Goetzl · Neurology 2015 · Case-control and longitudinal biomarker study · n=62 patients plus controls

Altered lysosomal proteins in neural-derived plasma exosomes in preclinical Alzheimer disease.

Cited 473 times in the scientific literature.

Level 4 - case-series / case-control

Case-control and retrospective longitudinal biomarker comparison

PubMed 26062630 · doi:10.1212/WNL.0000000000001702 · record verified 2026-08-31

What was done

Neural-derived blood exosomes were isolated via anti-L1CAM immunoabsorption from plasma samples across multiple patient groups: cross-sectional cohorts with Alzheimer disease (AD, n = 26), frontotemporal dementia (FTD, n = 16), and case controls, as well as a longitudinal cohort of 20 patients sampled both when cognitively normal (1 to 10 years prior to diagnosis) and at clinical AD diagnosis, along with matched controls. Autolysosomal proteins (cathepsin D, LAMP-1, ubiquitinylated proteins, and heat-shock protein 70) were quantified by ELISA and normalized to the exosomal marker CD81.

What was found

In cross-sectional comparisons, mean exosomal levels of cathepsin D, LAMP-1, and ubiquitinylated proteins were significantly higher, and heat-shock protein 70 was significantly lower, in AD patients compared to controls (p ≤ 0.0005). Cathepsin D, LAMP-1, and ubiquitinylated proteins were also significantly higher in AD than in FTD (p ≤ 0.006). Stepwise discriminant modeling achieved 100% correct classification of AD patients. In the longitudinal cohort, all protein levels differed significantly from matched controls both at diagnosis and 1 to 10 years before clinical onset (p ≤ 0.0003).

Why it matters

These findings suggest that neuronal autolysosomal dysfunction is reflected in peripheral blood exosomes up to a decade before dementia onset, supporting the development of blood-based preclinical biomarkers for Alzheimer disease.

Limits

The sample sizes were small across all subgroups, and the exact count of control subjects was not reported in the abstract. Perfect classification accuracy (100%) in discriminant modeling strongly suggests overfitting, requiring validation in larger independent prospective cohorts.

Cited by