Serum Metabolic Profiling Analysis of Chronic Gastritis and Gastric Cancer by Untargeted Metabolomics

Abstract

Purpose: Gastric cancer is a common tumor of the digestive system. Identification of potential molecules associated with gastric cancer progression and validation of potential biomarkers for gastric cancer diagnosis are very important. Thus, the aim of our study was to determine the serum metabolic characteristics of the serum of patients with chronic gastritis (CG) or gastric cancer (GC) and validate candidate biomarkers for disease diagnosis.

Experimental design: A total of 123 human serum samples from patients with CG or GC were collected for untargeted metabolomic analysis via UHPLC-Q-TOF/MS to determine characteristics of the serum. Principal component analysis (PCA), partial least squares discriminant analysis (PLS-DA), and heat map were used for multivariate analysis. In addition, commercial databases were used to identify the pathways of metabolites. Differential metabolites were identified based on a heat map with a t-test threshold (p < 0.05), fold-change threshold (FC > 1.5 or FC < 2/3) and variable importance in the projection (VIP >1). Then, differential metabolites were analyzed by receiver operating characteristic (ROC) curve to determine candidate biomarkers. All samples were analyzed for fasting lipid profiles.

Results: Analysis of serum metabolomic profiles indicated that most of the altered metabolic pathways in the three groups were associated with lipid metabolism (p < 0.05) and lipids and lipid-like molecules were the predominating metabolites within the top 100 differential metabolites (p < 0.05, FC > 1.5 or FC < 2/3, and VIP >1). Moreover, differential metabolites, including hexadecasphinganine, linoleamide, and N-Hydroxy arachidonoyl amine had high diagnostic performance according to PLS-DA. In addition, fasting lipid profile analysis showed the serum levels of total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C) and apolipoprotein A1 (Apo-A1) were decreased concomitant to the progression of the progression of the disease compared with those in the control group (p < 0.05).

Conclusions: Thus, this study demonstrated that lipid metabolism may influence the development of CG to GC. Hexadecasphinganine, linoleamide, and N-Hydroxy arachidonoyl amine were selected as candidate diagnostic markers for CG and GC.

Keywords: candidate biomarkers; chronic gastritis; gastric cancer; lipid metabolism; untargeted metabolomics.

Figure 1

Score plots of PCA model and the first three latent components of PLS- DA model. (A) Three-dimensional score plot of the PCA model of all samples in the ESI+ and ESI− ion modes. (B, C) Score plots for the first three latent components of the PLS-DA model for CG vs Control, CG vs GC, GC vs Control, and CG plus GC vs Control in the ESI+ and ESI− ion modes, respectively. Healthy volunteers, control, CG, chronic gastritis; GC, gastric cancer.

Figure 2

Permutation tests of the PLS-DA model for CG vs Control, CG vs GC, and GC vs Control in the (A) ESI+ and (B) ESI− ion modes. Random permutations (a total of 200) were used to evaluate whether a possibility of overfitting in the PLS-DA model. The statistical parameters R²X₀, Q²X₀, R²Y, and Q²Y were used for the analysis of the multivariate models.

Figure 3

Heat map of the 100 significantly differential metabolites in the serum in the control (purple), CG (red), and GC (green) groups. Metabolites were included based on VIP > 1, FC > 1.5 or FC < 2/3, and p < 0.05. The colors from blue to red indicate the relative contents of the metabolites in the three groups.

Figure 4

ROC curve analysis of the candidate biomarkers for CG or GC. Individual ROC curves and peak areas for hexadecasphinganine (A, D, G), linoleamide (B, E, H) and N-Hydroxy arachidonoyl amine (C, F, I). AUC (0.5−0.7), low accuracy; AUC (0.7−0.9), moderate accuracy; AUC (> 0.9), high accuracy. From the panel, hexadecasphinganine, linoleamide, and N-Hydroxy arachidonoyl amine displayed high efficiency for distinguishing patients with CG or GC from healthy control, but moderate efficiency when used to distinguish CG group from GC group.

Figure 5

Validation of the combination of the three lipid metabolites (AUC > 0.90, ^***p < 0.005) as potential biomarkers by ROC curve analysis. (A) The comparison of normalized intensity peak areas of three candidate biomarkers (hexadecasphinganine, linoleamide, and N-Hydroxy arachidonoyl amine) in the control (orange), CG (green), and GC (blue) groups. (B) Potential diagnostic performance of the three identified metabolites (AUC > 0.90, ^***p < 0.005). The sensitivity and specificity values were optimized by PLS-DA. As shown in the panels, the area under the ROC curve of the combination of the three candidate biomarkers is significantly increased, suggesting that the combination of the three parameters has the highest diagnostic accuracy, especially for distinguishing GC patients from CG patients.