Interrogating DNA methylation associated with Lewy body pathology in a cross brain-region and multi-cohort study

Abstract

Lewy body (LB) diseases are an umbrella term encompassing a range of neurodegenerative conditions all characterized by the hallmark of intra-neuronal α-synuclein associated with the development of motor and cognitive dysfunction. In this study, we have conducted a large meta-analysis of DNA methylation across multiple cortical brain regions, in relation to increasing burden of LB pathology. Utilizing a combined dataset of 1239 samples across 855 unique donors, we identified a set of 30 false discovery rate (FDR) significant loci that are differentially methylated in association with LB pathology, the most significant of which were located in UBASH3B and PTAFR, as well as an intergenic locus. Ontological enrichment analysis of our meta-analysis results highlights several neurologically relevant traits, including synaptic, inflammatory and vascular alterations. We leverage our summary statistics to compare DNA methylation signatures between different neurodegenerative pathologies and highlight a shared epigenetic profile across LB diseases, Alzheimer's disease and Huntington's disease, although the top-ranked loci show disease specificity. Finally, utilizing summary statistics from previous large-scale genome-wide association studies we report FDR significant enrichment of DNA methylation differences with respect to increasing LB pathology in the SNCA genomic region, a gene previously associated with Parkinson's disease and dementia with Lewy bodies.

Keywords: Alpha-synuclein (α-synuclein); Brain; DNA methylation; Dementia; Dementia with Lewy bodies (DLB); Epigenome-wide-association study (EWAS); Lewy body (LB); Meta-analysis; Parkinson’s disease (PD); SNCA.

Figure 1.. Cohort summary of samples used in DNA methylation association meta-analysis of Lewy bodies.

A) Brain regions and analysis protocol for meta-analyses. Brain regions include two cortical regions, the prefrontal cortex (PFC, BA9) and the anterior cingulate cortex (ACC, BA24). Analysis plan outlining steps in data generation, harmonization, within cohort linear association analyses and fixed effects meta-analysis. B) Summary of sample numbers in the UK Brain Bank Network (UKBBN), Netherlands Brain Bank (NBB) and Brains for Dementia Research (BDR) cohorts. Left hand bar plots show number of samples per Braak LB Stage, right hand plots show number of samples with coinciding AD pathology as measured by the Braak NFT stage. Components of figure created in https://BioRender.com

Figure 2.. DNA methylation associated with Lewy body Braak stage in a cross cortical meta-analysis.

A) Manhattan plot showing differentially methylated sites associated with Braak LB stage in a fixed effects inverse variance weighted meta-analysis of 1239 total samples (primary full cohort LB pathology meta-analysis). Y-axis shows -log10 transformed p-value of association, x-axis shows chromosomal positions (chromosomes 1–22). The black solid horizontal line shows the genome-wide significance threshold (p < 9 × 10⁻⁸), whilst the grey dashed horizontal line denotes the FDR-corrected significance threshold (q < 0.05). Points are colored to show chromosome number and for those that passed the significance thresholds annotated by UCSC gene symbols in the Illumina manifest file and where unavailable, Illumina CpG ID is displayed. B) Forest plots of the effect sizes for the three sites passing the genome-wide significance threshold in the primary full cohort LB pathology meta-analysis. Grey points show cohort specific estimates, sized by weight in the inverse variance weighted meta-analysis. Whiskers show 95% confidence interval (CI) of estimate. Diamonds show effect estimates and 95% CI for fixed effect estimates from the primary meta-analyses. The I² heterogeneity statistic and heterogeneity p-value are displayed for each site. C) Correlation plot of effect sizes between the primary full cohort LB pathology meta-analysis (y-axis) versus a subset analysis of samples without substantial coincident AD pathology ( Braak NFT stage ≤ II) (x-axis) for the 30 FDR-significant DMPs identified in the primary full cohort meta-analysis. Points are sized by -log10 transformed p-value significance in the primary full meta-analysis and colored by -log10 transformed p-value significance in the pure LB subset analysis. Effect sizes are displayed as change in methylation (beta value) per increasing LB Braak stage.

Figure 3.. A) Gene ontology (GO) analysis of primary full cohort Lewy body pathology meta-analysis.

The nine FDR-significant terms identified in the ontological analysis are shown, with -log10 transformed raw p-value shown on the x-axis. REVIGO prioritized gene parent terms are shown on the y-axis. B) Mini-Manhattan plots of the LTF gene region, C) NDRG4 gene region and D) MYO16 gene region. Y-axis shows -log10 transformed p-value for the primary full cohort LB pathology meta-analysis and x-axis shows the base-pair position. Points show probe coverage for the EPIC array and are colored by their direction of effect in association with LB-stage. The highlighted blue region shows the differentially methylated region (DMR) identified from comb-p analysis. Below track shows the annotated transcripts for the region. The black dashed horizontal line represents nominal significance (p < 0.05).

Figure 4.. Summary of comparisons of epigenomic signatures between different pathologies and accuracy of specific epigenetic pathology predictors.

A) Heatmap of overlap in direction of effect for the most significant LB-Braak associated sites (p < 1 × 10⁻⁵, left panel) and NFT-Braak associated sites (p < 1.25 × 10⁻⁷, right panel) with different neurodegenerative disease outcomes. Fill color is the overlap probability is direction of association between two overlapping summary statistics as calculated by the binomial sign test. Bolded tiles indicate an FDR significant test for significant overlap in directionality of associations. Braak NFT effects are taken from Smith et al. (2021) and Huntington’s Disease effects from Wheildon et al. (2025). B-D) Scatterplots of DNA methylation associated with different pathologies. B) Shown for the top LB-associated loci is the comparison of effect size in cortex for LB and NFT (Smith et al., 2021) for the 21 overlapping sites across summary statistics, and C) for LB and HD (Wheildon et al., 2025) for the 72 overlapping sites across summary statistics. D) shows the comparison of effect size in cortex for previously reported NFT associated loci (Smith et al., 2021) in our LB samples for the 208 sites overlapping across summary statistics. E) Contingency tables of multiple epigenetic based pathology prediction models for binary classification of cortical pathology. Pathological outcome used to train the model is displayed along the x-axis, with the pathological outcome tested along the y-axis. Tiles are colored by the proportion of the true group predicted and text shows number of cases (top) and proportion of true predicted (bottom). F) Bar-plots showing balanced accuracy for each prediction scenario. X-axis shows which model was employed; y-axis shows balanced accuracy of prediction (+/− Standard Error). True outcome tested shown above the top of each plot grid.