https://orcid.org/0000-0001-6137-6659
Abstract
Parkinson’s disease (PD) is the second most common neurodegenerative disorder, and aging and genetic and environmental exposure can contribute to its pathogenesis. DNA methylation has been suggested to play a pivotal role in neurodevelopment and neurodegenerative diseases. 5-hydroxymethylcytosine (5hmC) is generated through 5-methylcytosine (5mC) oxidization by ten-eleven translocation proteins and is particularly enriched in the brain. Although 5hmC has been linked to multiple neurological disorders, little is known about 5hmC alterations in the substantia nigra of patients with PD. To determine the specific alterations in DNA methylation and hydroxymethylation in PD brain samples, we examined the genome-wide profiles of 5mC and 5hmC in the substantia nigra of patients with PD and Alzheimer’s disease (ad). We identified 4119 differentially hydroxymethylated regions (DhMRs) and no differentially methylated regions (DMRs) in the postmortem brains of patients with PD compared with those of controls. These DhMRs were PD-specific when compared with the results of AD. Gene ontology analysis revealed that several signaling pathways, such as neurogenesis and neuronal differentiation, were significantly enriched in PD DhMRs. KEGG enrichment analysis revealed substantial alterations in multiple signaling pathways, including phospholipase D (PLD), cAMP and Rap1. In addition, using a PD Drosophila model, we found that one of the 5hmC-modulated genes, PLD1, modulated α-synuclein toxicity. Our analysis suggested that 5hmC may act as an independent epigenetic marker and contribute to the pathogenesis of PD.
Introduction
Parkinson’s disease (PD) is the second most common neurodegenerative disease after Alzheimer’s disease (AD). The characteristic motor symptoms of PD include tremor at rest, rigidity and bradykinesia (1), whereas AD is defined clinically by a progressive decline in cognitive functions, especially memory (2). PD is mainly caused by changes in the substantia nigra, with the presence of intracellular inclusions known as Lewy bodies (LBs), which are made up of α-synuclein (3), and the nigra neurons are more vulnerable to environmental toxins (4). Although the pathogenesis of neurodegenerative diseases is not fully understood, aging and genetic and environmental exposures may contribute to PD (5). The majority of PD cases are sporadic, and known monogenic forms combined explain only |$\sim$|30% of monogenic and 3–5% of genetically complex cases (6). Epigenetic changes are thought to mediate the relationship between genetic determinants and aging or the environment (7). Recently, the role of epigenetic regulation in the pathogenesis of neurodegenerative diseases, including PD, has attracted increased attention (8–15).
The demographic characteristics of PD patients and controls
| PD | Control | P-value | ||
|---|---|---|---|---|
| Number of subjects | 12 | 9 | 0.513 | |
| Gender | Male | 6 | 6 | 0.317 |
| Female | 6 | 3 | ||
| Age at death (yr) | 80.6 ± 5.1 | 76.3 ± 11.9 | 0.731 | |
| PMI (hr) | 17.3 ± 5.4 | 12.4 ± 7.1 | 0.368 | |
| Years in storage (yr) | 10.3 ± 4.9 | 14.6 ± 6.3 | 0.752 | |
| PD | Control | P-value | ||
|---|---|---|---|---|
| Number of subjects | 12 | 9 | 0.513 | |
| Gender | Male | 6 | 6 | 0.317 |
| Female | 6 | 3 | ||
| Age at death (yr) | 80.6 ± 5.1 | 76.3 ± 11.9 | 0.731 | |
| PMI (hr) | 17.3 ± 5.4 | 12.4 ± 7.1 | 0.368 | |
| Years in storage (yr) | 10.3 ± 4.9 | 14.6 ± 6.3 | 0.752 | |
PMI, postmortem interval; hr, hour; yr, year.
The demographic characteristics of PD patients and controls
| PD | Control | P-value | ||
|---|---|---|---|---|
| Number of subjects | 12 | 9 | 0.513 | |
| Gender | Male | 6 | 6 | 0.317 |
| Female | 6 | 3 | ||
| Age at death (yr) | 80.6 ± 5.1 | 76.3 ± 11.9 | 0.731 | |
| PMI (hr) | 17.3 ± 5.4 | 12.4 ± 7.1 | 0.368 | |
| Years in storage (yr) | 10.3 ± 4.9 | 14.6 ± 6.3 | 0.752 | |
| PD | Control | P-value | ||
|---|---|---|---|---|
| Number of subjects | 12 | 9 | 0.513 | |
| Gender | Male | 6 | 6 | 0.317 |
| Female | 6 | 3 | ||
| Age at death (yr) | 80.6 ± 5.1 | 76.3 ± 11.9 | 0.731 | |
| PMI (hr) | 17.3 ± 5.4 | 12.4 ± 7.1 | 0.368 | |
| Years in storage (yr) | 10.3 ± 4.9 | 14.6 ± 6.3 | 0.752 | |
PMI, postmortem interval; hr, hour; yr, year.
DNA methylation is a reversible process that converts cytosine to 5-methylcytosine (5mC) by adding a methyl group to the fifth carbon of cytosine in the presence of DNA methyltransferases, with S-adenosylmethionine (SAM) as the methyl group donor. 5mC can be gradually oxidized to 5-hydroxymethylcytosine (5hmC), 5-formylcytosine and 5-carboxylcytosine by the ten-eleven translocation (TET) protein family, which constitutes the endogenous active DNA demethylation pathway (16,17). 5hmC is enriched in the brain and increases markedly from the fetal to adult stage, suggesting that 5hmC-mediated epigenetic modification is critical in neurodevelopment and neurodegenerative disorders (18–20). Kaut et al. found that 5mC was significantly increased in the cortex of patients with PD, whereas 5hmC was elevated only in the cerebellar white matter (21). Stoger et al. used an enzyme-linked immunosorbent assay to show significantly higher 5hmC levels in the cerebellum of patients with PD, but without a significant change in the 5mC level (10). More recently, Marshall et al. showed a considerable increase in hydroxymethylation at enhancers in neurons of the prefrontal cortex in patients with PD (22). Overall, these studies failed to reach a consensus regarding the occurrence of overall DNA methylation and hydroxymethylation changes in the brain regions of patients with PD. In addition, little is known about the genome-wide patterns of 5mC and 5hmC, as well as their potential roles in the substantia nigra of PD.
This study primarily aimed to assess the genome-wide profiling of DNA methylation and hydroxymethylation levels in the substantia nigra of patients with PD and to identify whether the differentially methylated regions (DMRs) or hydroxymethylated regions (DhMRs) are PD specific. Our secondary aim was to study the biological functions of the genes contained within the identified regions in terms of functional biological pathways and to explore the possible mechanism of 5mC or 5hmC changes in the pathogenesis of PD.
Results
Genome-wide 5hmC alteration in the postmortem substantia nigra of patients with PD
We employed a previously established chemical labeling and affinity purification method coupled with high-throughput sequencing (hMe-Seal) to profile 5hmC, and methylated DNA immunoprecipitation (MeDIP) coupled with high-throughput sequencing (MeDIP-seq) to profile 5mC. We examined DNA methylation and hydroxymethylation in the substantia nigra of 12 patients with PD and 9 matched neurologically normal controls (Table 1). Using an FDR < 0.05, no DMRs were identified (Supplementary Material, Table S2). In contrast, at an false discovery rate (FDR) of < 0.05, 4119 DhMRs were detected, of which 1800 were hyperhydroxymethylated and 2319 were hypohydroxymethylated in PD compared with controls (Fig. 1A). The overall abundance of 5hmC was reduced in the PD samples (Fig. 1B and C; Supplementary Material, Fig. S1).
Characteristics of 5hmc in human PD brain samples. (A) Circos plots of the differentially DhMRs between PD patients and controls. Points show the distribution of DhMRs on the genome. The x-axis corresponds to the genomic coordinate, and the y-axis corresponds to the logFC values (log2 fold changes). The histograms show the DhMR enrichment with its P-value (−log transformed). Red indicates hypermethylated DhMRs and blue indicates hypomethylated DhMRs. Profile (B) and Heatmaps (C) plot with 5hmC in PD and control group. PD, Parkinson’s disease; CTL, control. (D) Percent of DhMRs that fall in gene model contexts as compared with a length-matched genomic region set.
Next, we investigated the distribution of the DhMRs across the genome. The identified DhMRs were mainly enriched in intragenic regions, including the 3′UTR (P = 3.72 E-12), exons (P = 8.78 E-27), introns (P = 1.09 E-10) and promoter regions (P = 1.34 E-22), but not in intergenic regions (Fig. 1D). Specifically, exons and promoter regions were the most significantly enriched. This finding is consistent with previously reported results (23,24).
5hmC changes associated with PD substantia nigra is specific
To determine whether the 5hmC changes observed were specific to the PD substantia nigra, we profiled genome-wide 5hmC from the substantia nigra of 13 patients with AD and nine matched controls (Supplementary Material, Table S1). Under an FDR < 0.05, 80 DhMRs were detected in the AD substantia nigra, with only 23 DhMRs located on autosomes (Supplementary Material, Table S3). There was no overlap between the 80 DhMRs and 4119 DhMRs detected in PD. These results suggest a lack of genome-wide 5hmC changes in AD substantia nigra and that the 5hmC changes in PD substantia nigra represent an epigenetic signature specific to PD.
Selective biological pathways associated with PD DhMRs
The 4119 DhMRs identified in the PD substantia nigra were annotated to 2933 unique genes. We investigated the biological functions of the annotated genes using Gene Ontology (GO) biological process and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses. We found that these genes were enriched for GO biological processes, including neurogenesis (FDR < 2.2 E-16), neuron differentiation (FDR < 2.2 E-16), neuron generation (FDR < 2.2 E-16), neuron development (FDR = 7.12 E-13), neuron projection development (FDR = 1.83 E-10) and axonogenesis (FDR = 7.77 E-09) (Fig. 2A). These results indicate that DhMRs potentially play an important role in the epigenetic regulation of neuronal and synaptic function. In addition, multiple KEGG pathways, such as the cAMP signaling pathway (FDR = 1.12 E-07), Rap1 signaling pathway (FDR = 5.03 E-06), axon guidance (FDR = 5.07 E-05), cGMP-PKG signaling pathway (FDR = 1.85 E-04), phospholipase D (PLD) signaling pathway, Ras signaling pathway (FDR = 2.35 E-04) and Hippo signaling pathway (FDR = 3.28 E-04), were significantly enriched (Fig. 2B). Interestingly, among these genes and pathways, the phospholipase D1 (PLD1) gene was involved in three of the top 10 significantly enriched pathways, which could serve as the node for points where these signaling pathways converge. Moreover, critical molecules in the PLD signaling pathway, including PLD1, RHO and RAF1, display altered DNA hydroxymethylation. These results suggest that the PLD signaling pathway, particularly PLD1, may be associated with PD pathogenesis. The details of the DhMRs in PLD1 are shown in Supplementary Material, Figure S2.
GO and KEGG pathway analysis of DhMRs identified in PD patients. (A) DhMRs associated genes-enriched GO categories. Significant GO categories identified by the functional annotation tool web-based gene set analysis toolkit. Red dashed line shows FDR = 0.05. (B) DhMRs associated genes-enriched KEGG pathways (Top 20). Rich factor is the ratio of enrichment.
We further searched for interactions among 248 DhMR-associated genes enriched in the top 10 KEGG pathways. A total of 248 DhMR-associated genes were used as seed genes for protein-protein interaction (PPI) analysis. The PPI network generated by STRING (25) consisted of 247 nodes and 2743 edges (Supplementary Material, Fig. S3). The number of edges was significantly larger than expected (PPI enrichment P ≤ 1 E-16). This suggests that the PPI network is highly interconnected, indicating that the DhMR-annotated genes within the top 10 pathways interact with each other and may be biologically relevant.
Genome-wide 5hmC changes in α-synuclein transgenic mice
To determine the role of 5hmC in PD pathogenesis, we profiled 5hmC using an existing PD mouse model, α-synuclein-transgenic mice (26,27). We used an approach similar to that described above for patients with PD to capture the distribution of 5hmC in nine transgenic mice and nine control mice. Using an FDR of < 0.2, 21 465 DhMRs were identified, of which five regions displayed an increased 5hmC signal (hyperhydroxymethylation) and 21 460 regions displayed a loss of 5hmC (hypohydroxymethylation) in the α-synuclein transgenic mice. The distribution of DhMRs across the genomic regions is shown in Figure 3.
Characteristics of 5hmc in α-synuclein mice. (A) Read count over all gene body in alpha-synuclein mice and wild-type mice. (B) Histogram of the log fold change from expected versus distinct genomic features. (C) The relationship between the probability of hydroxymethylation changes and distance to TSS. (D) Percentages of DhMRs across different genomic locations. TSS, transcription start site; TES, transcription end site; UTR, untranslated region.
These DhMRs were annotated to 5669 genes. GO and KEGG pathway analyses were performed to determine the biological roles of the genes associated with altered 5hmC expression. GO biological process analysis revealed that several biological processes were enriched, such as detection of chemical stimuli involved in sensory perception of smell (P = 4.28 E-43), nervous system development (P = 1.14 E-19), neurogenesis (P = 1.82 E-18) and neuron differentiation (P = 1.11 E-10) (Fig. 4A). KEGG pathway analysis also revealed important pathways, including the gonadotropin-releasing hormone receptor pathway (P = 4.20 E-04), vascular endothelial growth factor (VEGF) signaling pathway (P = 1.72 E-03), Ras pathway (P = 9.55E-03) and Notch signaling pathway (P = 1.15E-02) (Fig. 4B).
GO and KEGG pathway analysis of 5669 overlapping DEGs. (A) GO analysis of 5669 overlapping DEGs. (B) KEGG pathway analysis of 5669 overlapping DEGs. Significant GO categories and KEGG pathway were identified by subjecting regions of DhMRs to GREAT. Blue and red bars represent −log10 (binomial P value).
Overlap of 5hmC signals between the PD mouse model and patients with PD
To understand the overlap of 5hmC signals between brain tissues from patients with PD and α-synuclein transgenic mice, we overlapped the DhMR-annotated genes detected in the PD mouse model and patients with PD. Considering that nearly all the DhMRs identified in mice were hypohydroxymethylated, we next compared only the hypo-DhMR-annotated genes. A Venn diagram was used to show the overlap of hypo-DhMR-annotated genes (N = 5667) from the PD mouse model and hypo-DhMR-annotated genes (N = 1967) from patients with PD (Supplementary Material, Fig. S4A). Among these hypo-DhMR-annotated genes, 555 genes, including PLD1, overlapped in the PD mouse model and patients in the same direction.
To further examine the biological significance of overlapping DhMR-annotated genes (N = 555), we performed GO and KEGG analyses and found a significant enrichment of actin cytoskeleton and neuronal ontological from gene sets, including actin cytoskeleton organization (FDR = 1.95 E-04), neurogenesis (FDR = 2.59E-04), generation of neurons (FDR = 2.59E-04), actin filament-based process (FDR = 3.49E-04) and cytoskeleton organization (FDR = 5.23E-04) (Supplementary Material, Fig. S4B). KEGG pathway analysis identified a significant enrichment of signaling pathways involved in the Rap1 signaling pathway (FDR = 4.51E-02), MAPK signaling pathway (FDR = 4.51E-02), adherens junction (FDR = 4.51E-02), Ras signaling pathway (FDR = 4.51E-02), autophagy (FDR = 4.51E-02) and regulation of actin cytoskeleton (FDR = 4.65E-02) (Supplementary Material, Fig. S4C).
PLD1 modulates the toxicity associated with the expression of α-synuclein
Recently, Drosophila models have been widely used to study the molecular pathogenesis of neurodegenerative disorders such as PD (28). To determine whether PLD1 could modulate this pathogenesis of PD, we constructed Drosophila lines expressing α-synuclein and PLD in dopaminergic cells. The overexpression of α-synuclein can lead to locomotor activity deficits (29), so we performed a climbing test to compare the locomotor function of transgenic fly lines expressing α-synuclein, PLD, or α-synuclein/PLD. The climbing ability of the α-synuclein Drosophila model was significantly decreased on Days 20 and 30 (P < 0.01), which was rescued by the overexpression of PLD1 in dopaminergic cells (Supplementary Material, Fig. S5). Our observations suggest that PLD1 modulates α-synuclein toxicity.
Discussion
There are few studies on genome-wide methylation and hydroxymethylation in the brain tissues of patients with PD; although, total 5mC and 5hmC levels or methylation of PD-associated genes have been measured in several studies (30,31). Unchanged levels of 5mC immunoreactivity have been reported in different brain regions in PD; however, 5hmC was significantly upregulated in the cerebellar white matter and cerebellum (10,21). Marshall et al. identified 1799 differentially methylated cytosines in enhancers by bisulfite padlock probe sequencing in the prefrontal cortex of PD neurons compared with control neurons, and the majority of enhancers in PD neurons were hypermethylated (22). The widespread increase in cytosine modifications at enhancers in PD neurons coincides with a gain in hydroxymethylation by hMeDIP, suggesting that the increase in cytosine modifications may involve a prominent contribution to hydroxymethylation (22). In this study, we examined DNA methylation and hydroxymethylation in the substantia nigra of both PD and AD using hMe-Seal and MeDIP to distinguish 5hmC from 5mC. We did not observe any 5mC changes in the substantia nigra of patients with PD, whereas genome-wide 5hmC alteration DhMRs was identified. The 4119 DhMRs were annotated to 2933 unique genes involved in neuronal differentiation, generation, axon function and multiple signal transduction pathways. Our results are consistent with those of previous studies and suggest that DNA hydroxymethylation may play an important role in PD pathology.
To identify PD-specific DNA hydroxymethylation and hydroxymethylation changes, we also examined genome-wide 5hmC levels in the substantia nigra from AD postmortem brains. Only 80 DhMRs were identified in the AD tissues compared with the controls. There was no overlap between the PD 4119 DhMRs and the ad DhMRs, suggesting that the DhMRs identified in the PD substantia nigra were specific to PD. In a previous study, we generated genome-wide profiles of both 5mC and 5hmC in human frontal cortex tissues from patients with AD and cognitively normal controls and identified abundant DhMRs in AD (32). This difference in hydroxymethylome changes detected in the substantia nigra between PD and AD may be because the cortex is the brain region most affected by AD (33). In addition, Kaut et al. found that 5hmC levels did not differ between the neocortex of PD and control groups (21). The changes in 5hmC in the substantia nigra of PD may be an independent epigenetic characteristic of PD.
Aggregation of α-synuclein is the main component of LBs in patients with PD. The SNCA gene was the first to be linked to familial PD. Mice overexpressing α-synuclein have altered neuronal function and show α-synuclein aggregation, which replicates familial PD, making the model a useful tool to elucidate poorly understood α-synuclein function in PD (34). To explore the common changes in 5hmC between brain tissues from human PD and α-synuclein transgenic mice, we investigated the genome-wide DNA hydroxymethylation profiles of the substantia nigra of α-synuclein transgenic and control mice. Interestingly, we identified 555 overlapping DhMR-annotated genes that showed the same change in direction between mouse and human postmortem brains. We further examined the common biological roles of genes associated with 5hmC changes in both mice and humans. GO and KEGG pathway analysis revealed that differential hydroxymethylation was significantly enriched in neurobiological processes and clustered in functional signal pathways, some of which have been linked to PD pathogenesis previously (35–38).
It is worth noting that the PLD signaling pathway was significantly enriched in the DhMR-annotated gene sets of patients with PD. Moreover, PLD1 is significantly hypohydroxymethylated in both human brain tissues and transgenic mouse brains. PLDs are a class of phosphodiesterases that catalyze the hydrolysis of phospholipids and other amine-containing glycerophospholipids to generate phosphatidic acid and free head groups (39). PLDs play important cellular roles in membrane remodeling and biogenesis, such as vesicular transport, signal transduction, cytoskeletal dynamics, degranulation and cell cycle progression (39–41). The PLD signaling pathway has recently been proposed as an emerging therapeutic target for neurodegenerative disorders (40). The two classic mammalian isoenzymes, PLD1 and PLD2, have different regulatory properties and subcellular localizations (41). Recently, PLD1 was reported to interact with α-synuclein. Bae et al. found that specific inhibition of PLD1 or knockdown of PLD1 expression resulted in impaired autophagic flux and accumulation of α-synuclein aggregates in autophagosomes, and that the neuronal toxicity caused by α-synuclein accumulation could be rescued by overexpression of PLD1, but not mutant PLD1 (42). Conde et al. demonstrated that wild-type α-synuclein could also inhibit PLD1 expression, resulting in a decrease in neurofilament light chain (NFL) and altered actin cytoskeleton, and restoration of PLD1 expression promotes NFL recovery (43). In our study, we found significant hypohydroxymethylation at the PLD1 locus in PD brains compared with that in controls. Using a Drosophila model, we showed that the overexpression of PLD1 in dopaminergic neurons could rescue the locomotor ability of Drosophila expressing α-synuclein, suggesting that 5hmC-mediated epigenetic modulation of PLD1 may be involved in PD pathogenesis.
Nevertheless, we acknowledge that our study had some limitations. The sample size of our study was small, and repetition in a larger group of patients is warranted to validate the findings and conduct further hydroxymethylation analysis on the basis of aging stages and disease duration. We also acknowledge that analysis of PD mouse model-based DNA is a limitation, given that animal models may not fully reflect the true pathophysiology of PD. DNA hydroxymethylation analysis of large samples of PD cases and controls with post-mortem brain tissues should be a future focus of this field. In addition, multinational replication would identify ethnicity-specific and common alterations to the brain hydroxymethylome in PD. Further functional studies are needed to confirm our findings.
In conclusion, we profiled DNA methylation and hydroxymethylation in PD and detected abundant DhMRs rather than DMRs. These DhMRs were PD specific, and the differentially hydroxymethylated annotated genes were enriched in pathways associated with PD, including the PLD signaling pathway. We further performed validation test of gene PLD1 with PD Drosophila model. Overall, our study suggests that DNA hydroxymethylation plays an important role in the pathogenesis of PD and provides insights into the molecular mechanisms underlying PD.
Materials and Methods
Human brain tissue
Postmortem brains included in this study were obtained from the Harvard Brain Bank, Case Western Brain Bank and NIH Neurobiobank. This study was approved by the institutional review boards of the Case Western Reserve University and Central South University. All subjects signed a written consent form for brain donation. A total of 12 patients with PD, 13 patients with AD and 9 matched neurologically normal controls were enrolled. The demographic characteristics of patients and controls are summarized in Table 1 and Supplementary Material, Table S1. All patients with PD and ad were diagnosed by neurological physicians according to international diagnostic criteria (44,45). In addition, neuropathological examination of the patients and controls was completed by a neuropathologist after brain donation. The substantia nigra was excised from postmortem brains by a trained technician, followed by genomic DNA extraction (Thermo Fisher Scientific, K182001) according to the manufacturer’s instructions. The samples were store at −20°C before use.
Mouse brain tissue
Animal studies were approved by the Institutional Animal Care and Use Committee of the Case Western Reserve University. Mice were maintained under constant environmental conditions at the Animal Research Center of the Case Western Reserve University with free access to food and water. The development of α-synuclein transgenic mice appeared normal compared with that of non-transgenic littermates. Male 16-month-old α-synuclein transgenic [SN(+/+), n = 9] and non-transgenic littermates [SN(−/−), n = 9] were included. Mice were sacrificed, and the substantia nigra was isolated. Genomic DNA extraction was performed as described for the human brain samples.
Genome-wide profiling of brain DNA methylation and hydroxymethylation
hMe-Seal, a selective chemical labeling method, was used to determine the genome-wide distribution of 5hmC, as described previously (46). In brief, genomic DNA was ligated with Illumina sequencing adapters, and T4 bacteriophage β-glucosyltransferase was used to transfer an azide-modified glucose onto the hydroxyl group of 5hmC to selectively label 5hmC with biotin. DNA fragments containing biotin-modified 5hmC were pulled down, eluted and purified, and sequencing libraries were generated by polymerase chain reaction amplification. For 5mC, we employed a previously reported methylated DNA immunoprecipitation (MeDIP) protocol followed by high-throughput sequencing (MeDIP-seq) (32,47).
Sequence alignment, 5hmC/5mC qualification and DhMR/DMR analysis
Basic read quality was verified using summaries produced by the FastQC program. The data were cleaned and filtered to remove adapters and low-quality bases using Trimmomatic (48). The raw fastq files generated from the high-throughput sequencing were mapped to the human genome (hg19) and mouse genome (mm10) using the Burrows–Wheeler aligner (v0.7.10) (49). SAM files were converted to sorted BAM files and filtered to include only de-duplicated uniquely mapped reads using samtools (50). Principal component analysis was used to detect outliers. Qualified samples were included in further analyses.
To identify the DhMRs and DMRs, the reference genome was broken into 1 kb windows. The R package MEDIPS (MEDIPS v1.32.0) (51) was used to calculate the differential coverage between the control and PD groups. We preprocessed our data using the MEDIPS.createSet function (parameters: extend = 120, shift = 0, window size =1000, and unique = FALSE) and controlled for quality using saturation estimation and sequence pattern coverage. Differential windows were called for MEDIPS using the MEDIPS.meth function (parameters: p.adjust = fdr, diff.method = edgeR, MeDIP = FALSE, CNV = FALSE, minRowSum = 20). Windows with an FDR-adjusted P-value less than threshold 0.05 were considered DhMRs and DMRs. Annotation of DhMRs with respect to genomic features and nearby genes was performed using R package Goldmine (52).
Bioinformatic analysis
To provide insight into the biological function of the annotated DhMR genes, we conducted a functional enrichment analysis for the identified DhMR genes using a web-based gene set analysis toolkit (http://www.webgestalt.org/option.php). Gene symbols corresponding to DhMRs were used as input data for GO and KEGG pathway analysis. Profiles and heatmaps were generated using the ngs.plot tool (53) to validate the patterns and show the 5hmC signal within the detected DhMRs.
To identify potential interactions between the DhMR-mapped genes from the top 10 KEGG pathways, we employed the Search Tool for the Retrieval of Interacting Genes Database (STRING v11) (https://string-db.org) (25) to construct the PPI network. Using seed proteins as inputs, we constructed a PPI network containing only the query proteins. All interactions between query proteins were derived from active interaction sources including text mining, experiments, curated databases and co-expression. The species was limited to Homo sapiens and an interaction score ≥ 0.4 was applied.
Transgenic Drosophila and climbing assay
Fly lines UAS-α-synuclein/DDC-GAL4 and UAS-PLD/DDC-GAL4, which express α-synuclein and PLD in dopaminergic cells, respectively, were crossed. The climbing assay was carried out as described previously (29) with modifications: flies were placed in an empty vial and gently tapped until all flies fell to the bottom of the vial. The number of flies that climbed 19 cm above the vial was counted after 30 s of climbing. The data shown represent results from a cohort of flies tested on the 10th, 20th and 30th day. The experiment was repeated thrice using independently derived transgenic lines. Transgenic Drosophila was tested for the following genotypes: (1) a-syn/+, DDC-GAL4/+; (2) uPLD/+, DDC-GAL4/+; and (3) a-syn/uPLD; DDC-GAL4/+. Drosophila melanogaster strain w1118 was used as a control.
Statistical analysis
Statistical analyses were performed using R (version 3.5.0). Fisher’s exact test on a 2 × 2 contingency table was used to test the significance of the difference in sex between PD cases and controls. Student’s t-test was used to test the significance of the difference in age at death and postmortem interval between patients with PD and controls. Statistical significance was set at P < 0.05.
Acknowledgements
We thank all donors for their generous donation of tissue for this research.
Conflict of Interest Statement: The authors declare that they have no competing interests.
Funding
Hunan Innovative Province Construction Project (2019S K2335 to B.T.); National Natural Science Foundation of China (81361120404 to B.T.); U.S.-China Program for Biomedical Collaborative Research of National Institute of Neurological Disorders and Stroke (NS083498 to X.Z.).
References
Author notes
Shishi Min, Qian Xu and Lixia Qin authors contributed equally to this work and are co-first authors.
