Pursuing Experimental Reproducibility: An Efficient Protocol for the Preparation of Cerebrospinal Fluid Samples for NMR-based Metabolomics and Analysis of Sample Degradation

Abstract

NMR-based metabolomics investigations of human biofluids offer great potential to uncover new biomarkers. In contrast to protocols for sample collection and biobanking, procedures for sample preparation prior to NMR measurements are still heterogeneous, thus compromising the comparability of the resulting data. Herein, we present results of an investigation of the handling of cerebrospinal fluid (CSF) samples for NMR metabolomics research. Origins of commonly observed problems when conducting NMR experiments on this type of sample are addressed, and suitable experimental conditions in terms of sample preparation and pH control are discussed. Sample stability was assessed by monitoring the degradation of CSF samples by NMR, hereby identifying metabolite candidates, which are potentially affected by sample storage. A protocol was devised yielding consistent spectroscopic data as well as achieving overall sample stability for robust analysis. We present easy to adopt standard operating procedures with the aim to establish a shared sample handling strategy that facilitates and promotes inter-laboratory comparison, and the analysis of sample degradation provides new insights into sample stability.

Keywords: CSF; NMR metabolomics; SOP; cerebrospinal fluid; pH; sample degradation; sample preparation; standard operating procedures; time and temperature dependence.

Figure 1

Partial one-dimensional ¹H NMR spectra of the model solution (a,c) and an authentic CSF sample (b). The impact of a rise in pH can be clearly seen following the resonance positions of His, Phe, Tyr, Gln, Ala and Thr. TMX/XAN = xanthine/caffeine (tentative assignments, see Supplementary Materials for discussion), Ala^† = Alanine + unknown, * = lactate ¹³C satellite, Lac^‡ = lactate + Thr, unk = unknown. Note: For compound abbreviations see Table 2.

Figure 2

Proposed CSF sample preparation workflow detailing key aspects of sample handling and pH control (buffer composition described in step IV).

Figure 3

Stack plot of two spectral regions of the proton spectra of 10 CSF samples prepared as described in text showing consistency of metabolite resonances typically affected by differences in ionic strength and pH (600 MHz, 298 K, phosphate buffer 50 mM, pH 7.30–7.69, 10% D₂O, line broadening = 0.5 Hz). Note: For compound abbreviations, see Table 2.

Figure 4

Sample degradation analysis of the room temperature (‘RT’) and low temperature (‘LT’) samples. (a) Table of partial-least squares (PLS) model performance and quality parameters. (b) ¹H NMR spectra of the ‘RT’ sample acquired at different time intervals. (c) (left) Cross-validated predictions for sample ‘RT’, and (right) loading plot derived from the PLS model of sample ‘RT’. (d) (left) Cross validated predictions for sample ‘LT’, and (right) loading plot derived from PLS model of sample ‘LT’. Note: For compound abbreviations, see Table 2; here: * = VAL, LEU, aHBA and bHMB; # = regions where signals of ethanol/isopropanol were deleted.

Figure 5

Time dependence of the relative concentration profiles of selected metabolites in the CSF sample ‘RT’ (error bars represent a 5% quantification uncertainty).