Microarray Analysis of the Molecular Mechanism Involved in Parkinson's Disease

Purpose This study aimed to investigate the underlying molecular mechanisms of Parkinson's disease (PD) by bioinformatics. Methods Using the microarray dataset GSE72267 from the Gene Expression Omnibus database, which included 40 blood samples from PD patients and 19 matched controls, differentially expressed genes (DEGs) were identified after data preprocessing, followed by Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses. Protein-protein interaction (PPI) network, microRNA- (miRNA-) target regulatory network, and transcription factor- (TF-) target regulatory networks were constructed. Results Of 819 DEGs obtained, 359 were upregulated and 460 were downregulated. Two GO terms, “rRNA processing” and “cytoplasm,” and two KEGG pathways, “metabolic pathways” and “TNF signaling pathway,” played roles in PD development. Intercellular adhesion molecule 1 (ICAM1) was the hub node in the PPI network; hsa-miR-7-5p, hsa-miR-433-3p, and hsa-miR-133b participated in PD pathogenesis. Six TFs, including zinc finger and BTB domain-containing 7A, ovo-like transcriptional repressor 1, GATA-binding protein 3, transcription factor dp-1, SMAD family member 1, and quiescin sulfhydryl oxidase 1, were related to PD. Conclusions “rRNA processing,” “cytoplasm,” “metabolic pathways,” and “TNF signaling pathway” were key pathways involved in PD. ICAM1, hsa-miR-7-5p, hsa-miR-433-3p, hsa-miR-133b, and the abovementioned six TFs might play important roles in PD development.


Introduction
Parkinson's disease (PD) is one of the most common agerelated neurodegenerative diseases [1]. e age at PD onset is approximately 55 years, and the incidence in the population aged > 65 years is approximately 1% [1][2][3]. PD mainly occurs because of the death of dopaminergic neurons in the substantia nigra [4]. Patients with PD present with symptoms such as bradykinesia, resting tremor, rigidity, and postural instability [5]. e current therapy for PD is targeted at its symptoms rather than at dopaminergic neuron degeneration [1]. e diagnosis of PD at the early stage is challenging, and successfully managing PD is di cult at its later stages [4]. To date, the cause of PD remains unknown; however, it appears to involve the intricate interplay of environmental and genetic factors [1,4].
Much e ort has been spent in investigating PD pathogenesis, and the misfolding, aggregation, and aberrance of proteins are considered to be some of the main causes [1,4,5]. Some key genes such as hydrogen sul de, chromobox 5 (CBX5), and transcription factor 3 (TCF3) are related to PD [6,7]. Several pathways have also been identi ed to be related to PD. Activation of the protein kinase B (Akt)/glycogen synthase kinase 3 beta/(GSK3β) pathway by urate reportedly protects dopaminergic neurons in a rat model of PD [8]. In addition, the E2-related factor 2 (Nrf2)/antioxidant response element pathway reportedly counteracts mitochondrial dysfunction, which is a prominent PD feature [9]. e ubiquitin, lipid, nigrostriatal, autophagy-lysosome, and endosomal pathways are also involved in PD [10][11][12][13][14][15]. Furthermore, a recent study revealed several microRNAs (miRNAs) associated with PD; miR-205 suppresses LRRK2 expression and miR-205 expression levels in the brains of patients with PD decreases [16]. Furthermore, miR-34b and miR-34c are downregulated in the brains of patients with PD, which is related to the reduction in the expression of DJ-1 and PARKIN [17], and miR-133 and miR-7 are also associated with PD [18][19][20]. Numerous reports that have described the roles of transcription factors (TFs) in PD have also been published. e TF paired-like homeodomain 3 has roles in developing and maintaining dopaminergic neurons [21,22], and engrailed l, which is downregulated in the rat models, plays a role in the apoptosis of dopaminergic neurons and the symptoms of PD [23]. Moreover, Nrf2, nuclear factor kappa B (NF-κB), GATA2, and PHD nger protein 10 are TFs involved in PD [24][25][26][27]. However, understanding the key mechanisms underlying the development of PD remains unclear.
In a previous study, the microarray dataset GSE72267 generated by Calligaris et al. [7] was used to identify key di erentially expressed genes (DEGs) such as CBX5, TCF3, dedicator of cytokinesis 10, and mannosidase alpha class 1C in the blood of patients with PD compared with those of healthy controls. Moreover, crucial pathways related to chromatin remodeling and methylation were revealed. In the current study, we downloaded this microarray dataset to comprehensively analyze DEGs in patients with PD compared with those in matched controls by bioinformatics approaches and to describe their functional annotations. Compared with the previous analysis conducted by Calligaris et al. [7], we performed additional analyses, including those for the protein-protein interaction (PPI), miRNAtarget regulatory, and TF-target regulatory networks, to further elucidate the key mechanisms underlying PD. Our   HC01  HC02  HC03  HC04  HC05  HC06  HC07  HC08  HC09  HC10  HC11  HC12  HC13  HC14  HC15  HC16  HC17  HC18  HC19  PD01  PD02  PD03  PD04  PD05  PD06  PD07  PD08  PD09  PD10  PD11  PD12  PD13  PD14  PD15  PD16  PD17  PD18  PD19  PD20  PD21  PD22  PD23  PD24  PD25  PD26  PD27  PD28  PD29  PD30  PD31  PD32  PD33  PD34  PD35  PD36  PD37  PD38  PD39  PD40   2   4   6   8   10   12   14 Data normalization    (version 1.50.0) [29] in R language, including background correction, normalization, and expression calculation. Annotations to the probes were performed, and probes that were not matched to the gene symbol were excluded. e average expression values were taken if di erent probes mapped to the same gene. DEGs in patients with PD compared with those in healthy matched controls were analyzed using the limma package (version 3.10.3) [30] in R language. e cuto threshold was set to a p value of <0.05.

Pathway Enrichment Analysis.
Gene ontology (GO) (http://www.geneontology.org/) analysis is commonly used for functional studies of large-scale genomic or transcriptomic data and classi es functions with respect to three aspects: molecular function (MF), cellular component (CC), and biological process (BP) [31,32]. e Kyoto Encyclopedia of Genes and Genomes (KEGG; http://www.kegg.jp/) pathway database [33] is widely used for systematic analysis of gene functions, linking genomic data with higher order functional data. e database for annotation, visualization, and integrated discovery (DAVID) is an integrated biological knowledgebase with analytical tools used for systematic and integrative analysis of large gene lists [34]. In this study, GO terms and KEGG pathway enrichment analyses for up-and downregulated DEGs were performed using DAVID (version 6.8). e cuto thresholds were as follows: an enrichment gene number count of ≥2 and a super geometry inspection signi cance threshold p value of <0.05.

PPI Network Analysis. Search Tool for the Retrieval of
Interacting Genes/Proteins (STRING; http://www.string-db. org/) [35] is an online database that assesses and integrates PPIs. In this study, DEGs were mapped into the STRING database for PPI analysis, with a PPI score of 0.4 as the parameter setting. e PPI network established by DEGs was constructed using the Cytoscape software (version 3.2.0) [36], and the topology scores of the nodes, including node degree in the PPI network, were analyzed using the CytoNCA plugin (version 2.1.6; http://apps.cytoscape. org/apps/cytonca) [37] (parameter setting: without weight). Degree was used for describing importance of protein nodes in network. e higher the degree was, the more important the nodes were in network. In addition, subnetworks were identi ed using the MCODE plugin [38] in the Cytoscape software, and subnetworks with a score of >5 were identi ed as key subnetworks. Finally, KEGG pathway enrichment analyses for the genes in the key subnetworks were performed.

miRNA-Target Regulatory Network Analysis.
e miR2disease (http://www.mir2disease.org/) database [39] is a manually curated database that provides a comprehensive resource of miRNA deregulation in various human diseases. miRWalk2.0 (http://zmf.umm. uni-heidelberg.de/apps/zmf/mirwalk2/) [40] is a comprehensive database that presents predicted and validated data, regarding miRNA targets in human, mouse, and rats. In this study, miRNAs related to PD were extracted from the miR2disease database, and experimentally veri ed miRNA-gene regulatory pairs were obtained by searching miRWalk2.0. Finally, a miRNA-target regulatory network was constructed by comparing DEGs with obtained miRNA-gene regulatory pairs using the Cytoscape software.

TF-Target Regulatory Network Analysis.
e genes in the PPI network described above were further analyzed to identify TF-target interaction pairs that were then used to construct a TF-target regulatory network. e iRegulon plugin (version 1.3; http://apps.cytoscape.org/apps/iRegulon) [41] in the Cytoscape software collects multiple human TF-target interaction databases such as Transfac, Jaspar, and Encode using two computational methods: Motif and Track. In this study, we analyzed the TF-target pairs using the iRegulon plugin and compared them with TFs with DEGs in the PPI network, followed by a TF-target regulatory network construction. e parameter settings were as follows: minimum identity between orthologous genes, 0.05 and maximum false discovery rate on motif similarity, 0.001. e normalized enrichment score (NES) indicates the reliability of the results, and the cuto threshold was NES of >3. Degree was used for describing the importance of protein nodes in network. e higher the degree was, the more important the nodes were in network.

Analysis of DEGs.
e boxplot of the preprocessed data indicated good normalization (Figure 1). In total, 22,277 probes were obtained, among which 971 probes were di erentially expressed. After annotation, 819 DEGs in patients with PD compared with those in healthy matched controls were identi ed (Supplementary Table 1), including 359 upregulated DEGs and 460 downregulated DEGs. Table 2). e signi cant GO terms and KEGG pathways are shown in Figure 2. e upregulated DEGs were signi cantly enriched in four KEGG pathways, namely, metabolic pathways, inositol phosphate metabolism, mRNA surveillance pathway, and RNA degradation, and GO terms such as transcription, DNAtemplate processing, and rRNA processing (Figure 2(a)).  e downregulated DEGs were enriched in pathways such as those of in uenza A, viral myocarditis, and TNF signaling and GO terms such as cytoplasm, cell surface, and interferon gamma-mediated signaling pathway (Figure 2(b)).

PPI Network Analysis.
e PPI network, including 605 nodes and 1937 PPI pairs, is shown in Figure 3. e top 10 DEGs with the highest degree included ve upregulated DEGs such as estrogen receptor 1 (ESR1), mechanistic target of rapamycin (MTOR), ATM serine/threonine kinase (ATM), CD40 molecule (CD40) and thymidine kinase 2, mitochondrial (TK2), and ve downregulated DEGs such as mitogen-activated protein kinase 14 (MAPK14), phosphatase and tensin homolog (PTEN), intercellular adhesion molecule 1 (ICAM1), aurora kinase A (AURKA), and protein kinase, DNA-activated, catalytic polypeptide (PRKDC) ( Table 1). ree subnetworks were identi ed (subnetworks a-c). Subnetwork a (Figure 4(a)) included nine nodes and 36 PPI pairs, and these genes were signi cantly enriched in three KEGG pathways (Table 2), including neuroactive ligand-receptor interaction, chemokine signaling pathway, and cytokinecytokine receptor interaction. Subnetwork b (Figure 4(b)) included seven nodes and 21 PPI pairs, and these genes were not enriched in any KEGG pathway. Subnetwork c (Figure 4(c)) included 27 nodes and 81 PPI pairs, and these genes were enriched in 12 KEGG pathways (Table 2), such as cell cycle, herpes simplex infection, and NF-κB signaling pathways.  In addition, ICAM1 was involved in six KEGG pathways of subnetwork c, such as viral myocarditis, cell adhesion molecules (CAMs), and NF-κB signaling pathways (Table 2). e detailed information existed in PPI network, and three subnetworks are shown in Supplementary Table 3.

miRNA-Target Regulatory Network Analysis.
According to the data from the miR2disease database, six miRNAs were identi ed to be associated with PD and 698 miRNA-gene pairs were obtained by searching miRWalk2.0. A total of 40 miRNA-target interaction pairs were obtained by comparing miRNA-gene pairs with DEGs, and subsequently, the miRNA-target regulatory network was constructed. e network ( Figure 5) contained 40 miRNA-target interaction pairs and 43 nodes (Supplementary Table 4), among which three miRNAs (hsa-miR-7-5p, hsa-miR-433-3p, and hsa-miR-133b) were included.

TF-Target Regulatory Network Analysis.
According the information of TF-target interaction databases such as Transfac, Jaspar, and Encode in the Cytoscape software, a total of 83 TFs were identi ed from the PPI network, forming 5371 TF-gene pairs. Among the 83 TFs, six were di erentially expressed: three upregulated ones, that is, zinc nger and BTB domain-containing 7A (ZBTB7A), ovo-like transcriptional repressor 1 (OVOL1), and GATA-binding protein 3, and three downregulated ones, that is, transcription factor dp-1 (TFDP1), SMAD family member 1 (SMAD1), and quiescin sulfhydryl oxidase 1 (QSOX1). e TF-target regulatory network ( Figure 6) was constructed and included 166 nodes and 288 interaction pairs (Supplementary Table 5). e top 20 nodes with the highest degree are listed in Table 3, including the six TFs described above and 14 other DEGs, such as ectodermal-neural cortex 1, bronectin type III domain-containing 3A, and midline 1, which were coregulated by the six TFs.

Discussion
PD is the second most common age-related neurodegenerative disease. However, the pathogenesis and genes involved in PD are not well known [42]. In this study, we performed a comprehensive bioinformatics analysis of the blood gene expression pro le using the GSE72267 dataset.
Our results revealed that the upregulated DEGs were enriched in the KEGG pathway "metabolic pathways" and the GO term "rRNA processing," and the downregulated DEGs were enriched in the KEGG pathway "TNF signaling pathway" and the GO term "cytoplasm." A previous study [43] demonstrated that some metabolic patterns were altered in patients with advanced PD. Multiple metabolic pathways are also involved in PD [44], which supports our study results. Cytoplasmic inclusions are a pathological hallmark of PD [45]. Lewy body pathology is involved [46,47], and glial cytoplasmic inclusions are associated with Lewy bodies [48]. us, the GO term "cytoplasm" may play a role in PD. Furthermore, TNF receptor-associated protein is excluded from the nucleolus and is sequestered to the cytoplasm by TNF receptor-associated factor 6, thereby altering ribosomal RNA (rRNA) biogenesis [49]. e TNF signaling pathway is also involved in PD [50], and rRNA transcription is repressed in patients with PD [51]. erefore, the GO term "rRNA processing" and the KEGG pathway "TNF signaling pathway" may play important roles in PD. Altogether, the metabolic pathways, TNF signaling pathway, rRNA processing, and cytoplasm are essentially involved in PD pathogenesis.
ICAM1 was among the top 10 DEGs in the PPI network. Moreover, ICAM1 gene was involved in six KEGG pathways for subnetwork c. ICAM1 is involved in the adhesion and transmigration of leukocytes across the endothelium, promoting brain in ammation and resulting in brain diseases [52]. T helper 17 cells can exert a neurotoxic e ect in the brain parenchyma of patients with PD by interacting with ICAM1 and leukocyte function-associated antigen 1 [53]. In addition, ICAM1 is involved in persistent in ammation in PD [54]. Our results from the KEGG pathway analysis for genes in subnetworks revealed that ICAM1 might play roles in viral myocarditis and CAMs and thus contributed to PD. e miRNA-target regulatory network analysis identi ed three miRNAs involved in PD, namely, hsa-miR-7-5p, hsa-miR-433-3p, and hsa-miR-133b. A study described miR-7-2 dysregulation (the stem loop of hsa-miR-7-5p) in Parkinson's patient's leukocytes [55] and revealed that hsa-miR-7-5p expression decreased in PD, possibly upregulating α-SYN, a PD-related gene [56]. e variation of the hsa-miR-433-(the stem loop of hsa-miR-433-3p-) binding site of broblast growth factor 20 can lead to α-SYN overexpression, increasing the risk for PD [57]. hsa-miR-133b expression is increased in the cerebrospinal uid of patients with PD [58]; however, its expression levels in serum is decreased, which is related to low serum ceruloplasmin levels [59]. hsa-miR-133b is also de cient in the midbrain tissue of patients with PD and is associated with the maturation and function of midbrain dopaminergic neurons [60]. Notably, reduced circulating levels of miR-433 and miR-133b are considered as promising biomarkers for PD [61]. erefore, we speculate that the three miRNAs, including hsa-miR-7-5p, hsa-miR-433-3p, and hsa-miR-133b may play important roles in PD.
TFs are important regulators of target gene expressions [53,62]. In this study, we analyzed DEGs in the PPI network to screen TFs involved in PD. Among the 83 TFs identi ed in the PPI network, six were found to be di erentially expressed. ZBTB7A, OVOL1, and GATA3 were upregulated in patients with PD compared with those in healthy matched controls, whereas TFDP1, SMAD1, and QSOX1 were downregulated. ZBTB7A is a tumor suppressor, which is involved in several cancers such as prostate and nonsmall cell lung cancers [63][64][65]. OVOL1, encoding a zinc nger protein, is expressed in embryonic epidermal progenitor cells and is an inducer of mesenchymal-to-epithelial transition in human cancers [66,67]. GATA3, a member of the GATA family, is a regulator of T-cell development and plays roles in endothelial cells [68,69]. TFDP1 is involved in the cell cycle and contributes to hepatocellular carcinomas [70,71], SMAD1 is involved in multiple pathways [72,73], and QSOX1 plays roles in some cancers such as breast cancer and neuroblastoma [74][75][76]. However, there are few reports regarding the involvement of these TFs in PD. Hence, further studies regarding the associations between the TFs identi ed in this study and PD are warranted.
In conclusion, our data demonstrated that the metabolic pathways, TNF signaling pathway, rRNA processing, and cytoplasm play important roles in PD pathogenesis; ICAM1 might also play a vital role. Besides six TFs, three miRNAs, including hsa-miR-7-5p, hsa-miR-433-3p, and hsa-miR-133b, may be involved in PD. However, because of the study limitations, further investigation remains to be performed in the future.

Conflicts of Interest
e authors declare that there are no con icts of interest regarding the publication of this article.