Transcriptomic and epigenetic analyses of environmental exposure in mouse models

Abstract

Environmental exposures are known to contribute to the development and progression of human diseases. For example, air pollution is a modifiable risk factor for dementia, but the underlying biological pathways linking inhaled pollutants to amyloid-beta accumulation and neuroinflammation are not well defined. The lung-brain axis has been proposed as a route whereby toxins, inflammatory mediators, and immune cells may enter the circulation and access the brain through a compromised blood-brain barrier, but empirical evidence remains limited. The transcription factor Nrf2 regulates antioxidant defence pathways that counteract environmentally induced oxidative stress, though its broader regulatory functions require further characterisation. In addition, the genomic occurrence of DNA methylation, which is a repressive epigenetic mark occurring at cytosine-phosphate-guanine (CpG) sites, can be altered by air pollution exposures, however, the impacts on cell function and health remain unclear. This PhD study aims to define molecular pathways altered by environmental pollutants in mouse models and to evaluate their translational relevance to human disease. Heme oxygenase (HO)-based oxidative-stress reporter mice carrying either Nrf2-wildtype or Nrf2-knockout genotypes were exposed in vivo to polychlorinated biphenyl (PCB) compounds, which are long-lasting industrial chemical pollutants. Two Alzheimer’s disease (AD) mouse models carrying humanized amyloid-beta-precursor protein (APP)-mutations were subjected to in vivo inhalation exposures to either woodsmoke or propane-derived ultrafine particles. Multi-omic approaches were performed following these exposure experiments, including bulk-RNA sequencing, Infinium DNA methylation BeadChip profiling, and validation by Nanopore long-read sequencing for DNA methylation, and mass spectrometry-based proteomics. In the transcriptomic analysis, differentially expressed genes (DEG) were filtered (|log2 fold change| > 1.5 & FDR < 0.05), followed by over-representation analysis using Gene Ontology and KEGG databases. Weighted gene correlational network analysis (WGCNA), an unsupervised learning algorithm, was employed to cluster co-expressed genes related to exposure conditions, and then identify putative pathway regulators. Gene set enrichment analyses (GSEA) were applied to assess cross-species consistency with human datasets. Differentially methylated regions (DMRs) were identified as the genomic segments with consistent methylation changes spanning adjacent CpGs. To examine their relationship with gene expression, DMRs were mapped to genomic regulatory features, including enhancers, promoters, and gene bodies. Random forest method was further used to select key CpG loci most discriminative of exposure conditions. In addition, analogous to WGCNA, co-methylated CpG clusters were identified, which may reflect coordinated pathway-level responses. Key findings of this PhD study: (1)Using the oxidative stress reporter mouse model, we identified a previously unrecognised Nrf2-dependent cell-cycle response in the liver following PCB exposure. (2)Building on the emerging concept of the lung-brain axis in neurodegeneration, we demonstrated that woodsmoke exposure in Alzheimer’s disease (AD) model mice induces markedly exacerbated pulmonary immune responses and suppresses hippocampal ion-channel activity. Notably, reduced hippocampal ion-channel activity was also observed in a separate familial AD mouse model exposed to propane-derived ultrafine particles. (3)Although direct relationship between DNA methylation and gene expression were limited across these exposure experiments, methylation alterations were highly enriched within open-chromatin regions, suggesting that these domains are susceptible to environmental insults. Together, these findings demonstrate how controlled exposure models can uncover novel molecular pathways and highlight potential neurotoxic effects of inhaled pollutants. The identified gene sets show strong correspondence with human molecular signatures, underscoring the translational potential of this work in contributing understanding to human diseases.