Synaptic Activity and (Neuro)Inflammation in Alzheimer’s Disease: Could Exosomes be an Additional Link?

Main hallmarks of pathogenesis

Genetic studies have identified mutations in the amyloid-β (Aβ) precursor protein (AβPP), presenilin-1 and -2 (PS1, PS2) genes that cause the rare familial forms of AD (FAD). The proteolysis of AβPP by BACE (β-site AβPP cleaving enzyme) yields βCTF (β-secretase generated C-terminal fragment) which is followed by the PS-containing γ-secretase complex cleavage of βCTF yielding Aβ peptides of different sizes (i.e., composed of 38 to 42 amino acids: Aβ_1–38 to Aβ_1–42). The accumulation of monomeric and oligomeric forms, as well as conformational misfolding changes, leads to extracellular deposition of Aβ peptides in insoluble amyloid plaques (Aβ plaques). The amyloid hypothesis [1] posits that Aβ peptides play a pivotal role in AD pathogenesis by triggering the pathological “amyloid cascade” which consists in early synaptic dysfunction, microglial and astrocyte activation and hyperphosphorylation of tau proteins. The intracellular accumulation of abnormally hyperphosphorylated tau (ptau) triggers neurofibrillary tangles (NFTs) formation, and further contributes to synaptic and neuronal dysfunctions leading to neuronal loss and macroscopic atrophy, which culminates in extensive neurodegeneration. This hypothesis further posits that synaptic dysfunction precedes alterations in neurotransmitter release, and thus impairment of memory and cognitive functions [1].

The reasons for the latency between Aβ deposition (amyloidosis) and clinical symptom onset are currently unknown [2]. Importantly, alterations of neuronal activity appear already during early stages of AD. For example, the risk of epileptic seizure is increased in sporadic AD and more than 40% of AD patients display subclinical epileptiform activity according to a recent clinical study [3]. This increased risk has been also reported in AD-like mouse models of amyloidosis during the pre-symptomatic stage (i.e., prior to Aβ-plaque deposition and cognitive impairment, reminiscent of the pre-clinical stage in humans). These seizure-associated phenomena might be due to the alteration of neuronal membrane properties by soluble Aβ, resulting in hyperexcitability of hippocampal pyramidal cells and subsequent epileptiform activity [4]. The latter may be further related to inhibition of long-term potentiation (LTP) [5], a form of synaptic plasticity required for learning and memory. Besides, the impact of tau protein on synaptic activity has been studied much less than the impact of Aβ. Nevertheless, the recent evidence strongly suggests that soluble forms of tau alter neuronal (including synaptic) functions during the early stages of AD (reviewed [6]). Other mechanisms can also contribute to early neuronal dysfunction including, for instance, high release of reactive oxygen species (ROS) known for their impact on neuronal excitability and synaptic activity [7].

Diagnosis and factors of pathogenesis: Focus on neuroinflammation

AD can be currently diagnosed earlier than in the past due to a specific phenotype as determined by neurocognitive testing combined with analysis of biomarkers [8]. However, diagnosis is still made at the stage when neuronal lesions are already advanced, making the development of curative treatments highly challenging. In spite of the relatively late diagnosis, considerable progress has been made recently so that modification of biomarkers can be now detected at the stage of mild cognitive impairment (MCI), corresponding to the prodromal stage of AD, though before major cognitive and executive function disability occur [8].

In addition to age, which is a major risk factor for developing AD, it is generally recognized that a complex interplay between genetic and environmental factors impact AD pathogenesis [9]. Among genetic factors, the expression of ɛ4 allele of apolipoprotein E (APOE4), which plays a crucial role in regulating cholesterol metabolism, severely increases the risk of AD. Other significant genetic risk factors for AD are mutations in the Triggering Receptor Expressed on Myeloid cells 2 (TREM2), Complement Receptor-1 (CR1), CD33, among others, which are associated with microglia dystrophy, decreased phagocytosis, and an increased pro-inflammatory phenotype [9]. Most importantly, more than 2/3 of identified risk single nucleotide polymorphism mutations reported by genome-wide association studies are exclusively or mostly expressed by microglia. This strongly suggests that microglia, the major mediator of innate immunity in the brain, play a key role in AD pathogenesis which has been previously underestimated [10]. Overall, the majority of identified risk factors are related to inflammatory-, cholesterol metabolism-, and endosomal-vesicle recycling pathways.

In agreement, AD has been associated with chronic innate inflammation in the central nervous system (CNS). Such neuroinflammation involves activated microglia and reactive astrocytes, increased production of cytokines, chemokines, and other inflammatory factors, as well as infiltration of immune cells followed by secondary neurodegeneration [11]. Interestingly, peripheral T-cells in old subjects and AD patients display higher Aβ-reactivity than those coming from neurologically normal subjects [12]. Enhanced T-cell infiltration was observed in the brain of APPPS1 mouse model, while peripheral modulation of T-cell subsets was reported to impact on local CNS innate neuroinflammatory responses [13]. Systemic inflammatory factors have also been suggested to modulate AD-related neuroinflammation [14], as well as peripheral innate immune cells such as neutrophils [15]. Altogether these data point to a key role of CNS innate neuroinflammatory responses in the pathophysiology of AD, as well as the complex interplay between local (innate) and peripheral (both innate and adaptive) immunity.

Importantly, recent clinical studies suggested that neuroinflammatory-like changes (i.e., glial activation, release of pro-inflammatory factors, and neuronal damage) occur very early (i.e., from pre-clinical stages) in the course of AD progression [16]. These changes are associated with microglia activation and appear beneficial [17], but microglia progressively switches from homeostatic to disease-associated microglia (DAM) phenotype [18]. Such AD-related functional switch alters not only the phagocytic function of these cells and their ability to restrict cerebral Aβ accumulation, but also their steady-state surveying and regulatory functions including cytokine, chemokine, and growth factor production [18]. Consistently, reduction of chronic neuroinflammation by decreasing the pro-inflammatory cytokine levels in APPPS1 mice improves cognitive performance both at the onset and in advanced stages of AD-like pathology [19]. Moreover, the synaptic hyperexcitability observed in early stages of AD pathogenesis [4] is related to the neuroinflammatory-like changes, such as TNFα induction [20]. Most importantly, by using XPro1595 antagonist to block TNFα actions during the pre-symptomatic stages of AD, neuronal hyperexcitability, LTP and cognitive impairments were prevented at the later, advanced stages in TgCRND8 mice [21]. These data demonstrate the causal relationship between early pro-inflammatory cytokine production and cognitive dysfunctions.

Pathogenic spread of AD

A growing body of evidence suggests that AD, besides other neurodegenerative disorders, propagates in the brain via prion-like intercellular induction of protein misfolding. Neurotoxic proteins Aβ and ptau share properties with classical prions, including their ability to spread within the brain and the periphery [22]. This capacity is of fundamental importance since it corroborates the possibility that sporadic AD could be triggered in vulnerable brain regions by Aβ/ptau seed that may be imported, and not only by in situ production of these proteins. In the next two sections, we will focus on Aβ since the prion-like mode of ptau spread, although convincingly demonstrated (for recent review, see [23]), is likely secondary to Aβ propagation because Aβ changes are detectable before those of tau, at least in FAD ([24, 25]; see also discussion in [1]).

Direct neuron-to-neuron transfer of soluble oligomeric Aβ has been also observed in primary cultures of hippocampal rat neurons [26]. Moreover, in vivo transfer of Aβ aggregates by passive, extracellular diffusion from neurons has been suggested (reviewed in [27]), pointing to their putative propagation along interconnected regions in the brain.

In addition to such neuron-mediated propagation, activated microglia and astrocytes may also be involved. Both types of glia cells are usually present near Aβ plaques and neurofibrillary tangles in the hippocampus of AD patients. The impaired clearance of Aβ deposits by glial cells contributes to their cerebral accumulation and subsequent spread in the brain, in line with the reported lowered clearance abilities of AD-associated microglia [28]. In addition, increase in Aβ is concomitant with microglia activation which is in turn accompanied with increased production of pro-inflammatory cytokines (TNFα, IL-1β, IL-6, IL-12 ...) [14, 28]. These brain-born cytokines may act synergistically with peripheral cytokines [28] to further impair the integrity of the blood-brain barrier (BBB) and thus implement a feed-forward vicious cycle of amplification in terms of production and propagation of cytotoxic proteins [16].

Definition and general considerations

Extracellular vesicles (EVs) are nano-sized structures released by all mammalian cell types and found in different body fluids, including cerebrospinal fluid (CSF) and blood [29]. EVs are surrounded by a lipid bilayer and comprise vesicles ranging from 30 to over 1,000 nm in size. They carry a complex cargo composed of specific proteins such as signaling molecules (integrins, cytokines ...) and their receptors, bioactive lipids, nucleic acids including RNAs (mRNAs, miRNAs, small non coding RNAs), and DNAs [29]. Nature of molecular species and their relative proportion in the cargo are highly diversified among EVs, likely reflecting heterogeneity of their cellular origin and microenvironment in which they are generated. The precise characteristics, as well as regulatory mechanisms of biogenesis, sorting and degradation or secretion of EVs are not completely understood so far.

EVs can be divided in two major families: 1) microvesicles released by budding of the cellular membrane and 2) exosomes secreted via exocytosis from multivesicular bodies (MVBs) that are formed along the endocytic pathway [29]. Currently, it remains difficult to discriminate between EVs and exosomes and there is no consensus about the specific markers allowing to distinguish between EV subtypes. The International Society for Extracellular Vesicles (ISEV) thus recommends to denote exosomes as “small EVs” based on their size (30–100 nm) [30]. For the purpose of this review, the terms “exosomes” and “EVs” are used as per publication to which each particular citation refers to. It should be however kept in mind that in majority of publications using term “exosome”, this denomination is based mainly on the size criterion without demonstrating the presence of endosomal markers to confirm their endocytic origin.

All EVs contain different proteins involved in their transport and fusion during biogenesis [29]. These proteins can be common to all EVs, comprising transmembrane proteins, tetraspanins, heat shock proteins, lipid-related proteins, and phospholipases. Alternatively, cargo proteins can be specific for a given class of exosomes depending on a donor cell in which they are generated [29]. For example, miRNA-141 is specific for the cargo of exosomes derived from metastatic cells [31].

EVs are also characterized by specific surface marker proteins, such as CD9 and CD81 [32] (Table 1). Exosomes, as a particular class of EVs and in relation to their endosomal origin, express the common markers as, for instance, CD81, ALIX (ALG-2 interacting protein), Tsg 101 (Tumor susceptibility gene 101), and tetraspanins [29] (Table 1). The content of the cargo and the macromolecule composition differ depending on the physiological state of the donor cell, and can be altered in diseases [29].

Biogenesis, sorting, and degradation of exosomes has been extensively discussed in excellent recent reviews (e.g., [29, 33]) and will not be detailed here. The exosome’s life cycle can terminate by degradation of their components that may be used by the recipient cells for the biogenesis of their own constituents or in their intermediate metabolism [29]. Even though, molecules contained in exosomes can also escape degradation and act as bioactive factors to regulate the cell target functions, including in different cell types in the nervous system [34].

Physiological roles of exosomes in the brain

Classically, the exosome secretion is seen as an excretion of unnecessary or toxic molecules by donor cells [35]. However, as already discussed above, growing body of evidence suggests that exosomes play also a role in intercellular communication involving neighboring and distant cells in the same or different organs [29, 36] (Fig. 1).

OPEN IN VIEWER

Fig.1

Hypothetical involvement of the cross-talk between peripheral and brain-derived exosomes in AD pathogenesis. EVs cross the BBB in both directions. By consequence, AD pathogenesis may be influenced by peripheral EVs transporting inflammatory factors, and generated for instance, by lymphocytes or gut microbiota. Circulatory EVs are able to reach the brain and impact glial and neuronal activity. It has been proposed that during the pre-symptomatic stage of AD, EVs participate in Aβ clearance and maintain of neuronal and synaptic functions. However, during advanced stages of the disease, EVs may spread AD-related proteins (Aβ, ptau) and pro-inflammatory factors (TNFα, IL-1β, ROS) through direct and indirect communications with the recipient cells. The propagation of toxic molecules by peripheral and brain-derived exosomes thus participates to neuroinflammation and neurodegeneration, and consistently to AD pathogenesis. Conversely, brain-derived exosomes can also cross the BBB and are detectable in the peripheral circulation. Analysis of their cargo may turn to be useful, non-invasively detectable diagnostic tool.

In the brain, exosomes are produced by oligodendrocytes, neurons, astrocytes, microglia, and Schwann cells, as well as by endothelial cells of the brain blood vessels [37]. Their role in reciprocal communication between neurons and glial cells, synaptic plasticity, and neuronal activity has recently attracted much interest. For instance, delivery of exosomes carrying myelin-associated proteins derived from oligodendrocytes to neurons establishes communication between these cells, contributing to myelination and maintenance of neuronal integrity [38]. Additional physiological roles of exosomes in the control of neuron-glia interactions include microglia-mediated regulation of synaptic pruning [38]. Regarding the regulation of neuronal activity, exosomes have also been reported to deliver to neurons the enzymes involved in energy metabolism [38] (Fig. 1).

In addition to the common canonical markers (Table 1), neuron-derived exosomes (NDE) contain the specific adhesion molecules NCAM (Neural Cell Adhesion Molecule) and CD171 or L1CAM (L1 Cell Adhesion Molecule), which were used initially by Dr. Edward Goetzl’s group to detect NDE in the plasma from a patient at the pre-clinical stage of AD ([39]; for review, see [40]; see also other contributions by Dr. Goetzl and his team cited in this review). In addition, NDE contain lipid raft protein flotillin-2, sub-units of α-amino-3-hydroxy-5-methyl-isoxazolepropionic acid receptors (AMPARs) of glutamate, and Microtubule-Associated Protein 1B (MAP-1B). Their cargo typically contains proteins involved in synaptic neurotransmission like synaptogamin and synaptophysin, among others [40, 41]. Interestingly, it has been recently demonstrated that exosomes released from cortical neurons upon activation of glutamatergic synapses bind selectively to other neurons instead of being internalized by glial cells [42] thus underscoring the molecular substrate for neuron-to-neuron communication via exosomes.

Astrocyte-derived exosomes (ADE) cargo is enriched in GFAP (Glial Acidic Fibrillary Protein), Glutamine Synthetase (GlySyn), FGF2 (Fibroblast Growth Factor-2), VEGF (Vascular Endothelial Growth Factor), and extracellular matrix protein endostatin. Synapsin-1 is also comprised in the ADE cargo reflecting the role of astrocytes in synaptic neurotransmission [41]. In addition, the components of amyloidogenic pathway (AβPP, BACE1, γ-secretase) have recently been identified in ADE [43] thus suggesting that in pathological conditions, ADE can contribute to amyloidogenic pathway (see next Section: Pathological role of exosomes in neurodegenerative diseases: focus on AD).

Oligodendrocytes-derived exosomes (ODE) carry the cargo containing myelin-associated proteins (proteolipids and glycoproteins) and oligodendrocyte-specific cyclic-nucleotide phosphodiesterase [44], required for the phospho-diester hydrolysis of the 2^′3^′-cyclic nucleotide to 2’-nucleotide during myelin biosynthesis. ODE play a major physiological role in neuron protection against the adverse effects of the transient increase in synaptic glutamate concentration in the course of neurotransmission. The evidence supporting such role of ODE comes from elegant studies using neuron/oligodendrocyte co-cultures allowing exclusively for exchange of the particles with a diameter lower than 1μm [45]. Furthermore, via their capacity to transfer superoxide dismutase and catalase to neurons, ODE participate in the control of neuronal oxidative stress [46].

In addition to the markers common to all exosomes (Table 1) and non-specific cargo proteins shared with dendritic cell- and B lymphocyte-derived exosomes (e.g., tetraspannins, chaperons, etc.), the cargo of microglia-derived exosomes (MDE) contain some specific proteins such as aminopeptidase CD13 [47]. Reminiscent of physiological roles of microglia, MDE are involved in the control of neurite outgrowth, coordination of the innate immune response in the brain, and modulation of neuronal activity [38]. Regarding the modulation of neuronal activity, it has been demonstrated that MDE can increase miniature Excitatory Post-Synaptic Potential (mEPSP) via selective enhancing the sphingolipid metabolism to control synaptic release of neurotransmitter containing synaptic vesicles [48].

Exosomes in AD pathogenesis

Among the neurotoxic proteins involved in neurodegenerative diseases, the role of exosomes in clearance and spread of Aβ and tau in AD has been extensively studied. Thus, it has been reported that the enlargement of exosomal compartment in pyramidal neurons of the neocortex and the presence of Aβ in NDE sorted from the plasma of AD patients could be detected up to 10 years before the onset of the clinical symptoms (for recent review, see [49]). In addition, exosomal markers were found to be enriched in Aβ plaques of postmortem AD brains [51]. Besides, exosome production decreases in old AD mice, further suggesting that downregulation of exosomes may be related to increased plaque deposition and AD pathogenesis [52]. Interestingly, the presence of AD-related APOE isoform ɛ4 compromises exosomal biogenesis and secretion [53], thus suggesting that the impairment of toxic protein clearance via exosomal route may indeed contribute to AD pathogenesis, in line with the recent hypothesis [54]. The involvement of exosomes in the spread of Aβ and tau proteins will be discussed in the next section because exosomes can indirectly control synaptic activity. Hence, by controlling the content of these proteins in the brain parenchyma, which in turn impacts the synaptic activity ([6, 52]; see also the end of subsection Main hallmarks of AD pathogenesis), exosomes can contribute to the regulation of synaptic and neuronal functions.

Role of exosomes in Aβ and tau clearance and spread

Binding of Aβ to NDE through glycosphingolipid glycans or cellular prion protein (PrP^C) on the extracellular vesicle surface may serve to remove extracellular Aβ [49, 55]. A recent study has explicitly demonstrated the association between Aβ and exosomes by identifying a fraction of Aβ⁺/CD68⁺ double-positive exosomes in AD plasma samples [56]. Such capacity of exosomes to trap extracellular Aβ may promote its clearance by microglia. In agreement, exogenously added, labeled exosomes were found to colocalize with lysosomal/late endosomal markers (e.g., Lamp1) in the cytoplasm of the cultured microglia cells [54]. Moreover, in addition to their capacity to directly promote Aβ clearance by uptake and subsequent lysosomal digestion, microglia can contribute to this process indirectly. Thus, adding mixture of NDE and Aβ to primary cultures of cortical neurons promotes formation of Aβ fibrils in the extracellular space at the expense of toxic soluble oligomers. The NDE-associated fibrillar Aβ is more prone to microglia uptake by a mechanism which depends on phospholipid expression and activity of sphingolipid-metabolizing enzymes on the surface of exosomes. This mode of exosome-mediated uptake of fibrillar Aβ appears immunologically silent, in contrast to direct ingestion of Aβ by microglia [57]. Of note, the lipid expression at the surface of exosomes may be crucial for Aβ clearance. Indeed, ADE express less glycosphingolipids than NDE and bind to Aβ with lower affinity than NDE [52].

Besides Aβ, exosomes can carry AβPP and its metabolites, such as CTF fragments [58]. In this light, AD-associated mutations in both AβPP and PS1 have been correlated with impairment of endo-lysosomal pathway. It has been though hypothesized that lysosomal dysfunction in neurons yields increase in AβPP-CTFs in their endosomal compartment, which in turn triggers endosomal-lysosomal dysfunctions yielding a vicious feed-forward loop of amplification [59].

As already discussed, NDE isolated from blood, CSF, and culture medium of cellular lines overexpressing Aβ, all contain different (oligomeric and fibrillar) forms of Aβ (see [49] for review). In a recent elegant study, Sardar Sinha and colleagues demonstrated that exosomes isolated form human AD brain can be transferred from neuron-to-neuron in SH-SY5Y cell line and human induced pluripotent stem cells (iPSC). When biogenesis, secretion or uptake of NDE coming from AD patients was inhibited in cultured donor cells, the spread of Aβ oligomers and their toxicity in recipient cells were decreased too [55]. Interestingly, by using either coverslips (allowing for direct contact between neurites of donor and recipient cells) or transwell system (where no neuritic contact is possible) to grow recipient cells, these authors demonstrated that NDE-mediated Aβ transfer does not require direct neuritic contact, although the transfer is more efficient when such contact is possible [55].

The mechanisms behind tau secretion and spread via exosomes have been less studied in the past than those relevant for Aβ. It is however known for a while that tau can be secreted via exosomes from tau-overexpressing neuronal cell lines (reviewed in [49]). Similar to Aβ, NDE from AD patients display up to 20-fold increase in pathology-related tau phosphorylation at threonine-181 and serine-396, as compared to tau in NDE isolated form neurologically normal, age-matched controls. In addition, the level of threonine 181-hyperphosphorylated tau is higher in NDE isolated from AD patients at late stages of pathology than in prodromal MCI subjects, in agreement with the proposed causal role of exosomes in the onset and progression of neurodegenerative diseases, including AD [49, 50].

Remarkably, using neuronally-differentiated human iPSC, it has been demonstrated that exosomes containing AD-related ptau remain aggregation-competent after transfer into the recipient neuron-differentiated iPSC [60]. In an analogous approach, neuronally-differentiated human iPSC, further engineered to express repeat domain of tau P301 L and V337M mutations, were used to generate NDE. After subsequently transferring these NDE into the hippocampus of the wild-type mice, Winston and collaborators reported tau inclusions, increase in ptau and extensive neurite degeneration in the recipient brain [61] thus explicitly demonstrating spread of tau pathology via EVs. Most importantly, EVs secreted by neuronally-differentiated human iPSC coming from a patient with FAD and harboring an A246E mutations in PS1 encoding gene, triggered increased tau expression and aberrant phosphorylation in vivo after intra-hippocampal injection in wild-type mice [62]. In addition to providing further evidence for the seeding capacity of AD brain-derived EVs, this study brought the first in vivo demonstration of the possible link between amyloid and tau pathology in which EVs may play a critical role [62].

Altogether, these recent studies suggest that exosomes might be among key mediators of pathogenic progression related to Aβ and tau spread in AD (Fig. 1).

Putative role of exosomes in the control of synaptic activity

There are currently only a few published studies in which the role of exosomes in the control of synaptic activity has been assessed directly. Indeed, the majority of studies in this context assessed the content of pre- and post-synaptic proteins in the cargo of NDE sorted from plasma of AD patients. These studies are concordant in terms of reporting the presence of synaptic proteins in NDE that probably reflects the synaptic loss. For instance, Goetzl and collaborators reported that the NDE content of pre-synaptic proteins pentraxin-2 and neurexin 2α and their post-synaptic ligands glutamate receptor AMPA4 and neuroligin-1, respectively, decreases progressively with advancement of AD pathology [63]. Because these two ligand-receptor pairs are expressed specifically at excitatory synapses, and, given that the observed decline in the level of AMPA4 in NDE was further positively correlated with cognitive dysfunction [63], it is likely that NDE containing excitatory synapse-specific receptors (i.e., AMPA4) may be associated with the regulation of the relevant synaptic activity, even if in fine, they reflect the failure of this regulation yielding synaptic loss.

In a more direct approach, intra-cerebroventricular infusion of exosomes derived from either neuroblastoma cell line N2a or human CSF from healthy donors, could counteracts LTP impairment induced by subsequent infusion of soluble Aβ species (obtained either by oligomerization of synthetic Aβ or purified from human AD brain) [64]. The observed protective effect of exosomes against synaptic toxicity of Aβ was attributed to the sequestering and immobilization of Aβ on the surface of exosomes [64].

Regarding the modulation of synaptic activity, it has been furthermore demonstrated that MDE can increase mEPSP via selectively enhancing the sphingolipid metabolism to control synaptic release of neurotransmitter-containing synaptic vesicles [48]. These data point to an additional, exosome-dependent mechanism for regulation of neuronal activity by microglia.

Interestingly, a recent study reported that EV derived from the bone marrow Mesenchymal Stem Cells (MSCs) exert protective effects on cultured primary hippocampal neurons exposed to Aβ-induced synaptic damage [65]. Besides, the protective effect of MSC-derived EVs encompassed also Aβ-induced oxidative stress via EV-mediated catalase delivery to the recipient neurons. MSC-mediated rescue from Aβ-triggered decrease in post-synaptic density protein 95 (PSD95) and pre-synaptic marker synaptophysin was explicitly demonstrated by co-culturing primary hippocampal neurons and MSCs in the transwell system which allows for inter-cellular communication exclusively by soluble factors [65]. These findings were further extended by demonstrating that in vitro beneficial effects of MSC-derived EVs involve also inhibition of nitrosative stress by preventing Aβ-induced iNOS induction [66]. Such beneficial effects of MSC-derived EVs were in addition confirmed in vivo, by intra-cerebroventricular injection of MSC-derived EVs in APP/PS1 mouse model. This manipulation could efficiently alleviate the impairment of both Aβ-related pre-synaptic function, as attested by rescue of EPSP, and memory-related LTP which was further correlated with improvement of cognitive performance in new-object recognition and Morris water maze tests [66].

In addition to the above discussed in situ impact of EVs (including exosomes) generated by the cells of the CNS (neurons and microglia, but also astrocytes) or provided exogenously by intracerebral injection, the peripheral-born EVs may hypothetically also contribute to the control of synaptic activity. This hypothesis is based on the analogy to EVs derived in the peripheral circulation from immune cells that are known to vehicle immunoregulatory molecules [67]. Thus, EVs secreted during peripheral inflammatory responses are thought to promote inflammation in endothelial cells, neutrophils, hepatocytes, macrophages, and monocytes via different mechanisms. These mechanisms include delivery of inflammasome’s component neprylisin (e.g., NLRP3), pro-inflammatory cytokines (IL-1β, TNFα, IL-6, IL-8, INF-γ) and chemokines [68], induction of prostaglandins or lipid-inflammatory mediators, to name a few [67]. As BBB is permeable to EVs, they may penetrate the brain and contribute to impairment of synaptic activity by triggering neuroinflammation. The latter could be in turn related to the cross-talk between systemic and CNS inflammation which is currently well recognized, although the underlying mechanisms remain poorly understood [69]. Of note, recent evidence suggests that circulating peripheral EVs may be underlying this cross-talk [70]. In agreement with this hypothesis, a pioneer study by Li and colleagues demonstrated last year that exosomes isolated from the serum of endotoxin-treated mice and transferred to naive recipient mice by i.v. injection, induce dramatic microglia and moderate astrocyte activation which was accompanied by the induction of TNFα and IL-6 transcripts, and pro-inflammatory microRNAs including miR-155, in both blood and brain [71]. Relevantly, the capacity of pro-inflammatory cytokines to regulate synaptic and neuronal functions is currently well recognized [72].

Indirect, neuroinflammation-mediated, contribution of exosomes to the control of synaptic activity

As already discussed in the section dealing with Complexity of AD, microglia play a central role in neuroinflammation based on their capacity to produce immunomodulatory cytokines/chemokines [14] to subsequently impact neuronal activity [72], but also by sensing the neuronal activity in response to these immune mediators. Among the most studied underlying neuron-microglia interactions, there is the binding of ligands CD200, fractalkine (CX3CL1) and CCL2 expressed by neurons with their cognate microglia receptors CD200R, CXCL1R and CCL2R, respectively. Of note, CD200, CCL2, and CX3CL1 were identified respectively in neuroblastoma-, MSC-, and fibroblast-derived EVs, thus strongly suggesting that the interaction of these ligands carried by EVs may impact neuron-microglia interactions. Such impact could translate into release from the inhibitory tonus that neurons exert on microglia to keep it in a homeostatic state (reviewed in [41]). In this light, identification of TGF-β in neuroblastoma-derived EVs [73] is of particular importance, given that this growth factor plays a central role in the control of homeostatic phenotype of microglia [18].

Besides, EVs derived from microglia in inflammatory environment display cargo content which is distinct from the one in EVs derived from homeostatic microglia. Thus, it has been reported that EVs generated in the inflammatory conditions contain pro-inflammatory cytokines TNFα, IL-1β, and IL6 as well as miRNA (e.g., miR-155) (for recent review, see [74]). In particular, miRNAs in EVs shaded from inflammatory microglia may impact the expression of synaptic proteins in recipient neurons yielding synaptic dysfunction and loss [75] thus pointing to putative role of MDE in the control of synaptic activity. Most importantly, such MDE coming from the inflammatory environment may further trigger activation of additional microglia and astrocytes, and though contribute to implementing neuroinflammation [41]. Moreover, the expression of pro-inflammatory cytokines TNFα, IL-1β, and IL6 is higher in ADE coming from AD than age-matched neurologically normal subjects [76]. Glia EVs containing cytokines appear particularly important in terms of the role that cytokines exert in the control of synaptic activity via, for instance, synaptic up-scaling [72, 78].

Moreover, ADE obtained from plasma of AD patients contain higher level of complement proteins, including C3. Given that: 1) genetic inhibition of C3 rescues the age-related decline in synaptic function as assessed by EPSP recording and quantification of pre- and post-synaptic proteins expression [79] and 2) reported capacity of Aβ to trigger C3 production by astrocytes [80], it is plausible that Aβ sequestration and secretion via ADE could corroborate further impairment Aβ-mediated synaptic dysfunction.

Circulating exosomes: Diagnostic markers and read-out for monitoring of AD progression and treatment?

From the evidence discussed in the previous sections, it clearly appears that plasma EVs may become useful diagnostic markers for AD. Indeed, EVs (including exosomes) carry toxic proteins (e.g., Aβ and tau [51, 60]) or intracellular components (e.g., lysosomal proteins, HSP70, and ubiquitinylated proteins) whose levels differ between patients and control subjects [81]). These alterations are detectable in plasma up to 10 years before the clinical onset of the disease [81], thus pointing to EVs as attractive candidates for early and more reliable AD diagnosis. In agreement, the level of specific proteins such as advanced glycation end products in the plasma exosomes was successfully used to differentiate early from moderate stage of AD [82].

EVs could also be used as targets for development of new therapeutic approaches aimed on neuroprotection by, for instance, boosting the microglia capacity to generate EVs involved in clearance of Aβ and tau [38] or by providing healthy EVs to locally alleviate neurodegeneration by stimulating neurogenesis [59]. In this regard, use of MSC-derived EVs may turn to be particularly promising. In the work published in 2017, Cui and coworkers reported that EVs coming from MSCs and i.v. injected bimonthly for 4 months to 7-month-old APP/PS1 mice reduced plaque deposition and cerebral Aβ load, microglia and astrocyte activation, and pro-inflammatory cytokines TNFα and IL-1β. These alterations were concomitant with increase in anti-inflammatory cytokines IL-4 and IL-10 and improved cognitive performance. Interestingly, all these beneficial effects were increased if EVs were derived from MSCs cultured in hypoxia conditions [83]. These data are of paramount importance since they bring the pre-clinical proof of concept for the efficacy of using EVs for the treatment of AD-like pathology in mice at the overt stage of the disease.