Page 23 - Read Online

Eitan et al. Extracell Vesicles Circ Nucleic Acids 2023;4:133-150 https://dx.doi.org/10.20517/evcna.2023.13 Page 147

associations with cognition across the entire range of AD severity since all participants had the early disease
and their cognitive score range was narrow. It is also interesting that, unlike other synaptic proteins, NRGN
was higher in early AD compared to controls. Whether this finding represents a specific biological process
(such as ongoing synaptic damage) in AD remains to be determined by mechanistic studies.

Reproducibility between cohorts is of high importance for biomarker development [55,56] . Here, we used
multiple cohorts to examine the robustness of NDEV isolation. While most differences between early AD
and control groups were consistent across cohorts, the absolute levels of several biomarkers varied widely
between cohorts. We consider it likely that these differences were partly due to inconsistencies in
preanalytical parameters (blood draws, plasma processing, etc.), which are known to influence EV
analyses [57-59] . This highlights the importance of controlling and monitoring preanalytical conditions and
incorporating quality controls during EV biomarker development. The development of preanalytical
conditions or controls to overcome variability is a remaining challenge for the analytical validation of
NDEV-based diagnostic assays, and we are actively working to overcome it.

Two biomarkers, proBDNF and GluR2, demonstrated significant differences between controls and early AD
individuals across cohorts, which were maintained even after Bonferroni correction. Of note, these were
also the only biomarkers that yielded significant correlations with MMSE and CDR-SOB alongside p181-
Tau. ROC analysis pointed to p181-Tau as the best separator, as could be expected given its importance as a
leading AD biomarker in all biofluids; however, its performance was closely followed by those of GluR2 and
proBDNF, while GAP43, NRGN, Syntaxin-1, and PSD95 also showed significant AUC values. A model that
combined multiple NDEV biomarkers accurately classified 94.7% of early AD patients but only 61.5% of
controls. The lower classification accuracy for controls may be partly due to the lack of longitudinal follow-
up, which may have led to mislabeling pre-symptomatic AD cases (as high as 10%-15% can be expected in
this age group) as controls.

In summary, we introduce a novel methodology for NDEV isolation and a range of novel biomarkers for
synaptic dysfunction in AD [55,56] . Despite limitations, e.g., the variance of absolute values for some analytes,
we believe that the diagnostic promise of NDEVs and the method presented here were convincingly
articulated. The development of effective AD treatments is costly and remains elusive, partly due to the lack
of an efficient toolbox of minimally invasive blood biomarkers. Our results demonstrate the potential of
novel NDE biomarkers reflecting synaptic dysfunction for diagnosis and, possibly, monitoring AD
progression and treatment responses.

DECLARATIONS
Author’s contributions
Conceptualized, designed, and spearheaded the study and method development, participated in the
biomarker data analysis, and performed the bulk of the manuscript’s writing and editing: Eitan E
Participated in conceptual discussions, data analysis, and manuscript editing: Thornton-Wells T
Performed the bulk of the data acquisition and analysis: Elgart K, Erden E, Gershun E
Performed the lipidomic analysis: Levine A
Participated in the experimental design, data analysis, figure design, manuscript writing, and editing:
Volpert O
Participated in conceptual discussions, data analysis, and manuscript editing: Azadeh M, Smith DG
Played a key role in the conceptual design of the study, provided samples, performed data analysis, and
extensive manuscript writing: Kapogiannis D