Plasma extracellular vesicles and phosphorylated tau 181 as early biomarkers of cognitive impairment in Alzheimer's dementia

Abstract

Background: Timely and accurate diagnosis of Alzheimer’s disease (AD) in clinical practice is a great challenge, especially during early disease stages with subtle or mild symptoms of cognitive decline. Moreover, robust and more accessible blood-based screening tests for early diagnosis are needed. In this study, we investigated the core AD blood biomarkers — amyloid beta 42 (Aβ₄₂) and 40 (Aβ₄₀) peptides, phosphorylated tau 181 (p-Tau₁₈₁), neurofilament light chain (NfL), and total tau (t-Tau) — and extracellular vesicle (EVs) size and concentration in individuals characterized by different stages of cognitive decline to identify biochemical markers of dementia for early diagnosis.

Methods: A total of n = 800 human plasma samples were analyzed. Plasma levels of NfL, t-Tau, p-Tau₁₈₁, Aβ₄₂, Aβ₄₀ and plasma EVs were evaluated in n = 217 elderly healthy subjects (CTRL), in individuals with subjective cognitive complaints (SCC, n = 48), pre-mild cognitive impairment (pre-MCI, n = 58) and mild cognitive impairment (MCI, n = 426), and in n = 51 probable AD dementia patients (AD-dem), using ultrasensitive Single Molecule Array technology (Simoa®) and nanoparticle tracking analysis (NTA). Logistic regression and Receiver Operating Characteristic (ROC) analyses were employed.

Results: Plasma NfL displayed increased levels in AD-dem and MCI patients, while p-Tau₁₈₁, Aβ₄₂/Aβ₄₀ ratio, Aβ₄₂/p-Tau₁₈₁ ratio, and EVs plasma levels were altered since the early stages of the pathology: in particular, p-Tau₁₈₁ levels increased as cognitive symptoms worsened, already in the SCC and pre-MCI groups compared to CTRL, while the ratio of EVs concentration and size (EVs ratio) was decreased in all groups compared to CTRL. Plasma p-Tau₁₈₁ best classified AD-dem patients from CTRL with an area under the curve (AUC) equal to 0.87, while EVs ratio best differentiated SCC from CTRL (AUC = 0.78). Combining p-Tau₁₈₁ and EVs ratio with Aβ₄₂/Aβ₄₀ ratio and NfL, respectively, significantly improved the classification of pre-MCI and MCI from CTRL (AUC_comb = 0.79 and AUC_comb = 0.85). Combining biomarkers did not improve accuracy in discriminating MCI from SCC, pre-MCI and AD-dem.

Conclusions: p-Tau₁₈₁ and EVs ratio are promising biomarkers for the identification of individuals at risk of degenerative dementia. Combining the core AD plasma biomarkers with EVs ratio can aid in diagnosing the early stages of AD dementia.

Supplementary Information: The online version contains supplementary material available at 10.1186/s13195-025-01897-2.

Keywords: Alzheimer’s dementia; Amyloid beta peptide; Early diagnosis; Extracellular vesicles; Mild cognitive impairment; Neurofilament light; Phosphorylated Tau 181; Plasma biomarkers; Preclinical stages; Ultrasensitive assays.

Fig. 1

Plasma biomarker concentrations in participants by clinical diagnosis. a NfL, (b) p-Tau₁₈₁, (c) ratio of Aβ₄₂/p-Tau₁₈₁, (d) ratio of Aβ₄₂/Aβ₄₀, and (e) EVs ratio (ratio between EVs concentration and size). CTRL, controls; SCC, subjective cognitive complaints; pre-MCI, pre-mild cognitive impairment; MCI, mild cognitive impairment; AD-dem, probable Alzheimer’s disease dementia. Violin plots show raw data distribution with median (thick dashed black line) and interquartile range (25th and 75th percentile; thin dotted black line). *p < 0.050, **p < 0.010, ***p < 0.001 after multiple-comparisons with Tukey’s post-hoc test, adjusted for age and sex (generalized linear model)

Fig. 2

ROC curves showing the improvement in classification accuracy of combined predictors of multiple logistic models compared to single predictors. Only significant AUC comparisons are shown. ROC curves of single and combined predictors are depicted in blue and red, respectively. a CTRL vs. pre-MCI, (b) CTRL vs. MCI, (c) SCC vs. AD-dem, (d) pre-MCI vs. AD-dem

Fig. 3

ROC curve comparisons of the best biomarker combination in multiple logistic models, including or excluding EVs ratio. Only significant AUC comparisons are shown. ROC curves of models including EVs ratio (_comb) are depicted in red, while ROC curves of models excluding EVs ratio (_{EVs ratio excl.}) are depicted in gray. Panel (a) shows the CTRL vs. pre-MCI comparison, where the AUC of the model including EVs ratio, p-Tau₁₈₁, and Aβ₄₂/Aβ₄₀ ratio is significantly greater (p = 0.018) than the AUC of the model including p-Tau₁₈₁ and Aβ₄₂/Aβ₄₀ ratio, but not EVs ratio. Panel (b) shows the significantly improved performance (p < 0.001) of the model including EVs ratio, NfL, and p-Tau₁₈₁ compared to the model including NfL, p-Tau₁₈₁, and also Aβ₄₂/Aβ₄₀ ratio, but not EVs ratio, for the CTRL vs. MCI comparison. P-values estimated by DeLong test comparing the AUCs of the best predictor combination, including or excluding EVs ratio, for the same diagnostic comparison