Hindawi BioMed Research International Volume 2021, Article ID 6616434, 46 pages https://doi.org/10.1155/2021/6616434

Research Article Weighted Coexpression Network Analysis Uncovers Critical and Pathways for Multiple Brain Regions in Parkinson’s Disease

Jianjun Huang ,1 Li Liu ,2 Lingling Qin,3 Hehua Huang,4 and Xue Li5

1Department of Neurology, Youjiang Medical University for Nationalities, Affiliated Hospital of Youjiang Medical University for Nationalities, Baise, 533000 Guangxi, China 2Department of Cardiology, Youjiang Medical University for Nationalities, Affiliated Hospital of Youjiang Medical University for Nationalities, Baise, 533000 Guangxi, China 3Department of Medical Quality Management, Affiliated Hospital of Youjiang Medical University for Nationalities, Baise, 533000 Guangxi, China 4Department of Anatomy, Youjiang Medical University for Nationalities, Baise, 533000 Guangxi, China 5Department of Electrophysiology, Affiliated Hospital of Youjiang Medical University for Nationalities, Baise, 533000 Guangxi, China

Correspondence should be addressed to Jianjun Huang; [email protected]

Received 4 December 2020; Revised 21 January 2021; Accepted 8 February 2021; Published 15 March 2021

Academic Editor: Parameshachari B. D.

Copyright © 2021 Jianjun Huang et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Objective. In this study, we aimed to identify critical genes and pathways for multiple brain regions in Parkinson’s disease (PD) by weighted gene coexpression network analysis (WGCNA). Methods. From the GEO database, differentially expressed genes (DEGs) were separately identified between the substantia nigra, putamen, prefrontal cortex area, and cingulate gyrus of PD and normal p ∣ ð Þ ∣ : samples with the screening criteria of value < 0.05 and log2fold change FC >0 585. Then, a coexpression network was presented by the WGCNA package. Gene modules related to PD were constructed. Then, PD-related DEGs were used for construction of PPI networks. Hub genes were determined by the cytoHubba plug-in. Functional enrichment analysis was then performed. Results. DEGs were identified for the substantia nigra (17 upregulated and 52 downregulated genes), putamen (317 upregulated and 317 downregulated genes), prefrontal cortex area (39 upregulated and 72 downregulated genes), and cingulate gyrus (116 upregulated and 292 downregulated genes) of PD compared to normal samples. Gene modules were separately built for the four brain regions of PD. PPI networks revealed hub genes for the substantia nigra (SLC6A3, SLC18A2, and TH), putamen (BMP4 and SNAP25), prefrontal cortex area (SNAP25), and cingulate gyrus (CTGF, CDH1, and COL5A1) of PD. These DEGs in multiple brain regions were involved in distinct biological functions and pathways. GSEA showed that these DEGs were all significantly enriched in electron transport chain, proteasome degradation, and synaptic vesicle pathway. Conclusion. Our findings revealed critical genes and pathways for multiple brain regions in PD, which deepened the understanding of PD-related molecular mechanisms.

1. Introduction ogenesis, symptomatic treatment is mainly applied such as the replacement of dopamine [4]. Molecular biomarkers have Parkinson’s disease (PD) is the second most common neuro- been proven as promising clinical tools for PD diagnosis [5]. degenerative disease related to the loss of dopaminergic Thus, it is an urgent need to uncover new strategies for early neurons globally [1]. It is characterized by tremor and slow diagnosis and therapeutic intervention to improve the quality movement, affecting approximately 7 million people world- of life of the affected population. wide, most of whom are elderly [2]. Male and age are Understanding the mechanisms of PD at the molecular independent risk factors of PD [3]. Due to its complex path- levels is valuable for clinical treatment. With the widespread 2 BioMed Research International

Table 1: Dataset information from the GEO database.

Location Accession Platform Type Number Substantia nigra GSE20292 GPL96 Microarray 18 control vs. 11 PD Substantia nigra GSE7621 GPL570 Microarray 9 control vs. 16 PD Putamen GSE20291 GPL96 Microarray 20 control vs. 15 PD Prefrontal cortex area GSE20168 GPL96 Microarray 15 control vs. 14 PD Prefrontal cortex area GSE68719 GPL11154 RNA-seq 44 control vs. 29 PD Cingulate gyrus GSE110716 GPL11153 RNA-seq 8 control vs. 8 PD use of microarray and RNA-seq technologies, genes related 2.2. Differential Expression Analysis. Microarray expression to PD have been widely identified, which help decipher the data were used for differential expression analysis between complex pathogenesis of PD, thereby promoting the develop- the PD group and the control group using the limma package ment of effective drug targets and preventing the occurrence [13]. Before analyzing the expression differences, the probes of PD at an early stage [6–8]. Gene coexpression networks are were annotated. For the case where multiple probes corre- widely used for function prediction and identification of sponded to the same gene, the average value of multiple genes modules in a set of samples including PD [9]. As a probes was taken as the expression value of the gene. For method of bioinformatics research, WGCNA is usually the case where there were multiple datasets at the same site, applied to reveal the correlation between genes in different DEGs of multiple datasets were overlapped as the final signif- samples [10–12]. However, the candidate biomarkers for icant DEGs for downstream analysis. The high-throughput clinical gene therapy of PD are unclear. In this study, the sequencing data were utilized for DEGs between the PD microarray and RNA-seq datasets from GEO were used to group and the control group by the edgeR package [14]. identify DEGs between multiple brain regions of PD and The screening threshold for a significant difference in gene p ∣ normal samples. Then, through WGCNA, PD-related key expression was adjusted value < 0.05 and log2fold change modules were constructed. Further functional enrichment ðFCÞ ∣ >0:585. analysis was carried out to evaluate the potential functions of genes in key modules. 2.3. Weighted Gene Coexpression Network Analysis (WGCNA). In this study, the WGCNA package was used to realize WGCNA [15]. Through the goodGeneSample func- 2. Materials and Methods tion, a hierarchical clustering tree was constructed for all samples and outliers of which node height was significantly 2.1. Data Collection and Preprocessing. Expression profiles of higher than other samples were removed. The gene coexpres- PD (GSE20292, GSE7621, GSE20291, GSE20168, GSE68719, sion similarity matrix was composed of the absolute value of ffi and GSE110716) were downloaded from the Gene Expres- the Pearson correlation coe cient between two genes. The S ½S Š i sion Omnibus (GEO; http://www.ncbi.nlm.nih.gov/geo/) correlation matrix was constructed as follows: = i,j ( database (Table 1). The GSE20291 dataset included 11 PD and j indicate the ith and jth gene). Soft threshold β value substantia nigra samples and 18 normal samples on the was then calculated according to the following formula: ai j ff , GPL96 (A ymetrix U133A Array) plat- ðS βÞ jS jβ a = power i,j, = i,j ( i,j indicates the adjacency function form. The GSE7621 dataset was composed of 16 PD substan- ith jth tia nigra samples and 9 control samples on the GPL570 between the and genes). To follow the principle of non- R2 : (Affymetrix Human Genome U133 Plus 2.0 Array) platform. scale network, >08 was set. After determining the soft β The GSE20291 dataset covered 15 PD putamen samples and threshold through the pickSoftThreshold function, the cor- S ½S Š 20 control samples on the GPL96 (Affymetrix Human relation matrix = i,j was converted into adjacency matrix A ½A Š Genome U133A Array) platform. The GSE20168 dataset = i,j by the pickSoftThreshold function. Topological included 14 prefrontal cortex PD samples and 15 normal overlap measure (TOM) was performed to calculate the samples on the platform of GPL96 (Affymetrix Human degree of association between genes as follows: TOMIJ = Genome U133A Array). There were 29 prefrontal cortex ð∑uaiuauj + aijÞ/ðmin ðki, kjÞ +1− aijÞ (aij is ½0, 1Š). Gene samples and 44 normal samples in the GSE68719 dataset modules were divided based on the high topological over- on the GPL11154 (Illumina HiSeq 2000 (Homo sapiens)) lap between genes in the modules. The dynamic cutting platform. The GSE110716 dataset was composed of 8 cingu- tree algorithm was used to calculate gene modules. late gyrus PD samples and 8 normal samples on the platform of GPL15433 (Illumina HiSeq 1000 (Homo sapiens)). Raw 2.4. Protein-Protein Interaction (PPI) Network. PPI of the tar- data was standardized by log2 conversion. Principal compo- get gene list was analyzed using the STRING (https://string- nent analysis (PCA) was presented to detect and remove out- db.org/) online database [16]. The confidence of protein liers and to find samples with high similarity. Furthermore, interaction was set as combined score > 0:4. Then, the Cytos- the correlation of levels between samples cape software was utilized to visualize the PPI network [17]. was analyzed. By the cytoHubba plug-in [16], the degree of connectivity BioMed Research International 3

PCA

GSM606624 GSM606625

1.0 GSM606626

GSM508735GSM508729 GSM508728GSM508733GSM508731 GSM521253 0.8 GSM508730 GSM508720GSM508722GSM508715 GSM508725 0.7 GSM508708GSM508713 GSM508724GSM508718GSM508712GSM508717 GSM508726GSM508721 0.6 0.6 GSM508710GSM508711 GSM508714 0.5 GSM508723GSM508734GSM508716 PC3 (19.8%) 0.4 0.4 GSM508732 0.3

0.2 PC2 (21.72%) 0.2 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 PC1 (58.48%) Group Control PD (a) PCA

GSM184358 0.8 GSM184359GSM184354 0.7 GSM184378 GSM184361GSM184355GSM184371 GSM184372GSM184356GSM184373 0.6 GSM184362GSM184366GSM184365 GSM184370 GSM184357GSM184369GSM184368GSM184363 0.5 GSM184360 GSM184374GSM184377 0.4 GSM184375GSM184367 1.0 0.9

PC3 (20.66 % ) GSM184364 0.8 GSM184376 0.3 0.7 0.6 0.2 0.5 0.4 PC2 (35.33%) 0.1 0.3 0.3 0.4 0.5 0.6 0.7 0.8 0.9 PC1 (44%)

Group Control PD (b)

Figure 1: Continued. 4 BioMed Research International

PCA

1.0 GSM606623GSM606620GSM606621 GSM606622GSM606619 0.9

0.8

0.7 GSM508696GSM508702GSM508701GSM508699 GSM508695GSM508704 0.6 GSM508697GSM508690GSM508691GSM508705GSM508693GSM508609GSM508686GSM508594GSM508684 GSM508700GSM508610GSM508687 GSM508689GSM508688GSM508608GSM508661 0.9 GSM508692GSM508706GSM508683GSM508627 0.8

PC3 (22.09%) 0.5 GSM508611GSM508629 0.7 0.6 0.4 GSM508628 GSM508703 0.5 0.4 0.3 0.3 0.2 PC2 (29.43%) 0.2 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 PC1 (48.49%) Group Control PD (c) PCA

GSM506059GSM506063 GSM506062 0.40 GSM506061 GSM506060 GSM506068GSM506066 0.35 GSM506042 GSM506041GSM506064 0.30 GSM506040 GSM506055GSM506054 0.25 GSM506039GSM506043

0.20 GSM506053GSM506048 GSM506044GSM506045GSM506065GSM506056GSM506047 GSM506046

PC3 (3.94%) 1.0 0.15 GSM506037GSM506050 GSM506051 0.9 GSM506067GSM506057GSM506052 0.8 0.10 0.7 0.6 0.05 0.5 0.4 PC2 (44.13%) 0.00 0.3 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 PC1 (51.93%) Group Control PD (d)

Figure 1: Continued. BioMed Research International 5

PCA

GSM1679683 GSM1679702 0.9 GSM1679720GSM1679710GSM1679658 GSM1679657GSM1679715GSM1679701GSM1679662 0.8 GSM1679711GSM1679679GSM1679709 GSM1679653GSM1679659 GSM1679651GSM1679700GSM1679664GSM1679666 0.7 GSM1679654GSM1679698GSM1679695 GSM1679678GSM1679682 GSM1679673GSM1679690GSM1679706GSM1679707GSM1679719GSM1679650 GSM1679674GSM1679676GSM1679694GSM1679714GSM1679718GSM1679670GSM1679708 GSM1679689GSM1679704GSM1679703GSM1679649GSM1679672GSM1679652GSM1679712GSM1679661GSM1679677 0.6 GSM1679684GSM1679691GSM1679688GSM1679660GSM1679669GSM1679671GSM1679696GSM1679655 GSM1679716GSM1679681GSM1679717 GSM1679687GSM1679663GSM1679675GSM1679665 GSM1679648 0.5 GSM1679697GSM1679680GSM1679699GSM1679713 0.9 GSM1679685GSM1679705GSM1679692GSM1679668GSM1679667GSM1679656GSM1679686

PC3 (22.24%) 0.8 GSM1679693 0.7 0.4 0.6 0.5 0.3 0.4 0.3 PC2 (37.43%) 0.2 0.2 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 PC1 (40.34%)

Group Control PD (e) PCA

GSM3014964

0.25 GSM3014965

0.20

0.15 GSM3014973 0.10 GSM3014972GSM3014979 GSM3014967 GSM3014969 0.05 GSM3014970GSM3014966GSM3014978GSM3014976GSM3014977 GSM3014968GSM3014975 0.00 GSM3014971GSM3014974 PC3 (0.65%) −0.05

− 0.9 0.10 0.8 0.7 −0.15 0.6 0.5 − 0.4 0.20 PC2 (49.12%) 0.4 0.5 0.6 0.7 0.8 0.9 1.0 PC1 (50.23%)

Group Control PD (f)

Figure 1: Principal component analysis for multiple brain regions of PD samples and normal samples: (a, b) substantia nigra (GSE20292 and GSE7621); (c) putamen (GSE20291); (d, e) prefrontal cortex area (GSE20168 and GSE68719); (f) cingulate gyrus (GSE110716). There are three principal components (PC1, PC2, and PC3). Blue indicates control normal samples, and green indicates PD samples. 6 BioMed Research International

GSM606626 1.0 GSM606625 GSM606624 GSM521253 GSM508735 GSM508734 0.9 GSM508733 GSM508732 GSM508731 GSM508730 GSM508729 GSM508728 0.8 GSM508726 GSM508725 GSM508724 GSM508723 GSM508722 0.7 GSM508721 GSM508720 GSM508718 GSM508717 GSM508716 GSM508715 0.6 GSM508714 GSM508713 GSM508712 GSM508711 GSM508710 0.5 GSM508708 GSM508708 GSM508710 GSM508711 GSM508712 GSM508713 GSM508714 GSM508715 GSM508716 GSM508717 GSM508718 GSM508720 GSM508721 GSM508722 GSM508723 GSM508724 GSM508725 GSM508726 GSM508728 GSM508729 GSM508730 GSM508731 GSM508732 GSM508733 GSM508734 GSM508735 GSM521253 GSM606624 GSM606625 GSM606626 (a)

Figure 2: Continued. BioMed Research International 7

GSM184378 1.00 GSM184377 GSM184376 GSM184375 GSM184374 GSM184373 GSM184372 GSM184371 0.95 GSM184370 GSM184369 GSM184368 GSM184367 GSM184366 GSM184365 0.90 GSM184364 GSM184363 GSM184362 GSM184361 GSM184360 GSM184359 GSM184358 0.85 GSM184357 GSM184356 GSM184355 GSM184354 GSM184354 GSM184355 GSM184356 GSM184357 GSM184358 GSM184359 GSM184360 GSM184361 GSM184362 GSM184363 GSM184364 GSM184365 GSM184366 GSM184367 GSM184368 GSM184369 GSM184370 GSM184371 GSM184372 GSM184373 GSM184374 GSM184375 GSM184376 GSM184377 GSM184378 (b)

Figure 2: Continued. 8 BioMed Research International

GSM606623 1.0 GSM606622 GSM606621 GSM606620 GSM606619 GSM508706 GSM508705 GSM508704 0.9 GSM508703 GSM508702 GSM508701 GSM508700 GSM508699 GSM508697 0.8 GSM508696 GSM508695 GSM508693 GSM508692 GSM508691 GSM508690 0.7 GSM508689 GSM508688 GSM508687 GSM508686 GSM508684 GSM508683 0.6 GSM508661 GSM508629 GSM508628 GSM508627 GSM508611 GSM508610 0.5 GSM508609 GSM508608 GSM508594 GSM508594 GSM508608 GSM508609 GSM508610 GSM508611 GSM508627 GSM508628 GSM508629 GSM508661 GSM508683 GSM508684 GSM508686 GSM508687 GSM508688 GSM508689 GSM508690 GSM508691 GSM508692 GSM508693 GSM508695 GSM508696 GSM508697 GSM508699 GSM508700 GSM508701 GSM508702 GSM508703 GSM508704 GSM508705 GSM508706 GSM606619 GSM606620 GSM606621 GSM606622 GSM606623 (c)

Figure 2: Continued. BioMed Research International 9

GSM506068 1.00 GSM506067 GSM506066 GSM506065 GSM506064 GSM506063 GSM506062 0.96 GSM506061 GSM506060 GSM506059 GSM506057 GSM506056 GSM506055 GSM506054 0.92 GSM506053 GSM506052 GSM506051 GSM506050 GSM506048 GSM506047 0.88 GSM506046 GSM506045 GSM506044 GSM506043 GSM506042 GSM506041 GSM506040 0.84 GSM506039 GSM506037 GSM506037 GSM506039 GSM506040 GSM506041 GSM506042 GSM506043 GSM506044 GSM506045 GSM506046 GSM506047 GSM506048 GSM506050 GSM506051 GSM506052 GSM506053 GSM506054 GSM506055 GSM506056 GSM506057 GSM506059 GSM506060 GSM506061 GSM506062 GSM506063 GSM506064 GSM506065 GSM506066 GSM506067 GSM506068 (d)

Figure 2: Continued. 10 BioMed Research International

GSM1679720 GSM1679719 1.0 GSM1679718 GSM1679717 GSM1679716 GSM1679715 GSM1679714 GSM1679713 GSM1679712 GSM1679694 GSM1679711 GSM1679710 GSM1679709 GSM1679708 GSM1679707 GSM1679706 GSM1679705 GSM1679704 GSM1679703 GSM1679702 GSM1679701 GSM1679700 GSM1679699 0.8 GSM1679698 GSM1679697 GSM1679696 GSM1679695 GSM1679693 GSM1679692 GSM1679691 GSM1679690 GSM1679689 GSM1679688 GSM1679687 GSM1679686 GSM1679685 GSM1679684 GSM1679683 GSM1679682 GSM1679681 GSM1679680 GSM1679679 GSM1679678 GSM1679677 0.6 GSM1679676 GSM1679675 GSM1679674 GSM1679673 GSM1679672 GSM1679671 GSM1679670 GSM1679669 GSM1679668 GSM1679667 GSM1679666 GSM1679665 GSM1679664 GSM1679663 GSM1679662 GSM1679661 GSM1679660 GSM1679659 GSM1679658 GSM1679657 GSM1679656 0.4 GSM1679655 GSM1679654 GSM1679653 GSM1679652 GSM1679651 GSM1679650 GSM1679649 GSM1679648 GSM1679714 GSM1679716 GSM1679717 GSM1679718 GSM1679719 GSM1679720 GSM1679694 GSM1679712 GSM1679713 GSM1679715 GSM1679697 GSM1679698 GSM1679699 GSM1679700 GSM1679701 GSM1679702 GSM1679703 GSM1679704 GSM1679705 GSM1679706 GSM1679707 GSM1679708 GSM1679709 GSM1679710 GSM1679711 GSM1679691 GSM1679693 GSM1679695 GSM1679696 GSM1679648 GSM1679649 GSM1679650 GSM1679651 GSM1679652 GSM1679653 GSM1679654 GSM1679655 GSM1679656 GSM1679657 GSM1679658 GSM1679659 GSM1679660 GSM1679661 GSM1679662 GSM1679663 GSM1679664 GSM1679665 GSM1679666 GSM1679667 GSM1679668 GSM1679669 GSM1679670 GSM1679671 GSM1679672 GSM1679673 GSM1679674 GSM1679675 GSM1679676 GSM1679677 GSM1679678 GSM1679679 GSM1679680 GSM1679681 GSM1679682 GSM1679683 GSM1679684 GSM1679685 GSM1679686 GSM1679687 GSM1679688 GSM1679689 GSM1679690 GSM1679692 (e)

Figure 2: Continued. BioMed Research International 11

GSM3014979 1.0 GSM3014978 GSM3014977 GSM3014976 GSM3014975 GSM3014974 0.9 GSM3014973 GSM3014972 GSM3014971 GSM3014970 GSM3014969 0.8 GSM3014968 GSM3014967 GSM3014966 GSM3014965 GSM3014964 GSM3014964 GSM3014965 GSM3014966 GSM3014967 GSM3014968 GSM3014969 GSM3014970 GSM3014971 GSM3014972 GSM3014973 GSM3014974 GSM3014975 GSM3014976 GSM3014977 GSM3014978 GSM3014979 (f)

Figure 2: Heat maps depicting sample correlation analysis between PD and normal samples: (a, b) substantia nigra (GSE20292 and GSE7621); (c) putamen (GSE20291); (d, e) prefrontal cortex area (GSE20168 and GSE68719); (f) cingulate gyrus (GSE110716). The correlation coefficient indicates the similarity between samples. The closer the correlation coefficient is to 1, the higher the similarity between the two samples.

of the node was calculated and hub genes in the PPI network from multiple brain regions of PD, including the substantia were determined [18]. nigra (GSE20292 and GSE7621), putamen (GSE20291), pre- frontal cortex area (GSE20168 and GSE68719), and cingulate 2.5. Functional Enrichment Analysis. (GO) gyrus (GSE110716). Before downstream analysis, all samples enrichment analysis including biological process, cellular were assessed by PCA. The results showed that PD substantia component, and molecular function was carried out through nigra samples (Figures 1(a) and 1(b)), putamen (Figure 1(c)), the Gene Ontology database [19]. Moreover, Kyoto Encyclo- prefrontal cortex area (Figures 1(d) and 1(e)), and cingulate pedia of Genes and Genomes (KEGG) pathway enrichment gyrus (Figure 1(f)) were distinctly distinguished from normal analysis was also presented by KEGG PATHWAY DATA- samples. Furthermore, based on these gene expression data, BASE [20]. Fisher’s exact test was used to find out which we calculated the correlation coefficients between the two items or pathways were significantly related to a set of genes. samples. There was a significant high correlation between p value < 0.05 indicated significant enrichment. different samples for PD substantia nigra samples (Figures 2(a) and 2(b)), putamen (Figure 2(c)), prefrontal 2.6. Gene Set Enrichment Analyses (GSEA). The clusterProfi- cortex area (Figures 2(d) and 2(e)), cingulate gyrus ler package [21] was used to perform GSEA on the transcrip- (Figure 2(f)), and corresponding normal samples. tional data of multiple brain regions in PD from the GSE20295 dataset [17, 22]. Using the gene set file wikipath- 3.2. Differentially Expressed Genes for Multiple Brain Regions ways-20180810-gmt-Homo_sapiens.gmt from the cluster- of PD. With the screening criteria of p value < 0.05 and ∣ fi ∣ : Pro ler package, GSEA was presented based on the default log2FC >0 585, DEGs between multiple brain regions of parameters. PD samples and normal samples were identified. In the GSE20292 dataset, there were 191 upregulated and 369 3. Results downregulated genes between the substantia nigra of PD and normal samples (Figure 3(a)). 530 upregulated and 590 3.1. Principal Component Analysis for Multiple Brain Regions downregulated genes were screened for the substantia nigra of PD Samples. In this study, we obtained expression profiles of PD samples compared to normal samples in the 12 BioMed Research International

PD vs control

PACS2 6

LOC100272216 ZBTB16

4 SV2C SLC18A2 MAN2A1 TMEM255A

( p _value) FGF13 10

−log NEFL SLCO4A1

2

0

−2 −1 0 1 2

log2(fold change)

Up (191) Middle (12677) Down (369) (a) PD vs control

6

SLC18A2 AXINI

LOC100996756 4 SV2C DDC RSPO2 TH FZD9 ( p _value) 10

−log NEFL DDX3Y 2

XIST KDM5D

0

−2 0 2 log2(fold change) Up (530) Middle (21706) Down (590) (b)

Figure 3: Continued. BioMed Research International 13

PD vs control

4 DDIT4

CALB1 HSPB1

3

COCH GPR6 PENK MT1M

( p _value) TAC1 SLCO4A1 10 2

−log UNC119

1

0

−2 −1 0 1 2 log2(fold change)

Up (317) Middle (12603) Down (317) (c) PD vs control

6 OXR1

LPPR4

4 ATP8B1 SYT1 MAN2A1 ( p _value) TAC1 10 ITPKBMKNK2 −log VSNL1

HSPB1 2

0

−2 −1 0 1 2 log2(fold change) Up (90) Middle (12820) Down (327) (d)

Figure 3: Continued. 14 BioMed Research International

PD vs control

7.5

CDKL5 HSPA6 5.0 NXPH2KRT5 CRH CXCL1 ( p _value) 10 NPAS4 STC1 −log

SERPINA3 2.5 IL1RL1

0.0

−2 −1 0 1 2

log2(fold change) Up (1271) Middle (15217) Down (1030) (e)

Figure 3: Continued. BioMed Research International 15

PD vs control

PRLR

RTP5 6

ONECUT2 4 ADAMTS18 FAM90A12 ( p _value) 10 HLA−DRB6 −log

WFIKKN2 2 AGAP7P LOC100996723 MT1H

0

−5.0 −2.5 0.0 2.5 5.0

log2(fold change) Up (116) Middle (14534) Down (292) (f)

Figure 3: Volcano plots depicting differentially expressed genes between multiple brain regions of PD and corresponding normal tissues: (a, b) substantia nigra (GSE20292 and GSE7621); (c) putamen (GSE20291); (d, e) prefrontal cortex area (GSE20168 and GSE68719); (f) cingulate gyrus (GSE110716). Red suggests upregulation, and blue suggests downregulation. The top five most significant upregulated genes and downregulated genes were labeled separately.

GSE7621 dataset (Figure 3(b)). After overlapping the results 3.3. Construction of WGCNA for the Substantia Nigra of PD. from the two datasets, 17 upregulated and 52 downregulated 11 substantia nigra PD and 18 control samples were used genes were identified for the substantia nigra of PD. In the for coexpression analysis in the GSE20292 dataset. Coex- GSE20291 dataset, 317 upregulated and 317 downregulated pression module analysis was easily affected by outlier genes were identified between PD putamen and normal tis- samples, so removing outlier sample data before construct- sues (Figure 3(c)). Following the intersections of DEGs from ing the network was especially important for obtaining the GSE20168 dataset (90 upregulated and 327 downregu- meaningful analysis results. Herein, there were no outliers lated genes; Figure 3(d)) and the GSE68719 dataset (1271 among them (Figure 5(a)). Thus, no samples were upregulated and 1030 downregulated genes; Figure 3(e)), 39 removed. By dynamic cutting tree method, gene modules upregulated and 72 downregulated genes were identified for were divided, and highly similar modules were merged PD prefrontal cortex area tissues in comparison to normal (Figure 5(b)). Finally, nine modules were constructed. tissues. In the GSE110716 dataset, there were 116 upregu- Among them, the purple module was significantly related lated and 292 downregulated genes between PD cingulate to PD (r = −0:44 and p =0:02) (Figure 5(c)). Heat maps gyrus and normal tissues (Figure 3(f)). Heat maps depicted showed that there was a high correlation between different that these DEGs could significantly distinguish PD substantia genes (Figure 5(d)). We further assessed the coexpression nigra samples (Figures 4(a) and 4(b)), putamen (Figure 4(c)), similarity of modules. These modules were divided into prefrontal cortex area (Figures 4(d) and 4(e)), and cingulate two main clusters, which were validated by adjacency heat gyrus (Figure 4(f)) from the corresponding normal samples. maps (Figure 5(e)). 16 Group GSM508710 PD Control GSM508711 GSM508712 GSM508713 GSM508714 GSM508715 GSM508716 GSM508718 GSM508728 GSM508731 GSM508732

Figure GSM508708 GSM508717 GSM508720 :Continued. 4:

(a) GSM508721 GSM508722 GSM508723 GSM508724 GSM508725 GSM508726 GSM508729 GSM508730 GSM508733 GSM508734

GSM508735 International Research BioMed GSM521253 GSM606624 GSM606625 GSM606626 −4 −2 0 2 4 iMdRsac International Research BioMed Group GSM184363 PD Control GSM184364 GSM184365 GSM184366 GSM184367 GSM184368 GSM184369 GSM184370 GSM184371 GSM184372 Figure GSM184373 GSM184374 :Continued. 4:

(b) GSM184375 GSM184376 GSM184377 GSM184378 GSM184354 GSM184355 GSM184356 GSM184357 GSM184358 GSM184359 GSM184360 GSM184361 GSM184362 −4 −2 0 2 4 17 18 Group GSM508608 PD Control GSM508609 GSM508610 GSM508611 GSM508627 GSM508628 GSM508629 GSM508661 GSM508684 GSM508695 GSM508697 GSM508700 GSM508701

Figure GSM508704 GSM508706 GSM508594

:Continued. 4: GSM508683 (c) GSM508686 GSM508687 GSM508688 GSM508689 GSM508690 GSM508691 GSM508692 GSM508693 GSM508696 GSM508699 GSM508702 GSM508703 GSM508705 iMdRsac International Research BioMed GSM606619 GSM606620 GSM606621 GSM606622 GSM606623 −4 −2 0 2 4 iMdRsac International Research BioMed Group GSM506039 PD Control GSM506040 GSM506041 GSM506042 GSM506043 GSM506044 GSM506045 GSM506046 GSM506048 GSM506059 GSM506062

Figure GSM506063 GSM506066

:Continued. 4: GSM506068 (d) GSM506037 GSM506047 GSM506050 GSM506051 GSM506052 GSM506053 GSM506054 GSM506055 GSM506056 GSM506057 GSM506060 GSM506061 GSM506064 GSM506065 GSM506067 −4 −2 0 2 4 19 20 Group GSM1679692 GSM1679693

PD Control GSM1679694 GSM1679695 GSM1679696 GSM1679697 GSM1679698 GSM1679699 GSM1679700 GSM1679701 GSM1679702 GSM1679703 GSM1679704 GSM1679705 GSM1679706 GSM1679707 GSM1679708 GSM1679709 GSM1679710 GSM1679711 GSM1679712 GSM1679713 GSM1679714 GSM1679715 GSM1679716 GSM1679717 GSM1679718 GSM1679719 Figure GSM1679720 GSM1679648 GSM1679649 GSM1679650 GSM1679651 GSM1679652 :Continued. 4: GSM1679653 (e) GSM1679654 GSM1679655 GSM1679656 GSM1679657 GSM1679658 GSM1679659 GSM1679660 GSM1679661 GSM1679662 GSM1679663 GSM1679664 GSM1679665 GSM1679666 GSM1679667 GSM1679668 GSM1679669 GSM1679670 GSM1679671 GSM1679672 GSM1679673 GSM1679674 GSM1679675 GSM1679676 GSM1679677 GSM1679678 GSM1679679 GSM1679680 GSM1679681 International Research BioMed GSM1679682 GSM1679683 GSM1679684 GSM1679685 GSM1679686 GSM1679687 GSM1679688 GSM1679689 GSM1679690 GSM1679691 −6 −4 −2 0 2 4 6 iMdRsac International Research BioMed usatanga(S222adGE61;()ptmn(S221;(,e rfotlcre ra(S218adGE81) (f) GSE68719); and (GSE20168 area cortex prefrontal e) (d, downregulation. (GSE20291); suggests blue putamen and upregulation, (c) suggests GSE7621); Red (GSE110716). and gyrus cingulate (GSE20292 nigra substantia Figure :Ha assoigepeso atrso di of patterns expression showing maps Heat 4: Group

PD Control GSM3014972

GSM3014973

GSM3014974

GSM3014975

GSM3014976

GSM3014977 ff rnilyepesdgnsbtenP n orsodn omltsus a b) (a, tissues: normal corresponding and PD between genes expressed erentially GSM3014978

GSM3014979 (f)

GSM3014964

GSM3014965

GSM3014966

GSM3014967

GSM3014968

GSM3014969

GSM3014970

GSM3014971 −3 −2 −1 0 1 2 3 21 22

Height 0.5 0.6 0.7 0.8 0.9 Module Merged module

Height 10000 12000 2000 4000 6000 8000 Group

GSM508732 GSM606626 GSM606624 GSM606625 GSM508728 GSM508730 GSM508729 Figure Gene dendrogramandmodulecolors GSM508715 Sample dendrogramandtraitheatmap GSM508731 GSM508723

:Continued. 5: GSM508726 (b) (a) GSM508724 GSM508725 GSM508721 GSM508720 GSM508712 GSM508718 GSM508733 GSM508735 GSM521253 GSM508708 GSM508722 GSM508714 GSM508717 GSM508734 GSM508711 International Research BioMed GSM508710 GSM508713 GSM508716 BioMed Research International 23

Module−trait relationships Network heatmap plot, all genes 1 0.32 MEtan (0.1)

0.28 MEblue (0.1)

0.35 0.5 MEpink (0.06)

0.096 MEblack (0.6)

−0.44 MEpurple (0.02) 0

−0.19 MEred (0.3)

−0.14 MEsalmon (0.5) −0.5

−0.19 MEbrown (0.3)

−0.33 MEturquoise (0.08) −1 Group

(c) (d) 1.2 1.0 0.8 0.6 0.4 MEblack Group 0.2 MEtan MEpink MEblue MEsalmon MEred MEbrown MEpurple MEturquoise 1

0.8

0.6 Group

0.4

0.2

0

Group (e)

Figure 5: Construction of WGCNA for the substantia nigra of PD. (a) Sample hierarchical clustering tree to detect outliers. (b) Dynamic cutting tree method was utilized to determine gene modules. (c) Module-trait relationship network. The color of the square indicates the correlation between the module and the clinical traits. The p value is in brackets. (d) Hierarchical clustering dendrogram. The branches correspond to each module. The module memberships colored by different colors are shown in the color bar below and to the right of the tree diagram. Shades of color are proportional to coexpression interconnectedness. (e) Clustering of module eigengenes and eigengene adjacency heat map. Red represents high correlation and blue represents low correlation.

3.4. Construction of WGCNA for the Substantia Nigra of PD. Finally, fifteen coexpression modules with high similarity In the GSE20291 dataset, 15 putamen PD and 20 normal were merged. Among them, the blue module was distinctly samples were utilized for WGCNA. No outliers were detected correlated to PD (r = −0:37 and p =0:03) (Figure 6(c)). Heat and removed among them (Figure 6(a)). Using dynamic cut- maps showed that there was a high correlation between ting tree method, gene modules were built (Figure 6(b)). different genes in different modules (Figure 6(d)). The 24

Height 0.5 0.6 0.7 0.8 0.9 Module Merged module

Height Group 4000 6000 8000 10000 14000 18000

GSM606623 GSM508699 GSM508702 GSM508696 GSM508701 GSM508695 GSM508697 GSM508700 GSM606620 GSM606621 Sample dendrogramandtraitheatmap Figure Gene dendrogramandmodulecolors GSM606619 GSM606622

hclust (*,"average") GSM508611

:Continued. 6: GSM508628 (b) (a) GSM508703 dist(expt) GSM508692 GSM508689 GSM508690 GSM508691 GSM508688 GSM508687 GSM508684 GSM508686 GSM508594 GSM508609 GSM508704 GSM508705 GSM508610 GSM508661 GSM508627 GSM508683

GSM508629 International Research BioMed GSM508706 GSM508608 GSM508693 BioMed Research International 25

Module−trait relationships 1 0.023 MEdarkgrey (0.9) −0.1 MEorange (0.5) 0.27 MEgreen (0.1) MElightyellow 0.17 (0.3) 0.5 −0.18 MEsteelblue (0.3) −0.12 MElightcyan (0.5) −0.28 MElightgreen (0.1) −0.37 MEblue (0.03) 0 −0.14 MEdarkturquoise (0.4) 0.079 MEsalmon (0.7) −0.24 MEmagenta (0.2) −0.31 −0.5 MEmidnightblue (0.07) −0.25 MEdarkred (0.2) 0.2 MEpaleturquoise (0.3) 0.27 MEviolet (0.1) −1 Group (c) Network heatmap plot, all genes

(d)

Figure 6: Continued. 26 BioMed Research International

1.0

0.6 Group

0.2 MEdarkred MEblue MEgreen MEviolet MEorange MEsalmon MElightyellow MEsteelblue MEdarkgrey MEdarkturquoise MElightcyan MEmagenta MElightgreen MEpaleturquoise MEmidnightblue 1

0.8

Group 0.6

0.4

0.2

0

Group (e)

Figure 6: Construction of WGCNA for the putamen of PD. (a) Sample hierarchical clustering tree to detect outliers. (b) Gene modules were determined by dynamic cutting tree method. (c) Module-trait relationship network. (d) Hierarchical clustering dendrogram. (e) Clustering of module eigengenes and eigengene adjacency heat map. coexpression similarity of modules was analyzed, as shown in tive correlations to PD. The medium purple (r = −0:65 and Figure 6(e). p =0:006) and salmon (r = −0:55 and p =0:03) modules had negative correlations to PD in Figure 8(c). In the network 3.5. Construction of WGCNA for the Prefrontal Cortex of PD. heat map plot, each module was independent of others 14 prefrontal cortex PD and 15 normal samples were ana- (Figure 8(d)). Moreover, their coexpression similarity was lyzed by WGCNA in the GSE20168 dataset. There was no evaluated by adjacency heat maps (Figure 8(e)). outlier sample among them (Figure 7(a)). Gene modules were divided using dynamic cutting tree method (Figure 7(b)). After merging, 25 modules were constructed. 3.7. PPI Networks for DEGs in Multiple Brain Regions of PD. Among them, the green (r = −0:43 and p =0:02), magenta DEGs for the substantia nigra, putamen, prefrontal cortex (r = −0:52 and p =0:004), and bisque (r = −0:54 and p = area, and cingulate gyrus of PD were extracted for PPI net- 0:002) modules were negatively correlated to PD works by the STRING database. There were 69 nodes in the (Figure 7(c)). Also, salmon (r =0:42 and p =0:02) and dark PPI network of substantia nigra PD, including 17 upregu- orange (r =0:49 and p =0:006) modules were positively lated and 52 downregulated genes (Figure 9(a)). Among related to PD. According to the network heat map plot, each them, SLC6A3 (degree = 6), SLC18A2 (degree = 6), and TH module was independent of others (Figure 7(d)). Further- (degree = 6) had the highest degree, which were considered more, their coexpression similarity was quantified by adja- as hub genes. In Figure 9(b), there were 317 upregulated cency heat maps (Figure 7(e)). genes and 317 downregulated genes in the PPI network of putamen. Among them, BMP4 (degree = 14) and SNAP25 3.6. Construction of WGCNA for the Cingulate Gyrus of PD. (degree = 13) were two hub genes. As shown in Figure 9(c), From the GSE110716 dataset, 8 cingulate gyrus PD and 8 there were 111 nodes in the PPI network of the prefrontal control samples were obtained for WGCNA. No outlier sam- cortex area, including 39 upregulated and 72 downregulated ples were detected among them (Figure 8(a)). Gene modules genes. SNAP25 was identified as a hub gene (degree = 26). were divided via dynamic cutting tree method (Figure 8(b)). There were 408 nodes in the PPI network of the cingulate Following merging, 40 coexpression modules were con- gyrus, composed of 116 upregulated and 292 downregulated structed. Among them, the orange red (r =0:65 and p = genes in Figure 9(d). CTGF (degree = 3), CDH1 (degree = 3), 0:006) and thistle (r =0:51 and p =0:04) modules had posi- and COL5A1 (degree = 3) were considered as hub genes. iMdRsac International Research BioMed

Height 0.65 0.70 0.75 0.80 0.85 0.90 0.95 Module Merged module

Height Group 20 30 40 50 60 70

GSM506060 GSM506061 GSM506063 GSM506059 GSM506062 GSM506042 GSM506041 GSM506068 Sample dendrogramandtraitheatmap

Figure GSM506066 Gene dendrogramandmodulecolors GSM506064 GSM506057

:Continued. 7: GSM506067 (b) (a) GSM506037 GSM506050 GSM506052 GSM506051 GSM506040 GSM506043 GSM506048 GSM506053 GSM506056 GSM506054 GSM506055 GSM506039 GSM506044 GSM506065 GSM506045 GSM506046 GSM506047 27 28 BioMed Research International

Module−trait relationships

−0.43 1 MEgreen (0.02) −0.52 MEmagenta (0.004) −0.53 MEviolet (0.003) −0.27 MEdarkmagenta (0.2) −0.073 MEplum2 (0.7) −0.54 MEbisque4 (0.002) −0.33 0.5 MElightcyan1 (0.08) −0.26 MEdarkorange2 (0.2) 0.0037 MElightsteelblue1 (1) 0.098 MEforalwhite (0.6) 0.16 MEblack (0.4) 0.42 MEsalmon4 (0.02) 0.29 MEdarkslateblue (0.1) 0 0.079 MEthistle1 (0.7) 0.29 MEpalevioletred3 (0.1) 0.27 MEplum1 (0.2) 0.26 MEsteelblue (0.2) 0.25 MEblue (0.2) 0.25 MEbrown (0.2) −0.5 0.49 MEdarkorange (0.006) 0.34 MEgreenyellow (0.07) −0.14 MEorangered4 (0.5) −0.015 MEwhite (0.9) 0.19 MEbrown4 (0.3) MEcyan −0.26 (0.2) −1 Group (c) Network heatmap plot, all genes

(d)

Figure 7: Continued. BioMed Research International 29 Group 0.2 0.6 1.0 MEwhite MEblack MEforalwhite MEcyan MEblue MEgreen MEorangered4 MEviolet MEplum2 MEplum1 MEbrown MEthistle1 MEbrown4 MEsalmon4 MEbisque4 MEsteelblue MEmagenta MEpalevioletred3 MEdarkorange2 MElightcyan1 MEdarkorange MElightsteelblue1 MEgreenyellow MEdarkmagenta MEdarkslateblue 1

0.8

0.6

0.4

0.2 Group

0

Group (e)

Figure 7: Construction of WGCNA for the prefrontal cortex of PD. (a) Sample hierarchical clustering tree to detect outliers. (b) Dynamic cutting tree method was utilized to determine gene modules. (c) Module-trait relationship network. (d) Hierarchical clustering dendrogram. The branches correspond to each module. (e) Clustering of module eigengenes and eigengene adjacency heat map.

3.8. Functional Enrichment Analysis of DEGs in the and chloride transmembrane transporter activity Substantia Nigra of PD. DEGs for the substantia nigra, puta- (Figure 10(c)). KEGG enrichment analysis results revealed men, prefrontal cortex area, and cingulate gyrus of PD were that they were significantly related to a variety of PD- used for functional enrichment analysis. DEGs in the sub- related pathways such as cocaine addiction, dopaminergic stantia nigra of PD were mainly enriched in PD-related bio- synapse, amphetamine addiction, serotonergic synapse, logical processes such as aminergic neurotransmitter loading ECM-receptor interaction, alcoholism, tyrosine metabolism, into synaptic vesicle, neurotransmitter transport, chemical Parkinson’s disease, PPAR signaling pathway, and synaptic synaptic transmission, amine transport, neurotransmitter vesicle cycle (Figure 10(d)). loading into synaptic vesicle, dopamine biosynthetic process, phytoalexin metabolic process, cell-cell signaling, and isoqui- 3.9. Functional Enrichment Analysis of DEGs in the Putamen noline alkaloid metabolic process (Figure 10(a)). These DEGs of PD. GO enrichment analysis results showed that DEGs in were involved in various key cellular components including the putamen of PD were involved in the regulation of ossi- neuron projection, cell projection, axon, plasma membrane fication, cell population proliferation, cartilage develop- bounded cell projection, postsynaptic membrane, synaptic ment, kidney morphogenesis, mesonephros development, membrane, presynapse, dendrite, dendritic tree, and neuro- chondrocyte differentiation, detection of abiotic stimulus, nal cell body (Figure 10(b)). Also, they had several key molec- nephron morphogenesis, and cartilage development ular functions like monoamine transmembrane transporter (Figure 10(e)). They were significantly involved in integral activity, sodium : chloride symporter activity, dopamine component of plasma membrane, intrinsic component of binding, cytoskeletal adaptor activity, cation : chloride sym- plasma membrane, amino acid transport complex, endo- porter activity, neurotransmitter : sodium symporter activity, plasmic reticulum lumen, cell periphery, plasma mem- spectrin binding, ammonium transmembrane transporter brane, extracellular space, cell surface, cell leading edge, activity, protein serine/threonine kinase inhibitor activity, and cytoplasmic side of plasma membrane (Figure 10(f)). 30

Height 0.5 0.6 0.7 0.8 0.9 Module Merged module

Height 100000 150000 200000 250000 50000 Group 0

GSM3014964

GSM3014974

GSM3014967

GSM3014971 Figure Gene dendrogramandmodulecolors GSM3014979 Sample dendrogramandtraitheatmap

GSM3014965 :Continued. 8: (b) (a) GSM3014972

GSM3014977

GSM3014973

GSM3014970

GSM3014978

GSM3014966

GSM3014969

GSM3014975 iMdRsac International Research BioMed GSM3014968

GSM3014976 BioMed Research International 31

Module−trait relationships

−0.29 1 MElightyellow (0.3) 0.23 MEdarkturquoise (0.4) 0.23 MEdarkred (0.4) 0.43 MEdarkseagreen4 (0.1) 0.25 MElavenderblush3 (0.3) 0.45 MEdarkorange2 (0.08) 0.65 MEorangered4 (0.006) 0.24 MEgreenyellow (0.4) 0.51 MEthistle2 (0.04) 0.45 MEbrown (0.08) 0.26 0.5 MEgreen (0.3) 0.14 MEplum2 (0.6) 0.28 MEsalmon (0.3) −0.23 MEblue (0.4) 0.21 MEgrey60 (0.4) 0.041 MElightpink4 (0.9) −0.65 MEmediumpurple3 (0.006) −0.29 MEforalwhite (0.3) −0.039 MEhoneydew1 (0.9) −0.27 MEblack (0.3) −0.28 MEmidnightblue (0.3) 0 −0.086 MEda rkgrey (0.8) −0.37 MEdarkgreen (0.2) −0.3 MEorange (0.3) −0.43 MEpalevioletred3 (0.1) −0.55 MEsalmon4 (0.03) −0.058 MEbrown4 (0.8) −0.27 MEnavajowhite2 (0.3) −0.04 MEwhite (0.9) 0.25 MEmagenta (0.4) 0.13 MEsienna3 (0.6) −0.5 0.26 MEred (0.3) 0.061 MEdarkorange (0.8) −0.088 MElightcyan (0.7) −0.4 MEbisque4 (0.1) 0.23 MEcyan (0.4) 0.32 MEmaroon (0.2) −0.28 MElightgreen (0.3) 0.051 MElightsteelblue1 (0.9) −0.46 MEcoral1 (0.07) −0.27 MEpurple (0.3) −1 Group (c) Network heatmap plot, all genes

(d)

Figure 8: Continued. 32 BioMed Research International

1.0

0.6 MEblack MEblue MEmagenta MElightyellow MEdarkred 0.2 MEdarkturquoise Group MEmidnightblue MElightgreen MEred MEbisque4 MEthistle2 MEplum2 MEsienna3 MEcoral1 MEpurple MEcyan MEbrown4 MEsalmon MEgreen MEwhite MEbrown MEorange MEdarkgrey MEgrey60 MEmaroon MEforalwhite MEsalmon4 MEgreenyellow MEhoneydew1 MElightcyan MElightpink4 MEdarkgreen MElightsteelblue1 MEorangered4 MEmediumpurple3 MEdarkorange2 MEdarkorange MEdarkseagreen4 MElavenderblush3 MEnavajowhite2 MEpalevioletred3 1

0.8 Group

0.6

0.4

0.2

0

Group (e)

Figure 8: Construction of WGCNA for the cingulate gyrus of PD. (a) Sample hierarchical clustering tree to detect outliers. (b) Dynamic cutting tree method was utilized to determine gene modules. (c) Module-trait relationship network. The color of the square indicates the correlation between the module and the clinical traits. The p value is in brackets. (d) Hierarchical clustering dendrogram. The branches correspond to each module. The module memberships colored by different colors are shown in the color bar below and to the right of the tree diagram. Shades of color are proportional to coexpression interconnectedness. (e) Clustering of module eigengenes and eigengene adjacency heat map. Red represents high correlation and blue represents low correlation.

Also, they possessed a variety of molecular functions like 3.10. Functional Enrichment Analysis of DEGs in the heparin binding, cytokine binding, prostaglandin receptor Prefrontal Cortex of PD. GO enrichment analysis of DEGs activity, NAD-dependent histone deacetylase activity, cyto- in the prefrontal cortex of PD revealed that detoxification kine activity, phosphoric diester activity, cytokine of copper ion, stress response to metal ion, chemical synaptic receptor activity, actin-dependent ATPase activity, inor- transmission, cellular response to zinc ion, cellular response ganic anion exchanger activity, and protein membrane to copper ion, nervous system development, cellular zinc anchor (Figure 10(g)). In Figure 10(h), these DEGs partic- ion homeostasis, cell communication, zinc ion homeostasis, ipated in key signaling pathways like cytokine-cytokine and cellular response to cadmium ion were mainly enriched receptor interaction, basal cell carcinoma, hematopoietic in Figure 10(i). They could regulate various cellular compo- cell lineage, calcium signaling pathway, mTOR signaling nents like neuron projection, axon, synaptic membrane, pathway, aldosterone synthesis and secretion, viral protein plasma membrane bounded cell projection, cell projection, interaction with cytokine and cytokine receptor, signaling postsynapse, presynaptic membrane, presynapse, GABA- pathways regulating pluripotency of stem cells, NF-kappa ergic synapse, and cell body (Figure 10(j)). In Figure 10(k), B signaling pathway, and central carbon metabolism in they had a variety of molecular functions like calcium ion cancer. binding, adiponectin binding, structural constituent of BioMed Research International 33

SDC1 COL9A1 FGF13

GABRB1 RET

COBL TH

DDC NRIP3

KCNJ6

MAGED1 SLC18A2 SLC6A3 RGS4

ALDH1A1 TUBB2A SYNGR3 AGTR1 CUX2 CBLN1 KIAA0319

DYNC1I1 CHRM2 SV2C LMO3 RALYL ANK1

TRIM36 DLK1 ACTR10

RNF220 PEG10

(a)

AAAS PYGB

DDX31

CASC3 HP1BP3 SLC3A2 EBNA1BP2 SLC7A5

MGAT1 GPR52 SLCO4A1

PTBP1 ALG12 LAS1L CAMTA1 GPR88 SLC22A8

WNT10B TUSC3 AXIN1 CRYBB3 UNC119 WNT11 PDE6G LAMP5 GJB4 FZD9 RBFOX1 GJA8 GLI3 FOXJ1 T ELAVL3 PITX3 SV2C NHLH1 BMP4 NES CRYBB1 NEFM AQP9 ACTR6 AMHR2 CHGB SCG5 PTGIR CGA MEPE TXNDC9 GDF5 LIF HK3 CRLF2 CHAC1 SNAP25 SSTR3 CRYM FSTL1 SCG2 FFAR3 PFKFB1 CSF3 ADAMTS3 IL4R PTGER1 CCR5 TAC1 PDGFRB CCL22 ZBTB16 UCHL3 MYH9 IL1R2 CCL2 NMNAT2 COL5A3 CD33 NTSR1 SCAMP1 TNFRSF1A LYL1 SLC12A5 PREPL HDAC7 NFKBIA PCMT1 SERPINH1 LTA

CD5 NOS1AP MYO1C TRIM25 RFK SYNJ1

HDAC11 MAT1A SLC4A1 PLCD1 PRKCB ATP1B1 CD200 GRAP2 PPP3CA KIR3DL3 KEL BHMT KCNE2 ATP2B1 NCAPH BIN2

ITGB7

TSPYL5 ILK RECQL4 RASGRP1 CAV3 KIFC1 LSP1 OBSCN ITGA5 EHD2

CACNA1F TUBB6 KIF1C EFCC1 ARID3B PHLDB1 POLD4 GPRC5A SIPA1 RCN2

PAMR1 ARID5A NXPE4 GTF2H5 SMYD5 PLEK2 (b)

Figure 9: Continued. 34 BioMed Research International

FBN1 DOT1L

ADAMTS14 COL5A1 CTGF CDH1

ADAMTS18 DSP SCN1B DPP4

KCNC3 CD9

IGSF9B

(c)

FBN1 DOT1L

ADAMTS14 COL5A1 CTGF CDH1

ADAMTS18 DSP

SCN1B DPP4

KCNC3

CD9

IGSF9B

(d)

Figure 9: PPI networks for of DEGs in multiple brain regions of PD: (a) substantia nigra; (b) putamen; (c) prefrontal cortex area; (d) cingulate gyrus. Red expresses upregulation and blue expresses downregulation. presynaptic active zone, G-protein alpha-subunit binding, 3.11. Functional Enrichment Analysis of DEGs in the calmodulin binding, benzodiazepine receptor activity, Cingulate Gyrus of PD. GO enrichment analysis of DEGs in GABA-gated chloride ion channel activity, inositol 1,4,5-tris- the cingulate gyrus of PD was performed. These genes could phosphate binding, inhibitory extracellular ligand-gated ion regulate a variety of biological processes like ion transmem- channel activity, and 1-phosphatidylinositol binding. KEGG brane transport, heart rate by cardiac conduction, ion enrichment analysis results demonstrated that mineral transport, atrial cardiac muscle depolariza- absorption, IL-17 signaling pathway, TNF signaling path- tion, cell communication involved in cardiac conduction, way, adipocytokine signaling pathway, synaptic vesicle cell-cell junction organization, cofactor transport, platelet cycle, mTOR signaling pathway, insulin secretion, gap junc- aggregation, monovalent inorganic cation transport, and tion, neuroactive ligand-receptor interaction, and phos- bundle of His cell to Purkinje myocyte communication phatidylinositol signaling system (Figure 10(l)). (Figure 10(m)). As shown in Figure 10(n), they could BioMed Research International 35

Signifcant GO−BP

Aminergic neurotransmitter loading into − synaptic vesicle

Neurotransmitter transport

Chemical synaptic transmission

Amine transport

Neurotransmitter loading into synaptic vesicle

Dopamine biosynthetic process

Phytoalexin metabolic process

Cell−cell signaling

Isoquinoline alkaloid metabolic process

Catecholamine biosynthetic process

56 Enrichment score Count p_value 5 1e−05 3e−05 10 2e−05 4e−05 (a) Signifcant GO−CC

Neuron projection

Cell projection

Axon

Plasma membrane bounded cell projection

Postsynaptic membrane

Synaptic membrane

Presynapse

Dendrite

Dendritic tree

Neuronal cell body

56789 Enrichment score Count p_value 12 1e−05 16 2e−05 20 3e−05 24 4e−05 (b)

Figure 10: Continued. 36 BioMed Research International

Signifcant GO−MF

Monoamine transmembrane transporter activity

Sodium:chloride symporter activity

Dopamine binding

Cytoskeletal adaptor activity

Cation:chloride symporter activity

Neurotransmitter:sodium symporter activity

Spectrin binding

Ammonium transmembrane transporter activity

Protein serine|threonine kinase inhibitor activity

Chloride transmembrane transporter activity

2.5 3.0 Enrichment score Count p_value 2.00 0.002 2.25 0.004 2.50 0.006 2.75 3.00 (c)

Signifcant pathway

Cocaine addiction

Dopaminergic synapse

Amphetamine addiction

Serotonergic synapse

ECM−receptor interaction

Alcoholism

Tyrosine metabolism

Parkinson disease

PPAR signaling pathway

Synaptic vesicle cycle

234 Enrichment score

Count p_value 2 0.01 3 0.02 4 0.03 5

(d)

Figure 10: Continued. BioMed Research International 37

Signifcant GO−BP

Detoxifcation of copper ion

Stress response to metal ion

Chemical synaptic transmission

Cellular response to zinc ion

Cellular response to copper ion

Nervous system development

Cellular zinc ion homeostasis

Cell communication

Zinc ion homeostasis

Cellular response to cadmium ion

7.0 7.5 8.0 8.5 9.0 9.5 Enrichment Score

Count p_value 10 2.5e−08 20 5.0e−08 30 7.5e−08 40 1.0e−07 50 60 (e)

Signifcant GO−CC

Neuron projection

Axon

Synaptic membrane

Plasma membrane bounded cell projection

Cell projection

Postsynapse

Presynaptic membrane

Presynapse

GABA−ergic synapse

Cell body

67891011 Enrichment score

Count p_value 10 1e−06 15 2e−06 20 3e−06 25 4e−06 30 (f)

Figure 10: Continued. 38 BioMed Research International

Signifcant GO−MF

Calcium ion binding

Adiponectin binding

Structural constituent of presynaptic active zone

G−protein alpha−subunit binding

Calmodulin binding

Benzodiazepine receptor activity

Inositol 1,4,5 trisphosphate binding

GABA−gated chloride ion channel activity

Inhibitory extracellular ligand−gated ion channel activity

1−phosphatidylinositol binding

345 Enrichment score

Count p_value

4 0.001

8 0.002

12 0.003

16 0.004

(g)

Figure 10: Continued. BioMed Research International 39

Signifcant pathway

Mineral absorption

IL−17 signaling pathway

TNF signaling pathway

Adipocytokine signaling pathway

Synaptic vesicle cycle

MTOR signaling pathway

Insulin secretion

Gap junction

Neuroactive ligand−receptor interaction

Phosphatidylinositol signaling system

23456 Enrichment score Count p_value 3 0.005 4 0.010 5 0.015 6 0.020 (h)

Signifcant GO−BP

Positive regulation of ossifcation

Regulation of ossifcation

Regulation of cell population proliferation

Positive regulation of cartilage development

Kidney morphogenesis

Mesonephros development

Positive regulation of chondrocyte diferentiation

Detection of abiotic stimulus

Nephron morphogenesis

Regulation of cartilage development

4.0 4.5 5.0 5.5 6.0 Enrichment score

Count p_value 10 0.00005 20 0.00010 30 0.00015 40 (i)

Figure 10: Continued. 40 BioMed Research International

Signifcant GO−CC

Integral component of plasma membrane

Intrinsic component of plasma membrane

Amino acid transport complex

Endoplasmic reticulum lumen

Cell periphery

Plasma membrane

Extracellular space

Cell surface

Cell leading edge

Cytoplasmic side of plasma membrane

2.0 2.5 3.0 3.5 4.0 Enrichment score Count p_value 25 0.0025 50 0.0050 75 0.0075 0.0100 (j)

Signifcant GO−MF

Heparin binding

Cytokine binding

Prostaglandin receptor activity

NAD−dependent histone deacetylase activity (H3−K14 specifc)

Cytokine activity

Phosphoric diester hydrolase activity

Cytokine receptor activity

Actin−dependent ATPase activity

Inorganic anion exchanger activity

Protein membrane anchor

1.9 2.0 2.1 2.2 Enrichment score Count p_value 2 0.008 3 0.010 4 0.012 5 6 7 8

(k)

Figure 10: Continued. BioMed Research International 41

Signifcant pathway

Cytokine−cytokine receptor interaction

Basal cell carcinoma

Hematopoietic cell lineage

Calcium signaling pathway

MTOR signaling pathway

Aldosterone synthesis and secretion

Viral protein interaction with cytokine and cytokine receptor

Signaling pathways regulating pluripotency of stem cells

NF−kappa B signaling pathway

Central carbon metabolism in cancer

2.0 2.5 3.0 3.5 4.0 Enrichment Score Count p_value 4 0.003 6 0.006 8 0.009 10 0.012 12 (l)

Signifcant GO−BP

Regulation of ion transmembrane transport

Regulation of heart rate by cardiac conduction

Regulation of ion transport

Regulation of atrial cardiac muscle cell− membrane depolarization

Cell communication involved in cardiac conduction

Cell−cell junction organization

Cofactor transport

Negative regulation of platelet aggregation

Monovalent inorganic cation transport

Bundle of His cell to Purkinje myocyte communication

3.5 4.0 4.5 5.0 Enrichment score Count p_value 2 1e−04 3 2e−04 4 3e−04 5 6 7 8 (m)

Figure 10: Continued. 42 BioMed Research International

Signifcant GO−CC

Intercalated disc

Cell−cell junction

Cell−cell contact zone

Plasma membrane protein complex

Voltage−gated potassium channel complex

Cell periphery

Adherens junction

Potassium channel complex

Cation channel complex

Cell−cell adherens junction

2.7 3.0 3.3 3.6 Enrichment score Count p_value 5 0.0005 10 0.0010 15 0.0015 20

(n)

Figure 10: Continued. BioMed Research International 43

Signifcant GO−MF

Channel activity

Passive transmembrane transporter activity

Glycosaminoglycan binding

Ion channel activity

Cell adhesion molecule binding

Voltage−gated potassium channel activity

Voltage−gated ion channel activity

Hyaluronic acid binding

Voltage−gated channel activity

Monovalent inorganic cation transmembrane− transporter activity

3.0 3.5 Enrichment score Count p_value 2 0.0005 3 0.0010 4 0.0015 5 0.0020 6 7 (o)

Signifcant pathway

Protein digestion and absorption

Rap1 signaling pathway

1.2 1.4 1.6 1.8 Enrichment Score

Count p_value 2 0.02 0.04 0.03 0.05 (p)

Figure 10: GO and KEGG enrichment analysis results of DEGs for the substantia nigra, putamen, prefrontal cortex area, and cingulate gyrus of PD. GO terms included biological process (BP), cellular component (CC), and molecular function (MF). (a–d) DEGs in the substantia nigra of PD. (e–h) DEGs in the putamen of PD. (i–l) DEGs in the prefrontal cortex area of PD. (m–p) DEGs in the cingulate gyrus of PD. 44 BioMed Research International

0.0

−0.2

−0.4 Running enrichment score −0.6

1

0

Ranked list metric −1

2500 5000 7500 10000 Rank in Ordered Dataset

Electron transport chain (OXPHOS system in mitochondria)%WikiPathways_20180810%WP111%Homo sapiens Proteasome Degradation%WikiPathways_20180810%WP183%Homo sapiens Synaptic Vesicle Pathway%WikiPathways_20180810%WP2267%Homo sapiens

Figure 11: GSEA results of DEGs in multiple brain regions of PD. distinctly participate in cellular components of intercalated 4. Discussion disc, cell-cell junction, cell-cell contact zone, plasma mem- brane protein complex, voltage-gated potassium channel Herein, we identified critical genes and pathways for multiple complex, cell periphery, adherens junction, potassium chan- brain regions including the substantia nigra, putamen, pre- nel complex, cation channel complex, and cell-cell adherens frontal cortex area, and cingulate gyrus in PD by WGCNA, junction. They could significantly have molecular functions which deepened the understanding of PD-related molecular of channel activity, passive transmembrane transporter mechanisms. activity, glycosaminoglycan binding, ion channel activity, In this study, we screened DEGs for the substantia nigra cell adhesion molecule binding, voltage-gated potassium (17 upregulated and 52 downregulated genes), putamen channel activity, voltage-gated ion channel activity, hya- (317 upregulated and 317 downregulated genes), prefrontal luronic acid binding, voltage-gated channel activity, and cortex area (39 upregulated and 72 downregulated genes), monovalent inorganic cation transmembrane transporter and cingulate gyrus (116 upregulated and 292 downregulated activity (Figure 10(o)). Furthermore, our KEGG enrich- genes) of PD based on microarray and RNA-seq expression ment analysis results demonstrated that protein digestion profiles. The regulatory relationship between genes is specific and absorption and signaling pathway were signifi- in time and space. In different organs and tissues, this regula- cantly enriched (Figure 10(p)). tory relationship changes accordingly, which determines the occurrence and development of PD. To achieve specific bio- 3.12. GSEA of DEGs in Multiple Brain Regions of PD. GSEA logical functions of living organisms, the modularization of was carried out based on DEGs in multiple brain regions of biological networks was conducted. WGCNA provides us PD in the GSE20295 dataset. As depicted in Figure 11, these with a simple and effective method to understand the regula- DEGs were most significantly enriched in electron transport tory relationship between genes, which is an indispensable chain, proteasome degradation, and synaptic vesicle method in systems biology research. Using WGCNA, gene pathway. modules were separately built for multiple brain regions of BioMed Research International 45

PD. Based on PD-related DEGs, we visualized the PPI net- electron transport chain, proteasome degradation, and syn- works for the substantia nigra, putamen, prefrontal cortex area, aptic vesicle pathway. These findings could provide novel and cingulate gyrus of PD by the Cytoscape software. The typ- insights into the molecular mechanisms of PD. ical feature of the PPI network is that most of the nodes in the network are connected to only a few nodes, and there are very Abbreviations few nodes connected to a very large number of nodes. These nodes connected to many nodes are important nodes called as PD: Parkinson’s disease hub genes in the network. These hub genes could be involved WGCNA: Weighted gene coexpression network analysis in regulating many biological processes. In this study, hub genes DEGs: Differentially expressed genes in the PPI networks were identified for the substantia nigra FC: Fold change (SLC6A3, SLC18A2, and TH), putamen (BMP4 and SNAP25), GEO: Gene Expression Omnibus prefrontal cortex area (SNAP25), and cingulate gyrus (CTGF, PPI: Protein-protein interaction CDH1, and COL5A1) of PD through the cytoHubba plug-in. GO: Gene Ontology Among them, SLC6A3 gene polymorphism has been found to KEGG: Kyoto Encyclopedia of Genes and Genomes be related to dopamine overdose in PD [23]. It has been identi- GSEA: Gene Set Enrichment Analyses. fied as a hub gene for PD progression in a previous study, which is consistent with our study [24]. SLC6A3 genotype may affect Data Availability cortical striatum activity in PD [25]. A meta-analysis reveals that SLC6A3 is a risk factor for PD [26]. SLC18A2 functions The datasets analyzed during the current study are available abnormally in the human PD brain. Improving SLC18A2 levels from the corresponding author on reasonable request. can increase the efficacy of levodopa [27]. SNAP25 gene poly- morphism may prevent PD and mediate the severity of disease Conflicts of Interest [28]. Furthermore, CDH1 expression is related to substantia nigra degeneration in a PD mouse model [29]. However, the The authors declare no conflicts of interest. functions of most of genes should be further explored in PD. These DEGs in multiple brain regions were involved in dis- Authors’ Contributions tinct biological functions and pathways. GSEA showed that theseDEGswereallsignificantly enriched in electron transport Jianjun Huang and Li Liu contributed equally to this work. chain, proteasome degradation, and synaptic vesicle pathway, which have been widely accepted to be related to PD progres- Acknowledgments sion. For example, Coenzyme Q10 as a component of the elec- tron transport chain may prevent neurodegeneration in This study was supported by the Middle-Aged and Young response to mitochondrial deficiency and oxidative stress, Teachers in Colleges and Universities in Guangxi Basic Abil- which possesses potential as a target for treatment and interven- ity Promotion Project (2017KY0512), the Starting Research tion of PD [30]. Proteasome degradation induced by misfolding Projects for Introducing Dr. of Youjiang Medical University could contribute to the development of PD [31]. Also, abnormal for Nationalities (2015bsky001), and the First Batch of fi ffi accumulation of synaptic vesicle-associated protein is related to High-Level Talent Scienti c Research Projects of the A li- PD [32]. Thus, these critical pathways enriched by DEGs may ated Hospital of Youjiang Medical University for Nationali- be involved in the pathogenesis of PD. ties in 2019 (R20196315). Our results identified biologically significant gene mod- ules by WGCNA and discovered clinical information- References related hub genes, which were consistent with literature “ ’ ” reports, thereby proving the accuracy and effectiveness of [1] S. G. Reich and J. M. Savitt, Parkinson s disease, The Medical – our WGCNA analysis results. Further excavation of gene Clinics of North America, vol. 103, no. 2, pp. 337 350, 2019. “ fi ff module information may assist us to have an in-depth under- [2] L. M. Chi, L. P. Wang, and D. Jiao, Identi cation of di eren- standing on the role and significance of hub genes and signal tially expressed genes and long noncoding RNAs associated with Parkinson’s disease,” Parkinson’s Disease, vol. 2019, Arti- pathways during PD progression. In our future studies, we cle ID 6078251, 7 pages, 2019. will continue to validate the biological functions of these [3] M. T. Hayes, “Parkinson’s disease and parkinsonism,” The hub genes and key pathways in PD progression by a series American Journal of Medicine, vol. 132, no. 7, pp. 802–807, 2019. of in vivo and in vitro experiments. [4] P. M. Antony, N. J. Diederich, R. Krüger, and R. Balling, “The hallmarks of Parkinson’s disease,” The FEBS Journal, vol. 280, 5. Conclusion no. 23, pp. 5981–5993, 2013. “ fi [5] A. S. Chen-Plotkin, R. Albin, R. Alcalay et al., Finding useful Taken together, this study identi ed hub genes for multiple biomarkers for Parkinson’s disease,” Science Translational brain regions including the substantia nigra (SLC6A3, Medicine, vol. 10, no. 454, article eaam6003, 2018. SLC18A2, and TH), putamen (BMP4 and SNAP25), prefron- [6] T. Kakati, D. K. Bhattacharyya, P. Barah, and J. K. Kalita, tal cortex area (SNAP25), and cingulate gyrus (CTGF, “Comparison of methods for differential co-expression analy- CDH1, and COL5A1) in PD based on WGCNA. Further- sis for disease biomarker prediction,” Computers in Biology more, PD-related key pathways were identified including and Medicine, vol. 113, article 103380, 2019. 46 BioMed Research International

[7] H. Patel, R. J. B. Dobson, and S. J. Newhouse, “A meta-analysis [23] B. D. Robertson, A. S. al Jaja, A. A. MacDonald et al., “SLC6A3 of Alzheimer’s disease brain transcriptomic data,” Journal of polymorphism predisposes to dopamine overdose in Parkin- Alzheimer's Disease, vol. 68, no. 4, pp. 1635–1656, 2019. son’s disease,” Frontiers in Neurology, vol. 9, p. 693, 2018. [8] P. Wu, X. Zuo, H. Deng, X. Liu, L. Liu, and A. Ji, “Roles of long [24] J. Li, Y. Sun, and J. Chen, “Identification of critical genes and noncoding RNAs in brain development, functional diversifica- miRNAs associated with the development of Parkinson’s dis- tion and neurodegenerative diseases,” Brain Research Bulletin, ease,” Journal of Molecular Neuroscience, vol. 65, no. 4, vol. 97, pp. 69–80, 2013. pp. 527–535, 2018. [9] Y. H. Chuang, K. C. Paul, J. M. Bronstein, Y. Bordelon, [25] C. Habak, A. Noreau, A. Nagano-Saito et al., “Dopamine S. Horvath, and B. Ritz, “Parkinson’s disease is associated with transporter SLC6A3 genotype affects cortico-striatal activity DNA methylation levels in human blood and saliva,” Genome of set-shifts in Parkinson’s disease,” Brain, vol. 137, no. 11, Medicine, vol. 9, no. 1, p. 76, 2017. pp. 3025–3035, 2014. [10] M. Niemira, F. Collin, A. Szalkowska et al., “Molecular signa- [26] D. Zhai, S. Li, Y. Zhao, and Z. Lin, “SLC6A3 is a risk factor for ture of subtypes of non-small-cell lung cancer by large-scale Parkinson’s disease: a meta-analysis of sixteen years’ studies,” transcriptional profiling: identification of key modules and Neuroscience Letters, vol. 564, pp. 99–104, 2014. genes by weighted gene co-expression network analysis “ ’ ” [27] K. M. Lohr and G. W. Miller, VMAT2 and Parkinson s dis- (WGCNA), Cancers, vol. 12, no. 1, p. 37, 2020. ease: harnessing the dopamine vesicle,” Expert Review of Neu- [11] Q. Wan, J. Tang, Y. Han, and D. Wang, “Co-expression mod- rotherapeutics, vol. 14, no. 10, pp. 1115–1117, 2014. ules construction by WGCNA and identify potential prognos- [28] C. Agliardi, F. R. Guerini, M. Zanzottera et al., “SNAP25 gene tic markers of uveal melanoma,” Experimental Eye Research, ’ – polymorphisms protect against Parkinson s disease and mod- vol. 166, pp. 13 20, 2018. ulate disease severity in patients,” Molecular Neurobiology, [12] L. Yin, Z. Cai, B. Zhu, and C. Xu, “Identification of key path- vol. 56, no. 6, pp. 4455–4463, 2019. ways and genes in the dynamic progression of HCC based on [29] S. Yeo, K. S. An, Y. M. Hong et al., “Neuroprotective changes WGCNA,” Genes, vol. 9, no. 2, p. 92, 2018. in degeneration-related gene expression in the substantia nigra “ ff [13] M. E. Ritchie, B. Phipson, D. Wu et al., limma powers di er- following acupuncture in an MPTP mouse model of parkin- ential expression analyses for RNA-sequencing and microar- sonism: microarray analysis,” Genetics and Molecular Biology, ” ray studies, Nucleic Acids Research, vol. 43, no. 7, p. e47, 2015. vol. 38, no. 1, pp. 115–127, 2015. “ [14] M. D. Robinson, D. J. McCarthy, and G. K. Smyth, edgeR: a [30] X. Yang, Y. Zhang, H. Xu et al., “Neuroprotection of coenzyme ff Bioconductor package for di erential expression analysis of Q10 in neurodegenerative diseases,” Current Topics in Medic- ” digital gene expression data, Bioinformatics, vol. 26, no. 1, inal Chemistry, vol. 16, no. 8, pp. 858–866, 2016. pp. 139-140, 2010. [31] R. S. Samant, C. M. Livingston, E. M. Sontag, and J. Frydman, “ [15] P. Langfelder and S. Horvath, WGCNA: an R package for “Distinct proteostasis circuits cooperate in nuclear and cyto- ” weighted correlation network analysis, BMC Bioinformatics, plasmic protein quality control,” Nature, vol. 563, no. 7731, vol. 9, no. 1, p. 559, 2008. pp. 407–411, 2018. [16] C. von Mering, M. Huynen, D. Jaeggi, S. Schmidt, P. Bork, and “ “ [32] D. J. Busch and J. R. Morgan, Synuclein accumulation is asso- B. Snel, STRING: a database of predicted functional associa- ciated with cell-specific neuronal death after spinal cord tions between proteins,” Nucleic Acids Research, vol. 31, ” – injury, The Journal of Comparative Neurology, vol. 520, no. 1, pp. 258 261, 2003. no. 8, pp. 1751–1771, 2012. [17] B. Zheng, Z. Liao, J. J. Locascio et al., “PGC-1α, a potential therapeutic target for early intervention in Parkinson’s dis- ease,” Science Translational Medicine, vol. 2, no. 52, article 52ra73, 2010. [18] C. H. Chin, S. H. Chen, H. H. Wu, C. W. Ho, M. T. Ko, and C. Y. Lin, “cytoHubba: identifying hub objects and sub- networks from complex interactome,” BMC Systems Biology, vol. 8, Suppl 4, p. S11, 2014. [19] M. Ashburner, C. A. Ball, J. A. Blake et al., “Gene ontology: tool for the unification of biology. The Gene Ontology Consor- tium,” Nature Genetics, vol. 25, no. 1, pp. 25–29, 2000. [20] M. Kanehisa and S. Goto, “KEGG: Kyoto Encyclopedia of Genes and Genomes,” Nucleic Acids Research, vol. 28, no. 1, pp. 27–30, 2000. [21] G. Yu, L. G. Wang, Y. Han, and Q. Y. He, “clusterProfiler: an R package for comparing biological themes among gene clus- ters,” OMICS, vol. 16, no. 5, pp. 284–287, 2012. [22] Y. Zhang, M. James, F. A. Middleton, and R. L. Davis, “Tran- scriptional analysis of multiple brain regions in Parkinson’s disease supports the involvement of specific protein process- ing, energy metabolism, and signaling pathways, and suggests novel disease mechanisms,” American Journal of Medical Genetics. Part B, Neuropsychiatric Genetics, vol. 137b, no. 1, pp. 5–16, 2005.