Introduction
Currently, Alzheimer’s disease (AD) remains the only leading cause of death without an available disease-modifying therapy. It is characterized by the co-existence of aberrantly accumulated amyloid-β (Aβ) and hyperphosphorylated tau [
1]. According to the latest diagnostic frameworks [
2], individuals exhibiting evidence of brain Aβ deposition have already entered the Alzheimer’s continuum, indicating a high risk of AD. Due to its incurable and irreversible nature, it is of great importance to recognize AD patients at the ultra-early stage and carry out specific interventions [
3,
4]. Although cerebrospinal fluid (CSF) detection and positron emission tomography (PET) imaging have made great progress [
2], there is still an urgent and unmet need for convenient and cost-effective early diagnostic biomarkers.
The discovery of extracellular vesicles (EVs) has greatly improved our understanding of cell-to-cell communication. EVs facilitate the accumulation and spread of AD-associated toxic cargo, while enhancing intercellular communication [
5]. During this process, some EVs are likely to cross the blood–brain barrier into the peripheral blood [
6], making them potential carriers of biomarkers. Additionally, EVs can reflect the state of their source cells [
7], and the successful isolation of blood–brain-derived EVs further enhances this possibility [
8,
9]. In our previous reviews [
5,
10], we have summarized the role of EVs as AD biomarkers. Briefly, brain-derived EVs, such as neuronal-derived EVs (nEVs), which are present in the blood, carry many different types of cargo, including Aβ [
9], phosphorylated tau [
9], synapse-related proteins [
11], and other molecules [
8,
12], and can be used to diagnose AD. Furthermore, a recent study suggested that Aβ
42, tau phosphorylated at threonine 181 (p-tau181), and t-tau levels in nEVs are closely related to those in the CSF [
13]. Similar results have been obtained in AD mouse models, where biomarkers in circulating nEVs were strongly and positively correlated with their levels in the brain [
14]. Additional file
1: Table S1 lists previous studies on nEV Aβ and tau as biomarkers of AD. However, several unresolved issues remain. First, the findings regarding the preclinical stage of AD are conflicting [
9,
12]: it remains unclear whether the cargos (particularly Aβ) in nEVs have really changed at this early stage. Second, no study has explored the relationship between nEVs and neuroimaging (amyloid-PET, structural magnetic resonance imaging [sMRI], etc.) or cognition. Exploring these questions will facilitate early diagnosis of AD and prediction of outcome events, which are particularly meaningful for clinical research.
The goals of this study were as follows: (1) to explore the dynamic changes in AD-related proteins, such as Aβ, carried in nEVs, in the Alzheimer’s continuum, with a focus on cognitively normal controls (NCs) with high brain Aβ loads (Aβ+) and (2) to evaluate the relationships between nEV cargo and brain Aβ deposition (reflected by amyloid-PET), brain regional volume (reflected by sMRI), and cognition. In addition, we quantified the plasma Aβ levels of some participants to make horizontal comparisons.
Discussion
Using a Chinese community-based population, we analyzed the AD-related cargos of EVs and found gradually increasing concentrations of Aβ
42 along the Alzheimer’s continuum (from Aβ− NCs, through Aβ+ NCs, aMCI, to ADD). In contrast, the plasma Aβ concentrations did not change among the groups. More specifically, our study verified previous findings that the nEV Aβ
42 assay indeed provides high diagnostic accuracy in identifying patients with cognitive impairment [
9,
13]. Moreover, our study attempted to add new evidence for the preclinical stage of AD [
12] and proved that the concentration of nEV Aβ
42 is already increased in Aβ+ NCs, although its diagnostic efficacy was not marked (AUC with
APOE genotype = 0.705). Furthermore, nEV Aβ
42 levels were strongly associated with global and regional AV45 SUVR, suggesting that they reflect brain Aβ deposition. In addition, baseline nEV Aβ
42 levels predicted longitudinal changes in cognition and entorhinal volume.
The specific enrichment of nEVs using immunoprecipitation is a pioneering discovery [
30], opening a “window into the brain.” The blood-isolated total EVs are derived from a wide range of sources and cannot specifically reflect the changes in neuronal function, unlike the CD171+ EVs. Enrichment of CD171+ EVs of neuronal origin is mainly based on the fact that they contain higher levels of multiple neuronal markers, as shown in our study and previous studies [
12,
31]. Our isolation protocol and its validation followed the MISEV2018 recommendations [
25], further verifying the reliability. However, the quality control results in this study were unexpected. Neither p-tau181 nor markers of neuronal damage, including t-tau (in the pre-experiment) and NFL, were included in the subsequent analysis. Theoretically, compared to the traditional enzyme-linked immunosorbent assay (ELISA) method used in previous studies [
8,
9,
11,
13], electrochemiluminescence and Simoa immunoassays are more sensitive. In addition, although the electrochemiluminescence method was fully suitable for the detection of plasma Aβ, its sensitivity was not sufficient to detect nEV Aβ (in the pre-experiment), which also violates the previous findings in which nEV Aβ could be quantified using ELISA [
9,
13]. Fortunately, consistent with a recent study [
12], the Simoa method reliably detected Aβ in nEVs. Several factors may account for the differences between our study and previous studies. First, there were differences in plasma volume, experimental procedures, and quality controls among studies. Second, besides the differences in assay platforms, different kits may be equipped with different antibody pairs. Additionally, the NFL detection kit was newly developed, and to the best of our knowledge, this kit and the p-tau181 detection kit were applied to nEVs here for the first time, thus requiring further verification. Third, it should be noted that some researchers have found that the nEV t-tau was largely undetectable using ELISA or Luminex immunoassays and proposed that previously reported quantifications may have resulted from contamination [
32]. Considering our electrochemiluminescence results of nEV t-tau, we consider that its concentrations are probably too low to be detected. P-tau181 is derived from t-tau [
33]; thus, it may be reasonable to speculate that nEV p-tau181 is also undetectable.
The ability of plasma Aβ to assist in AD diagnoses has been questioned over the years [
10,
34]. The results are easily influenced by detection platforms [
35], and blood Aβ is not necessarily brain-derived [
36], making it difficult for it to replace CSF Aβ as a reliable biomarker. The electrochemiluminescence method has been reliably applied to the detection of CSF Aβ [
37]; however, our study found that plasma Aβ had limited roles in the diagnosis of AD patients, let alone those in the preclinical stage, and it was not correlated with amyloid-PET results. In comparison, nEV Aβ
42 can distinguish cognitively impaired patients from NCs. However, we admit that its ability to recognize Aβ+ NCs is not outstanding. According to our recent study on the discrimination of Aβ+ NCs [
16], we believe that the values are reasonable because large-scale neurodegeneration has not yet occurred, and this stage represents the earliest identifiable preclinical stage [
38]. Our study partially proved previous conclusions that nEV Aβ
42 can predict MCI conversion [
39] and that its concentrations increased with disease progression [
9]. However, a recent study found that the assay was not helpful for the construction of a diagnostic model of preclinical AD, and the reasons for this may be complex [
12]. We found that the nEV Aβ
40 and Aβ
42 were closely correlated, which was contrary to a previous study on nEV Aβ
40 [
40] and our viewpoint about their ratio. The reasons for this remain unclear after discussion with Quanterix™. Currently, EV normalization methods are not unified; either surface markers [
13] or particle numbers can be used [
12]. Recent findings suggest that the surface markers that are typically used are present in only a fraction of EVs and are not particularly enriched in smaller EVs in the exosome range [
41]; therefore, normalization by particle number is likely to be more accurate.
The strong correlations between baseline nEV Aβ
42 and AV45 SUVR indicated that nEVs have the potential to reflect brain pathological changes and are emerging as liquid biopsy tools. Furthermore, nEV Aβ
42 was also associated with cognitive deterioration, as well as with entorhinal atrophy, suggesting that the blood assay could also serve as a predictor of disease progression, and thus could be used to select individuals most likely to progress during typically short clinical trial periods. We analyzed multiple brain regions; however, only the entorhinal cortex remained significant after correcting for confounding factors. The entorhinal cortex region is characteristically affected by tau pathology at an early stage in AD [
1]. Compared to NCs, individuals with subtle cognitive difficulties demonstrate faster atrophy of the entorhinal cortex [
42], and the volume, glucose metabolism, blood flow, and texture features of this region all play roles in predicting cognitive decline [
24,
43‐
45]. Subgroup analysis suggested that the correlation between nEV Aβ
42 and entorhinal atrophy only existed in Aβ+ NCs, but not in Aβ− NCs, indicating that the more severe the pathological damage, the more severe the atrophy. However, elevated brain Aβ deposition alone is probably insufficient to produce neuronal damage and cognitive changes [
46], and their associations are likely mediated by neurofibrillary tangles, with a temporal delay [
47,
48]. Nevertheless, based on the “amyloid cascade hypothesis” or real-world studies [
1,
49], Aβ is the actual initiating factor of downstream pathological changes in AD. The dual effects of Aβ and tau aggravated the deterioration of cognition more than tau alone [
50], and Aβ is an independent risk factor for cognitive impairment [
51].
Limitations
This study has several limitations. First, the small sample size limited the statistical power of our data, and not all participants underwent amyloid-PET. Second, CD171 is not absolutely expressed in the brain, but also in other tissues, and CD171+ EVs are thus not absolutely of neuronal origin only [
31]. Third, longitudinal follow-up data are still lacking, and the follow-up time varied significantly. Fourth, due to the lack of tau pathology, it is unknown whether the relationship between nEV Aβ and neurodegeneration is mediated by tau. Fifth, extracting nEVs and quantifying their cargo are complex and expensive, and NTA tests are time-consuming. These factors limit their clinical application. Sixth, there is a lack of cognitively impaired patients with other neurodegenerative diseases. Seventh, recent findings have suggested that, although plasma Aβ detected by the Simoa method could not predict amyloid status among NCs well, it has a certain value in the diagnosis of symptomatic AD [
29,
35,
52‐
54]. In the future, we will use Simoa assays instead of the electrochemiluminescence method to detect plasma Aβ for a matched and more meaningful horizontal comparison of nEV Aβ. Recently, a series of studies have suggested that phosphorylated tau proteins in the blood are reliable biomarkers of AD [
10]. Their diagnostic effects have been verified in the Alzheimer’s continuum from multiple perspectives, including CSF, PET, autopsy, and clinical follow-ups. This weakens the significance of our study to a certain extent. However, blood tau is less stable than Aβ [
55]. Additionally, EVs are essential for intercellular communication [
6]. Thus, the extraction, validation, and cargo detection of nEVs may provide a basis for subsequent functional studies [
56,
57]. Considering the shortcomings of our research and the limitations in this field, multicenter collaboration to include more pathology-identified participants is required in the future.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.