Differential gene expression and immune profiling in Parkinson's disease: unveiling potential candidate biomarkers

Abstract

Background: Parkinson's disease (PD) represents a common neurodegenerative disorder characterized by a multifaceted interaction with immune infiltration. Despite a well-defined clinical diagnosis, the misdiagnosis rate of PD remains around 20%. The aim of this study is to discover new diagnostic biomarkers for PD and investigate their pathogenesis to improve early intervention and effective management of patients with PD.

Methods: Five PD-related GEO datasets were used: four for training (GSE7621, GSE8397, GSE20186, and GSE20292) and one for validation (GSE26927). Gene expression analysis included batch correction and "RobustRankAggreg" (RRA) methods. Differentially expressed genes (DEGs) were linked to functions via Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG). Hub genes were identified using CytoHubba in Cytoscape and validated with ROC analysis. Real-time quantitative polymerase chain reaction (RT-qPCR) confirmed hub gene expression in PD patients' substantia nigra. CIBERSORT, along with the Wilcoxon test and Least Absolute Shrinkage and Selection Operator (LASSO) regression, analyzed differences in immune cell abundance between PD patients and healthy controls (HC). Spearman's rank correlation in R explored the link between biomarkers and immune cells.

Results: The intersection of two methods identified 124 DEGs in PD. GO analysis revealed enrichment in neurotransmitter transport, while KEGG analysis identified involvement in the dopaminergic synapse pathway. Three hub genes (DDC, NEFL, and SLC18A2) were identified using the "UpSet" R package, and their expression was significantly lower in PD patients than in the HC group (all p < 0.05), as confirmed by RT-qPCR. LASSO regression and ROC analysis demonstrated that SLC18A2 could diagnose PD with high specificity and sensitivity in both training (0.85 and 0.84) and validation sets (1.00 and 0.75). CIBERSORT analysis showed increased memory B cells, activated mast cells, NK cells, and CD8⁺ T cells in PD, with notable differences in the abundance of memory B cells and activated mast cells between PD and HC.

Conclusion: The study identifies SLC18A2 as a potential candidate biomarker for PD and emphasizes the involvement of memory B cells and activated mast cells in the onset and progression of the disease.

Keywords: Bioinformatics; Diagnostic biomarkers; Memory B cells; Parkinson's disease; SLC18A2.

Fig. 1

Data preprocessing. Box plot and principal component analyses (PCA) were conducted to address batch correction in the datasets GSE7621, GSE8397, GSE20186, and GSE20292. Panels (A, B) depict the data before batch correction, while panels (C, D) show the data after batch correction. TPM, Transcripts Per Million

Fig. 2

Identification of Differentially Expressed Genes (DEGs). (A) The top 20 upregulated and downregulated DEGs across the four datasets were identified using the “Robust Rank Aggregation” (RRA) method. (B-C) A heatmap (B) and a volcano plot (C) were employed to visualize the final DEGs identified through the “Batch Correction” methods. (D) A Venn diagram was utilized to determine the intersection of DEGs identified by the two methods. HC, healthy controls (n = 50). PD, Parkinson’s disease (n = 46)

Fig. 3

The functional analysis was conducted on the DEGs common to both the Parkinson’s disease (PD) group (n = 50) and the healthy controls (HC) (n = 46). (A) The Gene Ontology (GO) enrichment analysis was performed on these overlapping DEGs. (C) Similarly, the Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis was carried out for the same set of DEGs. (B, D) The outcomes of the GO (B) and KEGG (D) analyses are depicted using circular charts

Fig. 4

Identification of Hub Genes. (A) A correlation network of the overlapping DEGs was constructed using Cytoscape. (B) Ten algorithms were employed to identify hub genes utilizing the R package “UpSet”. (C) The expression levels of three hub genes were visualized through a heatmap in the combined microarray dataset. HC, healthy controls (n = 50). PD, Parkinson’s disease (n = 46)

Fig. 5

Validation of Hub Genes. The RNA-Seq dataset GSE26927 was used to validate the expression levels of SLC18A2, NEFL, and DDC. The results were presented as both a heatmap (A) and a violin plot (B-D). (E-G) RT-qPCR analysis was conducted to assess the mRNA expression of SLC18A2, NEFL, and DDC in the substantia nigra of HC (n = 6) and PD (n = 10) patients. *p < 0.05, **p < 0.01. HC, healthy controls. PD, Parkinson’s disease

Fig. 6

Evaluation of biomarker diagnostic efficacy for PD. (A) Receiver Operating Characteristic (ROC) curves were employed to evaluate the diagnostic performance of three hub genes within the training datasets. Furthermore, the diagnostic effectiveness of these three hub genes was compared with that of other published biomarkers. (B-C) LASSO regression model l-screening. (D, G) ROC curves were generated for the training and testing models. (E, H) Confusion matrix for the training and testing sets. The GSE26927 dataset served as the validation dataset. (F) ROC curves were utilized to assess the diagnostic performance of three hub genes in the testing datasets.

Fig. 7

The relative abundance of immune cells within the substantia nigra of patients with PD was investigated. The composition of 22 distinct immune cell types in each sample was depicted using both a histogram (A) and a heatmap (B). (C) The correlations among these 22 immune cell types within the substantia nigra tissues of PD patients were assessed, with positive correlations indicated in red and negative correlations in blue. (D) Principal Component Analysis (PCA) was conducted to differentiate the infiltrating immune cells between PD-affected and normal substantia nigra tissues. HC, healthy controls (n = 16). PD, Parkinson’s disease (n = 31)

Fig. 8

The study aimed to identify immune cells that exhibit significant differences in abundance in PD and to explore their correlations with relevant biomarkers. (A-C) The Wilcoxon test (A) and LASSO regression (B-C) were employed to analyze the abundance variations of immune cells between individuals with PD and healthy controls. (D) Correlation among three effective biomarkers and two significantly differentially abundant immune cell types. HC, healthy controls (n = 16). PD, Parkinson’s disease (n = 31)