Astrocytes‐derived extracellular vesicles in motion at the neuron surface: Involvement of the prion protein

Abstract Astrocytes‐derived extracellular vesicles (EVs) are key players in glia‐neuron communication. However, whether EVs interact with neurons at preferential sites and how EVs reach these sites on neurons remains elusive. Using optical manipulation to study single EV‐neuron dynamics, we here show that large EVs scan the neuron surface and use neuronal processes as highways to move extracellularly. Large EV motion on neurites is driven by the binding of EV to a surface receptor that slides on neuronal membrane, thanks to actin cytoskeleton rearrangements. The use of prion protein (PrP)‐coated synthetic beads and PrP knock out EVs/neurons points at vesicular PrP and its receptor(s) on neurons in the control of EV motion. Surprisingly, a fraction of large EVs contains actin filaments and has an independent capacity to move in an actin‐mediated way, through intermittent contacts with the plasma membrane. Our results unveil, for the first time, a dual mechanism exploited by astrocytic large EVs to passively/actively reach target sites on neurons moving on the neuron surface.

the environment surrounding the donor cell by changing the molecular composition of the extracellular fluid (Iraci et al., 2017). EVs were often classified into exosomes, which originate from endocytic pathway, microvesicles or ectosomes, that shed directly from the plasma membrane (Cocucci & Meldolesi, 2015) of healthy cells and apoptotic bodies, which are released by cells undergoing apoptosis. Currently EVs are preferentially distinguished based on physical characteristics (size and density), biochemical composition and cell of origin (Thery et al., 2018), due to difficulties in isolating EVs on the basis of their mechanisms of biogenesis. Small EVs (sEVs) are < 100-200 nm in size, while medium/large EVs range from 200 to 1,000 nm in diameter (Thery et al., 2018). Despite small EVs (exosomes) were the most widely studied EV population, conveying the false impression to be more important than other EV types (Tkach et al., 2018), the need is emerging to comprehensively study all type of secreted EVs to fully define the activity of these highly heterogeneous structures.
In this study we focused on this less studied population of large EVs, whose interaction with neurons can be monitored through bright field microscopy. Exploiting optical manipulation to study single EV-neuron dynamics, we show that most large EVs move on the neuron surface in anterograde or retrograde directions, and provide a dual mechanism for EV motion. In particular, we show that the majority of large EVs exploits the prion protein (PrP) and its neuronal receptors to passively navigate along neuronal projections while a minor fraction of EVs has an independent capacity to move on the neuron surface. These findings represent first evidence for the existence of a regulated trafficking of large EVs outside the neuron surface and provide initial evidence on the molecular mechanism controlling large EV motion.

. Cytotoxicity assay
Aliquots of conditioned medium (KRH) were collected from 96-well plated astrocytes maintained in basal conditions or stimulated with ATP for 30 min for the evaluation of lactate dehydrogenase (LDH) activity using the cytotoxicity assay kit (CyQUANT LDH Cytotoxicity Assay Kit, Invitrogen, Carlsbad, CA, USA). LDH levels were measured via a coupled enzymatic assay, which resulted in the formation of a coloured product (Formazan), the absorbance of which was recorded at 490 nm and 680 nm (SPECTROstar Nano, BMG Labtech, Ortenberg, Germany). To measure maximum LDH activity, cells were treated with lysis buffer, according to the manufacturer instructions.

. EV characterization by TRPS
Size distribution of medium/large EVs (10,000 × g pellet) has been evaluated through Tunable Resistive Pulse Sensing (TRPS) technique, using a qNano instrument (IZON, Christchurch, New Zealand) which provides accurate size measurements of EVs by detecting current blockage from EV passing through a nanopore. Cell supernatant of 7 × 10 6 astrocytes was partitioned in two aliquots, after clearing from cell debris (300 × g × 10 min twice). One aliquot was centrifuged at 10,000 × g for 30 min, the pellet resuspended with Izon EV reagent kit and analysed using an NP400 pore. The other aliquot was further centrifuged at 2,000 × g for 20 min, to eliminate bigger EVs, before being pelleted at 10,000 × g, resuspended and analysed using an NP200. Removal of bigger EVs limited NP200 pore obstruction. Analysis by NP400 pore was used to determine the concentration of EVs > 200 nm, while by NP200 pore we evaluated < 200 nm EV concentration (results were filtered for > 200 nm and < 200 nm respectively).

. F-actin and G-actin isolation from neurons and EVs
F-actin/G-actin ratio was assessed as previously described (Pyronneau et al., 2017). Briefly, cultured neurons or freshly isolated EVs, produced by 75 × 10 6 astrocytes, were resuspended in cold lysis buffer (10 mM K 2 HPO 4 , 100 mM NaF, 50 mM KCl, 2 mM MgCl 2 , 1 mM EGTA, 0.2 mM DTT, 0.5% Triton X-100, 1 mM sucrose, pH 7.0) and centrifuged at 15,000 × g for 30 min. The G-actin supernatant was transferred to a fresh tube, and the F-actin pellet was resuspended in lysis buffer plus an equal volume of a second buffer (1.5 mM guanidine hydrochloride, 1 mM sodium acetate, 1 mM CaCl 2 , 1 mM ATP, 20 mM tris-HCl, pH 7.5) and then incubated on ice for 1 h to convert F-actin into soluble G-actin. Samples were centrifuged at 15,000 × g for 30 min, and the supernatant (containing the F-actin, which was converted to G-actin) was transferred to a fresh tube. F-actin and G-actin samples were loaded with equal volumes and analysed by Western blotting.

. Western blotting
Astrocytes and neurons were lysed with a buffer containing 1% SDS, 2 mM EDTA pH 7.4, 10 mM Tris-HCl pH 7.4 and proteases inhibitor (1:500). A modified version of the Laemmli buffer was then added to a final 1× concentration (15% SDS, 575 mM sucrose, 325 mM Tris-HCl pH 6.8, 0.5% β-mercaptoethanol, 0.01% bromo-phenol blue). EVs released from 20 × 10 6 astrocytes were lysed with the same Laemmli buffer. 10 μg of total lysates of neurons or astrocytes and the whole pellet of EVs were loaded on a polyacrylamide gel. Proteins were then separated by electrophoresis, blotted on nitrocellulose membrane and probed using mouse anti-PrP (

. Cryo-EM of EVs and actin analysis
Cryo-EM allows imaging of samples without the addition of any heavy metals or fixatives, which might cause artefacts, with the drawback of yielding a lower contrast. The sample is frozen so rapidly that the water vitrifies forming no ordered crystals, and the native structure of the sample is preserved. Freshly prepared EVs in the 10,000 × g pellet, released from 7.5 × 10 7 astrocytes, were resuspended in saline and vitrified by applying a 3.5 μl droplet onto a holey carbon grids (Copper 300-mesh Quantifoil R2/1) glow discharged for 45 s at 40 mA using a GloQube system (Quorum Technologies, East Sussex, UK). After 60 s incubation, the grid was plunge-frozen in liquid ethane using a Vitrobot Mk IV (Thermo Fischer Scientific, Waltham, MA, USA) operating at 4 • C and 100% RH. Images of the vitrified specimen were acquired using a Talos Arctica transmission electron microscope (Thermo Fisher Scientific, Waltham, MA, USA) operating at 200 kV and equipped with a Falcon 3EC direct electron detector (Thermo Fischer Scientific, Waltham, MA, USA). Images were acquired with an exposure time of 1 s and a total accumulated dose of 80 electrons per A2 at a nominal magnification of 45,000 ×, corresponding to a pixel size of 2.29 Å/pixel at the specimen level, with applied defocus values between 2 and 4 μm. Images acquired with Volta phase-plate were recorded at defocus values between 0.5 and 1 μm. Contrast transfer function (CTF) estimation was performed using CTFFIND4 (Rohou & Grigorieff, 2015). All EVs contained within the image were measured using ImageJ software. EVs were classified in single round EVs, oval EVs, tubular EVs, irregular EVs or double/multilamellar EVs. Oval EVs and tubular EVs were measured along their long axis. Actin filament picking, segments extraction and analyses were performed using RELION-3 software (Zivanov et al., 2018). Filaments were manually picked using RELION's helix picker and a total of 1,381 segments were extracted using a box size of 100 pixel and inter-box distance of ∼10% (Rohou & Grigorieff, 2015;Zivanov et al., 2018).

. Subcellular localization of large mCLING-labelled EVs
EVs in the 10,000 x g pellet were resuspended in 500 μl of sterile and 0.1 μm filtered PBS and incubated with 400 nM mCLING-ATTO 647N-labeled (Synaptic System, Goettingen, Germany) in a black tube on ice for 5 min. The reaction was quenched by adding 500 μl of 1% BSA in PBS. Then EVs were diluted in 10 ml of PBS, re-pelleted at 10,000 x g to eliminate the dye excess, resuspended in neuronal medium and added to membrane-targeted GFP-transfected hippocampal neurons for 1 h before fixing the cells with 4% paraformaldehyde for 15 min. Coverslips were then mounted on a microscope slide and Z-stacks were acquired with a Zeiss LSM800 confocal microscopy (Oberkochen, Germany). Analysis of EV localization on neuron processes and cell bodies was developed with ImarisViewer 9.7.0 in the Z-projection. EVs interacting with thin processes (≤ 2 μm), onto we usually placed EVs by optical manipulation, were distinguished from those localized on primary dendrites in continuity with the cell body.

. Optical tweezers (OT)
An IR laser beam (1064 nm, CW) for trapping was collimated into the optical path of an inverted microscope (Axiovert 200 M, Zeiss, Oberkochen, Germany) through the right port of the microscope. The trapping beam was directed to the microscope lens (Zeiss 63X objective, oil immersion, NA 1.4) by the corresponding port mirror (100%) and the tube lens. Optical trapping and manipulation of large EVs was performed following the approach previously described (Prada et al., 2016). Immediately before recording, neurons were washed to remove EVs constitutively released by neurons, and large EVs produced by 2.5 × 10 6 astrocytes pre-loaded with Calcein (FilmTracer Calcein Green Biofilm Stain, Life Technologies, Carlsbad, CA, USA; 26 μM for 1 h) were added to neurons plated on glass coverslips and maintained in 400 μl of neuronal medium in a 5% CO 2 and temperaturecontrolled recording chamber at 37 • C. Calcein positive large EVs are about 77% of the EVs labelled by mCLING (n = 53) and 72% of EVs detectable in bright field (n = 134). As soon as a Calcein-positive EV appeared in the recording field (Supplementary Figure 1a), it was trapped and positioned on a selected neuron by moving the cell stage horizontally and the microscope lens axially. EVs larger than 200 nm were trapped and placed in contact with neurons although we cannot exclude that even EVs < 200 nm, due to the diffraction limit, could be imaged as EV with size = 200 nm and monitored in live imaging.

. Tracking of single EV
EV position was determined for each video frame (2 frames every 5 s) using a custom MATLAB code (it.mathworks.com). To characterize the EV displacement on the neuron process, 2 distances were calculated: the maximum distance the EV reached from the initial position (in both directions), and the length of the path travelled by the EV in 20 min. Mean velocity and distances were extracted from EV coordinates using a custom R code that exploits the Pythagorean theorem to reconstruct the EV path point-to-point (www.r-project.org). As the code works well only for straight path, the distance from the starting point was calculated using ImageJ software (www.imagej.nih.gov/ij/), when the EV trajectory was not linear, tracing the path manually. The R custom code used for EV analysis is available at https://doi.org/10.6084/m9.figshare.12808211.v1. We classified as 'static EVs' i) the EVs with net displacement < of the EV diameter and ii) EVs showing only random (Brownian) motion.

. Fura- videomicroscopy
A total of 14 DIV hippocampal neurons were loaded with 2 μM Fura-2 (Invitrogen, Carlsbad, CA, USA) pentacetoxy methylester for 45 min at 37 • C, washed and transferred to the recording chamber of an inverted microscope (Axiovert 100; Zeiss, Oberkochen, Germany) equipped with a calcium imaging unit Polychrome V (TILL Photonics GmbH, Planegg, Germany). Images were collected with a CCD Imago-QE camera (TILL Photonics GmbH, Planegg, Germany) and analysed with FEI Live Acquisition 2.6.0.14 software (FEI GmbH, Munich, Germany). After excitation at 340 and 380 nm wavelengths, the emitted light was acquired at 505 nm at 1 Hz. Calcium concentration was expressed as F340/F380 fluorescence ratio. The ratio values in selected region of interest corresponding to neuronal cell bodies were calculated from sequences of images to obtain temporal analysis.

. ATP measurements in EVs
The biological assays for ATP detection were performed modifying previously described methods (Coco et al., 2003;De Marchi et al., 2020), using fura-2 loaded oligodendrocytes or melanoma B16-F10 expressing pmeLUC as ATP sensor cells. For the fura-2 assay, large EVs, derived from 5 × 10 6 cells under ATP stimulation or constitutively released by 10 × 10 6 cells were resuspended in 200 μl of KRH and split into two aliquots before testing on oligodendrocytes. Only one aliquot was pretreated with Apyrase (30 units/ml, Sigma, St. Louis, MO, USA) for 15 min to assess whether the calcium responses evoked by the EVs in oligodendrocytes were due to EVs-associated ATP. The other aliquot was used as a positive control. To damage the EV membrane and favour ATP leakage into KRH, one aliquot of EVs was subjected to 3 cycles of freeze and thaw: for every cycle, EVs (resuspended in 100 μl of KRH) were immersed in liquid nitrogen for 30 s (freeze) and then put in a warm bath at 37 • C for 30 s (thaw).
To deplete large EVs from their luminal content, EVs were broken or not by hypo-osmotic stress (via suspension in 500 μl of 7.3 mM PBS at 4 • C for 30 min) and re-pelleted at 100,000 × g for 1 h before being resuspended in 100 μl of KRH and tested on oligodendrocytes. For the pmeLUC assay (De Marchi et al., 2019Pellegatti et al., 2005), 10 × 10 3 B16-F10 pmeLUC cells were seeded into a 96-wells plate in complete RPMI-1640 medium (Sigma-Aldrich), let adhere and grow for 24 h. Subsequently, 15 × 10 6 astrocytes were incubated with 100 μM BzATP (Sigma-Aldrich, St. Louis, MO, USA) for 30 min in 7 ml of KRH. Large EVs were collected as previously described and resuspended in sterile PBS (Euroclone, Milan, Italy). D-Luciferin (Promega, Madison, Wisconsin, USA) was added to the wells containing B16-F10 pmeLUC cells at a 60 μg/ml concentration. The basal luminescence emission was acquired at medium binning for 10 min with an IVIS Lumina luminometer (Perkin Elmer, Milan, Italy). Cells were then incubated with 10 μl of EV suspension or 500 μM ATP (Sigma-Aldrich), and luminescence emission was acquired for extra 10 min. Luminescence data were analysed with the Living image Software (Perkin Elmer, Milan, Italy) and expressed as photons/second of acquisition.

. Immunocytochemistry
Hippocampal neurons from FVB mice, plated on 18 mm glass coverslips were fixed with 4% paraformaldehyde for 15 min. Samples were then incubated with GSDB blocking solution (Goat Serum Dilution Buffer; containing goat serum, phosphate buffer and sodium chloride) in non-permeabilizing conditions for 30 min at RT. Coverslips were then incubated with primary mouse antibodies anti-PrP (W226) and anti-NCAM (AB5032, Merk Millipore) diluted in GSDB for 1 h, followed by speciesspecific fluorochrome-conjugated secondary antibodies: anti-Mouse Alexa Fluor 488, anti-Rabbit Alexa Fluor 594 and DAPI (4′,6-diamidino-2-phenylindole, Invitrogen), to stain nuclei. Coverslips were mounted on microscope slides and images were acquired with a Zeiss LSM800 (Oberkochen, Germany) confocal microscope and processed with ImageJ software.

. Data analysis
Statistical analysis was performed on all datasets. If not otherwise specified, data are expressed as means ± SEM. Data were first tested for normal distribution with GraphPad Prism 6 software, then the appropriate statistical test has been used (see figure legends). The accepted level of significance was P ≤ 0.05, indicated by one asterisk; those at P ≤ 0.01 are indicated by double asterisks, the ones at P ≤ 0.001 are indicated by triple asterisks and P ≤ 0.0001 are indicated by four asterisks.

. Large EVs produced from primary rat astrocytes have different morphologies
Primary rat astrocytes were exposed to ATP for 30 min to promote EV shedding and medium/large EVs were isolated by ultracentrifugation at 10,000 × g after pre-clearing of cell supernatants from cells and debris at 300 × g 13 . ATP stimulation for a brief period in extracellular saline ensured higher yield of medium/large EVs (Figure 1a), similar in size to constitutive EVs, as indicated by tunable resistive pulse sensing (TRPS), a technique suitable to analyse highly polydisperse EVs (Figure 1b and Supplementary Figure 1a). TRPS analysis revealed that EVs released under ATP stimulation are highly heterogeneous in size and have a mean diameter of 291.39 ± 3.58 nm ( Figure 1b). According to this method, medium/large EVs (> 200 nm) were ∼58% of the whole EV population. Short ATP stimulation also ensured high purity of EV preparation without the need of further purification steps. This was indicated by Western blot analysis of medium/large EVs for positive and negative EV markers compared to cell lysate/small EV fraction, isolated at 100,000 × g ( Figure 1c). Analysis of lactate dehydrogenase (LDH) revealed no difference between the medium collected at the end of ATP stimulation and under constitutive conditions (Figure 1d), excluding a link of F I G U R E  Ultrastructure of medium/large EVs released from ATP-stimulated astrocytes. (a) Histogram shows basal and ATP-induced production of EVs in the 10,000 × g pellet from cultured astrocytes. The number of ATP-induced EVs detected by TRPS with a qNano in 100 μl of 0.1 μm-filtered Krebs-Ringer solution were normalized to constitutive EVs (ATP-EVs vs. constitutive EVs t test P = 0.033; N = 3). (b) Size distribution of EVs in the 10,000 × g pellet according to TRPS. Note that medium/large EVs (> 200 nm) are ∼58% of EV population (n = 2507). (c) Representative immunoblot of astrocytes (15-EVs released upon short ATP stimulation to an apoptotic fate of the cells. Morphological analysis using high resolution cryoelectron microscopy (cryo-EM) (Yuana et al., 2013;Zabeo et al., 2017) revealed that most EVs in the 10,000 × g pellet were rounded in shape (83%, Figure 1e, g) and made by a single lipid bilayer (80%), although very heterogeneous in size (Figure 1e-f). Their size range was from 20 nm to 1300 nm (mean ± SEM 217 ± 12 nm, Figure 1f; number of EVs n = 307) with large EVs (> 200 nm) being 34% of the whole EV population, a lower percentage compared to TRPS. However, cryo-EM may underestimate large EV presence in the sample, as large EVs may not be included in the ∼100 nm ice film for cryo-EM detection. A subpopulation of EVs were multi-lamellar, i.e. contained two or more vesicles in their lumen (20%, Figure 1e panel C), while a minor percentage of EVs had a tubular (3.6% Figure 1e panel I and Figure 1g) or irregular shape (1%, Figure 1e panel G-H and Figure 1g). Other peculiar features of some EVs were rough surface (2.6%, Figure 1e panel A), likely due to the presence of transmembrane proteins, electron dense content (24%, Figure 1e panel A, C) and presence of actin filaments in their lumen (2%, Figure 1e panel D-I), a typical feature of larger EVs from our preparation, with a tubular/elongated shape (mean major axis = 435 ± 70 nm; n = 12), whose frequency may be under underestimated by cryo-EM. In tubular/elongated EVs actin filaments were oriented in parallel to one another and extended along the tubule (Figure 1e panel H-I and Figure 1h). Actin filaments in tubular EVs displayed canonical width of ∼7 nm. Inter-filament distance was ∼10 nm (Figure 1i) with inter-subunit distance of ∼60 Å (Figure 1i). In round EVs, filaments were randomly distributed, and their structure was not sufficiently defined to resolve their molecular composition (

. Single-particle tracking of astrocytic EVs on the neuron surface
To elucidate whether astrocytic EVs in the 10,000 × g pellets interact with neurons at preferential sites, we gently placed single EVs on different compartments of hippocampal neurons (cell body, dendrite and axon) by using optical manipulation (Prada et al., 2018) and examined EV-neuron interaction through live microscopy, starting from the observation that EVs loaded in the cell medium can randomly attach both to the soma and the neurites. EVs in the 10,000 x g pellet were added to neuronal medium, trapped by infrared laser tweezers and positioned on the selected compartment. Thirty seconds after contact the trapping laser was switched off and time-lapse images were collected at 2-20 Hz for 20 min to monitor EV-neuron dynamics (Figure 2a). Medium/large EVs (> 200 nm, accounting for about 60% of the 10,000 × g EV population according to TRPS analysis) were mostly imaged, as small EVs may be too tiny to be visualized in light microscopy and delivered to neurons. We found that this population of EVs, for simplicity called large EVs from now on, adhered more efficiently to the cell body (81% of adhesion) than neurites (62% of adhesion) (Figure 2b-d, Movie S1 and Movie S2 online with this article). This difference might be due to differential expression of EV interacting molecules on the two neuronal compartments or the fact that the soma offers a larger surface around EV contact point, making easier for the EV to find their specific surface interactors and/or undergo an higher number of interactions. However, we cannot exclude that higher adhesion of large EVs to the cell body might also result from better accuracy in placing the EVs on a larger cell surface. About 50% of adherent large EVs were round in shape, while the remaining were tubular (9%) or oval/irregular (41%), as analysed after turning the infrared laser off, to avoid possible EV distortions (Supplementary Figure 2, n = 212). Surprisingly, after adhesion a large fraction of EVs (58% ± 3.25%, n = 266, number of independent experiments N = 107) displayed net movement from the contact site, surfing on the plasma membrane of neurons. Although phase contrast resolution did not always allow to discriminate between intra-and extra-cellular EV movement, EV recapturing by the laser tweezers in close proximity to contact site of the EV on neuron blocked 100% of EV motion (Table 1, n = 25/25) (Movie S3) while didn't hamper the transport of intracellular organelles (Movie S3), indicating that EVs were moving outside neurons. In addition, some EVs (n = 11/212) jumped from one process to another (Movie S4), even belonging to different neurons, allowing neuron-to-neuron EV transition (Movie S5), confirming extracellular EV motion. A total of 70% ± 3.34% of EVs were in motion along neurites (n = 212, N = 95) (Figure 2b, e, Movie S1) and run sizable length (mean length = 143 ± 9.95 μm/20 min) to a maximal distance of 9 ± 0.51 μm from the contact point (Figure 2f). A lower percentage of large EVs moved on the cell body (44 ± 10.50%, n = 54, 2 μg of cell lysate), medium/large EVs (lEVs) and small EVs (sEVs) pelleted at 10,000 × g and 100,000 × g respectively, released from 20 × 10 6 cells, for the EV marker Alix, Flotillin 1, Annexin A2 and the negative Golgi marker GS28. (d) Basal and ATP-induced LDH release in the astrocyte supernatant (one-way ANOVA P = 0.0002, Dunn's multiple comparisons test: ATP vs. Maximum LDH release P = 0.0011, Spontaneous vs. Maximum LDH release P = 0.0012). Maximum LDH release was measured by exposing astrocytes to the lysis buffer. (e) Representative cryo-electron microscopy micrographs of round EVs (A-F), made by the characteristic phospholipid bilayer, as well as of less abundant irregular or tubular EVs (G-I) in the 10,000 × g pellet (low-pass filtered at 40 Å, except for the first micrograph). Note the presence of filaments (arrows) in elongated/tubular EVs (G-I) but also in a subpopulation of round EVs (D-F). Scale bars = 100 nm. (f) Size distribution of astrocyte-derived EVs, detected by cryo-EM in the 10,000 × g pellet. Note that medium/large EVs (>200 nm) are ∼34% of the EV population. (g) Prevalence of astrocytic EVs with different shape and morphologies in the 10,000 × g pellet.   (Figure 2e). Their motion was restricted to the cell body, with no EVs passing from cell body to neurites (Figure 2c, Movie S2). After placement on the cell body, large moving EVs could be recaptured by the laser trap (Table 1, n = 10/10), proving that they were outside neurons, whereas only 27% of 'static' EVs, that remained adherent to the place of delivery, were recaptured by optical tweezers, being transiently attracted towards the laser (Table 1, n = 4/15). Although we could not exclude that the laser trapping force (maximum 50 piconewton for our setup) might not be able to re-trap EVs strongly adhering to the neuron surface (Ashkin, 1992;Bernecker et al., 2021), this suggested that a fraction of large EVs may undergo internalization into the neuron cell body. To assess this hypothesis, we labelled large EVs with the fluorescent dye mCLING (Brenna et al., 2020) and analysed by confocal microscopy the localization of mCLING-labelled large EVs 1 h after in bulk addition to neurons transfected with membrane-targeted GFP. This analysis confirmed that a major fraction of large EVs (94%, n = 33) are taken up by neurons in the cell bodies (Supplementary Figure 3). Run length (102 ± 22.37 μm/20 min) and maximal distance from the contact point (6 ± 0.88 μm, Figure 2f) of large EVs moving on the cell bodies were shorter compared to neurites (run length soma vs. neurites, Mann-Whitney test, P = 0.030). Thus, EVs moving on neurites, either dendrites or axons, were selected for further analysis. Axons were distinguished from dendrites by their longer length and their smaller size in developing cultures (≤12 days in vitro), and by the absence of spines in mature cultures transfected with GFP (≥ 13 days in vitro) (Figure 2g).

. Directional movement of large EVs along neuronal processes
By a custom MATLAB code, we analysed the dynamics of all large EVs which adhered to neurites, including those that remained anchored to the contact site. Among the latter (∼35% of total EVs) only few EVs were virtually immobile (EV displacement < EV diameter) while the remaining fraction showed random Brownian motion, restrained by a tether to the neuron membrane (Figure 2h). The mean radial displacement from the attachment point was 1.8 ± 1.1 μm. For simplicity, both virtually immobile EVs and EVs displaying only radial motion are referred to as 'static EVs' hereafter ( Figure  the typical trajectory of an EV that retraced several times the same regions of a growing axon while exploring filopodia on the way to the growth cone. Visualization of EV trajectories in neurons cultured in vitro for 2-17 days revealed that most EVs moved along axons and dendrites in a prevalent direction (run length > 2 × length in the opposite direction), either retrograde, towards the neuron cell body, or anterograde, away from the soma. However, a significant fraction of EVs (n = 29/84) were characterized by truly bidirectional movement, running similar length along neurites in both directions, as indicated by the plot of the maximal positive and negative distances from the contact point ( Figure 2m). Importantly, along mature axons half of large EVs (50%) moved in an anterograde direction (n = 14, N = 9). The speed of large EVs along neurites ranged from 0.011 to 0.412 μm/sec (mean speed = 0.126 ± 0.008 μm/sec, n = 135). EV velocity was similar on axons and dendrites in immature neurons, up to 11 days in vitro (Figure 2n). With age, and in parallel to increased complexity of dendrites in terms of morphology and composition, both EV speed (Figure 2n) and the percentage of moving EVs decreased (Figure 2o) on dendrites. Conversely, EV speed did not change on axons, with a significant difference between the two processes in neurons ≥12 days in vitro (Figure 2n).

. Stable contact of large EVs at the neuron surface promotes formation of novel protrusions
Next we asked what happens at preferential contact points, where large EVs stop moving and at places of stable adhesion, where EVs immediately establish a stable interaction with neurites. We found that 50% of large EVs (n = 12/24) could be recaptured at sites of persistent contact on neurites, as indicated by transient shift of EVs towards the laser trap (Table 1, Supplementary Figure  4). This suggested that at least half large EVs were still outside neurites at the end of recording (20 min). Further analysis of mCLING-labelled large EV localization in neurons transfected with membrane GFP showed that the vast majority of EVs (94% n = 69) was still outside thin (≤ 2 μm) neurites 1 h after EV addition (Supplementary Figure 3), while along primary dendrites most EVs were internalized (82%, n = 38). Collectively, these data point to large EV size/neurite thickness ratio as key factor retaining EVs at the neuron surface. They also suggest that not all EVs outside neurons can be re-captured by the laser trap. Interestingly, time-lapse imaging revealed that astrocyte-derived EVs have the ability to induce formation of new protrusions at or around the sites of persistent contact (Figure 3a-c, n = 9/32 N = 9). Protrusions formed after 3-20 min of persistent contact. Novel protrusions were not always detectable in bright field while were easily visible by confocal microscopy, which allowed to capture thin filopodia budding from RFP transfected dendrites (Figure 3d, n = 6/30 N = 12). In one particularly striking case (Figure 3c), a moving EV induced formation of multiple filopodia around the stop site on the dendritic shaft. One filopodium budded exactly from the contact site and grew under the EV. Other filopodia appeared around the stop site and then converged towards the EV, ultimately leading to a single branch/protrusion. Carboxylated microsynthetic particles, covalently coupled with BSA, were used as negative control for filopodia formation driven by astrocyte large EVs. New filopodia were very rarely observed upon delivery of BSA-coated beads (BSAb) in bright field (BSAb n = 2/42 N = 8 vs. EVs n = 9/32 N = 9, t test P = 0.0017) and by confocal microscopy (BSAb n = 1/38 N = 5 vs. EVs 6/30 N = 5) (Figure 3e-f), ensuring the specificity of astrocytes-derived EVs in inducing new filopodia. Taken together these findings suggest that large astrocytic EVs may scan the neuron surface in search for specific spots of the plasma membrane where new protrusions can be induced.

. Motion of EVs depends on intact actin cytoskeleton in neurons
Mechanistically, binding to a neuronal receptor coupled to a moving cytoskeletal network could drive EV motion (Figure 4a). This hypothesis was suggested by two independent observations. First, we found that carboxylated microsynthetic particles covalently coupled with prion protein (PrP), a GPI anchor protein highly enriched in brain EVs (Brenna et al., 2020;Faure et al., 2006) (Figure 4b, Fevrier et al., 2004), were more efficiently transported along neurites compared to control beads, covalently coupled with albumin, when delivered to neurons by optical tweezers, mimicking the trafficking of astrocytic EVs outside neurons (Figure 4c-d). This observation suggested that binding of PrP to a neuronal receptor enhances passive movement of the synthetic particle on the neuronal surface. Interestingly, a lower percentage of PrP-coated beads, similar to control beads, were transported along the surface of PrP knock out (PrP -/-) neurites (Figure 4c-d), suggesting that neuronal PrP may act as receptor for PrP coupled to the synthetic beads. Indeed, PrP is capable to undergo homophilic interaction with a PrP molecule in trans and can elicit contact formation Malaga-Trillo & Sempou, 2009;Nguyen et al., 2019). No differences were observed in the fraction of PrP-coated beads and control beads adhering to either wild-type or PrP -/neurites (adhesion of PrP-coated beads on wild-type neurons 100%, N = 7; PrP-coated beads on PrP -/neurons 100%, N = 9; BSA-coated beads on wild-type neurons 79%, N = 8; BSA-coated beads on PrP -/neurons 76%, N = 7), excluding a major contribution of neuronal PrP to EV-neuron interaction. Second, we found examples of EVs placed on very thin neuronal processes which were induced to move on the neuron surface by the transit of large intracellular vesicles, actively transported along the neuron processes (Figure 4e; Movie S7 and S8; n = 5/135). These observations suggested that the large intraneuronal vesicles, which exceeded the diameter of the neurites, may occasionally drag the cytosolic portion of the EV receptor during their transport along the cytoskeleton, generating simultane- To assess whether energy-dependent cytoskeleton rearrangements could drive EV motion on the neuron surface, we first intoxicated neurons with the metabolic inhibitor Rotenone (2μM) for 1 h. Rotenone-treated neurons exhibited increased cytosolic calcium concentration, revealed by increased F340/F380 fluorescent ratio in cultures loaded with the calcium dye Fura-2 F I G U R E  Actin cytoskeleton dynamics drive extracellular motion of large EVs. (a) The scheme shows possible coupling of EV to the moving cytoskeleton of neuron via binding to a neuronal receptor. It also shows that transport of large intracellular vesicles along the cytoskeleton may alter extracellular EV motion by dragging the cytosolic portion of EV receptor or adaptor protein. (b) Immunoblot of astrocytes (at decreasing loading amounts), large EVs (lEVs) and small EVs (sEVs) pelleted at 10,000 x g and 100,000 × g respectively from 20 × 10 6 cells for PrP, the EV markers alix, flotillin 1 and annexin A2 (Ann-A2) and the Golgi and mitochondrial markers GS28 and TOM20.  (Figure 4f), indicating an impairment of plasma membrane ATPase activity. In addition, Rotenone-treated neurons displayed reduced energy-dependent trafficking of LysoTracker labelled lysosomes inside the cells (Figure 4g), confirming depletion of energy reserves. Importantly, only 33% of EVs were in motion on intoxicated neurites, where lysosome transport was blocked, while 75% of EVs drifted on the surface of control neurons (Figure 4h). In line with this evidence, acute application of the mitochondria inhibitor Sodium Azide (20 mM) (Svitkina et al., 1986) blocked the movement of a few EVs (n = 3) along neurites (Supplementary Figure 5, Movie S10). To corroborate these findings, we mildly fixed neurons with cold methanol (at 4 • C) and placed single EVs on the surface of dead neurons. Only 8% of EVs exhibited a net displacement from the contact point at the surface of dead cells (Figure 4i). Altogether these findings showed that EV motion largely depends on neuron energy metabolism.
To assess the involvement of the cytoskeleton in extracellular EV motion we next blocked actin filaments, a component of the cytoskeleton involved in various types of cellular motility (Rodriguez et al., 2003), by exposing neurons to 1 μM cytochalasin D (CytoD) for 60 min, a drug that binds to actin fast growing end, or to 100 μM Blebbistatin for 30 min, an inhibitor of motor dependent restructuring of actin network (Straight et al., 2003). CytoD-dependent depolymerisation of actin filament was indicated by increased ratio of globular (G-) versus filamentous (F-) actin, as revealed by Western blot analysis (Figure 4j,  N = 3). Importantly, by positioning EVs on CytoD-treated neurites (CytoD) we found no changes in EV adhesion (48% Ctrl n = 57; 47% CytoD n = 79) but a decrease of EV extracellular motion (Figure 4k). Inhibition of actin filaments rearrangements by Blebbistatin caused a significant reduction of EV motion too (Figure 4l), confirming the involvement of the actin cytoskeleton (Supplementary Figure 6). By contrast, depolymerisation of the microtubule cytoskeleton by Nocodazole (10-30 μM for 60 min), did not alter EV motility (Figure 4m), despite a trend to decrease of EV motion.

. Involvement of PrP in EV motion
Drift of astrocytic EVs on neuronal surface strikingly resembles movement of adenoviral particles on target cells, which enhances infection by transporting viruses to sites competent for endocytosis (Ewers et al., 2005;Lehmann et al., 2005;Schelhaas et al., 2008). Indeed, a parallel between virus and EV extracellular motion has been recently established by a study showing that EV surfing along filopodia facilitates cell entry similarly to virus infection (Heusermann et al., 2016). Binding of viral particles to plasma membrane coxsackievirus adenovirus receptor (CAR), a GPI anchor-protein, was shown to give rise to extracellular virus motion, by coupling the viruses with the actin cytoskeleton of target cells (Burckhardt et al., 2011). Among surface neuronal receptors which may elicit motion of large glial EVs, our experiments with PrP-coated synthetic particles (Figure 4c-d) pointed to the GPI anchor-protein PrP, as putative linker between EV and the moving actin cytoskeleton. The protein is expressed in astrocytes, in addition to neurons, and highly enriched in EVs thereof (Guitart et al., 2016), especially in large EVs, as indicated by Western blot analysis (Figure 4b).
To explore the involvement of PrP in astrocytic EV trafficking along the neuron surface, we first silenced PrP in both neurons and donor astrocytes using specific smart pool siRNAs. Western blot analysis showed that PrP was downregulated by 60% and 40% in neurons and astrocytes respectively, compared to vehicle-treated cultures (Figure 5a). By positioning single large EVs produced by PrP-silenced astrocytes onto the processes of PrP-silenced neurons, we found an increase in the percentage of static EVs (Figure 5b, D-siRNA), suggesting the involvement of vesicular and/or neuronal PrP in the extracellular motion. Similar results were obtained with control EVs delivered to PrP-silenced neurons (Figure 5b, siRNA). To corroborate these findings, we isolated large EVs from PrP knock out (Leshchyns'ka & Sytnyk, 2016) astrocytes (PrP -/--EVs) (Figure 5c). TRPS analysis revealed no significant changes in the average diameter of PrP -/--EVs in the 10,000 × g pellet compared to wild-type EVs (PrP +/+ -EVs) nor in the mode of the diameter (Figure 5d; N = 3). By placing single PrP -/--EVs on neurites of PrP knock out neurons (PrP -/neurons) (Figure 5e-f) we didn't observe significant changes in EV-neuron adhesion (67% of adhesion of PrP -/--EVs on PrP -/neurons compared to 76% in wild type conditions), excluding that homophilic PrP interaction is necessary for EV-neuron contact. However, this caused a significant decrease in the percentage of moving EVs (Figure 5f), confirming the involvement of neuronal and/or vesicular PrP in extracellular EV motion. To specifically address the involvement of homophilic PrP interaction in extracellular EV motion we next delivered large PrP +/+ -EVs on PrP -/neurons and large PrP -/--EVs on WT neurons. Lack of PrP in either EVs or neurons decreased the percentage of EVs in motion, albeit the difference was significant only upon removal . (e) Representative confocal images of WT and PrP -/neurons stained for PrP (green), N-CAM (red) and the nuclear marker DAPI (blue) in non-permeabilizing conditions. Note that PrP shows a punctate pattern and appears equally distributed in the plasma membrane of neurites and the cell body. Scale bar = 10 μm. (f) Percentage of moving EVs from WT astrocytes on WT neurons (PrP +/+ EVs/WT, n = 107, N = 17), from PrP -/astrocytes on PrP -/neurons (PrP -/-EVs/PrP -/-, n = 117, N = 14), from WT astrocytes on PrP -/neurons (PrP +/+ EVs/PrP -/-, n = 33, N = 10) and from PrP -/astrocytes on WT neurons (PrP -/-EVs/WT, n = 33, N = 8); One-way ANOVA P = 0.0028, Holm-Sidak's multiple comparisons test: PrP +/+ EVs/WT vs. PrP -/-EVs/PrP -/-P = 0.0039, PrP +/+ EVs/WT vs. PrP -/-EVs/WT P = 0.0290. Data have been normalized to PrP +/+ EVs/WT condition of vesicular PrP (Figure 5f, Supplementary Figure 7). These results suggest that vesicular PrP controls EV transport at the neuron surface by interacting with either PrP, in agreement with PrP-coated beads data, and/or other PrP binding proteins on neurons.

. A fraction of EVs actively moves on neuronal surface
Being equipped with actin filaments, as indicated by cryo-EM (this study, Cvjetkovic et al., 2017;Zabeo et al., 2017) as well as actin binding proteins (Drago et al., 2017), a fraction of astrocyte-derived large EVs may possess independent, active motile capability, similar to that produced by actin in cells . Support to this hypothesis (Figure 6a) was provided by time lapse video recordings of single EV dynamics at the neuron surface displaying large EVs jumping from one neurite to another, as described above (Movie S4 and S5). To the best of our knowledge, there is no direct communication between the cytoplasm of adjacent neurites in differentiated neurons. Therefore, the saltatory motion could only rely on the capability of large EVs to change their shape and to interact with surface EV receptors on adjacent neurites. Following this hypothesis, active EV movement should be sensitive to the block of actin dynamics inside EVs. Western blot analysis for F-and G-actin confirmed the presence of actin filaments in large astrocyte-derived EVs (Figure 6b). Importantly, F I G U R E  Active motion of astrocytes-derived EVs. (a) Schematic representation of active motion of EVs, through intermittent contact with neuronal surface receptors. (b) Representative Western blot (left) of F-actin and G-actin in control and CytoD-treated medium/large EVs. F-and G-actin fractions are positive for the EV markers Alix, annexin A2 (Ann-A2) and negative for the Golgi and mitochondrial markers GS28 and TOM20. On the right, the histogram shows the ratio between G-actin and F-actin in control and CytoD-treated large EVs (Mann Whitney test P = 0.05, N = 3). (c) Percentage of moving EVs from control and CytoD-treated EVs on the neuron surface (Ctrl n = 18; CytoD n = 32; N = 10; unpaired t test, one-tailed P = 0.0461). (d) Temporal analysis of both F-and G-actin fractions isolated from large EVs were negative for markers of intracellular organelles, thus excluding contamination with cytosolic actin. EV treatment with CytoD for 60 min caused an increase in the ratio between G-and F-actin in large EVs (Figure 6b, N = 3), indicating F-actin depolymerization, and lowered the percentage of EVs in motion at the neuron surface (Figure 6c). This suggests that actin rearrangements inside large EVs may produce motion in at least a fraction of EVs.

. EVs contain ATP as energy source
To modify their shape and to actively move along the neuron surface, large EVs should possess an energy source that supports actin cytoskeleton rearrangements.
To assess whether large astrocyte-derived EVs contain ATP, the main cellular energy source, we used two different bioassays previously established to measure ATP (Coco et al., 2003;Pellegatti et al., 2005). In the first assay, oligodendroglial cells, which respond to low ATP concentration (in the nM-μM range) with rapid intracellular Ca2+ elevations (Fumagalli et al., 2011), were used as ATP sensor cells. Addition of large EVs to Fura-2-loaded oligodendrocytes evoked calcium transients in ∼37% of the cells (n = 84) (Figure 6d). Calcium responses were completely abrogated by treatment of large EVs with the ATP degrading enzyme apyrase (3 units/100 μl) (Figure 6d), indicating that ATP was the compound evoking Ca 2+ transients. Constitutive large EVs, released by astrocytes in the absence of ATP stimulation, also evoked apyrase-sensitive Ca 2+ transients (Figure 6e, Supplementary  Figure 8), excluding that externally added ATP, used to evoke EV shedding, may mediate the Ca 2+ response in sensor cells. Given that ATP packaged inside EVs could not activate ATP receptor in sensor cells nor being degraded by apyrase, these data suggest that large EVs release or leak ATP into the extracellular fluid. To probe this hypothesis constitutive EVs were maintained in control condition or subjected to freeze and thaw, to facilitate ATP leakage from the EVs, before addition to sensor cells. Disruption of EV membrane integrity produced clear Ca 2+ transients in sensor cells, even when responses to control EVs were barely detectable ( Figure 6f). Consistently, much lower Ca 2+ responses were evoked by large EVs, depleted of their cytosolic cargo by hypo-osmotic shock (Gabrielli et al., 2015) (Figure 6g), revealing that ATP comes from the EV lumen. The second ATP bioassay exploited B16-F10 cells-pmeLUC (De Marchi et al., 2019, which stably express a chimeric firefly luciferase (pmeLUC). The chimeric protein localizes to the outside of the plasma membrane and is therefore able to sample extracellular ATP (Pellegatti et al., 2005). By this assay we could measure the ATP content of large EVs released from astrocytes stimulated with BzATP, an agonist of the P2X7 ATP receptor, which is not detected by firefly luciferase. The pmeLUC cells-based assay confirmed that large EVs released from stimulated astrocytes contain and release ATP in the close proximity of the plasma membrane, and revealed that EVs can release ATP at concentrations in the high nM range (Figure 6h).
Finally, the presence of ATP inside astrocyte-derived EVs was confirmed by EV labelling with 1 μM quinacrine, a fluorescence dye which stains intracellular ATP stores (Coco et al., 2003;Falchi et al., 2013) (Figure 6i). Most EVs were quinacrine positive, confirming that they store ATP (Figure 6i).

 DISCUSSION
Astrocyte-derived EVs signal to neurons via multiple mechanisms, including transfer of RNA and protein cargoes and activation of signalling events at the neuronal surface, influencing synaptic activity at both pre-and post-synaptic sites (Antonucci et al., 2012;Prada et al., 2018). In addition, a role for EVs released by glial cells, especially microglia, upon ATP activation is emerging in brain diseases (Lombardi et al., 2021), such as traumatic brain injury (Liu et al., 2017) and tauopathy, where ATP-induced EVs have been implicated in the spreading of tau protein (Ruan et al., 2020). However, how large glial EVs released upon ATP stimulation move in the extracellular space to reach target neurons and whether large EVs interact with neurons at preferential sites to influence synaptic transmission is unknown. By using optical tweezers (OT) combined to time-lapse imaging we hereby show that large glial EVs adhere to any neuronal compartment (cell body, dendrites, axon) and that, after adhesion, most large EVs move from the contact site scanning the calcium changes in fura-2 loaded oligodendrocytes induced by EVs released upon ATP stimulation in the presence and in the absence of apyrase. Each trace is from a distinct oligodendroglial cell. (e) Representative calcium responses evoked by constitutively released EVs (Constitutive EVs). (f) Calcium responses evoked by the same sample of constitutive EVs kept in control condition or exposed to freeze and thaw to disrupt the EV membrane (Constitutive broken EVs) before being probed on ATP-sensor cells. (g) Representative calcium responses evoked in oligodendrocytes by intact EVs and EVs depleted of luminal cargo. (h) Luminescence of B16-F10 cells-pmeLUC after 10 min exposure to vehicle (Ctrl), large EVs from 5 × 10 6 astrocytes or 500 μM ATP. On the Right corresponding quantification of the relative pmeLUC activity expressed as photons/second (One-way ANOVA P < 0.0001, Dunn's multiple comparisons test: Ctrl vs. large EVs P = 0.0311; Ctrl vs. ATP P = 0.0025; large EVs vs. ATP P = 0.3871; Ctrl n = 5, EVs n = 9, ATP n = 3). (i) Representative fluorescent images of quinacrine (green) positive astrocytes (top) and EVs (bottom) merged with bright field images (Scale bar = 10 μm) neuron surface to reach preferential interaction sites in the somato-dendritic compartment, where EVs stop moving. On the cell body, extracellular EV motion is confined to the somatic region while along dendrites and axons of developing neurons large EV motion occurs in both retrograde and anterograde direction, towards the cell soma or the periphery to reach connected neurons. EVs stop along their path especially when exploring dendritic spine protrusions on mature dendrites, resulting in lower EV speed on dendrites compared to axons in fully differentiated cultures. At preferential interaction sites EVs can elicit the formation of new filopodia. Previous studies pointed to a role for microglia in the formation of filopodia or spine head filopodia at sites of microglia-synapse contacts (Miyamoto et al., 2016;Weinhard et al., 2018). This suggests that large EVs shed from the glial cell surface retain the capacity of glial processes to induce filopodia formation at sites of persistent contact with neurons. Further work would be necessary to define whether preferential interaction points of EVs have a synaptic localization. Importantly, along developing axons, EVs move from neuron-to-neuron. In addition, in differentiated cultures EV speed remains elevated on axons and 50% of large EVs move in anterograde direction. This suggests that large astrocytic EVs may use axons as routes to spread their cargoes between synaptically-coupled neurons. As the cargoes of astrocyte-derived EVs include misfolded proteins, such as tau and amyloid-β (Aβ) (Chiarini et al., 2017;Eitan et al., 2016), extracellular motion of large EVs along projecting axons may underlie the propagation of Aβ/tau aggregates, which occurs through an anatomically defined pattern of connections, from the entorhinal cortex up to the hippocampal region and the neocortex in the brain of Alzheimer's disease (AD) patients (Braak & Braak, 1991). Previous evidence in AD cellular models shows that small EVs storing misfolded proteins can be internalized and transferred between neurons through axonal projections (Sardar Sinha et al., 2018;Wang et al., 2017). However, intra-and extracellular EV trafficking may be complementary, not mutually exclusive. Indeed, extracellular motion may represent a specific route to spread misfolded proteins for large EVs which exceed the axon diameter and cannot be taken up and transported intracellularly without impairing vesicle trafficking. Consistent with this hypothesis, in this study we show that the vast majority of mCLINGlabelled large EVs do not undergo internalization inside thin neurites. This may explain why large EVs move so much at the surface of thin neuronal processes. Conversely, a small fraction of large EVs contacting the cell body remained outside neurons, as proved by recapturing by optical tweezers and confocal analysis of mCLING-labelled EVs, suggesting that transport toward the soma may be followed by endocytosis.
EVs share many features with viruses, including physical and chemical characteristics, mechanisms of biogenesis and cellular uptake (Raab-Traub & Dittmer, 2017). Here we show that EV extracellular motion is an additional characteristic which EVs and viruses have in common. Indeed, both EVs (this study) and viruses (Burckhardt et al., 2011) are passively transported at the target cell surface by rearrangements of actin cytoskeleton. Furthermore, both virus and EV motion involves a GPI anchor protein: CAR receptor on the surface of infected cells (Burckhardt et al., 2011) and PrP on EV and, possibly, on the neuron surface (this study). In neurons, a diverse collection of receptors and ions channels are physically coupled to the actin cytoskeleton, which critically controls their localization and trafficking at the synapses (Shaw & Koleske, 2021). Like CAR receptor, PrP does not interact directly with the actin cytoskeleton. However, PrP interacts with a number of actin-binding intracellular partners, including the synaptic vesicle protein synapsin, and the adaptor protein critical for actin-based cell motility Growth factor receptor-bound protein 2 (Grb2) (Nieznanski, 2010) which may couple EVs to the neuron cytoskeleton ( Figure 4a). Alternatively, neuronal PrP may indirectly interact with actin via NCAM, an adhesion protein which binds to both PrP Slapsak et al., 2016) (in cis) and actin filament interactors, such as Ankyrin B and ezrin (Dickson et al., 2002;Nishimura et al., 2003). Importantly NCAM, by interacting in trans with vesicular PrP (Santuccione et al., 2005), may also act as EV receptor, contributing to EV transport at the neuron surface. Further experiments are required to assess this hypothesis.
A recent work showed that PrP at the surface of brain EVs inhibits EV uptake (De Marchi et al., 2019). This is in full agreement with our study, which unveils a role for vesicular PrP in EV retention at the neuron surface and promotion of extracellular EV transport. By providing first evidence for the involvement of vesicular PrP in extracellular EV motion, our study points at PrP and its neuronal interactor(s) as potential targets to limit diffusion of EV-associated toxic amyloid proteins in patients affected by neurodegenerative diseases.
The main accomplishment of our study is the demonstration that a fraction of EVs contains actin and has an independent capacity to move at the neuron surface, highlighting the active role of EVs in intercellular communication. EVs move on energy depleted or fixed neurons, and their motion is sensitive to block of actin dynamics by Cytochalasin D. Polymerization of G-actin monomers at the plus (or 'barbed') end of an actin filament, and depolymerisation of monomers from the minus (or 'pointed') end can generate forces to cause movement in the EVs similar to that produced by actin in cells . In support to our findings, a previous study showed that a part (albeit small) of EVs isolated from cell cultures, biological fluids or tissue displays morphological changes detectable by time-lapse fluorescence imaging . EVs could round up starting from an elongated structure, glide one along the other and stretch out flexible protrusions, morphological changes which may be induced by actin rearrangement. Active motion may be important for EVs to make their way through the limited interstitial space, which may affect extracellular EV motion in vivo. Importantly, tubular EVs which contain actin filaments strikingly resemble actin-rich membrane protrusions such as filopodia and isolated tunnelling nanotubes (TNTs), recently characterized plasma membrane structures which assemble together to form connections between cells (Sartori-Rupp et al., 2019). Similar to filopodia and isolated TNT, tubular EVs carry actin filaments oriented in parallel to one another, as evidenced by cryo-EM, suggesting that EVs could derived from either budding or fission from isolated TNT/filopodia (Sartori-Rupp et al., 2019). Future characterization of the membrane and cytoskeleton-associated protein(s) involved in EV budding at the plasma membrane and filopodia/TNT formation may help to clarify this intriguing hypothesis.
Overall, this study provides first evidence for the existence of a regulated trafficking of EVs at the cell surface which may help EVs to reach their destination sites in vivo. Extracellular motion may be a common feature for EVs of distinct cell origin, including cancer and immune cells, and may therefore have a general role in the propagation of pathophysiological signals.

A C K N O W L E D G E M E N T S
Authors are grateful to Dr Ottavio Arancio (Columbia University, New York, NY, USA) and Dr Francesca Navone (Institute of Neuroscience, Milan, Italy) for useful discussion, Dr Lothar Stitz (Institute of Immunology, Tuebingen, Germany) for providing W226 antibody and Dr Gabriele Leoni for technical support on R code. This work was supported by FISM under grant n