Untargeted Metabolomic Profiling Reveals Differentially Expressed Serum Metabolites and Pathways in Type 2 Diabetes Patients with and without Cognitive Decline: A Cross-Sectional Study

Abstract

Diabetes is recognized as a risk factor for cognitive decline, but the underlying mechanisms remain elusive. We aimed to identify the metabolic pathways altered in diabetes-associated cognitive decline (DACD) using untargeted metabolomics. We conducted liquid chromatography-mass spectrometry-based untargeted metabolomics to profile serum metabolite levels in 100 patients with type 2 diabetes (T2D) (54 without and 46 with DACD). Multivariate statistical tools were used to identify the differentially expressed metabolites (DEMs), and enrichment and pathways analyses were used to identify the signaling pathways associated with the DEMs. The receiver operating characteristic (ROC) analysis was employed to assess the diagnostic accuracy of a set of metabolites. We identified twenty DEMs, seven up- and thirteen downregulated in the DACD vs. DM group. Chemometric analysis revealed distinct clustering between the two groups. Metabolite set enrichment analysis found significant enrichment in various metabolite sets, including galactose metabolism, arginine and unsaturated fatty acid biosynthesis, citrate cycle, fructose and mannose, alanine, aspartate, and glutamate metabolism. Pathway analysis identified six significantly altered pathways, including arginine and unsaturated fatty acid biosynthesis, and the metabolism of the citrate cycle, alanine, aspartate, glutamate, a-linolenic acid, and glycerophospholipids. Classifier models with AUC-ROC > 90% were developed using individual metabolites or a combination of individual metabolites and metabolite ratios. Our study provides evidence of perturbations in multiple metabolic pathways in patients with DACD. The distinct DEMs identified in this study hold promise as diagnostic biomarkers for DACD patients.

Keywords: dementia; diabetes-associated cognitive decline; metabolomics; mild cognitive impairment; type 2 diabetes.

Figure 1

Differential expression of the plasma metabolome between the DACD and DM groups. (A) Volcano plot displaying the log2-fold change (x-axis) against the −log10 statistical p-value (y-axis) for all metabolites. Downregulated and upregulated metabolites in DACD (p < 0.05 and FC > 2) are shown in blue and red, respectively. Nonsignificant metabolites between the two groups are in black. (B) List of the differentially expressed metabolites with their FC and FDR values. (C) Supervised cluster analysis across the DM and the DACD groups using the top 30 differentially expressed metabolites. The blue color represents downregulated metabolites, while the brown color represents upregulated metabolites in the DACD group. * Denotes a compound whose identity hasn’t been officially verified using a standard, yet we possess a high level of confidence in its identification.

Figure 2

Chemometric analysis of metabolomic data sets of the DM and DACD samples. (A) Score plot of the orthogonal partial least squares discriminant analysis (OPLSDA) model. The pink and green dots represent the DM and DACD samples, respectively. (B) Results of the 1000-time permutation test of the OPLSDA model; the empirical p-values for R2Y and Q2 were all <0.001. (C) Top 30 metabolites important for the separation between the DM and DACD samples based on the VIP score. Red and blue colors on the scale on the right indicate upregulation and downregulation in the respective group.

Figure 3

The results of pathway analysis and enrichment analysis of the metabolomic data. (A) Enrichment analysis based on the Small Molecule Pathway Database (SMPDB). (B) Pathway analysis based on the KEGG database.

Figure 4

The ROC curves generated in the Biomarker Module of Metaboanalyst. (A–D) The top four biomarker candidate metabolites and metabolite ratios identified based on ROC curve analysis performed with 50 serum metabolic features, and their ratios estimated for their relative concentrations in the DM and DACD study groups. The computed AUC-ROC, 95% confidence interval (CI), and marker metabolites are shown for each model. The AUC-ROC is shown in red to highlight the diagnostic potential of the model. (E) The whisker plots shown on the right revealed significantly decreased serum levels of these metabolites and their ratios in the DACD (green) group compared to the DM (red) group. The boxes denote interquartile ranges, the average value are represented by the horizontal black line, and yellow diamond representing median (or 50th percentile value) within each box. The bottom and top boundaries of the boxes are the 25th and 75th percentiles, respectively. The lower and upper whiskers are the 5th and 95th percentiles, respectively. Outlier points beyond this range are indicated above or below the whiskers. Each box plot shows quantitative variations in metabolic concentrations. * Denotes a compound whose identity hasn’t been officially verified using a standard, yet we possess a high level of confidence in its identification. DACD: diabetes-associated cognitive dysfunction.