2.2. Quantitative NMR Spectroscopy

Raw NMR spectra were acquired using Bruker body fluid B.I. methods [53]. Sample preparation was performed following the included standards of the procedure to ensure reliable results. For quality control, the B.I. BioBankQC^TM module was applied. For quantification, the modules B.I. QUANT-PS^TM for metabolites and B.I. LISA^TM for lipoproteins, respectively, were applied (All B.I. modules: Bruker BioSpin GmbH, Ettlingen, Germany). Blood serum samples were thawed for approximately 30 min at room temperature before 400 µL of each aliquot was pipetted into a 1.5 mL PTFE container and mixed with 400 µL of a commercially prepared pH 7.4 sodium phosphate plasma buffer from Bruker. The mixture was then shaken gently for 1 min before extracting 600 µL of it to fill a Bruker 5 mm NMR tube. The SampleJet cooling setting was set to 279 Kelvin. Monodimensional ¹H-NMR spectra were acquired using a 5 mm triple resonance (TXI; ¹H, ¹³C and ¹⁵N) room temperature probe on a Bruker IVDr Avance III HD 600 MHz system, which was operated using Bruker’s standard NMR software TopSpin (Version 3.6.2, Bruker BioSpin GmbH, Ettlingen, Germany). Five monodimensional ¹H NMR spectra types were collected for each blood sample with water peak suppression and varied pulse sequences to selectively observe molecular components. First, a NOESY (Nuclear Overhauser Effect SpectroscopY) 32-scan NMR experiment was used to show the NMR spectrum quality (via the B.I. BioBankQC™) and to enable the quantification of metabolites (i.e., glucose, lactic acid and amino acids of the B.I. BioBankQuant-PS™) and high-molecular-weight compounds, such as lipoproteins (as shown in B.I. LISA™). Then, a 32-scan CPMG (Carr–Purcell–Meiboom–Gill, filtering out macromolecular resonance signals) program was run, as well as 32-scan DIFF (DIFFusion measurements of, primarily, macromolecular signal massifs [54]) and 64-scan PGPE (Pulsed Gradient Perfect Echo, used primarily for Glyc and SPC quantification [55]). Moreover, a two-dimensional NMR experiment, 2-scan JRES (J-RESolved spectroscopy), was included with the IVDr methods and was performed to analyze J coupling constants. Additionally, JRES can be useful for manual data look-up and represents a vital piece of the novel Bruker PhenoRisk PACS™ software (Bruker BioSpin GmbH, Ettlingen, Germany). The NMR experiments utilized a group of sample-dependent parameters, like the frequency offset O1 and 90° pulse P1 duration.

All recorded spectra were quantified in full automation. The parameters, GlycA, GlycB and SPC, were subsequently determined with the B.I. PACS^TM module. Previously published work has enabled the standardization of the approach used in this project and supports its reliability [26,56]. Exemplary annotations can be found in Supplementary Material (Figure S1).

To obtain a meaningful and high-quality data set, we performed quality control prior to analysis, resulting in the exclusion of 25 samples due to a linewidth of >2.3 Hz. The vast majority of measurements yielded a high ρ, which quantifies the correlation of the calculated fit with the metabolite signal. An overview of this can be found in Supplementary Material (Figure S2). Accordingly, regarding the B.I. QUANT-PS^TM module, no cut-off for ρ was defined.

To enable the interpretation of the acquired metabolomics profiles in the serum of COVID-19 patients using B.I. QUANT-PS^TM and B.I. LISA^TM, Bruker BioSpin GmbH kindly provided corresponding serum data of 305 healthy sex- and age-matched individuals (without further information according to donor conditions). Accordingly, in Section 3.1 and Section 3.2, where comparisons are made with healthy controls, we work only with the parameters from these modules.