Plasma Metabolomic and Lipidomic Profiling of Metabolic Dysfunction-Associated Fatty Liver Disease in Humans Using an Untargeted Multiplatform Approach

Abstract

Metabolic dysfunction-associated fatty liver disease (MAFLD) is a complex disorder that is implicated in dysregulations in multiple biological pathways, orchestrated by interactions between genetic predisposition, metabolic syndromes and environmental factors. The limited knowledge of its pathogenesis is one of the bottlenecks in the development of prognostic and therapeutic options for MAFLD. Moreover, the extent to which metabolic pathways are altered due to ongoing hepatic steatosis, inflammation and fibrosis and subsequent liver damage remains unclear. To uncover potential MAFLD pathogenesis in humans, we employed an untargeted nuclear magnetic resonance (NMR) spectroscopy- and high-resolution mass spectrometry (HRMS)-based multiplatform approach combined with a computational multiblock omics framework to characterize the plasma metabolomes and lipidomes of obese patients without (n = 19) or with liver biopsy confirmed MAFLD (n = 63). Metabolite features associated with MAFLD were identified using a metabolome-wide association study pipeline that tested for the relationships between feature responses and MAFLD. A metabolic pathway enrichment analysis revealed 16 pathways associated with MAFLD and highlighted pathway changes, including amino acid metabolism, bile acid metabolism, carnitine shuttle, fatty acid metabolism, glycerophospholipid metabolism, arachidonic acid metabolism and steroid metabolism. These results suggested that there were alterations in energy metabolism, specifically amino acid and lipid metabolism, and pointed to the pathways being implicated in alerted liver function, mitochondrial dysfunctions and immune system disorders, which have previously been linked to MAFLD in human and animal studies. Together, this study revealed specific metabolic alterations associated with MAFLD and supported the idea that MAFLD is fundamentally a metabolism-related disorder, thereby providing new perspectives for diagnostic and therapeutic strategies.

Keywords: NMR; lipidomics; mass spectrometry; metabolic dysfunction-associated fatty liver disease; metabolomics; multiblock analysis.

Figure 1

The associations between clinical characteristics and MAFLD: (A) a simplified diagram of the study design; (B) the correlation between plasma and liver TGs; (C–E) the results of the liver histological analysis using hematoxylin–eosin staining on the different groups of patients ((C) a normal liver (HC) from a 30-year-old woman, magnification 100× (scale bars: 100 μm); (D) steatosis (ST) from a 30-year-old woman showing “isolated” macrovacuolar steatosis affecting 60–90% of hepatocytes without inflammation, ballooning and fibrosis, magnification 100× (scale bars: 100 μm); (E) non-alcoholic steatohepatitis (NASH) from a 40-year-old woman showing macrovacuolar steatosis affecting 60–90% of hepatocytes with inflammation (black arrows) and ballooning (blue arrows), magnification 100× (scale bars: 100 μm), inset 400× (scale bars: 25 μm)); (F–J) the clinical and biochemical characteristics of the patients ((K) the PCA results for all participants based on seven clinical characteristics; (L) a correlation matrix highlighting the most significant contributing variables for each principal component. The Benjamini and Hochberg adjusted p-values were computed using Dunn’s multiple comparison test [42], which is a post-hoc Kruskal–Wallis test. Note: ns, non-significant; * p < 0.05; ** p < 0.01; *** p < 0.001; **** p < 0.0001; AST, aspartate transaminase; ALT, alanine transaminase; GGT, gamma-glutamyl transferase; HOMA-IR, homeostatic model assessment for insulin resistance; TGs, triglycerides; HC, health control (normal liver); ST, non-alcoholic fatty liver or steatosis; NASH, non-alcoholic steatohepatitis; a.u., arbitrary unit; PC, principal component.

Figure 2

The NMR-based plasma metabolomics of MAFLD: (A) the OPLS-DA score plot (HC versus NASH; 1D NOESY data); (B) the 200 times permutation test for the OPLS-DA model (HC versus NASH; 1D NOESY data); (C) the OPLS-DA covariance plot (HC versus NASH; 1D NOESY data). The red hydrogen atom indicates the ¹H detected by ¹H NMR.

Figure 3

Our MAFLD metabolome-wide association study: (A) MWAS highlighted 241 metabolomic features associated with the disease at FDR < 25%. (B) 787 lipidomic features were associated with the disease at FDR < 20%. The figure represents the retention time (s) of LC-MS features detected in positive (above the blue line) and negative (below the blue line) ionization mode plotted against −log₁₀(p-value), respectively. Each dot represents a unique LC-MS feature: red dots are features that positively associated with MAFLD, while blue dots are features that negatively associated with MAFLD. The “dashed” red horizontal line indicates the raw p-Value of 0.05 and the “dot–dash” horizontal line shows the false discovery rate-adjusted q-Value of 0.2 for lipidomics and 0.25 for metabolomics; (C,D) the PCA score plots for the metabolomic and lipidomic data, respectively; (E–G) the three main lipids that matched the top 10 ranked features based on the regression p-value and correlation with liver TG content.

Figure 4

The classification of HC, ST and NASH patients based on MS data: (A) the ROC curve and AUC from the sPLS-DA for the metabolomic data of the first component; (B) the ROC curve and AUC from the sPLS-DA for the lipidomic data of the first component; (C) the confusion matrix of the test set sample (n = 24) from the sPLS-DA for the metabolomic data; (D) the confusion matrix of the test set sample (n = 25) from the sPLS-DA for the lipidomic data. Note: balanced accuracy = (sensitivity + specificity)/2.

Figure 5

Our multiblock integrative analysis for clinical, metabolomic and lipidomic data: (A) the circos plot showing the integrative frameworks and the positive (red lines) and negative (blue lines) correlations (cutoff r = 0.7) between selected variables from each block, represented on the side quadrants (the selected feature names were coded as “metxxx” for metabolomic data and “lipxxx” for lipidomic data); (B) the clustered image map (CIM) for the variables selected by our multiblock sPLS-DA of the first component in which the Euclidean distance and complete linkage methods were used (the CIM shows the samples in the rows (as indicated by their stages of MAFLD on the left-hand side of the plot) and the selected features in the columns (as indicated by their data type at the top of the plot); (C–E) the three main lipids that matched the selected features from the multiblock analysis and correlation with liver TG content. Note: a.u., arbitrary unit.

Figure 6

The potential cross-talk between different factors and biological pathways involved in the development of MAFLD. MAFLD is a complex disease that involves dysregulations in multiple biological pathways, specifically amino acid, arachidonic acid, bile acid and lipid metabolism, which is modulated by interactions between genetic predisposition, metabolic syndromes and environmental factors. Note: acetyl-CoA, acetyl coenzyme A; G6P, glucose 6-phosphate; PFAS, perfluorinated alkyl substances; COXs, cyclooxygenases; LOXs, lipoxygenases; PL, phospholipase; PNPLA3, patatin-like phospholipase domain-containing protein 3 gene; TM6SF2, transmembrane 6 superfamily member 2 gene; GCKR, glucokinase regulatory protein gene. Partially inspired by Masoodi et al. [18]. The mitochondrion and lipid droplets illustrations were adapted (modified with layer and text) from Servier Medical Art. Servier Medical Art by Servier is licensed under a Creative Commons Attribution 3.0 Unported License (https://creativecommons.org/licenses/by/3.0/, accessed on 1 October 2022).