Identification of copper metabolism-related markers in Parkinson’s disease

Abstract

Objectives

This study aimed to identify key genes related to copper metabolism in Parkinson’s disease (PD), providing insight into their roles in disease progression.

Methods

Using bioinformatic analyses, the study identified hub genes related to copper metabolism in PD patients. Differentially expressed genes (DEGs) were identified using the limma package, and copper-metabolism-related genes (CMRGs) were sourced from the Genecard database. Immune cell-related genes were derived through immune infiltration and Weighted Gene Co-expression Network Analysis (WGCNA). Hub genes were pinpointed by integrating DEGs, CMRGs, and immune cell-related genes. Functional analyses included Receiver Operating Characteristic (ROC) analysis, Ingenuity Pathway Analysis (IPA), and networks for miRNA-mRNA-transcription factor (TF), Competitive Endogenous RNA (ceRNA), and hub gene-drug interactions. Validation was performed in cerebrospinal fluid (CSF) samples from PD patients, while in vitro experiments utilized GBE1- overexpressing SH-SY5Y cells to examine cell proliferation, migration, and viability.

Results

Nine hub genes (HPRT1, GLS, SNCA, MDH1, GBE1, DDC, STXBP1, ACHE, and AGTR1) were identified from 753 CMRGs, 416 DEGs, and 951 immune cell-related genes. ROC analysis showed high predictive accuracy for PD, and principal component analysis (PCA) effectively distinguished PD patients from controls. IPA identified 20 significant pathways, and various networks highlighted miRNA, TF, and drug interactions with the hub genes. Hub gene expression was validated in PD CSF samples. GBE1-overexpressing cells displayed enhanced proliferation, migration, and viability.

Conclusions

The study identified nine copper metabolism-related genes as potential therapeutic targets in PD, highlighting their relevance in PD pathology and possible treatment pathways.

1. Introduction

Parkinson’s disease (PD) is a chronic neurodegenerative disease that predominantly affects older individuals [1]. The clinical manifestation of PD is characterized by resting/stationary tremor, bradykinesia, muscle rigidity. Severe cases of PD may also present cognitive impairment, autonomic dysfunction and depression even occur, which can impose significant financial burden on patients and their families [2]. Currently, studies have shown that there are 5.8 million PD patients worldwide, and the number is projected to increase to over 9 million by 2030 due to the aging global population [3]. Additionally, PD is now being diagnosed in younger individuals. Clinical diagnosis relies on Parkinson’s scale (rating scale), patient complaints and physician’s clinical experience. However, subjective factors can lead to misdiagnosis and underdiagnosis [2]. Although levodopa is the primary treatment for PD patients, it only relieves symptoms by increasing dopamine levels [4]. In this context, it is essential to identify PD-related biomarkers that can be used for potential treatment.

Copper is an essential element that binds to proteins or enzymes and participates in various physiological processes, including energy metabolism and mitochondrial respiration [5, 6]. In the brain, it functions to bind with proteins or enzymes that are fundamental to neural development, antioxidant defense and neurotransmitter synthesis. The homeostatic balance of copper, governed by the regulation of its intake and excretion, is essential for maintaining neuronal integrity and function [7]. The imbalance in copper metabolism can lead to toxic accumulation or deficiency, each contributing to neurotoxicity through various mechanisms, including oxidative stress, impaired mitochondrial function, and aberrant protein aggregation, inducing neuroinflammatory signaling pathway [8]. Recent studies also have highlighted that microglia-mediated neuroinflammation plays a crucial role in the pathogenesis of Parkinson’s disease [9]

H S Pall et al. reported patients with PD exhibited an increased concentration of copper in their cerebrospinal fluid (CSF), coupled with reduced copper levels in both serum and SN [10, 11], underscoring a distinct copper concentration when compared to healthy individuals. Additionally, studies have found the level of Cu and Cu transporter protein 1 (Ctr1), were statistically decreased in the surviving neurons in SN and locus coeruleus (LC), emphasizing the role of copper homeostasis in PD pathology [12]. Furthermore, previous studies demonstrated that both Cu2+ and Cu + had the unique ability to promote the aggregation of α-synuclein, which could lead to motor dysfunction in a PD mouse model. Inhibition of α-synuclein aggregation by copper transporter 1 (Ctr1) alleviated motor dysfunction in the dopaminergic cell-specific Ctr1 conditional knock-out (Ctr1-CKO) mouse model [13]. Disruptions in copper metabolism may further aggregate the pathological changes of PD, particularly by participating in the oxidation of dopamine [14, 15]. This underlines the crucial role of copper homeostasis in maintaining neuronal integrity and function. Given these disturbances, accurately delineating copper metabolism in PD patients is critical—not only to deepen our understanding of the disease’s pathophysiology but also to aid in the discovery of biomarkers for diagnosis and disease progression. Consequently, our research is directed at identifying key hub genes involved in copper metabolism within PD patients. This could pave the way for novel targeted therapeutic interventions and establish groundwork for alleviating the neurodegenerative mechanisms involved in PD.

The current work utilized bioinformatic analysis to discover hub genes related to copper metabolism in PD patients compared with the control group. The microarray data of the samples were downloaded from the GEO database. The hub genes were identified through a comparative analysis of DEGs, CMRGs and immune cells-related genes. Subsequently, the biological function of hub genes was evaluated using GSEA and IPA and ROC analysis. In addition, networks were constructed to examine interaction between hub gens and miRNA, TFs and drugs. Hub genes were validated by qRT-PCR by using CSF sample from PD samples. GBE1, the potential hub gene, was further examined its role by using the PD cellular model.

2. Materials and methods

2.1. Data source

The GSE8397 and GSE7621 dataset were downloaded from the GEO database. 24 PD samples and 13 samples as control were from GSE8397 dataset. In GSE7621 dataset used 16 PD patients and 9 samples for control were included. In addition, copper metabolism-related genes (Supplementary Table 1) (CMRGs, relevance score > 5) were downloaded from the genecard database (https://www.genecards.org/).

2.2. Identification of differentially expressed copper metabolism-related genes in PD samples

Differentially expressed genes (DEGs) between PD group and control group in the GSE8397 dataset were screened using the ‘limma’ package (p < 0.05, |log2FC| > 1) [16]. DE CMRGs were obtained in comparison of DEGs and CMRGs. Functional enrichment analysis of DE CMRGs based on gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) was subsequently conducted by using the ‘clusterProfiler’ package [17].

2.3. Immune infiltration analysis

Studies shown that immune cells could infiltrate brain in PD patients and may contribute to neuroinflammation and neuronal damage. Targeting immune infiltration in PD may be a promising avenue for developing new treatments. To investigate which immune cells contribute to PD pathology, GSE8397 dataset was utilized to analyze immune infiltration. The proportion of 28 immune cells was estimated through the single-sample Gene set enrichment analysis (ssGSEA) algorithm. Then the immune cells were compared between PD and control group by rank sum test.

2.4. Identification of immune cell-related genes

To investigate immune cell-related genes in PD patients, weighted gene co-expression network analysis (WGCNA) was performed on the GSE8397 dataset. Firstly, the samples were clustered (distance metric such as Euclidean distance and a clustering algorithm such as hierarchical clustering or k-means clustering), and outliers were excluded to avoid bias for further analysis. Next, a soft threshold (β) was selected to construct a co-expression network by scale-free R² value close to 0.85 and an average connectivity value close to 0, ensuring that constructed network corresponded more closely to scale-free topology.

Next, genes were compared for similarity based on their proximity, and a phylogenetic tree of those genes was constructed. Dynamic tree cutting algorithm was employed to partition modules with a minimum requirement of 30 genes per module. The correlation coefficients between the identified gene modules and the clinical traits were analyzed using Pearson’s correlation test. Finally, module with p < 0.05 and correlation with all immune cells greater than 0.6 were identified as key modules. The key module genes were then obtained from key module-trait (highest coeffcience value) with Gene Significance, |GS| > 0.2 and Module membership, |MM| > 0.8 for further study.

2.5. Identification and evaluation of hub genes

The candidate genes were obtained by comparing CMRGs, DEGs and immune cell-related genes. To explore the interactions between hub genes, a protein-protein interactions (PPI) network (Confidence = 0.15) was constructed based on hub genes by using the STRING database (https://string-db.org). In addition, the diagnostic value of the hub genes was explored by plotting Receiver Operating Characteristic (ROC) curve. Subsequently, the expression of hub genes was extracted from the GSE8397 dataset and principal component analysis (PCA) was performed to analyze whether hub genes could identify PD samples from control group. The diagnostic value of hub genes and the PCA were also repeated in the GSE7621 dataset.

2.6. Single gene GSEA

In single gene GSEA, genes of interest are tested for enrichment based on their biological function or regulatory pathways. In this study, the hub genes in GSE8397 dataset were used as the genes of interest. The correlation coefficients between the expression of all genes and the target gene were calculated. Next, GO and KEGG pathways were used for functional enrichment analysis, followed by GSEA enrichment analysis on expression of hub genes via the ‘clusterProfiler’ R package. The criteria was developed with |NES| > 1, P.adjust < 0.05 and q < 0.25.

2.7. Ingenuity pathway analysis (IPA)

To explore the interaction network between hub genes and the disease or functional pathways that are affected by hub genes, IPA analysis was performed in the GSE8397 dataset. First, DEGs were obtained by developing the criteria of (|log2 (fold change)| > 1 and p < 0.05). Statistically significant pathways (p < 0.05) were obtained by the qualitative method of IPA. Next, interaction networks were constructed and extracted, which revealed important genes and other relevant molecules, such as chemicals or drugs.

2.8. Construction of miRNA-mRNA-transcription factors network, competitive endogenous RNA network and drug-hub genes network

The expression of mRNA was regulated by both miRNA and TFs. First, the miRNA interaction of the hub genes was predicted using the miRwalk website (http://mirwalk.umm.uni-heidelberg.de/). The enriched TFs binding sites of the hub gene were identified through ChEA3 database (https://amp.pharm.mssm.edu/chea3/). The miRNA-mRNA-TF network was subsequently constructed based on miRNA-mRNA and mRNA-TF network. Next, the Starbase website (http://starbase.sysu.edu.cn/) was used to study the interaction of long non-coding RNAs (lncRNAs) interaction with relevant miRNAs. ceRNA network was built based on hub genes, LncRNAs with chipExpNum > 1 and miRNAs with energy < −30. Furthermore, the Comparative Toxicogenomics Database (http://ctdbase.org/) provide information on the relationship between hub genes and drugs that affect the expression of hub genes. Drug-hub genes network was constructed based on the drug-gene relationship.

2.9. RNA extraction and validation of hub genes by qRT-PCR

The Ethics Committee of Liaocheng People’s Hospital approved all the experiments of human samples used in the study (No. 2023029). The total RNA from the cerebral spinal fluid (CSF) of PD patients was extracted by using RNeasy kit (Qiagen 74104). Reverse transcription was conducted subsequently by a reverse transcription kit (AG11706, AG, CHINA) based on the manufacture’s instructions. SYBR premixed Ex Taq kit (AG11718, AG, CHINA) was used to amplify cDNA with Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) as the internal reference. The 2(-Delta CT) technique calculated mRNA expression. All the primers applied were listed in the Table 1, and the general clinical information of the subjects was presented in the Table 2.

Table 1.

Sequence of primers of RT-PCR of human brain DDC, GBE1, SNCA, AGTR1, ACHE, STXBP, HPTR1, MDH1 and GLS. Primers for the qRT-PCR test.

Gene	Forward 5’ to 3’	Reverse 5’ to 3’
DDC	TGGGGACCACAACATGCTG	TCAGGGCAGATGAATGCACTG
GBE1	CAAAGTATGTGGTTCGTGAAGGT	GATTGCCATCAACTGAATGCAG
SNCA	AAGAGGGTGTTCTCTATGTAGGC	GCTCCTCCAACATTTGTCACTT
AGTR1	GGCTATTGTTCACCCAATGAAGT	TGGGACTCATAATGGAAAGCAC
ACHE	CCTGTCCTCGTCTGGATCTATG	AAGAAGCGGCCATCGTACAC
STXBP	AGAAGTCCGTCCACTCTCTCA	GCTCGGGATTTTACCAGTTCAT
HPRT1	CCTGGCGTCGTGATTAGTGAT	AGACGTTCAGTCCTGTCCATAA
MDH1	TTTGGATCACAACCGAGCTAAAG	ACATCTGGATACTGAGTCGAGG
GLS	ATTCAGTCCCGATTTGTGGGG	AGAAGGGAACTTTGGTATCTCCA

Table 2.

General clinical characteristics of the participants.

Samples	Gender	Age	Sample type
PD	F	67	Cerebral spinal fluid
PD	F	66	Cerebral spinal fluid
PD	M	59	Cerebral spinal fluid
PD	F	58	Cerebral spinal fluid
PD	M	68	Cerebral spinal fluid
PD	F	63	Cerebral spinal fluid
control	M	62	Cerebral spinal fluid
control	M	58	Cerebral spinal fluid
control	F	59	Cerebral spinal fluid
control	F	65	Cerebral spinal fluid
control	M	66	Cerebral spinal fluid
control	F	62	Cerebral spinal fluid

2.10. Establishment of PD cellular model and cell culture

SH-SY5Y cells were purchased from FuHeng Biology (Shanghai, China) and cultured in Dulbecco’s Modified Eagle Medium (DMEM) (Hyclone, Logan, USA) supplemented with 10% fetal bovine serum (Gibco10099-141C, Australia), 100 U/mL penicillin (Sigma-Aldrich, USA) at the 37 °C, 5% CO₂ incubator. A cellular model of PD was established by treating SH-SY5Y cells with rotenone (10 μmol/L) for 24 h [18].

2.11. Cell transfection

The GBE1 overexpression plasmid and the corresponding empty vector were obtained from Genechem (Shanghai, China). The full-length human GBE1 cDNA was cloned into Lentiviral Vector Ubi-MCS-3FLAG-SV40-IRES-puromycin to generate the overexpression construct. HEK-293T cells were used to package the lentivirus by co-transfecting the GBE1 overexpression plasmid (oe-GBE1) or an empty vector control (oe-NC), along with the packaging plasmids psPAX2 and pMD2.G. Following a 48- and 72-hour incubation period after transfection, the viral supernatants were collected and concentrated. The ROT-treated SH-SY5Y cells were then transduced with the resulting GBE1-overexpressing or control lentivirus. Transduction efficiency was verified through quantitative RT-PCR (qRT-PCR) analysis of GBE1 expression levels.

2.12. Cell proliferation assay

The transduced SH-SY5Ycells with intensity of 1000 cells/well were seeded onto the 96-well plate. After attachment, cell proliferation was quantitatively assessed at each time points (24h, 48h and 72h) using CCK-8. The optical density (O.D.) of each well was measured at an absorbance wavelength of 450 nm.

2.13. Cell migration capability in Wound-Healing assay

Cell migration experiment was performed in accordance with the established protocol. Briefly, SH-SY5Y cells transduced with GBE1 at the density of 1*10⁵cells/ml were seeded onto the 6-well plate and allowed to grow to confluence. A 10 µl sterile pipette tip was utilized to create scratch, forming a uniform physical gap within the cell monolayer. Cell debris were then gently removed by medium wash, followed by addition of complete medium. The process of cell migration into the gap was tracked with live cell imaging and documented at the time point of 24 h [19].

2.14. Cell viability analysis

To evaluate cell viability in transduced SH-SY5Y cells, the Live/Dead Cell staining Kit (Beyotime, China) was employed, which utilized calcein-AM to stain live cells green and propidium iodide to stain dead cells red. Post staining, cells were imaged to capture fluorescence signals through a fluorescence microscope (OLYMPUS, Japan). The cell viability was quantitatively calculated by the software.

3. Results

3.1. Identification of DE CMRGs

In the GSE8397 dataset, 416 DEGs were identified in PD group in comparison with control group, with 67 upregulated genes and 349 downregulated genes (Figure 1A, Supplementary Table 2). In the heatmap, orange color indicated higher level of gene expression and blue color indicated lower value of gene expression. 67 upregulated DEGs were visualized in Figure 1B.

Figure 1.

Identification of 21 DE CMRGs. (A) Volcano plots showing 416 DEGs in total between control and PD groups in the GSE8397 dataset, red dots indicate up-regulation and blue dots indicate down-regulation. (B) Heat map of the expression of DEGs. (C) Venn diagram of 21 DE CMRGs by comparison between DEGs and CMRGs; (D) Heat map of the expression of 21 DE CMRGs. (E) GO analysis of 21 DE CMRGs. Y-axis represents the enriched GO terms; X -axis represents the amounts of genes enriched in GO terms. (F) KEGG pathway analysis of 21 DE CMRGs. Y-axis represents the KEGG signaling pathways. X-axis represents amounts of genes enriched in KEGG pathways.

A total of 21 DECMRGs were consequently obtained by comparison between 416 DEGs and 753 CMRGs. 15 DE CMRGs (GCH1, DRD2, SLC6A3, TH, AGTR1, ALDH1A1, HPRT1, GBE1, DDC, SNCA, CCK, MDH1, ACHE, GLS and STXBP1) were highly expressed in the control group. However, the expression of 6 DE CMRGs (CP, S100A12, AZGP1, LRP2, MT1G and NFKBIA) were upregulated in PD group (Figure 1C and D). Next, the enrichment analysis of 21 DE CMRGs revealed 28 GO terms and 8 KEGG pathways. Three GO terms (dopamine metabolic process, catecholamine metabolic process and cellular amine metabolic process) (Figure 1E, Supplementary Table 3) and four KEGG pathways (dopaminergic synapse, folate biosynthesis, parkinson’s disease and tyrosine metabolism) (Figure 1F, Supplementary Table 4) were related to PD.

3.2. Immune infiltration analysis

Immune infiltration analysis provided a comprehensive view of 28 immune cells’ landscape. The heatmap of 28 immune cells was shown in Figure 2A, Supplementary Table 5. In PD patients, proportion of 28 immune cells were estimated (Figure 2B). 8 immune cells (CD56 bright natural killer cells, central memory CD8T cells, type 17 T helper cells, plasmacytoid dendritic cells, neutrophils, monocytes, memory B cells and mast cells) with higher proportion were identified. However, 3 immune cells (type 2 T helper cells, gamma delta T cell and immature dendritic cell) with lower proportion were confirmed as well.

Figure 2.

Analysis of 28 immune cell infiltration according to ssGSEA (A) Heat map of the abundance of 28 immune cell types. Each square in each row represents each sample. The color of square represents p value. Each column represents the amount of 28 cell types in each sample. (B) Box plot of the percentage of 28 immune cell types between PD and control samples.

3.3. Identification of immune cell-related genes

No outlier sample in the GSE8397 dataset were detected (Figure 3A). The sample clustering and clinical trait heat map were shown in Figure 3B. The scale-free fitting index suggested 4 as the soft threshold (Figure 3C). 13 modules were subsequently identified by constructing co-expression matrix, among which blue, red, yellow, green modules were the key modules (Figure 3D and E, Supplementary Table 6). 951 key genes in total were identified from the key modules (Figure 3F). Taken together, the 951 genes were considered as key genes (immune cell-related genes) and they were the most relevant to immune function and regulation in PD patients.

Figure 3.

Detection of immune cell-related genes in GSE8397 dataset. (A) Sample clustering dendrogram. (B) Clustering dendrogram of samples with trait heatmap. (C) Analysis of network topology for various soft-thresholding powers. The left panel shows the scale-free fit index (y-axis) as a function of the soft-thresholding power (x-axis). The right panel displays the mean connectivity (degree, y-axis) as a function of the soft-thresholding power (x-axis). (D) Dendrogram of all samples clustered based on the measurement of dissimilarity (1-TOM). (E) Heatmap of the correlation between the module eigengenes and immune features of PD. (F) Royalblue module, black module, darkgrey module, and green module were identified to have the highest correlations with PD-related immune features.

3.4. Identification and evaluation of hub genes

By comparing 753 CMRGs, 416 DEGs and 951 immune cell-related genes, we confirmed nine hub genes (HPRT1, GLS, SNCA, MDH1, GBE1, DDC, STXBP1, ACHE, and AGTR1) were downregulated (Figure 4A). An interrelated PPI network with 15 edges was constructed based on eight of these genes (Figure 4B). The PPI network elucidated the interaction between HPRT1, GLS, SNCA, MDH1, GBE1, DDC, STXBP1 and ACHE. MDH1 was significantly correlated to GLS, HPRT1, GBE1, ACHE, STXBP1, while SNCA was related to HPRT1, GLS, DDC, ACHE, STXBP1. In the GSE8397 dataset, the AUC values of the ROC curves for the nine hub genes were greater than 0.8, suggesting a strong diagnostic potential when these genes are considered individually (Figure 5A). Moreover, the AUC value for the set of nine hub genes was 1, indicating a better diagnostic efficacy as a group than the individual genes (Figure 5B). Additionally, PCA revealed that the combination of all nine hub genes could better distinguish control and PD samples than individual genes (Figure 5C). In GSE7621 dataset, the AUC values also demonstrated the superior diagnostic effect of the hub genes compared to individual genes as well. However, the capability of the entire set of nine genes did not perform as robustly as in the training set, signifying a potential variance in the gene expression profiles or patient cohorts between two datasets (Figure 5D–F).

Figure 4.

Confirmation of hub genes. (A) Venn diagram of nine hub genes by comparison of immune cell-related genes, DEGs and CMRGs. (B) Protein-protein interaction network of hub genes (HRPT1, GLS, SNCA, MDH1, STXBP1, ACHE, DDC and GBE1).

Figure 5.

Evaluation and verification of the diagnostic values of the nine hub genes. (A) ROC curves of nine hub genes in the GSE8397 dataset. (B) ROC curve of the diagnostic model in GSE8397 dataset. (C) Principal component analysis (PCA) on PD and control samples from the GSE8397 dataset based on nine hub genes. (D) ROC curves of nine hub genes in the GSE7621 dataset. (E) ROC curve of the diagnostic model in GSE7621 dataset. (F) PCA on PD and control samples from the GSE7621 dataset based on nine hub genes.

3.5. Single gene GSEA

By integrating the results of GSEA of hub genes, nine hub genes were enriched to KEGG pathways of cytokine-cytokine receptor interaction, complement and coagulation cascades, cardiac muscle contraction, axon guidance, autoimmune thyroid disease, amyotrophic lateral sclerosis (ALS), amino sugar and nucleotide sugar metabolism, Alzheimer’s disease, alanine aspartate and glutamate metabolism (Figure 6).

Figure 6.

Single gene GSEA. Merged KEGG pathways enriched for nine hub genes using GSEA enrichment analysis.

3.6. IPA

IPA analysis suggested 51crucial pathways, with the top 20 pathways shown in Figure 7A. Notably, SNARE signaling pathway and synaptogenesis signaling pathway exhibited a negative associated with hub genes, while dopamine receptor signaling and cAMP-mediated signaling showed a positive association. Furthermore, our results indicated that the nine hub genes were involved in six IPA networks. For instance, SV2C interacted with SNCA indirectly, while Rab5 directly interacted with AGTR1, which could interact indirectly with AP3B2. Finally, ACHE had an indirect impact on the nicotinic acetylcholine receptor (Figure 7B–G).

Figure 7.

IPA of hub genes from the GSE8397 dataset. (A) Diseases and functional pathway analysis of nine hub genes. While the Y-axis is the pathway terms, the X-axis is the log (p-value). (B) Interaction network was constructed between hub genes (SNCA and GLS) by using IPA. (C) Interaction network was constructed for gene AGTR1 by using IPA. (D) Interaction network was constructed between hub genes (ACHE, DDC, and HPRT1) by using IPA. (E) Interaction network was constructed between hub genes (MDH1 and DDC) by using IPA. (F) Interaction network was constructed for gene DDC by using IPA. (G) Interaction network was constructed for gene GBE1 by using IPA. Solid lines indicate direct action, dashed lines indicate indirect action.

3.7. Construction of miRNA-mRNA-TF network, ceRNA network and drug-hub genes network

A miRNA-mRNA-TF network comprising 635 edges was constructed using 122 miRNAs and 243 TFs predicted by the nine hub genes (Figure 8A, Supplementary Table 7). Hsa-miR-6753-5p, YY1 and MYC were found to regulate the transcription of ACHE. 8 miRNAs and 214 lncRNAs were obtained using chipExpNum > 1 by chosing energy < −30. Three hub genes did not predict miRNAs. The ceRNA network was built according to the 8 lncRNAs and 214 miRNAs (Figure 8B, Supplementary Table 8). The results indicated that XIST regulated STXBP1 through hsa-miR-4739, while ACHE could be modulated by LINC01521 through hsa-miR-671-5p. Furthermore, the drug-gene network was presented with 270 nodes (9 hub genes and 261 drugs) and 641 edges (Figure 8C, Supplementary Table 9), The complexity of network underscored the potential for repurposing existing drugs as therapeutic agents for PD. For example, we found that expression of GBE1 and SNCA could be upregulated by 1-Methyl-4-phenylpyridinium (D015655). Thimerosal (D013849) exhibited to inhibit the expression of GBE1, while simultaneously upregulating the expression of GLS. Valproic Acid (D014635) could enhance the expression of HPRT1 but reduce the GLS production.

Figure 8.

Development of miRNA-mRNA-TF network, ceRNA network, and gene-drug network (A) a miRNA-mRNA-TF network with 635 edges based on nine hub genes, 122 miRNAs, and 243 TFs. (B) a mRNA-miRNA-lncRNA network with 225 edges based on three hub genes, eight lncRNAs, and 214 miRNAs. (C) gene-drug network with 641 edges based on nine hub genes and 261 drugs.

3.8. Validation of hub genes

To corroborate the predicted hub genes identified through in silico analysis, RNA samples from the CSF of the PD patients were extracted and analyzed. qRT-PCR was employed to verify the expression patterns of the nine hub genes. The results revealed that eight of the hub genes-HPRT1, SNCA, GBE1, DDC, STXBP1, ACHE, GLS and AGTR-were remarkably downregulated in PD patients, aligning with the in-silico predictions. In contrast, the expression of MDH1 was observed with no significant alterations (Figure 9, Supplementary Table 10).

Figure 9.

Validation of the expression levels of nine hub genes by qRT-PCR. (A) AGTR1. (B) MDH1. (C) GBE1. (D) HRRT1. (E) ACHE. (F) SNCA. (G) DDC. (H) GLS. (I) STXBP1.

3.9. GBE1 gene function experiment

GBE1 attracted attention as a gene of significant interest, evidenced by its highest AUC value in our analyses. Moreover, GBE1 represents a potentially novel gene within the context of Parkinson’s Disease (PD) pathology that has yet to be thoroughly investigated. Owing to its potential importance, GBE1 was chosen for subsequent experimental validation. The qRT-PCR results indicated successful overexpression of GBE1 in SH-SY5Y cells (Figure 10A). Subsequent cck8 experiment revealed that overexpression of GBE1 gene significantly enhanced the proliferative ability of SH-SY5Y cells (Figure 10B). Notably, the Wound Healing assay demonstrated an improved migratory ability in SH-SY5Y cells with GBE1 overexpression (Figure 10C), suggesting a role of in cellular dynamics relevant to PD. Live/dead staining further indicated a higher viability of SH-SY5Y cells post GBE1 overexpression (Figure 10D) (Supplementary Table 11), supporting the gene’s contributory role in enhancing cell survival.

Figure 10.

Functional analysis of GBE1 gene in SH-SY5Y cells (A)the qRT-PCR assay showed over-expression of GBE1 in the oe-GBE1 group; (B)The CCK-8 test showed higher cell proliferation in the oe-GBE1 group compared to the oe-NC group; (C)The Wound Healing assay indicated an increase in cell migration in the oe-GBE1 group compared to the oe-NC group; (D) Live/dead staining showed increased cell viability in theoe-GBE1 group compared to the oe-NC group. **, *** respectively represent P-value less than 0.01, less than 0.001.

4. Discussion

Parkinson’s disease is a chronic neurodegenerative disease for which effective cures have not been found in clinic. The molecular mechanisms underlying PD pathogenesis are complicated and have not been fully elucidated. In recent years, numerous studies on metabolism of metal ions have attracted scientists’ attention. Some groups have reported that chronic exposure to heavy metals, including copper, increase the prevalence of PD [20, 21]. Additional studies have shown that excess or deficiency of copper ions can catalyze the harmful redox of oxygen containing derivatives, leading to the occurrence of PD [22]. Meanwhile, previous studies demonstrated that protein-bound copper influenced the level of iron in the brain via ceruloplasmin, an iron oxidase enzyme. Therefore, reducing copper bound to proteins in the brain might increase iron buildup and consequent oxidative stress [23]. In this study, we focused on copper metabolism-related biomarkers in PD patients through bioinformatics analysis, providing potential biomarkers for treatment of PD.

Our investigation initially identified 21 DECMRGs enriched in significant pathways such as dopaminergic synapse, folate biosynthesis, Parkinson’s disease, tyrosine metabolism, dopamine metabolic process catecholamine metabolic process and cellular amine metabolic process. The degeneration of dopamine neurons in Parkinson’s disease originates from striatal axons and synaptic terminals. It’s reported that injection of 6-hydroxydopamine in rats lead to axonal loss following dopamine production in the striatum [24]. Moreover, lower level of plasma folate was detected in PD patients, and in vivo experiments demonstrated that folate deficiency in mice had been shown to facilitate dopaminergic neurodegeneration and exacerbate motor dysfunction [25, 26]. Additionally, phenylalanine and tyrosine serve as dopamine precursors, and a low tyrosine/phenylalanine ratio in serum of PD patients suggested that tyrosine may affect PD by regulating dopamine synthesis [27]. Complementary findings have highlighted that a significant increase of catechol compounds within the substantia nigra of the injury side following the loss of dopamine neurons in the substantia nigra-striatum of PD rat models. Together, these insights encouraged us to further explore the role of DE CMRGs in PD patients.

Through bioinformatics analysis, nine hub genes that in association with copper metabolism with immunological process in PD were identified: HPRT1, GLS, SNCA, MDH1, GBE1, DDC, STXBP1, ACHE, AGTR1. Notably, HPRT1 (hypoxanthine-guanine phosophoribosyltransferase) a crucial role in the purine salvage pathway. It is involved in cell cycle regulation by modulating guanine hypoxanthine [28, 29]. As a crucial factor in the purine rescue pathway, HPRT1 plays a role in mediating cancer cell proliferation, autophagy and apoptosis-related processes. HPRT deficiency can lead to neurobehavioral syndromes, dysregulation of Wnt/β-catenin signaling pathway, abnormal nigrostriatal axonal arborization and impaired function of dopaminergic neurons and dopamine pathways [30–32]. Furthermore, GLS is responsible for encoding brain-specific glutaminase. However, GLS deficiency could result in spastic disorders and optic atrophy [33]. Mutations in SNCA are well-recognized for contributing to pathological aggregation of α-synuclein to form Lewy bodies, which in turn lead to Parkinson’s disease, and a higher number of aberrant copies of the SNCA gene leads to an earlier age of onset and an increase in the severity of Parkinson’s disease [34, 35]. Moreover, MDH1 (Malate dehydrogenase 1), implicated in both PD and Alzheimer’s disease adds to the complexity of neurodegenerative process [33]. The branching enzyme GBE1 emerged as a key gene a key glycogen synthesis, and its deficiency damages skeletal muscle and the nervous system. Actually, research on GBE1 is not extensive in PD contexts. Our study, for the first time, revealed it reduced expression in the CSF of PD patients. Complementary cellular experiments in a PD cellular model affirmed that overexpression of GBE1 promoted the proliferation and migration ability and cell viability of SH-SY5Y [36–38]. Furthermore. Dopa-Decarboxylase (DDC) encodes levodopa in the brain, and the activity of L-DOPA decarboxylase in the striatum of PD patients was reduced [39]. The protein STXBP1 (syntaxin-binding protein 1) is now investigated to modulates the self-replicating aggregation of α-synuclein, a presynaptic protein implicated in various neurodegenerative disorders collectively known as synucleinopathies, including Parkinson’s disease [40]. Finally, AT1R (angiotensin II receptor 1) is one of the important coding genes of the RAS system. In the PD mouse model, the expression level of AT1R was significantly downregulated in nerve cells and neurons [41].

The single-gene GSEA enrichment analysis unveiled that nine hub genes were associated with cytokine receptor interaction. Interestingly, inflammatory cytokines, such as TGF-β1, IL-6 and IL-1β, were identified with increased expression to be highly expressed in the CSF of patients, consistent with a previous literature [42]. Further investigations demonstrated that IL-1β exacerbated progressive dopaminergic cell death [43]. Moreover, in a PD mouse model, specific axon guidance molecules and their receptors were implicated in the axon guidance of transplanted neurons, promoting the repair of degenerated nigrostriatal pathways [44]. These findings suggest a contributory role of hub genes in PD pathology through axon guidance mechanisms. Additionally, excessive glutamate can result in neurotoxicity by over activating (NMDAR) N-methyl D-aspartate. Intriguingly, memantine, an antagonist of aspartate receptors, has shown potential in ameliorating cognitive and memory deficits [45]. Glutamate is the major excitatory neurotransmitter in the central nervous system. Glutamate-induced excitotoxicity resulted from glial dysfunction contributed to the neurodegeneration in PD [46]. Therefore, further research is necessary to investigate the relationship between Alzheimer’s disease, alanine aspartate, glutamate metabolism and PD.

In the study, Hsa-miR-6753-5p, YY1 and MYC were identified as potential regulators of ACHE transcription. However, the precise mechanisms by which these factors modulate the expression of ACHE remain unclear. Future experiments should be conducted to unravel the complex regulatory pathways involved. Furthermore, previous studies have demonstrated the potential therapeutic effects of various compounds on PD [47]. For instance, chronic intraventricular administration of 1-methyl-4-phenylpyridinium for Parkinson’s disease has been demonstrated in a rat chronic model of PD [48]. Compounds such as thimerosal have been observed to regulate dopamine levels. Similarly, valproic acid (VPA) has also been reported to protect neurons and treat PD by upregulating trophic factors and protective proteins in the brain [49]. Elucidating the regulatory relationship in miRNA-mRNA-TF network, ceRNA network and drug-hub genes network can contribute to rational design of future PD research.

The decreased expression of hub genes was validated in human CSF samples, which was consistent with the bioinformatic analysis and underscoring the precision of our in-silico predictions. Furthermore, this validation suggested that individuals exhibiting lower expression of these hub genes were likely to be patients with Parkinson’s disease. While, GBE1 was distinguished by its highest AUC value, signaling its strong predictive power for disease progression. In the in vitro validation, Parkinson’s disease (PD) model (SH-SY5Y cells treated with ROT) with overexpression of GBE1 sustained viability but also exhibited enhanced proliferative and wound-healing capabilities. These results suggest that GBE1 plays a protective role in cellular mechanisms. The ability of GBE1 to promote cell survival and repair in adverse conditions might reflect its therapeutic potential in attenuating the progression of atherosclerosis by enhancing the regenerative capacity of vascular cells.

5. Conclusion

In summary, our study has identified nine hub genes and their enrichment analysis, which could potentially serve as biomarkers for the PD research. Nevertheless, some limitations need to be acknowledged. The hub genes were identified through bioinformatic analysis and were validated by using human samples and cells. However, future experimentation on animal models and molecular mechanisms is necessary to confirm their roles in biological network.

Funding Statement

Shandong Province Traditional Chinese Medicine Science and Technology Project (Q-2023200).

Authors contributions

JL: Conceptualization, Methodology, Writing - Original Draft. GZ: Data Curation, Supervision, Writing - Original Draft. BL: Data Curation, Writing - Original Draft. YS: Formal analysis, Investigation. XJ: Formal analysis, Data Curation. MW: Visualization, Conceptualization. JZ: Validation, Supervision, Writing - Review & Editing. ZX: Supervision, Writing - Review & Editing. All of the above authors reviewed the article and agreed to the final version.

Consent to participate

Informed consent was signed by all participants in this study.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Ethics statement

The present study received ethical approval from the Ethics Committee of Liaocheng People’s Hospital (Approval No. 2023029).

Data availability statement

The datasets used in this investigation are available in internet repositories. https://www.ncbi.nlm.nih.gov/, GSE8397 and GSE7621. The authors confirm that the data supporting the findings of this study are available within the article or its supplementary materials.