Extended Data Fig. 1: Analysis of sample distribution and genomic context of modified cytosine sites at enhancers.

a, Dendrogram showing unsupervised clustering of samples based on cytosine sites. DNA methylation at top 10,000 most variable cytosine sites. Clustering of corresponding technical (nuclei sorting and/or library preparation replicated for select DNA samples) and sequencing replicates (sequencing replicated across lanes/flow cells for select libraries) are shown in pink and purple, respectively. Pearson’s correlation of technical replicates is 0.972 ± 0.0012 s.e.m. and of sequencing replicates is 0.977 ± 0.0010 s.e.m., supporting a high technical reproducibility. b, The proportion of each CpH subtype for the differentially methylated cytosine sites in PD neurons (significant: orange) as compared to the total number of cytosine sites profiled (background: gray). CpH context for cytosine sites in enhancers in this study was similar to that of previous whole genome studies in neurons²⁴. c, The proportion of differentially modified cytosine sites in PD located within an enhancer or promoter. Differentially modified cytosines in neurons of PD patients relative to controls were identified by logistic regression after adjusting for age, sex, postmortem interval, and neuronal subtypes (n = 57 PD, 48 controls; q < 0.05, multiple testing corrected). d, The proportion of 21 specific neuronal subtypes present in the prefrontal cortex of PD patients (orange) and controls (gray). There are 12 glutamatergic and 9 GABAergic neuronal types in the NeuN+ nuclei from PD and controls (n = 57 and 48 individuals, respectively). Neuronal proportions were calculated by computational deconvolution, using neuronal subtype specific markers (n = 563 cytosine sites)²⁴. No significant differences were determined between PD and controls, as determined by a repeated-measures ANOVA and post hoc Tukey HSD test. Error bars represent s.e.m. e, Proportion of glutamatergic neurons and GABAergic neurons in PD and controls. Neuronal proportions determined by the sum of all glutamatergic or GABAergic neuronal subtypes identified by computational deconvolution in panel a. Averaged 76.95% glutamate neuronal subtype proportion ± 0.814 s.e.m. There were no significant differences in glutamatergic and GABAergic neuronal proportions between PD patients and controls, as determined by a repeated-measures ANOVA and post hoc Tukey HSD test. Boxplot center line is the mean, the bounds of the box are the lower and upper quartiles, and the whiskers extend to minimum and maximum data points within 1.5 times the interquartile range.