Serum Metabolomics for Biomarker Screening of Esophageal Squamous Cell Carcinoma and Esophageal Squamous Dysplasia Using Gas Chromatography-Mass Spectrometry

Abstract

Background: Esophageal squamous cell carcinoma (ESCC) is one of the most common malignancies with poor diagnosis. Esophageal squamous dysplasia (ESD) is considered as an immediate precancerous lesion of ESCC. Lack of biomarkers for discriminating ESCC and ESD from healthy subjects limits the early diagnosis and treatment of ESCC. Therefore, a serum metabolomic strategy was conducted to identify and validate potential metabolic markers for the screening of ESCC and ESD subjects.

Methods: A total of 74 patients with ESCC, 72 patients with ESD, and 75 normal control (NC) subjects were enrolled in this study. Gas chromatography-mass spectrometry was used to acquire serum metabolic profiles. Pathway analysis was conducted to uncover the fluctuated metabolic pathways during ESCC. Multivariate analyses were used to screen and validate the biomarkers.

Results: ESCC, ESD, and NC subjects revealed progressively altered metabolic profiles, in which amino acids globally increased, while fatty acids decreased in ESCCs compared with the control groups. Pathway analysis demonstrated the activated biosynthesis of amino acids and inhibited desaturation of saturated fatty acids. The panel constructed with propanoic acid, linoleic acid, glycerol-3-phosphate, and l-glutamine showed the area under the curve (AUC), sensitivity, and specificity of 0.817, 0.75, and 0.74, respectively, in the discrimination of ESCC/ESD patients from NC subjects. The panel constructed by propanoic acid, l-leucine, and hydroxyproline revealed the AUC, sensitivity, and specificity of 0.819, 0.76, and 0.72, respectively, in the discrimination of ESD from NC subjects. The combination of hypoxanthine, 2-ketoisocaproic acid, l-glutamate, and l-aspartate showed the AUC, sensitivity, and specificity of 0.818, 0.83, and 0.74, respectively, in the discrimination of ESCC patients from ESD subjects.

Conclusions: Our study revealed the systematic landscape for metabolic alterations in sera of ESD and ESCC patients. The defined metabolite markers showed reasonable performance in the discrimination of ESCC and ESD patients, and may provide helpful reference for clinicians and biologists.

Figure 1

Workflow for data analysis in this study.

Figure 2

Metabolic profiling alterations among ESCC, ESD, and NC subjects in the training set. (A) PLS-DA score plot presenting classification of ESCC, ESD, and NC subjects, red dots: ESCC subjects, yellow dots: ESD subjects, and blue dots: normal control subjects; (B) Cross-validation plot of the PLS-DA model after permutation for 100 times; (C) Loading plot of the PLS-DA model presenting scattering patterns of different categories of metabolites; (D) Heat-map diagram presenting relative amounts of various categories of metabolites in ESCC, ESD, and NC groups.

Figure 3

Pathway-network diagram constructed with identified metabolites. Pathway ①: arginine, proline, glutamate, aspartate, and asparagine metabolism; pathway ②: valine, leucine, and isoleucine degradation; pathway ③: galactose metabolism; pathway ④: TCA cycle, glycolysis, and gluconeogenesis; pathway ⑤: biosynthesis of saturated and unsaturated fatty acids; pathway ⑥: glycine, serine, alanine, and threonine metabolism; and pathway ⑦: purine metabolism.

Figure 4

Comparison of relative amounts of metabolites mapped in pathways ①, ②, ③, ⑤, and ⑥ in NC, ESD, and ESCC groups. The circle at the bottom-left presenting the alterations of ESD to NC subjects, circle at the top-middle presenting the alterations of ESCC to NC subjects, and circle at the bottom-right presenting the alterations of ESCC to ESD subjects. Circles in red and green stand for increases and decreases in ESCCs or ESDs compared with ESDs or NCs, respectively. Circles that are colorless stand for the metabolites nonsignificantly changed between corresponding groups.

Figure 5

Diagnostic performance of P1, P2, and P3 in the validation set. (A) AUC, sensitivity, and specificity of P1; (B) Relative values of P1 in ESCC/ESD and NC groups; (C) AUC, sensitivity, and specificity of P2; (D) Relative values of P2 in ESD and NC groups; (E) AUC, sensitivity, and specificity of P3; (F) Relative values of P3 in ESCC and ESD groups. Red dashed lines indicating the cutoff values at the corresponding sensitivity and specificity.