2.7. Untargeted Metabolomics Analysis

A total of 100 μL of each plasma sample was mixed with 700 μL of extractant containing internal standard (methanol: acetonitrile: water in the ratio of 4:2:1 v/v/v). After shaking for 1 min, the mixtures were placed in a −20 °C refrigerator for 2 h. After centrifugation at 25,000× g and 4 °C for 15 min, the supernatants were separated, and 600 μL of each supernatant was transferred into a new microcentrifuge tube. The transferred samples were then dried using a drying machine. The samples were reconstituted by adding 180 μL of methanol: pure water (1:1 v/v) and mixed for 10 min via vortexing. The samples were centrifuged at 25,000× g and 4 °C for 15 min, and the supernatants were then transferred to a new microcentrifuge tube. Lastly, 20 μL of each sample was mixed with the QC samples before proceeding to UPLC-MS analysis.

This experiment used Waters 2777c UPLC (Waters, Mildford, MA, USA) in series with the Q Exactive HF high-resolution mass spectrometer (Thermo Fisher Scientific, Waltham, MA, USA) to separate and detect the metabolites. Briefly, chromatographic separation was performed on a Waters ACQUITY UPLC BEH C18 column (1.7 μm, 2.1 mm × 100 mm; Waters, Mildford, MA, USA), and the column temperature was maintained at 45 °C. The mobile phase composition for the positive mode consisted of 0.1% formic acid (A) and acetonitrile (B); whereas, for the negative mode, it comprised 10 mM ammonium formate (A) and acetonitrile (B). The gradient conditions were as follows: 0–1 min, 2% B; 1–9 min, 2–98% B; 9–12 min, 98% B; 12–12.1 min, 98% B to 2% B; and 12.1–15 min, 2% B, respectively. The flow rate was 0.35 mL/min, and the injection volume was 5 μL. Then, primary and secondary mass spectrometry data acquisition was performed using Q Exactive HF (Thermo Fisher Scientific, USA). The full scan range was 70–1050 m/z with a resolution of 120,000, and the automatic gain control (AGC) target for MS acquisitions was set to 3 × 10⁶ with a maximum ion injection time of 100 ms. The top 3 precursors were selected for subsequent MS/MS fragmentation with a maximum ion injection time of 50 ms and resolution of 30,000, and the AGC was set to 1 × 10⁵, respectively. The stepped normalized collision energy was set to 20, 40, and 60 eV, respectively. The ESI parameters setting was as follows: the sheath gas flow rate was 40, the aux gas flow rate was 10, the positive-ion mode spray voltage(|KV|) was 3.80, the negative-ion mode spray voltage(|KV|) was 3.20, the capillary temperature was 320 °C, and the aux gas heater temperature was 350 °C, respectively.

After importing the off-line data of mass spectrometry into Compound Discoverer 3.3 (Thermo Fisher Scientific, San Jose, CA, USA) software and analyzing the mass spectrometry data in combination with the BGI metabolome database, mzCloud database, and ChemSpider online database, a data matrix containing information such as metabolite peak area and identification results was obtained.

The result files from the Compound Discoverer were transferred to R software package metaX (BGI Shenzhen, Guangdong, China) for data pre-processing and further analysis. During pre-processing, the data were normalized to obtain the relative peak areas by probabilistic quotient normalization (PQN). Then, the batch effects were corrected using quality control-based robust LOESS signal correction. Metabolites with a coefficient of variation larger than 30% on their relative peak area in QC samples were then removed from further analysis. Subsequently, multivariate statistical and univariate analyzes were used to screen different metabolites between the groups, with slight modifications from the previously described methods [20]. The pre-processed data was log-transformed and auto-scaled in the Pareto scale. Then, principal component analysis (PCA) was performed to reflect the actual distribution of samples and to observe the separation trend between the sample groups. Partial least squares-discriminant analysis (PLS-DA), a supervised statistical method used to enhance the differentiation between the classification groups, was subsequently conducted. The PLS-DA model was established between the comparative analysis groups (two groups of samples), and a 5-fold cross-validation was used to validate when building the model. Then, the orthogonal partial least squares discriminant analysis (OPLS-DA) was performed on two groups of biological samples. The purpose was to establish the relationship model between the metabolite expression and sample categories, thereby allowing for the modelling and prediction of these sample categories. At the same time, the ability of each metabolite to classify and distinguish each group of samples was measured by calculating the variable important for the projection (VIP). For the screening of metabolic biomarkers, it is generally considered that a VIP greater than 1 indicates that the variable significantly affects the differentiation of the sample categories.

For univariate analysis of the data, the differences in metabolite concentration were evaluated in terms of fold change, and statistical comparison was conducted using the independent t-test. The displayed metabolites with a p-value < 0.05 indicated significant differences in the fold-change values. Only the statistically significant metabolites were considered as differential metabolites. Metabolites with a VIP value ≥ 1, fold change ≥ 1.2 or ≤0.8, or p-value < 0.05 were considered as differential metabolites.

After the differential metabolites were screened, the expression patterns of the differential metabolites were analyzed by clustering analysis, correlation clustering, and network analysis. Additionally, biological functions were explored through pathway annotation and pathway enrichment analysis. Using the Euclidean distance method, Hierarchical cluster analysis (HCA) was applied to examine the expression levels of the differential metabolites in the two different groups. The identified metabolites’ taxonomic and functional annotations were performed by comparing the KEGG and HMDB databases. Afterward, enrichment and pathway analyzes were conducted using the MetaboAnalyst 5.0 software.