Common Factors in Neurodegeneration: A Meta-Study Revealing Shared Patterns on a Multi-Omics Scale

Abstract

Neurodegenerative diseases such as Alzheimer's disease (AD), Parkinson's disease (PD), Huntington's disease (HD), and amyotrophic lateral sclerosis (ALS) are heterogeneous, progressive diseases with frequently overlapping symptoms characterized by a loss of neurons. Studies have suggested relations between neurodegenerative diseases for many years (e.g., regarding the aggregation of toxic proteins or triggering endogenous cell death pathways). We gathered publicly available genomic, transcriptomic, and proteomic data from 177 studies and more than one million patients to detect shared genetic patterns between the neurodegenerative diseases on three analyzed omics-layers. The results show a remarkably high number of shared differentially expressed genes between the transcriptomic and proteomic levels for all conditions, while showing a significant relation between genomic and proteomic data between AD and PD and AD and ALS. We identified a set of 139 genes being differentially expressed in several transcriptomic experiments of all four diseases. These 139 genes showed overrepresented gene ontology (GO) Terms involved in the development of neurodegeneration, such as response to heat and hypoxia, positive regulation of cytokines and angiogenesis, and RNA catabolic process. Furthermore, the four analyzed neurodegenerative diseases (NDDs) were clustered by their mean direction of regulation throughout all transcriptomic studies for this set of 139 genes, with the closest relation regarding this common gene set seen between AD and HD. GO-Term and pathway analysis of the proteomic overlap led to biological processes (BPs), related to protein folding and humoral immune response. Taken together, we could confirm the existence of many relations between Alzheimer's disease, Parkinson's disease, Huntington's disease, and amyotrophic lateral sclerosis on transcriptomic and proteomic levels by analyzing the pathways and GO-Terms arising in these intersections. The significance of the connection and the striking relation of the results to processes leading to neurodegeneration between the transcriptomic and proteomic data for all four analyzed neurodegenerative diseases showed that exploring many studies simultaneously, including multiple omics-layers of different neurodegenerative diseases simultaneously, holds new relevant insights that do not emerge from analyzing these data separately. Furthermore, the results shed light on processes like the humoral immune response that have previously been described only for certain diseases. Our data therefore suggest human patients with neurodegenerative diseases should be addressed as complex biological systems by integrating multiple underlying data sources.

Keywords: Alzheimer’s disease; Huntington’s disease; Parkinson’s disease; amyotrophic lateral sclerosis; multi-omics; neurodegeneration.

Figure 1

Classification of candidate pathways contributing to neurodegeneration into three groups according to their cellular mechanisms or their primary site of action. The categorization is based on Ramanan’s pathways to neurodegeneration [15] and aims to help in classifying the current knowledge surrounding neurodegeneration.

Figure 2

Workflow Overview: Data acquisition was performed using the genome-wide association studies (GWAS) Catalog for genomic data, the European Bioinformatics Institute (EMBL-EBI) Expression Atlas and the Gene Expression Omnibus database for transcriptomic data, and literature research in PubMed and Google Scholar for proteomic data. After filtering these raw data tables and applying some data transformation, the processed data were used for the data analysis. For every omics layer, the intersections of all four analyzed NDDs were visualized as Venn diagrams. Common transcriptional patterns were searched with a hierarchical clustering approach and visualized as a heatmap showing the mean transcriptional direction of regulation per gene, and a dendrogram showing the clustering results. Finally, each set of genes after the intersections was used for the Kyoto Encyclopedia for Genes and Genomes (KEGG) pathway and GO-Term analyses.

Figure 3

Venn diagrams and hypergeometric test results for the overlap between significant single nucleotide polymorphism (SNP)-trait associations (genomic level) and significantly differentially expressed genes on the transcriptomic (middle) and proteomic (bottom) levels for AD, PD, HD and ALS. All tested intersections show highly significant enriched numbers of overlapping genes for the transcriptomic and proteomic data. The genomic data show significantly enriched numbers of overlapping genes for the AD-PD and AD-ALS intersections.

Figure 4

Heatmap with hierarchical clustering results of the mean regulation of all 139 genes that were maintained in all four NDD transcriptomic data. The clustering led to an inner cluster containing HD and AD transcriptomic data. This cluster was next clustered to the ALS transcriptomic data and finally these three NDD were clustered to PD. Colors represent the mean direction of regulation (see. Equation (1)).

Figure 5

Directed acyclic graph showing the significant results (false discovery ratio (FDR) ≤ 0.05) of the GO-Term ORA of the Cellular Component and Molecular Function of the transcriptomic overlap of all four NDDs (A,B), the Biological Process terms of the proteomic data (C) and the Biological Process terms of the transcriptomic data after affinity propagation (D). Blue shading indicates the value of the FDR. For better readability, all GO-Terms leading to the significant ones were hidden in (A,D).

Figure 5

Directed acyclic graph showing the significant results (false discovery ratio (FDR) ≤ 0.05) of the GO-Term ORA of the Cellular Component and Molecular Function of the transcriptomic overlap of all four NDDs (A,B), the Biological Process terms of the proteomic data (C) and the Biological Process terms of the transcriptomic data after affinity propagation (D). Blue shading indicates the value of the FDR. For better readability, all GO-Terms leading to the significant ones were hidden in (A,D).

Figure 6

Significant results (FDR < 0.05) of the biological process (BP) GO-Term analysis for the transcriptomic overlap in the four analyzed NDDs after performing affinity propagation. The names of the enriched sets are shown in the left column, followed by the contributing genes on their right. For each significant set, the FDR and enrichment is given. The enriched BP sets are categorized in the groups Intracellular Mechanisms, Local Tissue Environment and Systemic Environment based on Ramanan’s conceptual model of candidate pathways contributing to neurodegeneration [15], which is also depicted in Figure 1.