Mechanisms for biogenesis and release of neuronal extracellular vesicles

Abstract

Neurons release membrane-bound extracellular vesicles (EVs) carrying proteins, nucleic acids, and other cargoes to mediate neuronal development, plasticity, inflammation, regeneration, and degeneration. Functional studies and therapeutic interventions into EV-dependent processes will require a deep understanding of how neuronal EVs are formed and released. However, unraveling EV biogenesis and trafficking mechanisms is challenging, since there are multiple pathways governing generation of different types of EVs, which overlap mechanistically with each other, as well as with intracellular endolysosomal trafficking pathways. Further, neurons present special considerations for EVs due to their extreme morphologies and specialization for membrane traffic. Here, we review recent work elucidating neuronal pathways that regulate EV biogenesis and release, with the goal of identifying directed strategies for experimental and therapeutic targeting of specific types of EVs.

Graphical Abstract

Introduction

EVs are a heterogeneous population of 40–1000 nm membrane compartments that are released from donor cells, and carry diverse protein, nucleic acid, and lipid cargoes to recipient cells [1–4]. An explosion of recent evidence shows that EVs mediate intercellular communication between different cell types in the nervous system, disposal of unwanted neuronal and glial components, and propagation as well as clearance of toxic factors in the brain [5, 6]. Further, since circulating EVs are specific to both donor and recipient cells, they are being developed for the diagnosis and treatment of neurological disease [5–10].

To experimentally interrogate and therapeutically intervene in EV functions, we need well-validated approaches to specifically manipulate EVs in the nervous system. EVs fall into two broad categories: microvesicles, which form via direct budding of cargo-enriched subdomains from the plasma membrane, and exosomes, which arise when endosomal multivesicular bodies (MVBs) fuse with the plasma membrane, releasing their intralumenal vesicles (ILVs). Even within these broad categories, multiple pathways exist for cargo packaging and EV biogenesis, leading to the diversity of released EVs [1–4]. The field has developed rigorous standards for isolating EV subtypes, establishing their functional activities, and defining their biogenesis mechanisms [11–13]. However, several challenges remain. First, different EV biogenesis and release mechanisms share overlapping membrane trafficking machinery, and also intersect with canonical endocytic and endolysosomal transport pathways, making it difficult to design specific manipulations [1–4]. Second, the majority of EV research has been conducted on mixed populations of EVs collected from body fluids, homogenized tissues, and cell culture supernatants. These bulk approaches do not resolve the heterogeneity of EVs, and are poorly suited for understanding the cellular origin and biogenesis mechanisms of EVs derived from complex tissues like the brain [8, 9]. Third, most of our understanding of EV traffic comes from non-neuronal (primarily cancer) cells [3]. Presumably, many of the EV biogenesis and release pathways identified in non-neuronal cell types also function in neurons. However, neurons present unique considerations for EV trafficking due to their extreme morphologies and compartmentalization, and because they are highly specialized for membrane traffic. Open questions resulting from these unique properties include: Where are EVs generated within these highly elongated and compartmentalized cells? How do neuron-specific adaptations of membrane traffic machinery apply to different EV pathways? How is EV traffic controlled acutely and chronically in response to neuronal activity? This review will discuss new insights into neuron-specific EV biogenesis and release pathways. We refer readers to recent reviews on the related topics of EV uptake in recipient cells [4, 14], and distinct cellular mechanisms regulating glial EVs [15, 16].

Populating EV precursor compartments

Cargoes transit to EVs via the plasma membrane and endosomes [1, 17], and a critical step diverting these cargoes from canonical endolysosomal traffic is their accumulation at membranes permissive for EV formation. Neurons add an extra layer of complexity, since these EV precursors could be locally populated with cargo in axons, dendrites, or cell bodies, or trafficked to these locations after they are loaded with cargoes. Pioneering studies at the Drosophila neuromuscular junction (NMJ) identified membrane trafficking mutants, including Rab11 (a GTPase controlling endosome-to-plasma membrane recycling), that decrease EV levels of cargoes such as the signaling protein Evi/Wntless, the calcium sensor Synaptotagmin 4, and the retroviral Gag-like protein Arc [18–21]. Notably, loss of Rab11 also led to decreased levels of at least one of these cargoes (Arc) in the donor presynaptic terminal [21], suggesting that Rab11 could promote cargo loading into an EV-permissive membrane compartment, and/or protect cargo from lysosomal degradation (Fig. 1A). Similarly, in sympathetic neurons, MVB localization and activity-dependent EV release of the p75 neurotrophin receptor correlates with its accumulation in Rab11-positive compartments [22]. In cancer cells, Munc13–4 controls Ca²⁺-dependent interaction of Rab11-positive endosomes with MVBs, rendering them competent for EV release [23]; it will be interesting to test if similar mechanisms drive activity-dependent release of EVs in neurons. Together, these results suggest that traffic via the recycling endosome populates neuronal EV precursor compartments.

Figure 1. Membrane trafficking pathways controlling neuronal EV biogenesis and release.

Blue text indicates subcellular compartments; Red text indicates membrane traffic machinery; Black text indicates EV cargoes. (A) Packaging and release of EV cargoes at the Drosophila larval neuromuscular junction. (B) Diverse mechanisms for generating ILVs may work together or independently. It remains unknown if distinct EV subtypes form within a single MVB, or in MVBs dedicated to specific cargoes. (C) Alternative pathways for EV release from neurons. Abbreviations: Cargoes (ART: AXL, RAB18, and TMED10 EVs (contain SHH), APP-CTF: Amyloid precursor protein C-terminal fragment, 2-AG: endocannabinoid 2-arachadonoylglycerol, BMP: bis(monoacylglycero)phosphate). Compartments (EE: early endosome, MVB: multivesicular body, ILV: intralumenal vesicle, EV: extracellular vesicle, ER: endoplasmic reticulum).

Generation of intralumenal vesicles in endosomes

Most studies of neuronal EVs suggest that they are derived from the MVB-dependent exosome pathway, which utilizes multiple (possibly overlapping) mechanisms for budding of vesicles into the endosomal lumen [1–4] (Fig. 1B). The best-studied mechanisms rely on Endosomal Sorting Complex Required for Traffic (ESCRT) proteins [24]. In the canonical ESCRT pathway, ESCRT-0, -I, and -II components cluster ubiquitinated cargoes while curving membranes, and then recruit ESCRT-III components, which form a helical polymer that drives fission of the ILV bud into the MVB. The VPS4 ATPase then recycles these filaments [25]. An alternative mechanism for ILV generation bypasses early ESCRT proteins, by linking EV cargoes to ESCRT-III polymers through a complex consisting of ALIX (ALG-2-interacting protein) and syntenin, a cytosolic PDZ domain adaptor protein [26]. Though syntenin has known functions in the nervous system [27], its role in neuronal EVs, via the tetraspanin TSPAN6, has only recently been explored [28]. Overexpression of TSPAN6 in non-neuronal cells increases endosome size and ILV number in a syntenin-dependent manner, and increases the number of EVs containing amyloid precursor protein C-terminal fragments (APP-CTFs). Similarly, in neurons in culture and in vivo, TSPAN6 promotes accumulation of APP fragments [28]. Thus, ESCRT-III cargo adaptors may be valuable and as-yet-unexplored points of manipulation for specific EV trafficking pathways.

Related to this idea, a longstanding question has been why there are so many ESCRT III proteins (e.g. 12 in humans) [25]. An elegant in vivo study in mice [29] demonstrated a specific requirement for the ESCRT-III protein CHMP1A in generating brain EVs containing the morphogen Sonic Hedgehog (SHH). Loss of CHMP1A resulted in aberrant MVBs in the choroid plexus and Purkinje cell layer, reduced SHH release, and therefore diminished proliferation of cerebellar granule cell precursors. Further, knockout of Chmp1a in human iPSC-derived organoids led to loss of progenitors and premature neuron differentiation. Together, these results provide a molecular explanation for why loss of CHMP1A in humans causes microcephaly. Interestingly, SHH EVs were not enriched for the canonical EV markers CD63, CD9, and Tsg101, suggesting that they define a novel EV subtype termed ART-EV (AXL, RAB18, and TMED10 EV). Thus, diverse ESCRT-III filaments may specify unique EV subtypes. Given the existence of multiple distinct EV populations from a single cell type (e.g. CD63 versus APP-EVs in hippocampal neurons [37], CD63/CD9/syntenin versus SHH/ART-EVs in SVG-A cells [29], and Syt4 versus Evi-EVs at the Drosophila NMJ [20]), the question remains whether these cargoes are packaged in distinct ILVs within the same MVB [38], or in dedicated MVBs that could be released independently. Visualization of multiple EV cargoes by super resolution imaging or immunogold electron microscopy may resolve this question.

While ESCRT molecular activities are appealing for understanding ILV formation, ESCRT components are in some cases not necessary for EV release [24]. A lipid-directed pathway, mediated by neutral sphingomyelinase (nSMase), may operate together with or in parallel to ESCRT to generate EVs. nSMase cleaves sphingomyelin to liberate ceramide, which clusters into membrane microdomains that favor inward budding [1, 24, 30]. EV release of many neuronal cargoes is sensitive to nSMase depletion or inhibition; these include Tau [31], α synuclein [32], Prion protein (PrP) [33], CD63 [34], PTEN [35], and microRNAs [34]. In addition, pathological effects of EVs can be suppressed by inhibiting the nSMase pathway [36].

Finally, control of MVB formation depends on competition between the ILV-generating machinery and other cargo-sorting domains of the early endosome, such as those defined by the retromer complex [39]. Reduced ubiquitination of the late endosome GTPase Rab7, in patient-derived cells lacking the Parkinson’s Disease-associated ubiquitin ligase Parkin, leads to dissociation of retromer from endosomes, and increased ILV formation and EV release [40]. Future studies should resolve how ESCRT-directed, lipid-directed, and competing endosomal sorting mechanisms work in concert or independently in MVB formation, while also exploring additional pathways that may be involved (Fig. 1B).

A novel lysosome damage-dependent pathway for EV biogenesis

Due to their extreme morphologies and longevity, neurons face unique challenges in efficiently disposing of unwanted materials [41]. A novel pathway, distinct from that mediating EV release from healthy cells, relieves endolysosomal damage and stress via EV release. Lysosomal damage, induced by inhibition of the phosphatidylinositol-3-kinase Vps34 in neurons in culture and in vivo, results in release of nSMase-dependent neuronal EVs enriched for APP CTFs and the endosomal phospholipid bis(monoacylglycero)phosphate (BMP), which is not normally found in EVs [38]. This effect could be recapitulated by inhibiting lysosome acidification [42], but not by simple accumulation of APP-CTFs, or by inhibiting other Vps34-dependent pathways such as autophagy. Together, these data indicate that this EV release mechanism occurs specifically in response to lysosome dysfunction. However, though endolysosomal damage-induced release may be protective for the donor cell, the released EVs may propagate neurodegenerative-disease-associated pathological proteins to other cells; therefore, EV release through this mechanism could produce competing protective and neurotoxic effects.

Targeting and fusion of MVBs with the neuronal plasma membrane

Several recent studies provide insight into the question of how EV-specific MVBs are targeted for fusion with the plasma membrane rather than the lysosome. In one example, the adaptor muskelin associates with PrP transport carriers, and promotes their retrograde traffic in both axons and dendrites [43]. Loss of muskelin-dependent traffic decreases lysosomal degradation of PrP; instead it accumulates at the plasma membrane and is released in EVs, worsening PrP pathology in vivo. In another example, Rab27a, which promotes MVB fusion with the plasma membrane [44], is protected from ubiquitination and proteasomal degradation by the brain-enriched scaffolding protein KIBRA [45]. In the absence of KIBRA, MVBs are retained intracellularly, and KIBRA knockout mice exhibit fewer brain EVs. Finally, a novel membrane remodeling function for Hsp90 was recently discovered, independent of its canonical chaperone activity [46]. Hsp90 promotes fusion of MVBs with the plasma membrane in a reconstituted system, and is required for release of Evi/Wntless-containing EVs at the Drosophila NMJ, but notably has no effect on lysosomal degradation. Importantly, trafficking machinery that specifically affects the EV release step (such as KIBRA and Hsp90) will enable uncoupling of potentially separable roles of EV cargoes in donor cells versus recipient cells.

Alternative pathways for neuronal EV biogenesis and release

While most studies of neuronal EVs focus on MVB-derived exosomes, neurons exhibit several unique categories of EVs (Fig. 1C). One such example highlights a novel mechanism by which lipid signaling molecules such as endocannabinoids are released in EVs [47]. In response to cocaine administration, midbrain dopaminergic neurons release EVs containing the endocannabinoid 2-arachadonoylglycerol (2-AG), leading to retrograde inhibition of GABA release. Cocaine disrupts the interactions between sigma-1 receptor (Sig-1R), a chaperone protein that is tethered to mitochondrion-associated ER membrane (MAM), and the GTPase Arf6. Freeing Arf6 from MAM promotes a myosin light chain kinase (MLCK)-dependent EV release pathway. Interestingly, a similar Arf6-MLCK pathway is specifically attributed to microvesicle release in tumor cells [48], suggesting that 2-AG may also be released in microvesicles.

EVs are also released from primary cilia [49]. Ciliated sensory neurons in C. elegans release EVs that modulate male mating behavior, via an ESCRT-0, ESCRT-I, and Alixindependent mechanism, suggesting that they are not generated by the canonical MVB pathway [49, 50]. Interestingly, following their release these EVs require intraflagellar transport for movement away from the donor cell [50]. Conversely, failure to shed EVs from the specialized cilia of mammalian photoreceptors causes defects in their morphogenesis [51]. At a much larger scale, C. elegans neurons can also shed micron-sized “exophers” from their soma to dispose of toxic materials [52]. Whether these mechanisms are restricted to specialized or stressed neurons remains unknown.

Finally, two complementary studies, in mouse hippocampal neurons [53] and at the Drosophila NMJ [21], describe a novel EV pathway mediated by Arc (activity-regulated cytoskeleton-associated protein), a key regulator of synaptic plasticity and scaling [54]. Arc bears homology to retroviral Gag proteins, and can enclose and release its own and other abundant mRNAs in viral-like capsids that are taken up and translated in recipient cells. Little is known about Arc-EV biogenesis, but there may be some commonalities with other EV pathways; for example, Drosophila Arc mRNA and protein traffic at the NMJ depend on Rab11 [21]. It will be interesting to explore the EV-related roles of regulators of Arc oligomerization, such as Arc ligands [55] or phosphorylation [56]. Remarkably, vertebrate and Drosophila Arc arose from independent retroviral repurposing events, suggesting that this may be a common adaptation for intercellular communication, and that other such carriers may exist.

Subcellular location of neuronal EV release

Given the extreme morphology and compartmentalization of neurons, understanding EV functions requires knowing from where in the neuron EVs are released. Several studies demonstrate accumulation of EV cargoes and/or MVBs in the somatodendritic compartment. In C. elegans, manipulating the small GTPase RAB-5 causes release of EFF-1 positive EVs near the soma of mechanosensory neurons [57]. Mammalian cortical and hippocampal neurons accumulate the EV cargoes CD63, CD9, Flotillin, tetanus toxin heavy chain, and Tau in the somatodendritic compartment [17, 34, 58, 59]. Further, electron microscopy shows MVB accumulation within pyramidal, Purkinje, and cortical neuron dendrites, suggesting release from these sites [17, 29]. Notably, mammalian Arc mRNA is transported to and locally translated in dendrites in response to activity-dependent cues [54], suggesting a direct mechanism for its preferential release from dendrites.

There is also extensive evidence for release of neuronal EVs from other subcellular compartments, including cilia (described above), and axons. For example, though EV cargoes CD9 and Tau accumulate in the somatodendritic region of cultured neurons, microfluidic compartmentalization shows that they are also trafficked anterogradely and released from axons [58, 59]. In similar experiments, oligomers or aggregates of the ALS-associated protein TDP-43 (Transactive response DNA-binding protein-43) are released by axons, and efficiently transmitted in EVs to recipient cells [60]. Likewise, in Drosophila motor neurons, EV cargoes are released from axon terminals [18–21]. It will be important to define compartment-specific release mechanisms for EVs.

Next challenges

While it is clear that neurons release heterogenous EVs with important functions, we are only beginning to unravel the complexities of neuronal EV traffic. In particular, EVs are released from neurons in response to stimulation or depolarization [17, 53, 59, 61–67], but the mechanisms and (in most cases) functions of activity-dependent release are unknown. Understanding these diverse pathways and their activity-dependent regulation may allow manipulation of the specific steps in neuron-specific EV biogenesis and release without impacting other trafficking pathways or other cell types. This would be a significant step towards another future major challenge: deciphering EV-mediated crosstalk between different cells within the nervous system. EVs are targeted to and taken up by specific recipient cells [37, 59, 64, 68], but it remains unclear whether this is determined by the EV cell of origin, the target cell, the specific cargo carried by the vesicle, or a context-specific combination of these factors. To answer these questions, in vivo models for EV traffic and physiological function will be transformative for the field [13, 19, 29, 34, 69]. Sorting out the specific pathways of EV cargo sorting, biogenesis, release, targeting, and uptake in vivo will be essential for interrogating EV function for basic science, for developing EV-directed therapeutics, and for understanding the origins and significance of EV biomarkers.

Highlights.

• Extracellular vesicles (EVs) serve important functions in the nervous system, are misregulated in multiple neurological diseases, and are being developed as biomarkers and drug delivery vehicles to diagnose and treat these diseases.

• Growing evidence suggests that there are numerous distinct cargo trafficking and release pathways for EVs in neurons.

• Defining these specific trafficking routes will be critical for causal studies that distinguish the functions of EVs from other intertwined signaling and endosomal trafficking pathways.