Plasma miRNAs across the Alzheimer's disease continuum: Relationship to central biomarkers

Abstract

Introduction: MicroRNAs (miRNAs) play important roles in gene expression regulation and Alzheimer's disease (AD) pathogenesis.

Methods: We investigated the association between baseline plasma miRNAs and central AD biomarkers from the Alzheimer's Disease Neuroimaging Initiative (ADNI; N = 803): amyloid, tau, and neurodegeneration (A/T/N). Differentially expressed miRNAs and their targets were identified, followed by pathway enrichment analysis. Machine learning approaches were applied to investigate the role of miRNAs as blood biomarkers.

Results: We identified nine, two, and eight miRNAs significantly associated with A/T/N positivity, respectively. We identified 271 genes targeted by amyloid-related miRNAs with estrogen signaling receptor-mediated signaling among the enriched pathways. Additionally, 220 genes targeted by neurodegeneration-related miRNAs showed enrichment in pathways including the insulin growth factor 1 pathway. The classification performance of demographic information for A/T/N positivity was increased up to 9% with the inclusion of miRNAs.

Discussion: Plasma miRNAs were associated with central A/T/N biomarkers, highlighting their potential as blood biomarkers.

Highlights: We performed association analysis of microRNAs (miRNAs) with amyloid/tau/neurodegeneration (A/T/N) biomarker positivity. We identified dysregulated miRNAs for A/T/N biomarker positivity. We identified Alzheimer's disease biomarker-specific/common pathways related to miRNAs. miRNAs improved the classification for A/T/N positivity by up to 9%. Our study highlights the potential of miRNAs as blood biomarkers.

Keywords: Alzheimer's disease; amyloid; biomarkers; classification; microRNAs; neurodegeneration; plasma; tau.

FIGURE 1

Workflow of miRNA‐seq analysis with detailed tools and databases. A, amyloid; miRNA‐seq, microRNA sequencing; N, neurodegeneration; qPCR, quantitative polymerase chain reaction; T, tau.

FIGURE 2

Volcano and Venn plots of miRNAs related to the A/T/N biomarkers. A, Volcano plot for miRNAs identified related to the A biomarker. B, Volcano plot for miRNAs identified related to the T biomarker. C, Volcano plot for miRNAs identified related to the N biomarker (hsa‐miR‐1180‐3p has an adjusted p value smaller than smallest floating‐point value in R, so its adjusted p value is displayed as its closest adjusted p value listed [3.44E‐14]). Note: miRNAs with a baseMean < 10 are not shown in the volcano plot; red colored dots represent miRNAs with an absolute value of log₂FC ≥ 0.26 and adjusted p value < 0.05; blue colored dots represent miRNAs with an adjusted p value < 0.05; green colored dots represent miRNAs with an absolute value of log₂FC ≥ 0.26; gray colored dots indicate miRNAs that failed to pass these conditions. D, Venn plots of significant miRNAs detected in the three A/T/N biomarker groups. A, amyloid; log₂FC, log₂ fold change; miRNA‐seq, microRNA sequencing; N, neurodegeneration; T, tau.

FIGURE 3

Comparison of log₂ fold change and p values of significant miRNAs across the A/T/N biomarkers. Note: * indicates 0.01 ≤ adjusted p value < 0.05; ** indicates 0.001 ≤ adjusted p value < 0.01; *** indicates adjusted p value < 0.001. Red stars indicate the absolute value of log₂FC ≥ 0.26 and adjusted p value < 0.05. A, amyloid; log₂FC, log₂ fold change; miRNA‐seq, microRNA sequencing; N, neurodegeneration; T, tau.

FIGURE 4

MiRNA target gene network. The interaction networks were constructed using target genes of miRNAs significantly associated with either A positivity (A) or N positivity (B). MiRNAs highlighted in red indicate upregulation, while those in blue represent downregulation. The orange or purple solid line denotes miRNA‐target interactions predicted by both miRTarBase and TarBase, whereas the green dashed line represents protein–protein interactions sourced from STRING. The border color of target genes corresponds to the number of interactions: yellow for three interactions, orange for four interactions, green for five interactions, and red for six interactions. The network was constructed by the Cytoscape software. miRNA, micro RNA.

FIGURE 5

Pathway enrichment network. Top 20 clusters of enriched terms in Metascape were generated using target genes predicted by miRNAs significantly associated in A positivity (A) and N positivity (B). The color indicates their cluster ID, while the thickness of the edge reflects the similarity score. The 28 nodes with red borders represent terms that are enriched in both A and N positivity. A, amyloid; N, neurodegeneration.

FIGURE 6

Classification performance of the A/T/N biomarker positivity using the random forest machine learning algorithm. A, The ROC‐AUC curves of classification for amyloid positivity; (B) the ROC‐AUC curves of classification for tau positivity; (C) the ROC‐AUC curves of classification for neurodegeneration positivity. Note: The dark blue curve indicates the mean AUC curve from 5‐fold cross‐validation tests and the background curves indicate the ROC curve for each cross‐validation fold, while the gray shade indicates the standard deviation of the mean ROC curve; and (D) mean and standard deviation for the performance metrics based on different combinations of features, computed across three independent runs of 5‐fold cross‐validation. “miRNAs in the Model Features” indicates including all miRNAs in the analysis; “DE miRNAs in the Model Features” indicates including differentially expressed miRNAs for each biomarker in the analysis, respectively. A, amyloid; AUC, area under the curve; miRNA, micro RNA; N, neurodegeneration; ROC, receiver operating characteristic; T, tau.