White matter DNA methylation profiling reveals deregulation of HIP1, LMAN2, MOBP, and other loci in multiple system atrophy

Abstract

Multiple system atrophy (MSA) is a fatal late-onset neurodegenerative disease. Although presenting with distinct pathological hallmarks, which in MSA consist of glial cytoplasmic inclusions (GCIs) containing fibrillar α-synuclein in oligodendrocytes, both MSA and Parkinson's disease are α-synucleinopathies. Pathologically, MSA can be categorized into striatonigral degeneration (SND), olivopontocerebellar atrophy (OPCA) or mixed subtypes. Despite extensive research, the regional vulnerability of the brain to MSA pathology remains poorly understood. Genetic, epigenetic and environmental factors have been proposed to explain which brain regions are affected by MSA, and to what extent. Here, we explored for the first time epigenetic changes in post-mortem brain tissue from MSA cases. We conducted a case-control study, and profiled DNA methylation in white mater from three brain regions characterized by severe-to-mild GCIs burden in the MSA mixed subtype (cerebellum, frontal lobe and occipital lobe). Our genome-wide approach using Illumina MethylationEPIC arrays and a powerful cross-region analysis identified 157 CpG sites and 79 genomic regions where DNA methylation was significantly altered in the MSA mixed-subtype cases. HIP1, LMAN2 and MOBP were amongst the most differentially methylated loci. We replicated these findings in an independent cohort and further demonstrated that DNA methylation profiles were perturbed in MSA mixed subtype, and also to variable degrees in the other pathological subtypes (OPCA and SND). Finally, our co-methylation network analysis revealed several molecular signatures (modules) significantly associated with MSA (disease status and pathological subtypes), and with neurodegeneration in the cerebellum. Importantly, the co-methylation module having the strongest association with MSA included a CpG in SNCA, the gene encoding α-synuclein. Altogether, our results provide the first evidence for DNA methylation changes contributing to the molecular processes altered in MSA, some of which are shared with other neurodegenerative diseases, and highlight potential novel routes for diagnosis and therapeutic interventions.

Keywords: Brain tissue; EWAS; MSA; Neurodegeneration; Pathological subtypes; WGCNA.

Fig. 1

Flowchart of the analysis process in a multi-stage multiple system atrophy (MSA) study. Briefly, for the discovery cohort, DNA extracted from white matter tissue of three brain regions (N = 48 samples in total) was used for DNA methylation profiling using the Illumina MethylationEPIC arrays (covering > 850,000 methylation sites) (1). The EPIC array data were subject of thorough quality control (2, 47 samples and 756,224 CpG sites were used for downstream analyses). To identify DNA methylation MSA-associated loci, we have then performed an epigenome-wide association study (EWAS) comparing MSA mixed-subtype cases to controls, and accounting for possible confounder factors (e.g. age, sex, post-mortem delay, and neuronal proportions). The EWAS was done for individual CpGs (significance threshold FDR < 0.05) as well as for larger genomic regions comprising several CpGs (significance threshold Stouffer < 0.05). Some of the genes, which showed significant differences in beta values are indicated (3). We sought to replicate the results from the discovery cohort in the replication and follow-up cohort that comprised a second MSA mixed-subtype cohort as well as a set of the other MSA pathological subtypes (OPCA and SND subtypes). DNA extracted from cerebellar white matter tissue only (N = 48 samples in total) was used for DNA methylation profiling using the Illumina MethylationEPIC arrays (4), which was subject of thorough quality control checks (5, 46 samples and 758,752 CpG sites were used for downstream analyses). The EWAS comparing each MSA subtype to controls, and accounting for possible confounder factors (e.g. age, sex, post-mortem delay, and neuronal proportions), did not reveal significant DNA methylation changes at the genome-wide level (6). As our aim was to replicate the MSA mixed-subtype-associated loci identified in the discovery cohort in a new cohort of MSA mixed-subtype cases and to follow-up these loci in the other two MSA pathological subtypes (OPCA and SND), we then performed loci-specific analyses in the second cohort (7). Finally, we have used an alternative analysis approach by applying weighted gene correlation network analysis to the 10% most variable CpG sites (N = 75,798 CpG sites) from all cerebellar samples (N = 62, from both cohorts), to identify co-methylation MSA-associated signatures (8). MSA multiple system atrophy, OPCA olivopontocerebellar atrophy, SND striatonigral degeneration, CRBL cerebellum, FL frontal lobe, OL occipital lobe, WM white matter, QC quality control, EWAS epigenome-wide association study, FDR false discovery rate, WGCNA weighted gene correlation network analysis

Fig. 2

Semi-quantitative analysis of global DNA methylation by immunohistochemistry in multiple system atrophy (MSA) mixed-subtype cases compared to normal controls. a 5‐Methylcytosine (5mC) immunohistochemical staining in the white matter of an MSA mixed-subtype case and a normal control (scale bar 100 µm). b Boxplot showing no significant differences between MSA mixed-subtype cases (N = 9) and controls (N = 6) regarding the percentage of cells staining positive for 5mC in three brain regions with distinct degrees of pathology. Grey: control brain samples; red: brain region severely affected in the MSA mixed subtype; orange: brain region moderately affected in MSA; yellow: brain region mildly affected in MSA. CTRL control group, MSA MSA mixed group, CRBL cerebellar white matter, FL frontal lobe white matter, OL occipital lobe white matter

Fig. 3

Summary of CpG-specific DNA methylation changes identified during the epigenome-wide association study in multiple system atrophy (MSA) mixed-subtype cases compared to normal controls (discovery phase). a Manhattan plot showing the distribution of differentially methylated CpGs across different chromosomes using a powerful cross-region analysis (pooling together samples from cerebellum, frontal lobe and occipital lobe). b, c Number and distribution of differentially hypo- and hypermethylated CpGs (represented in blue and red bars, respectively) identified in cross-region and region-specific analyses (MSA mixed subtype vs controls, FDR < 0.05). d Venn diagram representing the overlap across the three brain regions (cerebellum CRBL, frontal lobe FL, occipital lobe OL), with blue and red numbers indicating the number of CpGs differentially hypo- and hypermethylated in MSA mixed subtype vs controls, respectively

Fig. 4

Differentially methylated regions (DMRs) in MSA mixed subtype compared to controls (discovery phase). a Number of hypo- and hypermethylated DMRs for the cross-region and region-specific analyses. b Top DMR in the cerebellum-specific analysis mapping to MOBP on chromosome 3. CTRL control group, MSA MSA mixed-subtype group, CRBL cerebellar white matter, FL frontal lobe white matter, OL occipital lobe white matter

Fig. 5

Boxplots of DNA methylation levels in the cerebellar white matter for the top 3 replicated CpGs in the MSA mixed-subtype across all sample groups. Green boxes highlight the discovery cohort (D), and orange boxes highlight the follow-up cohort (F), with light colours corresponding to controls, and dark colours to MSA cases. CTRL controls, MSA multiple system atrophy, Mixed mixed subtype, OPCA olivopontocerebellar atrophy subtype, SND striatonigral degeneration subtype

Fig. 6

Results from the cerebellar multiple system atrophy (MSA) co-methylation network analysis. a Module–trait relationships. The rows represent the co-methylation module eigengene (ME) and its colour, and the columns represent clinical/pathological traits. P values are presented within each cell and the colour scale at the right indicates the strength of the correlation (darker cells depict stronger correlations, with blue representing negative and red positive correlations). b Correlations between module membership (MM) and gene significance (GS) for MSA for the top 4 modules associated with the MSA status (r > 0.50). MSA multiple system atrophy, Mixed mixed subtype, OPCA olivopontocerebellar atrophy subtype, SND striatonigral degeneration subtype, GCI glial cytoplasmic inclusions