Brain-derived extracellular vesicle microRNAs in Lewy body and Alzheimer's disease

Abstract

Introduction: Robust plasma-based biomarkers to distinguish Lewy body disease (LBD) and Alzheimer's disease (AD) are currently lacking. We applied track-etch magnetic nanopore (TENPO) sorting for enrichment of brain-derived extracellular vesicle (EV) signatures as potential biomarkers to address this gap.

Methods: We analyzed plasma from 137 autopsy-confirmed patients [30 LBD, 31 AD, 30 AD/LBD, 19 AD with amygdala Lewy bodies (AD/ALB), and 27 controls], sequencing miRNAs from TENPO-isolated GluR2-positive (neuron-enriched) and GLAST-positive (astrocyte-enriched) EVs, and measuring plasma proteins (Aβ40, Aβ42, tau, p-Tau181, p-Tau231) via SIMOA.

Results: We identified 16 GluR2+, 8 GLAST+, and 4 protein biomarkers with differential expression (false discovery rate-corrected P value < .1) between LBD and AD. A multimodal 15-feature panel classified LBD versus AD with 10-fold crossvalidated accuracy = 0.95 and area under the curve (AUC) = 0.96.

Discussion: Brain-derived EVs offer accurate and accessible miRNA biomarkers for the differential diagnosis of LBD and AD.

Keywords: Alzheimer’s disease; Lewy body disease; astrocyte; extracellular vesicles; microRNA; microfluidics; neurodegenerative dementia; neuron; plasma biomarkers; protein biomarkers.

Figure 1:. Outline of the study and differential classifications performed.

(A) Outline of study time-course and sample processing for EVs. Antemortem plasma samples with postmortem pathology confirmation of neurological diagnoses were processed to isolate tissue-specific EVs via our microfluidic platform, alongside plasma protein biomarkers via commercial digital ELISA, for patients with AD, LBD, AD/LBD, AD/ALB and controls. (B) Left: schematic of operation of TENPO platform for tissue-specific EV isolation using antibody-labeled MNPs; right: SEM images of EVs isolated via GluR2 (top) and GLAST (bottom) pulldowns immobilized on TENPO. Abbreviations: LBD, Lewy body disease; AD, Alzheimer’s disease; AD/LBD, mixed Alzheimer’s and Lewy body disease; AD/ALB, Alzheimer’s disease with amygdala Lewy bodies; EV, extracellular vesicle; TENPO, track-etch magnetic nanopore; ELISA, enzyme-linked immunosorbent assay; MNP, magnetic nanoparticle; SEM, scanning electron microscopy.

Figure 2:. Differential expression of GluR2+ EV miRNAs, GLAST+ EV miRNAs, and plasma proteins in the LBD versus AD comparison.

(A) Heatmap of z-score of log₂(expression) for biomarkers with FDR-corrected P value < .1. Subjects (columns) are hierarchically clustered within cohort and biomarkers within each compartment (rows) are sorted by descending fold-change. (B) Volcano plot demonstrating differential expression of GluR2+ EV miRNAs, GLAST+ EV miRNAs, and plasma proteins. (C) Venn diagram showing overlap in FDR P value significant miRNAs between GluR2+ EVs and GLAST+ EVs. (D) Top 30 biomarkers in all compartments ranked by descending AUC. Error bars represent standard error from bootstrapping 10x. Abbreviations: LBD, Lewy body disease; AD, Alzheimer’s disease; EV, extracellular vesicle; FDR, false discovery rate; AUC, area under the curve.

Figure 3:. GO and KEGG analysis of FDR P value significant miRNAs in the LBD versus AD comparison.

GO and KEGG pathway analyses were performed via DIANA miRPath v4.0 using the TarBase v8.0 database. FDR P values were calculated using a one-sided Fisher’s exact test. The top 10 (ranked by number of target genes) terms within each of the three GO categories (BP, CC, MF) and top 10 (ranked by number of target genes) KEGG pathways were identified for each pulldown. (A) Top 10 terms within each GO category for GluR2+ EV miRNAs. (B) Top 10 KEGG pathways for GluR2+ EV miRNAs. (C) Top 10 terms within each GO category for GLAST+ EV miRNAs. (D) Top 10 KEGG pathways for GLAST+ EV miRNAs. In all panels, each bar is labeled to the right with the number of differentially expressed miRNAs associated with the given GO term or KEGG pathway. Abbreviations: LBD, Lewy body disease; AD, Alzheimer’s disease; EV, extracellular vesicle; FDR, false discovery rate; GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; BP, biological process; CC, cellular component; MF, molecular function.

Figure 4:. LASSO analysis for differentiating LBD (n = 30) from AD (n = 31).

(A) (A) Heatmap of z-score of log₂(expression) for LASSO-selected biomarkers. Subjects (columns) are hierarchically clustered within cohort and biomarkers within each compartment (rows) are sorted by descending AUC. (B) Kendall correlation staircase plots identifying the extent to which biomarker information was correlated between the LASSO-selected GluR2+ EV, GLAST+ EV, and protein biomarkers. Biomarkers are sorted within compartments by AUC. The inset shows the correlation distribution of Kendall’s τ, where the dotted line represents the median count. (C) LASSO panel accuracy versus panel size for classifying LBD versus AD, shown in blue; accuracy is assessed via 10-fold cross-validation, with error bars from 5 repeats of panel training on the LBD vs AD patient groups. Average accuracy and standard error for control experiments performed by scrambling patient labels 10x are shown in orange. (D) LASSO panel AUC versus panel size for classifying LBD versus AD, shown in blue with error bars as described in (C). Average AUC and standard error for the same control experiments described in (C) are shown in orange. (E) AUCs for the 15-marker LASSO panel and individual LASSO biomarkers, sorted by descending AUC. Error bars represent 95% confidence intervals, calculated from 5x repeats of 10-fold cross validation for the 15-marker panel or from bootstrapping 10x for individual markers. Abbreviations: LBD, Lewy body disease; AD, Alzheimer’s disease; EV, extracellular vesicle; LASSO, least absolute shrinkage and selection operator; AUC, area under the curve.