses discussed here we used a C18 reversed-phase UHPLC-MS/MS platform with an Agilent 6495 triple quadrupole mass spectrometer in MRM mode [5,6]. Plasma samples spiked with internal standards were extracted using a single-step, single-phase liquid-liquid extraction method [6]. Different quality control samples, e.g. blanks, batch/process (pooled) QCs, instrument performance QCs, blanks and response QCs were prepared. The NIST SRM 1950 reference plasma was used as reference material. Raw MRM data were integrated according to in-house SOPs using Agilent MassHunter. Data post-processing and visualizations were done using R and Python scripts, partially using reported [4] and in-house developed algorithms. The web apps were programmed in R/Shiny.

Results

We show that the established and broadly used approach of normalizing data with lipid-class specific internal standards (IS) may inflate analytical CV and introduce artefacts in some cases. Careful selection of IS and different IS-based normalizing approaches can however reduce analytical variations and partially compensate for systematic analytical drifts and batch-effects. However, we also show that drifts and batch effects can be lipid species-specific, in which case IS-normalization is not sufficient to remove the effects. Different drift and batch-correction methods, some based on QC samples, were explored. The frequency of QC samples required for dift/batch correction is often empirically defined. We consequently performed specific experiments to explore the QC frequencies required to for precise drift/batch correction and to estimate reliable analytical CV values. We also describe other aspect of QC workflows such as determining analytical saturation and sample matrix effects.

Conclusions & Discussion

We discuss the effects of established and commonly used QA/QC and data processing methods in our targeted large-scale lipidomics analyses and explore adapted and novel strategies. We also demonstrate the use of R scripts and in-house programmed web apps for efficient QA/QC and data exploration of our lipidomics analysis.

References:

[1] Yang K., and Han X. 2016. Lipidomics: Techniques, Applications, and Outcomes Related to Biomedical Sciences. Trends Biochem. Sci. 41: 954–969.

[2] Laaksonen R. et al. 2016. Plasma ceramides predict cardiovascular death in patients with stable coronary artery disease and acute coronary syndromes beyond LDL-cholesterol. Eur. Heart J. 37: 1967–1976.

[3] Wigger L. et al. 2017. Plasma Dihydroceramides Are Diabetes Susceptibility Biomarker Candidates in Mice and Humans. Cell Rep. 18: 2269–2279.

[4] Broadhurst D. et al. 2018. Guidelines and considerations for the use of system suitability and quality control samples in mass spectrometry assays applied in untargeted clinical metabolomic studies. Metabolomics 14: 72.

[5] Begum H., Torta F., Narayanaswamy P., Mundra P. A., Ji S., Bendt A. K., Saw W.-Y., Teo Y. Y., Soong R., Little P. F., Meikle P. J., and Wenk M. R. 2017. Lipidomic profiling of plasma in a healthy Singaporean population to identify ethnic specific differences in lipid levels and associations with disease risk factors. Clin. Mass Spectrom. 6: 25–31.

[6] Burla B., Muralidharan S., Wenk M. R., and Torta F. 2018. Sphingolipid Analysis in Clinical Research. Methods Mol. Biol. 1730: 135–162.

Acknowledgments:

MRW is supported by grants from the National University of Singapore via the Life Sciences Institute (LSI) and the National Research Foundation (NRFI2015-05).