To the Editors,

Multiple system atrophy (MSA) is a fatal neurodegenerative disease and its aetiology remains elusive. The pathological hallmark of MSA is the presence of glial cytoplasmic inclusions (GCIs) containing fibrillar α-synuclein in oligodendrocytes [3, 6], but the regional vulnerability of the brain to these GCIs remains poorly understood. We read with interest the paper by Rydbirk and colleagues recently published in Acta Neuropathologica Communications [5]. They investigated DNA methylation (5mC) and hydroxymethylation (5hmC) changes in prefrontal cortex samples from MSA patients. We reported previously [1] total DNA methylation (5mC + 5hmC) changes in MSA, and this work by Rydbirk et al. [5] further supports a contribution of epigenetic factors, namely DNA methylation, to MSA brain pathophysiology.

Given the differentially methylated CpGs (e.g. in AREL1 and KTN1), regions (DMRs) and blocks reported for the 5mC fraction by Rydbirk and colleagues [5], we performed additional loci-specific analysis of our MSA DNA methylation data [1]. We used data from our discovery cohort [1], which was composed of neuropathologically confirmed MSA mixed cases and controls, and investigated multiple brain regions characterized by different degrees of GCI burden in MSA, including the cerebellum, and the frontal and occipital lobes. In the frontal lobe, no changes were detected in AREL1 nor in the reported intergenic CpGs (Supplementary Table S1.1). Although with small effects (absolute delta betas < 5%), two CpGs in KTN1 were nominally significant (cg14002714 and cg21059882; p < 0.05). Regarding the block covering PHF3, two CpGs were nominally significant (cg16049132 and cg10435600; p < 0.05). Additionally, 52 CpGs in the DMR genes were nominally significant in our data, 14 of which (in 5 genes: FUT4, BCAR1, CTSZ, ZIC4 and FERMT3) demonstrated absolute delta betas of ≥5%. With the exception of cg18023065 in FUT4, none of these changes passed multiple testing correction (p < 9.07 × 10− 5 [0.05/551 CpGs]). Some of those CpGs and additional CpGs were also nominally significant in the other brain regions analysed (Supplementary Table S1.1). Interestingly, the DMR in the FUT4/PIWIL4 promoter (chromosome 11: 94278407–94,279,068), replicates a DMR we found in the frontal lobe and cross-region analyses of our previous study (Supplementary Tables S3.1 and S3.3 from [1]).

The study by Rydbirk et al. [5] and ours [1] have markedly different designs: a) Rydbirk et al. [5] included white and grey matter from the frontal lobe, whilst we carefully dissected white matter to enrich for oligodendrocytes and analysed different brain regions; and b) they investigated the contributions of 5mC and 5hmC separately. We are aware that in our data alterations in the 5mC and 5hmC proportions can counteract each other and mask the detection of significant changes in total methylation. As an example, the AREL1 shift from 5mC to 5hmC reported by Rydbirk et al. [5] is masked in our total DNA methylation data, highlighting an advantage of analysing 5mC and 5hmC separately. In addition, distinct cell type compositions in the brain tissue samples may contribute to discordant findings. According to RNAseq data from major brain cell types (data from Zhang et al. [7], Supplementary Fig. 1), our results support methylation changes in genes that are highly expressed in oligodendrocytes, including KTN1 and PHF3, or more highly expressed in microglia/macrophages, including FUT4, CTSZ, and FERMT3 (Supplementary Fig. 1). Conversely, genes more highly expressed in neurons and/or with low expression in oligodendrocytes, such as AREL1, were less susceptible to DNA methylation changes in our dataset.

Findings from Rydbirk et al. [5] also report increased AREL1 and MHC class I HLA gene expression in MSA brains. We therefore investigated in our MSA cerebellar white matter RNAseq data [4] gene expression changes in all of the genes reported by Rydbirk et al. [5]. Our study includes two independent cohorts of 66 MSA and 66 healthy controls and laser captured oligodendrocytes [4]. Although we did not find differential expression for AREL1 or PHF3, we found a nominally significant (p < 0.05) downregulation of KTN1 (log2 FC = − 0.465), and upregulation of CTSZ (log2 FC = 0.817), NCS1 (log2 FC = 0.813) and ZIC4 (log2 FC = 1.520) in some groups of our cohort 1 (Supplementary Table S2). The upregulation of ZIC4 was replicated in cohort 2 of our study (log2 FC = 1.551) and remained significant when accounting for multiple testing adjustments in the combined analysis of both cohorts 1 and 2 (log2 FC = 1.536; adjusted-p = 0.022). In our RNAseq data, MHC class I HLA genes have shown inconsistent results across cohorts/groups, with only HLA-A showing nominally significant upregulation in one group of cohort 1 (log2 FC = 1.156 in MSA-P; p = 0.002) and HLA-F in oligodendrocytes (log2 FC = 1.982; p = 0.032). Some of the MHC class I HLA genes, including HLA-A, have also shown nominally significant DNA methylation changes in our data (Supplementary Table S1.2).

Overall, we consider that these recent studies by Bettencourt et al. [1], Piras et al. [4], and Rydbirk et al. [5] are complementary, and bring important insights into the brain pathophysiology of MSA. All show changes in DNA methylation or in gene expression levels of genes that are more highly expressed in microglia/macrophages, therefore supporting previous studies highlighting the involvement of inflammatory processes in MSA (e.g. [2]).