MASLD is a staggering public health problem. According to the Global Burden of Diseases study, 1.27 billion people were diagnosed with the condition in 2021, and its progression to liver cancer represents one of the fastest-growing causes of mortality. No drug therapies are currently approved for MASLD, which makes identifying modifiable risk factors critically important. Environmental factors are thought to account for roughly 70 percent of the attributable risk, and among them, ambient PM2.5 has emerged as a major driver. In 2021, air pollution contributed to more than 90 percent of the 8.3 million deaths attributed to environmental pollutants. Yet while epidemiological studies have repeatedly reported associations between PM2.5 exposure and fatty liver disease, the biological mechanisms underlying that link have remained poorly characterized.

The research team, led by Shanshan Ran and colleagues, reasoned that circulating proteins could serve as a biologically informative bridge between environmental exposures and complex disease. Because the liver is both a major source and regulator of plasma proteins, environmentally induced perturbations in hepatic metabolism may be reflected in coordinated changes in the blood protein network before overt clinical disease appears. To capture those changes, the investigators drew on the UK Biobank Pharma Proteomics Project, which measured 2,923 unique plasma proteins using the Olink Explore 3072 proximity extension assay. After excluding participants with pre-existing liver disease or missing proteomic data, 52,190 adults with a mean age of 56.8 years remained for analysis.

Exposure assessment was unusually rigorous. The team geocoded every historical residential address for each participant and estimated annual mean PM2.5 concentrations using a 1-kilometer gridded dispersion model from the UK Department for Environment, Food and Rural Affairs, supplemented by the EMEP4UK model that resolves individual PM2.5 components, including elemental carbon, organic matter, sulfate, nitrate, and ammonium, at 3-kilometer resolution. Exposure was time-weighted according to how long participants lived at each address, yielding a cumulative average for the four years preceding baseline assessment. This approach captures real spatial and temporal variation in pollution rather than relying on crude regional averages.

To distill the proteomic signal, the researchers applied elastic net regression, a machine learning technique that combines the variable-selection power of Lasso with the stabilizing properties of Ridge regression. The analysis identified a PM2.5-related protein signature comprising 700 proteins. Key positive contributors included JAM2, SPOCK1, EGFR, CKAP4, and AXL, while PECAM1, CD99L2, LTBR, CEACAM1, and CANT1 carried negative weights. The signature proved longitudinally stable: correlations between the baseline protein profile and PM2.5 exposure measured one through four years earlier varied by a coefficient of only 4.12 percent, indicating that the choice of exposure window had minimal impact on the association.

Pathway enrichment analysis revealed what those 700 proteins were doing. The signature was significantly enriched in chemokine activity, cytokine-cytokine receptor interactions, the NF-κB signaling pathway, the IL-17 signaling pathway, and vascular endothelial growth factor signaling. Protein-protein interaction network analysis identified the inflammatory cytokines IL-6 and IL-1β as core hub nodes, connected to IL-4, EGFR, IL-10, CCL2, and ICAM1. Notably, a core set of 174 proteins shared across signatures built for all five PM2.5 components drove the same functional pathways, suggesting that the molecular response to particulate pollution is broadly consistent regardless of which constituent dominates the mixture.

The clinical payoff came when the team linked the signature to disease outcomes. During a median follow-up of 11.8 years, 624 participants developed incident MASLD. In fully adjusted Cox proportional hazards models, each standard deviation increase in PM2.5 exposure was associated with a 19 percent increase in MASLD risk, and the protein signature itself carried an identical 19 percent elevation. Individual proteins told a sharper story: 441 proteins within the signature were significantly associated with incident MASLD, with CDHR2 showing the strongest positive association and elevated levels of hub proteins including IL-6, IL-1β, CD28, CCL18, CSF3, EGFR, and HAVCR2 raising risk by 14 to 52 percent.

Because observational associations can be confounded by reverse causation, the researchers turned to two-sample Mendelian randomization, using genetic variants as instrumental variables to probe whether protein-MASLD links might be causal. Drawing protein genetics from the UK Biobank and MASLD genetics from the FinnGen biobank with 3,943 cases, the analysis supported associations for 117 PM2.5-related proteins. Hub proteins CD28, CSF3, IL18R1, and HAVCR2 maintained robust associations, with HAVCR2 showing a Mendelian randomization odds ratio of 2.28. The authors caution, however, that weak instrument bias and horizontal pleiotropy warrant careful interpretation of these genetic estimates.

Mediation analysis then quantified how much of the pollution-liver connection runs through the proteome. The protein signature as a whole statistically mediated 15 percent of the association between PM2.5 and MASLD, with 164 individual proteins acting as significant mediators. Proteins in the NF-κB/TNFα axis, including IL-18, CCL2, and TNFRSF11A, each mediated between roughly 5 and 7 percent of the risk, while the growth factor HGF mediated nearly 10 percent. Structural equation modeling extended the picture into metabolism: the protein signature was significantly associated with a previously identified PM2.5-related metabolic signature spanning lipid metabolism, glycolysis, and amino acid metabolism, and appeared to influence MASLD risk both directly and indirectly through those metabolic changes. The strongest protein-metabolite pairings, such as the coagulation zymogen F9 with triglyceride-enriched large HDL particles, provide molecular anchor points for future mechanistic work.

The authors are careful to note the limitations. Plasma proteins cannot be definitively assigned to liver tissue origin, the structural equation models reflect statistical dependencies rather than proven temporal causality, and the UK Biobank’s predominantly European ancestry and relatively low pollution levels limit generalizability to higher-exposure regions. Even so, the study delivers something the field has lacked: population-scale molecular evidence that air pollution is not merely correlated with fatty liver disease but is accompanied by a coordinated, measurable remodeling of inflammatory, immune, and endothelial proteins that tracks with metabolic disruption. If confirmed in more diverse cohorts, the 700-protein signature could eventually help identify which people living in polluted environments are most biologically vulnerable to liver disease, opening a window for prevention long before symptoms appear.

Subject of Research: Proteomic and Mendelian randomization analysis of the association between PM2.5 air pollution exposure and metabolic dysfunction-associated steatotic liver disease

Article Title: Association of protein signature of fine particulate matter with MASLD: the underlying pathways and Mendelian randomization analysis

Article References: Ran, S., Zheng, D., Yang, Z., Huang, Y., Wei, S., Hu, P., Zhang, J., Tian, F., & Lin, H. (2026). Association of protein signature of fine particulate matter with MASLD: the underlying pathways and Mendelian randomization analysis. Journal of Advanced Research. https://doi.org/10.1016/j.jare.2026.10.021

Image Credits: AI Generated

DOI: Not provided

Keywords: PM2.5, air pollution, MASLD, fatty liver disease, proteomics, UK Biobank, Mendelian randomization, NF-kB signaling, inflammation, metabolomics, mediation analysis, liver disease