Metabolic dysfunction-associated steatotic liver disease (MASLD, formerly NAFLD) increases in prevalence with age and is linked to worse outcomes including steatohepatitis, fibrosis, and hepatocellular carcinoma. Whether age-related molecular changes observed in MASLD reflect an acceleration of normal liver aging or represent a disease-specific divergence driven by chronic metabolic stress was unresolved. The authors aimed to test this by integrating public liver transcriptomic data across a broad age range and applying both linear and non-linear trajectory modeling to compare age-related transcriptional regulation in MASLD and control livers.
Public human liver transcriptome datasets were retrieved from GEO and GTEx. After multi-criteria filtering (human liver tissue, high-throughput transcriptomics) and manual metadata review to exclude drug interventions, unclear backgrounds, or very small studies, 13 datasets were selected, yielding 1,354 high-quality samples: 1,065 MASLD and 289 healthy controls. The sample age range was 9–92 years. Initial gene counts were reduced by removing mitochondrial, ribosomal, and low-expression genes, yielding 18,745 protein-coding genes for analysis.
To address inter-dataset batch effects, the authors applied ComBat_seq using dataset source as a covariate. Quality-control steps included low-expression filtering (edgeR filterByExpr) and alignment checks. Post-correction expression distributions aligned at the median and principal component analysis (PCA) showed markedly reduced dataset-specific clustering, supporting suitability for downstream analyses.
DEA between all MASLD and control samples (DESeq2; padj < 0.05 and |log2FC| > 1) identified 1,258 differentially expressed genes (DEGs): 956 upregulated and 302 downregulated in MASLD. Samples were then stratified into young (≤39 years), middle-aged (40–59 years), and elderly (≥60 years). DEG counts rose with age: 602 DEGs in the young group (318 up/284 down), 882 in middle-aged (562/320), and 1,693 in the elderly (1,249/444). An intersection analysis found 247 DEGs common to the global analysis and all three age strata, representing a stable MASLD core signature, while additional age-specific DEGs emerged—especially in the elderly group.
Focusing on MASLD samples, Spearman correlation of gene expression versus chronological age (VST-transformed values) and multiple-testing correction (BH) identified 180 age-associated genes using thresholds padj < 0.05 and |rho| > 0.2. Of these, 121 were positively correlated with age and 59 negatively correlated. By combining correlation direction with the global DE fold changes, the genes were classified into synergistic categories (e.g., upregulated in MASLD and increasing with age) and antagonistic categories (e.g., upregulated in MASLD but decreasing with age), indicating some genes have opposite age trends in MASLD compared with controls.
Because age-related processes are frequently non-linear, the study applied generalized additive models (GAM; mgcv package) with thin plate regression splines (basis dimension = 4) to VST-transformed gene expression as a smooth function of age. The effective degrees of freedom (edf) quantified non-linearity. Predicted curves and first derivatives (estimated across an age grid with simultaneous 95% confidence intervals) enabled classification of genes into four major trajectory patterns in MASLD: stable expression, early-life change, mid-life fluctuation, and late-life acceleration. Estimated trajectory inflection clusters centered near ages ~35 and ~75, highlighting potential critical transition intervals in MASLD aging.
Functional annotation (clusterProfiler, GO-BP enrichment) revealed contrasts between control and MASLD age-associated genes. Control-associated age genes mainly mapped to classical cell cycle processes. By contrast, MASLD age-associated changes were enriched for pathways related to protein deubiquitination, impaired proteostasis, and p53-associated stress and apoptotic signaling, implying activation of cellular stress-response programs and disturbances in protein quality control in MASLD with aging.
The authors examined clinical associations for age-associated MASLD genes using datasets reporting fibrosis stage and NAFLD activity score (NAS). Many age-associated genes were also associated with fibrosis or NAS in largely concordant directions after regression adjustment for age and dataset source (FDR < 0.05), suggesting that the age-related transcriptional remodeling in MASLD has links to histological disease severity.
Across a cross-age, multi-cohort liver transcriptome integration, MASLD displayed age-related transcriptional remodeling that differs from canonical physiological liver aging. Rather than simply accelerating normal aging programs, MASLD age-related changes emphasize dysregulation of proteostasis and deubiquitination pathways, activation of cellular stress and p53-mediated apoptotic signaling, and age-dependent heterogeneity in DEG burden. The identification of trajectory inflection points near ~35 and ~75 years and the association of age-linked genes with fibrosis and NAS underscore potential windows for age-stratified risk assessment and mechanistic targeting. Details on study limitations beyond cohort selection and analytical choices were not reported in the source.