Extracellular vesicles from neuronal cells promote neural induction of mESCs through cyclinD1

Extracellular vesicles (EVs) that are thought to mediate the transport of proteins and RNAs involved in intercellular communication. Here, we show dynamic changes in the buoyant density and abundance of extracellular vesicles that are secreted by PC12 cells stimulated with nerve growth factor (NGF), N2A cells treated with retinoic acid to induce neural differentiation and mESCs differentiated into neuronal cells. EVs secreted from in vitro differentiated cells promote neural induction of mouse embryonic stem cells (mESCs). Cyclin D1 enriched within the EVs derived from differentiated neuronal cells contributes to this induction. EVs purified from cells overexpressing cyclin D1 are more potent in neural induction of mESC cells. Depletion of cyclin D1 from the EVs reduced the neural induction effect. Our results suggest that extracellular vesicles regulate neural development through sorting of cyclin D1. Summary Extracellular vesicles (EVs) may mediate intercellular communication through the transfer of cargo molecules. Here, we report that cyclinD1 is sorted into EVs during neuronal differentiation and that these EVs help to promote the neural induction of mouse embryonic stem cells.


Introduction
Intercellular communication involves either direct contact between neighboring cells or indirect interaction via secreted factors including extracellular vesicles (EVs) (Hessvik and Llorente, 2018). Cells release two subtypes of EVs according to their cellular membrane origin, plasmamembrane-derived extracellular vesicles and endosome-derived exosomes (French et al., 2017). Plasma membrane-budded vesicles range from 30 to 1,000 nm whereas exosomes range from 30-150 nm in diameter. Intralumenal vesicles that accumulate within multivesicular bodies (MVBs) are secreted as exosomal EVs to the extracellular environment upon fusion of MVBs with the plasma membrane. EVs are enclosed by a lipid bilayer containing transmembrane proteins and luminal cytosolic proteins and nucleic acids (Raposo and Stoorvogel, 2013). Much interest has focused on the wide distribution of EVs in all biological fluids and their potential to trigger intercellular exchange of effector molecules, which may allow secreting cells to modulate gene expression in target cells and tissues (Mulcahy et al., 2014).
The lipid membrane of EVs ensures the stability of lumenal cargo as vesicles circulate in the extracellular space, potentially over long distances. The small size of EVs helps overcome various biological barriers, including the blood brain barrier (Van Niel et al., 2018). In the brain, EVs represent an ideal vehicle for intercellular transfer of information from neurons and glia to both neighboring and distal cells (Budnik et al., 2016). Multiple cell types in the mature central nervous system (CNS) release EVs, including neurons, astrocytes, and oligodendrocytes (Pascual et al., 2020). Neuronally-secreted EVs could control synaptic plasticity, and enhance the removal of degenerative neurites after internalization by microglial cells (Bahrini et al., 2015). EVs secreted by astrocytes display neuroprotection activity that is critical for neuronal cell survival (Verkhratsky et al., 2016). Oligodendrocytes secrete EVs that are endocytosed by neurons and increase their viability (Krämer-Albers et al., 2007). In addition to studies on EVs in the mature CNS, recent studies provide evidence that EVs from newly-differentiated neural cells promote neurogenesis (Sharma et al., 2019) .
In neural development, cell fate determination is tightly controlled by stepwise commitment, including neural induction and neurogenesis (Grow, 2018). Pluripotent stem cells convert into neural ectoderm progenitors, after which neural precursors further differentiate into nerve cells of defined function (Muñoz-Sanjuán and Brivanlou, 2002). Although EVs have been suggested to facilitate the later neurogenesis events, little is known about the role of EVs during the early stage of neural fate conversion. Furthermore, clear and direct evidence of such roles is lacking as the studies thus far have relied on highly impure, crude preparations of sedimented particles (Sharma et al., 2019).
Much of the literature on the proposed function of EVs has relied on differential sedimentation of slowly-sedimenting or crudely precipitated particles obtained from culture medium or other fluids. This method does not separate large shedding microvesicles from small exosomal like EVs, nor does it remove protein aggregates or ribonucleoprotein particles (Konoshenko et al., 2018). Further resolution can be achieved by a series of ultracentrifugation and density gradient centrifugation steps that removes cells, cell debris, microvesicles, and protein aggregates (Shurtleff et al., 2016(Shurtleff et al., , 2018Temoche-Diaz et al., 2019).
To investigate the role of purified EVs during neural development, we used buoyant density flotation to isolate EVs from the nerve growth factor (NGF)-induced PC12 cells and retinoic acid (RA)-induced neuro 2A cells (N2A). We then examined the effect of purified vesicles on mouse embryonic stem cells (mESCs) and found that the neuronal EVs accelerate aspects of mESC neural induction. We further demonstrated that a cell cycle-related factor, cyclin D1, was enriched within EVs derived from differentiated cells. Compared with the EVs from untreated cells, those purified from cells overexpressing cyclin D1 enhanced neural lineage gene expression in mESCs. Conversely, EVs from cyclin D1 knockout cells did not stimulate neural induction of mESCs. The chaperone protein Hsc70 facilitated packaging of cyclin D1 into EVs. Our results suggest that EVs contribute to neural fate determination through sorting of cyclin D1.

Dynamic changes of extracellular vesicles secreted by neuronal differentiated PC12 cells
We compared EVs secreted from undifferentiated neuronal progenitor-like cells with those secreted by differentiated neuronal cells. In an initial approach, we used PC12 cells, which differentiate after treatment with NGF. After 9 d of NGF treatment, long neurite extensions consistent with a neuronal fate were observed ( Fig. 1 A). Consistently, qPCR results showed that the neuronal marker genes, Tuj1 and Tau, were up-regulated concomitant with neurite extension. We tested two NGF concentrations and found that both elicited similar levels of expression of neuronal marker genes ( Fig. 1 B). As a result, the lower dosage of NGF, 50 ng/ml, was used in all subsequent experiments.
Next, to investigate if the protein composition of EVs changes during PC12 cell neuronal differentiation, we collected samples at different time points after adding NGF. Extracellular vesicles were obtained from the culture medium by differential centrifugation to concentrate small particles, followed by equilibrium sedimentation in an iodixanol linear density gradient. Each fraction was collected from top to bottom for further analysis (Fig. 1 C). Flot2, a lipid scaffolding protein, showed no dramatic change in the expression of EVs (in buoyant densities density ranging from 1.06 to 1.23 g/ml) from cells treated with NGF for 0, 3, 6 and 9 d. Thus, vesicles marked by this reporter protein appeared not to vary during the differentiation process (Fig. 1,D and E). However, the EV marker CD9, a tetraspanin enriched in MVBs, showed an obvious shift in the buoyant density peak in EVs from cells treated for 3 d. The prominent peaks changed from fractions corresponding to densities of~1.08 g/ml to~1.13 g/ml at the onset of differentiation. During the following 6 d, the CD9-containing EVs showed a broader density distribution (Fig. 1, D and E). Consistently, heat shock chaperone Hsc70, which is also a wellknown EV marker, displayed a similar expression pattern during differentiation (Fig. 1, D and E). Furthermore, the size of vesicles (in even-numbered fractions) analyzed with a NanoSight particle tracking device showed no dramatic difference between the sub-fractions. In contrast, the protein concentration appeared to increase gradually in fractions containing vesicles from low to high buoyant density (Fig. S1, A and B). These results indicated the physical properties of CD9containing EVs changed over the course of differentiation.

EV production increased during neuronal differentiation
EVs secreted from NGF-induced PC12 cells were further characterized by an independent isolation using differential centrifugation and sucrose gradient buoyant density flotation to obtain membranes at the 20%/40% interface, with a density corresponding to~1.12g/ml (Fig. 2 A). Membranes were sedimented, washed and resuspended to assess vesicle morphology. Negative stain electron microscope images displayed a characteristic cup-shaped, collapsed appearance in all four groups of EVs preparations (Fig. 2 B).
To determine if higher vesicle production was a common feature of differentiated neuronal cells, we employed another well-established neuronal differentiation system, the neuroblastoma Neuro2A cell line (N2A). Retinoic acid (RA,10 µM) induced the differentiation of N2A ceclls as confirmed by cell morphology and expression of differentiation-specific marker, Neurogenin 2 (Ngn2) (Fig. S2, A and B). Consistent with the observation of EVs from differentiated PC12 cells, after 6 d in RA-containing medium,~7000 EVs (RA6-EV) were produced/cell, whereas~500 EVs/cell (N2A-EV) were secreted from non-differentiated N2A cultures (Fig. 2 E).
In order to extend our observations to a more physiologic source of neuronal cells, EVs were collected during mESC neural differentiation. Pluripotent stem cells (ES D0) were cultured in suspension conditions for 8 days for conversion to embryonic bodies (EBs) which express neural progenitor markers (ES D8). EBs were trypsinized in N2 medium for another 4 days during which time they differentiated into Tuj1 + neurons (ES D12) (Fig. S2, C and D). During 12 days of differentiation,~400 EVs/cell (ES D12-EV) were released from neurons, whereas during 8 days of EB formation,~200 EVs (ES D8-EV) were produced from each neural progenitor cell. As a control, we cultured mESCs in pluripotency maintaining medium (N2B27 +2i +Lif) for 2 days and found~80 EVs (ES D0-EV) were secreted/cell (Fig. 2 F).
We examined the expression of multiple EV marker proteins in sucrose gradient-purified vesicles released from 2×10 7 cells incubated in control and neuronal differentiation conditions. CD9, Hsc70, and Flot2, which were detected in the whole EV density profiles shown in Fig.1, showed ã 3-fold increase in expression in the N6-EV compared with EVs from the untreated PC12 cells (PC12-EV). CD63, a tetraspanin important for EV biogenesis and Alix and Tsg101, essential components of the endosomal sorting complexes required for transport (ESCRT), were also up-regulated~3 fold in the N6-EV versus PC12-EV (Fig. 2, G and H). EV production increased more dramatically in RA6-EV compared to undifferentiated N2A cells, with a~14-fold increase in the same marker proteins (Fig. 2, I and J). These results suggested a possible role for the EVs produced during neuronal differentiation, at least as reflected in these cell lines.

Purified EVs are taken up by mESCs
Pluripotent mESCs may differentiate into various cell types, having the ability to commit to a specific lineage in response to external stimulation (Yu and Thomson, 2008). We used mESCs as recipient cells to evaluate the influence of buoyant density gradient-purified EVs. mESCs were trypsinized to disperse single cells and cultured in serum-free N2B27 medium with EVs collected from differentiating PC12 cells. EVs were fluorescently tagged with the lipophilic membrane dye, PKH67 washed in PBS, and incubated for 24h with mESCs. Labeled EVs from day 6 or 9 of NGF treatment appeared to be internalized in the recipient cells (Fig. 3, A and B).
We next tested if the uptake of a soluble, lumenal marker protein packaged into EVs could be internalized into recipient cells. For this purpose, we used lentivirus transfection to establish a GFP-overexpressing N2A cell line and then isolated EVs on a buoyant density gradient. The level of the EV marker CD9 and coincident GFP were detected by immunoblot in a linear range proportional to the number of EVs purified from RA-induced GFP-overexpressing N2A cells (Fig. 3 C). mESCs (2×10 5 ) were then incubated in 2 ml N2B27 medium for 24h with EVs purified from control or GFP-expressing cells. mESCs were then harvested by centrifugation and washed with PBS twice and the GFP signal was detected in the lysate of the receiving ES cells (Fig. 3 D). In incubations of mESCs in a wide range of EV concentrations, GFP was detected associated with cells at GFP-EVs levels of 2×10 9 or higher (Fig. 3 E). In a time course, we observed GFP-EVs uptake within 3 h and progressing with continued incubation (Fig. 3 F). We next compared the uptake of free GFP (10 ng) to GFP-EVs (~8×10 9 EVs contain~10 ng GFP) added in equivalent amounts, as detected by quantitative immunoblot, to mESCs incubated in 2 ml N2B27 medium. GFP-EVs was more rapidly and efficiently internalized (Fig. 3 G). These results indicate mESCs internalize lumenal soluble as well as membrane constituents of EVs in a time and a dosage dependent manner.

Buoyant density-purified EVs from differentiated neuronal cells promote mESC neural induction
mESCs default to a neural progenitor fate in serum-free growth medium (Ying et al., 2003). To explore the effects of differentiated neuronal EVs during this process, mESCs were treated with purified EVs in serum-free medium for 6 d. The medium containing EVs was changed every day with fresh EVs added each day. Cells were harvested for gene expression analysis by qPCR. We found that the neural stem cell marker nestin, and a neuronal marker gene Six3, were up-regulated 2-fold in comparison to an EV-free control by EVs from NGF-induced but not by EVs from uninduced PC12 cells (Fig. 4 A). EVs derived from PC12 cells that had been induced by NGF for 6 d (N6-EVs) caused the most robust upregulation of neural markers at mESC differentiation day 4 (Fig. 4, A and C). This time point was used in subsequent experiments. Neural marker upregulation by N6-EVs was dose-dependent ( Fig. 4 B). Furthermore, two other neural progenitor genes, Pax6 and Sox1, and neuronal marker Tuj1, were all up-regulated by N6-EVs ( Fig. 4 C). Although NGF is known to stimulate neural gene expression in mESCs, we found that addition of a neutralizing NGF antibody to an incubation of N6-EVs with mESCs did not affect the level of expression of these neural markers suggesting that possible residual NGF in the EV preparation did not account for this effect (Fig. 4C). The neutralizing activity of NGF antibody to NGF was confirmed by cell morphology and marker gene expression analysis (Fig. S4, A  Immunostaining showed an increase of Nestin and Pax6 expression in neural progenitor cells after treatment with RA-induced EVs (Fig. S4, E and F). These results suggest that EVs from neuronal cells promote a neural fate commitment of mESCs.
During the mESC serum-free neural differentiation process, we found that EVs from NGFinduced PC12 cells appeared to increase the size and density of cell clusters (Fig. 4 G). Likewise, an increase in mESC cell number from~5.2×10 4 /cm 2 to~7.5×10 4 /cm 2 was observed after 4 d of treatment with EVs from NGF-induced PC12 cells (Fig. 4 H). This result was affirmed using a BrdU cell proliferation assay (Fig. 4 I).

Cyclin D1 is sorted into EVs
We considered the possibility that EVs from differentiated PC12 and N2A cells may transfer proteins that could influence the fate and proliferation of stem cells. One candidate, cyclin D1, a cell cycle regulator, was reported to promote a neural fate conversion in human embryonic stem cells (Pauklin and Vallier, 2013;Pauklin et al., 2016). We examined the expression and sorting of the three paralogs of cyclin D (D1, 2 and 3) in differentiated PC12 cells and EVs. Cyclins D1-D3 were gradually up-regulated from day 0 to day 9 during NGF treatment (Fig. 5 A). Only cyclins D1and D2, but not D3, were detected in the EVs from differentiating cells (Fig. 5 B). Compared with the moderate increase of cyclin D proteins detected in cell lysates, a greater enrichment of cyclin D1, and to a lesser extent D2, in EVs was detected at day 6 ( Fig. 5, B and C). Specific cyclin D1 sorting was also enriched in EVs from RA-induced N2A cells, especially at day 6 ( Fig.  S5 A). Compared with EVs from non-differentiated mESC and EVs from 8 d differentiated EBs, significant cyclin D1 was enriched in EVs from 12 d differentiated neurons (Fig. 5 D). The cyclin D-dependent kinase family member CDK4 was observed in EVs (Fig. 5 E and Fig. S5 A). Flot2, which was used as a positive EV membrane protein control, was present in both EVs and cells, whereas the cis-Golgi matrix protein, GM130, was only detected within the cells (Fig. 5 E). Given that cyclin Ds drive the G1/S phase transition, we tested other cyclins which contribute to cell-cycle progression at S phase and M phase and found that, cyclin B1, cyclin E1 and cyclin A2, were expressed in cells but not in EVs (Fig. 5 E). Likewise, other cell-cycle related factors, including pRB, p57, p27, p21 and pErk, were detected in cells but not in EVs (Fig. S5 B). These results suggest that at least some factors specific to the G1/S phase transition were sorted into EVs, notably cyclin D1.
We used a standard protease protection assay to probe the localization of cyclin D1 with respect to the EV membrane. The EV marker proteins, Hsc70, Flot2, Tsg101 and Alix, were dramatically sensitive to proteolytic degradation by 10 μg/ml proteinase K in the presence of Triton X-100, but not in the absence of detergent (Fig. 5 F). We conclude that these proteins are enclosed within the interior of the EV. Similarly, cyclin D1, cyclin D2 and CDK4, were all sensitive to proteinase K but protected in the absence of Triton X-100 ( Fig. 5F, Fig. S5 C). Tetraspanin membrane protein, CD81, was proteolytic degradation by 5 μg/ml proteinase K (Fig. 5F). We then used an antibody directed to membrane exposed epitopes of the tetraspanin membrane protein, CD9, to immunoisolate intact vesicles and probe the coincident localization of cyclin D1 in EVs. An immobilized form of CD9 antibody co-immunoprecipitated cyclin D1 along with EV markers Hsc70, Tsg101, and Alix (Fig. 5 G). These results support the conclusion that cyclin D, especially cyclin D1, is sorted into the luminal interior of EVs produced by differentiating PC12 and N2A cells.

Hsc70 facilitates cyclin D1 package into EVs
We used ascorbic acid peroxidase (APEX) proximity labeling to detect proteins in contact with cyclin D1 during RA-induced differentiation of N2A cells (Hung et al., 2016). Cyclin D1-APEX stably expressing N2A cells were obtained by lentivirus-mediated gene delivery. Biotin-phenol (B) and H2O2 (H) were added to cells separately or in combination (B+H). Equal amounts of protein, as estimated by ponceau S staining of gels, were evaluated in samples from three incubations (Fig. 6 A). Streptavidin-HRP was used to label biotinylated proteins which were detected primarily in the sample incubated with biotin-phenol (B) and H2O2 (H) (Fig. 6 A). The biotinylated proteins were precipitated with streptavidin beads and bound proteins then analyzed by Mass Spectrometry (MS). We identified proteins that were enriched in the B+H group compared to the B-group, with an increase in at least 280 unique proteins ( Fig. 6 B, supplementary file 1). Unique peptides (312) of the heat shock protein, Hsp90b1, were identified in the B+H group with only 9 peptides in the B group. Other unique differences were seen among several Hsps (Fig. 6 B). We thus chose Hsp90 and Hsc70 to examine direct interactions with cyclin D1 and CDK4, the two proteins that we found to be selectively sorted into EVs during RAinduced differentiation of N2A cells.
We next tested the possibility that Hsc70 function may be important in the sorting of cyclin D1 into EVs. VER-155008 (VER) is a potent and selective inhibitor of Hsc70 ATPase activity that has been used to assess Hsc70 function in cellular processes (Massey et al., 2010). We purified EVs from 4 d differentiated N2A cells treated with or without 5 μM VER for another 2 d. EVs isolated from treated or untreated cells displayed roughly equivalent levels of Alix and CD9, indicating that VER treatment did not change the overall number of EVs being secreted, consistent results were obtained by the particle tracking assay. In contrast, cyclin D1 levels in EVs declined 2.2 fold during the period of treatment with VER ( Fig. 6 F), suggesting a potential role for Hsc70 in cyclin D1 packaging into EVs.
We then tested whether a non-functional form of Hcs70 would alter the loading of EVs with cyclin D1. Site-directed mutation of Asn to Asp-10 (D10N) abolishes the ATPase activity of Hsc70 (Huang et al., 1993). Correspondingly, the quantity of cyclin D1 in EVs was decreased in cells transiently expressing the Hsc70 D10N mutant compared to WT Hsc70 ( Fig. 6 G).
Finally, we examined whether EVs collected from differentiated N2A cells pretreated with or without VER had different effects on gene expression in mESCs. EVs (2×10 9 ) were collected and incubated with mESCs in serum-free medium for 4 d. The neural progenitor specific genes, Pax6 and Nestin, together with the neuronal marker, Six3, were upregulated by the EVs from control cells but not EVs from VER-pretreated cells (Fig. 6 H). In addition, we used lentivirus-mediated CRISPRi delivery to knock-down Hsc70 in N2A cells (Gilbert et al., 2014). A control of RA-EVs collected from dCas9 cells, but not the EVs from Hsc70 sgRNA transfected cells, up-regulated the expression of Pax6 and Nestin (Fig. 6 J). These results suggested that the heat shock protein Hsc70 contributes to cyclin D1 sorting and may serve a direct or indirect a role in promoting the function of EVs in the differentiation of mESCs.

Cyclin D1 is required for the EV mediated neural induction of mESCs
To determine whether EV cyclin D1 was taken up by mESCs, cyclin D1-GFP EVs were incubated with mESCs in N2B27 medium. After 24 h, green puncta were detected overlapping the endogenous CD9-labeled mESC cytoplasm, but rarely in the nucleus (Fig. 7 A). After 4 d of daily addition of cyclin D1-GFP EVs to mESC, about 30% of total cyclinD1 was detected by immunoblot of cell lysates at the position of migration of the hybrid protein. The level of endogenous cyclin D1 also increased during the 4 d incubation (Fig. 7 B). These results suggested EV cyclin D1 was internalized by mESCs in serum-free conditions of growth.
Fluorescence localization of the internalized cyclin D1-GFP offered inadequate spatial resolution to determine if the content of EVs was delivered to the cytoplasm of mESCs. As an alternative approach, we examined the proteins in contact with internalized cyclin D1-APEX in comparison to those found in isolated EVs. EVs were collected from cyclin D1-APEX expressing N2A cells (Fig. 7 C). In an incubation with isolated EVs, the biotin-phenol-and H2O2-treated EVs revealed multiple biotinylated proteins which were not detected in incubations containing biotin-phenol without H2O2 (Fig. 7 D). Cyclin D1-APEX EVs were then incubated with mESCs for 2 d in serum-free medium ( Fig. 7 E). Subsequently, these cells were incubated with APEX reagents which resulted in the appearance of biotinylated proteins in a range from~25 KD to~130 KD in the B+H group ( Fig. 7 F). Mass Spectrometry of mESC cell proteins from a streptavidin pulldown identified 116 proteins enriched in the B+H treatment group, with only 3 proteins detected in the biotin-phenol control group (Fig. 7 G). GO analysis indicated the specific enriched proteins were most related to mRNA and ribonucleoprotein binding partners (Fig. 7 H).
In order to distinguish the possibility that the biotinylated proteins were from cyclin D1-APEX in contact with mESC proteins as opposed to those in the EV donor vesicle, we analyzed two candidates, the primarily nuclear-localized proteins Lin28 and nucleolin, found in the mass spec analysis described above. Neither of these proteins was detected in the sample of isolated EVs ( Fig. 7 I). However, each was detected in biotinylated forms in the cells exposed for 2 d to cyclin D1-APEX EVs (Fig. 7 I). These results indicated that the cyclin D1 content of EVs isolated from RA-induced N2A cells is productively taken up and delivered to the cytoplasm or nucleoplasm of mESCs.
Next, to examine the contribution of EV cyclin D1 to mESC neural commitment, we generated cyclin D1-overexpressing N2A cells by use of a lentivirus vector. The empty vector without cyclin D1 overexpression was used as a control. The overexpression led to 2.1-fold more of cyclin D1 protein packaged into EVs (Fig. 7 J). We found that treatment of mESCs with high cyclin D1 EVs increased the expression of neural marker genes Pax6 and Six3 (Fig. 7 K), as well as of Pax6 expressed in neural progenitor cells (Fig. 7 L). To explore if EV cyclin D1 was necessary to promote neural differentiation, we generated a cyclin D1 knock-out N2A cell line using CRISPR/Cas9. The EVs purified from the knockout cells had similar levels of Tsg101 but lacked cyclin D1 (Fig. 7 M). The whole protein profile did not show significant changes after cyclin D1 knockout (Fig. S7 A). The expression of neural marker genes Pax6 and Six3 was significantly lower in cells treated with cyclin D1-depleted EVs than with EVs from control RAtreated N2A cells (Fig. 7 N). Pax6 positive cells were reduced in the cyclin D1 KO EVs treatment group than in the samples incubated with a control of EVs from RA-treated N2A cells (Fig. 7 P). Taken together, these results indicated that EV cyclin D1 accelerated the induction of neural fate in mouse embryonic stem cells.

Discussion
In this study, we report substantial changes in the abundance, physical and functional properties and content of extracellular vesicles produced during neuronal differentiation of neural cell lines PC12 and N2A. Cyclin D1 and CDK4 were selectively sorted into EVs during differentiation mediated by NGF (for PC12 cells) and retinoic acid (for N2A cells). EVs from differentiating cells fractionated by rate and buoyant density centrifugation promoted the expression of genes characteristic of neural induction in mESCs. Cyclin D1 was of particular importance in stimulating this differentiation and Hsc70, a constitutive component of EVs, played a role in the EV capture of cyclin D1.

The secretion of extracelluar vesicles in neuronal cells
Previous studies have shown that extracellular vesicles are released at different stages of neural cell development (Janas et al., 2016). Of particular importance are examples where cellular communication mediated by EVs produced by neural stem cells has been suggested to mediate cytokine signaling in target cells (Cossetti et al., 2014). Other studies have addressed the regulation of EV production, for example in the controlled release of EVs by primary undifferentiated cortical neurons (Fauré et al., 2006) and in a role for calcium influx in fully differentiated cortical and hippocampal neurons where glutamatergic synaptic activity promoted EV secretion (Lachenal et al., 2011). In more recent studies, the neuronal activity-dependent secretion of Arc protein enclosed within extracellular particles has been proposed to promote the intercellular transfer of RNA (Ashley et al., 2018;Pastuzyn et al., 2018). Little is known about the changes in EV content and function as they relate to the differentiation of neuronal precursor cells.
EVs derive from the cell surface by a process of membrane budding to produce particles that are termed microvesicles. A distinct population of EVs, termed exosomes, arise by invagination of membrane into endosomes where intralumenal vesicles accumulate in a structure called the generate multivesicular body (MVBs). MVBs have two possible fates: fusion with the lysosome which results in the degradation of the intralumenal vesicles or fusion with the cell surface membrane which results in the secretion of the intralumenal vesicles. In previous work from our lab, we showed that EVs can be broadly separated into two populations on an iodixinol density gradient with microvesicles sedimenting to a lower and exosomal EVs to a higher buoyant density (Temoche-Diaz et al., 2019).
Here we used the NGF-induced PC12 cells, retinoic acid-induced N2A cells and embryonic bodies derived from mESCs differentiated into Tuj1 + neurons as simple models that recapitulate neurite extension in vitro (Greene and Tischler, 1976;Tremblay et al., 2010;Bibel et al., 2007). Using a two-step fractionation procedure, we found dramatic increases in EV production and in the buoyant density profile of EVs produced during the differentiation of each cell type. Consistently, our results showed the EV marker proteins CD9 and Hsc70 became more heterogenous in terms of membrane buoyant density during differentiation in vitro (Fig. 1, D and E). These changes make it difficult to classify the EVs as either microvesicles or exosomal EVs. Whichever biogenesis path is used to produce the EVs from differentiating PC12 or N2A cells, their content of selected soluble cell cycle regulatory proteins and change in functional characteristics suggests an active role for cargo sorting in their production ( Fig. 5 and 5S).

Sorting of cyclin D1/CDK4 into EVs
Neurons are believed to have lost their capacity to proliferate once they are terminally differentiated (Frade and Ovejero-Benito, 2015;Ohnuma and Harris, 2003). Cyclin D is synthesized at the beginning of G1, and it binds and activates CDK4/6 when the cell leaves the quiescent state (van den Heuvel, 2005;Dehay and Kennedy, 2007). Here, we found cyclin D1 and CDK4 enriched in purified EVs from three different model sources of neuronal cells. In contrast, the regulators of G2/M phase transition, cyclin A2 and cyclin E1, were detected in differentiated cells but not in the EVs (Fig. 5 D). This sorting fidelity was reproduced during RAinduced differentiation of the neuroblastoma cell line N2A (Fig. 5S, A and C). Of possible relevance to our findings, many regulators of the G1/S transition are detected in the adult mouse brain. These may be deployed for cell cycle re-entry under pathological conditions such as in response to DNA damage and oxidative stress (Klein and Ackerman, 2003). Extracellular vesiclemediated secretion may be a protective response to reduce the possibility of abortive cell cycle reentry.
The cyclin D family may also play important roles in neural development (Lukaszewicza and Anderson, 2011). In the mouse cortex and hippocampus, overexpression of cylin D1/CDK4 delays neurogenesis and promotes expansion of basal neural progenitors by shortening the G1 phase (Lange et al., 2009;Pilaz et al., 2009). This may relate to our observation of an increase in the expression of cyclin D family members during NGF-induced differentiation (Fig. 5 A), consistent with other reports. The secretion of EVs enriched in cyclin D1 and their uptake in stem cells may reinforce this aspect of neuronal differentiation. Our findings on EV-mediated cyclin D1 secretion and transfer in a neuronal lineage may extend to astrocytes and oligodendrocyte lineages which have been shown previously to express cyclin D1 (Nobs et al., 2014;Ma et al., 2015;Bosone et al., 2001). Of course, other pathways almost certainly play a role in the differentiation of neural progenitor cells. Parthasarathy et al. (2014) demonstrated that Ntf3 acts as a feedback signal between postmitotic neurons and progenitors in the developing mouse neocortex. Our results suggest an additional role for intercellular communication through secreted vesicles in neuronal differentiation and maturation during mouse neocortex development.

Neuronal EVs promote mESC neural induction
Pluripotent mESCs are guided to alternative specific fates in ectoderm, mesoderm and endoderm tissues by extrinsic cues (Young, 2011). EVs may play a regulatory role in stem cell plasticity by supporting cell self-renewal, differentiation, and proliferation (Watt and Huck, 2013). Although there are reports of these effects, little if any molecular mechanistic insight has developed and in most cases the effects are observed with cells exposed to crude fractions of sedimentable particles, not isolated membrane vesicles. Here we report that EVs are produced in greater abundance during hormonal and chemically-induced neural differentiation of PC12 and N2A cells, and in the differentiation of mouse EBs into neurons, and that these EVs are productively taken up by target cells (Fig. 3 B to G). Buoyant density-purified EVs appear to accelerate aspects of early neural fate conversion of mESCs (Fig. 4 A to E).
In mESCs, cyclin Ds are expressed at a low level which increases in cells transferred to a serumfree medium (White and Dalton, 2005;Liu et al., 2019). In human ESCs, overexpression of cyclin D induces neuroectoderm differentiation (Pauklin and Vallier, 2013). Here, we found that EVs from neuronal differentiated, but not those secreted from undifferentiated cells, stimulated the expression of genes characteristic of mESC neural induction (Fig. 4 C and Fig. S4 A). Correspondingly, we found that cyclin D1 was selectively sorted into EVs secreted by differentiating neuronal EVs but not in EVs from undifferentiated PC12, N2A cells and mESCs either undifferentiated or in conversion to neural progenitor (ES-D8 EBs) (Fig. 5 B to D; Fig. S5 A). Furthermore, overexpression of EV-cyclin D1 increased the expression of neural markers in mESCs ( Fig. 7 J and K). Conversely, EVs from cells depleted of cyclin D1 showed a reduced effect on mESCs (Fig. 7 L and M).
In order to begin to explore the means by which internalized EVs may influence neuronal gene expression in recipient mESC cells, we applied a proximity labeling approach to detect possible intracellular targets of exogenous cyclin D1. EVs were isolated from donor differentiating N2A cells stably transfected with a cyclin D1-APEX gene fusion (Lobingier et al., 2017). Biotinylated target proteins included two nuclear proteins, Lin28 and nucleolin, which may be involved in the neural promoting effect of EV-cyclin D1. Lin28, for example, is a stem cell maintenance factor (Shyh-Chang and Daley, 2013). EV-cyclin D1 may contribute to the regulation Lin28 and other targets, driving mESC neural conversion.
Our results also raise the question of how proteins contained within EVs may become exposed to the cytoplasm/nucleus of target cells. Although many diverse effects have been attributed to the internalization of proteins and nucleic acids contained within EVs, the existence of a membrane fusion process or membrane channels that would allow such a topological transfer of macromolecules has not yet been demonstrated.  The cellular morphology of PC12 cells cultured in growth medium or low serum medium with NGF (50ng/ml) for 3, 6 and 9 days (d). Scale bars, 50 μm.
(B) Expression profiling of Tuj1 and Tau genes during neuronal differentiation of PC12 cells in low serum medium without (Control) or with different doses of NGF (50ng/ml, 100ng/ml).
Expression was normalized to Gapdh in this and all others by qPCR analysis. Data plotted were from three independent experiments. The values represent the mean ± SD. (*p<0.05; **p<0.01) (C) Schematic showing the fractionation of extracellular vesicles. Differential ultracentrifugation was followed by buoyant density flotation in a linear iodixanol gradient.
(D) Immunoblot of EV markers of the iodixanol linear gradient fractions purified from untreated PC12 cells or PC12 treated with NGF for 3, 6 and 9 d (N3, N6, N9). Pellet, 5% of 100,000g vesicle pellet fraction was loaded in gel. Cell, whole cell lysate (10 μg) was loaded in gel. The density of each fraction is indicated at the bottom.
(E) Relative expression of the CD9, Hsc70 and Flot2 from fraction 1 to fraction 24 that shown in Fig.1D. Data plotted represent the mean value from three independent experiments.  (B) Representative electron micrographs of negatively-stained samples of purified EVs at 9300X magnification. Purified EVs from untreated PC12 cells cultured for 3 days (PC12-EV) or PC12 cells treated with NGF for 3, 6 and 9 d (N3-EV, N6-EV, N9-EV). During PC12 differentiation, EVs were collected from 3 d cultured cells and fresh medium together with NGF were replaced every 3 d. Scale bar denotes 0.2 μm.
(C) Nanoparticle tracking analysis of the size and number distribution of purified EVs from 420 ml medium of untreated PC12 cells cultured for 3 days or PC12 cells treated with NGF for 3, 6 and 9 d.
(D) The number of EVs released per PC12 cell or PC12 cells treated with NGF for 3, 6 and 9 d. EV number quantified by nanoparticle tracking analysis. Cell number quantified by a hemocytometer. The values represent the mean ± SD, from three independent experiments.
(E) The number of EVs released per N2A cell cultured for 3 days or N2A cells treated with RA for 3 and 6 d. During N2A differentiation, EVs were collected from 3 d cultured cells and fresh medium together with RA were replaced every 3 d. The values represent the mean ± SD, from three independent experiments.
(F) The number of EVs released per ES cell or ES cell differentiated for 8 d and 12 d. During ES differentiation, EVs were collected from 2 d cultured cells and fresh medium was replaced every 2 d. The values represent the mean ± SD, from two independent experiments.
(H) Quantitative analysis of the immunoblots in G. The values represent the mean ± SD, from two independent experiments. The signal in PC12-EV group was set as 1.
(J) Quantitative analysis of the immunoblots in I. The values represent the mean ± SD, from two independent experiments. The signal in N2A-EV group was set as 1.  (B) Confocal images of differentiated mESCs incubated without EV (No-EV) or PKH6 dyelabeled N6 or N9 EV. Nuclei were stained with DAPI. Scale bar denotes 10 μm.
(C) Immunoblot of GFP and CD9 in EVs (GFP-EV) purified from GFP-expressing cells. mESCs (2×10 5 ) in 35mm dish were incubated in 2 ml of N2B27 medium for 24 h with EVs purified from control or GFP expressing cells.
(D) Immunoblot of GFP from whole cell lysate of mESCs treated with PBS or GFP-EV for 24 h.
(E) Immunoblot of GFP from whole cell lysate of mESCs treated with indicated doses of GFP-EV for 24 h.
(F) Immunoblot of GFP from whole cell lysate of mESCs treated with indicated time of incubation with GFP-EV.
(G) Immunoblot of GFP from whole cell lysate of mESCs treated with 10 ng GFP protein for indicated time or treated with GFP-EV that contained 10 ng GFP for 24 h. The GFP protein amount within the GFP-EV was detected by quantitative immunoblot. (F) Immunoblot analysis of cyclin D1/2 and multiple EV markers of the N6-EVs treated with different concentrations of proteinase K (PK), with or without 1% Triton.
(C) Immunoblots of cyclin D1, CDK4, and multiple EV markers from the N6-EVs treated with different concentrations of proteinase K (PK), with or without 1% Triton. (C) Co-immunoprecipitation analysis of Hsc70 and Hsc90 with cyclin D1 and CDK4 in N2A cells.
(F) Immunoblots of cyclin D1, Alix and CD9 in EVs collected from the differentiated N2A cells treated with VER-155008 (VER). N2A cells pretreated with RA-containing differentiation medium for 4 d, after which cells were exposed to fresh differentiation medium with or without 5 μM VER for 2 more d. EVs collected from 6 d differentiation of N2A cells.
(G) Immunoblots of cyclin D1, Alix and CD9 in EVs collected from the differentiated N2A cells transfected with wild type Hsc70 (WT) or D10N mutant Hsc70 (D10N) (>50% transfection efficiency). Wild type Hsc70 or D10N mutant Hsc70 were transfected by Lipofectamine™ 2000 in 7 plates of 70% confluency N2A cells in DMEM medium for 10 h, followed by a change to fresh differentiation medium for 3 d. EVs were collected from both cells.
(H) The expression analysis of Pax6, Nestin and Six3 in differentiated mESCs treated with 2×10 9 EVs from RA-induced N2A cells with (VER-EV) or without VER (RA-EV). EVs were collected as described in F. The values represent the mean ± SD, from three independent experiments. (*p<0.05; NS, non-significant difference between the two groups, P > 0.05) (I) Immunoblots of Hsc70 and actin in control or Hsc70 sgRNA transfected N2A cells. dCas9 was stably expressed in N2A cell by lentivirus (dCas9), Lentivirus was then used to introduce Hsc70 sgRNA1/2 by transfection of dCas9 cells.
(J) The expression analysis of Pax6 and Nestin in differentiated mESCs treated with 2×10 9 EVs from RA-induced N2A dCas9 cells or Hsc70 sgRNA transfection cells. The values represent the mean ± SD, from three independent experiments. (*p<0.05; NS, non-significant difference between the two groups, P > 0.05) (B) Immunoblots of cyclin D1, actin and GFP of differentiated mESC without or incubation for 4d with cyclin D1-GFP EVs. Quantification of fusion protein uptake was calculated as the ratio of exogenous cyclinD-GFP to endogenous cyclin D1.
(D) Streptavidin-HRP blotting analysis of biotinylated proteins in cyclin D1-APEX expressing EVs. EVs were treated by biotin-phenol together with H2O2 (B+H) or not (B). Biotinylated protein was detected by blotting with streptavidin-horseradish peroxidase (SA-HRP). Ponceau S staining (left of panel) of the same membrane serves as loading control.
(E) Schematic of mESC treated with cyclin D1-APEX EVs, and biotinylated proteins labeled in differentiated mESCs.
(G) Venn diagram of the MS results. MS sample was collected as described in method section. Immunoprecipitation with streptavidin was used to enrich the biotinylated proteins. Diagram generated by Venn diagram package in the R program for statistical computing.
(H) GO analysis of the MS results was shown in G. Go analysis generated by topGO package in the R program for statistical computing.
(I) After the treatment as described in E, F. Immunoblots of Lin28 and nucleolin in differentiated mESCs treated with cyclin D1-APEX EVs, and in Cyclin D1-APEX EVs.
(J) Cyclin D1 was increased in the EVs from N2A cells overexpressing cyclin D1 (OE). The protein level of cyclin D1 was detected in control and OE samples. Actin was used as the internal control of whole cell lysate, and Tsg101 was used as the loading control of EVs. (K) Gene expression level of Pax6, Six3 and Map2 was determined in differentiated mESC treated without EVs, with RA-EV or OE EVs. The values represent the mean ± SD, from three independent experiments. (*p<0.05; **p<0.01) (L) Quantitative analysis of the percentage of cells containing Pax6 normalized to DAPI stain in differentiated mESC treated without EVs, with RA-EV or OE EVs. The values represent the mean ± SD, from two independent experiments. (*p<0.05) (M) Cyclin D1 was absent from cyclin D1 knockout N2A cells and the EVs from cyclin D1 knockout (KO) N2A cells. The cyclin D1 protein was detected in control and KO samples.
(N) The expression of Pax6, Six3 and Map2 was analyzed in differentiated mESCs treated without EVs, with RA-EV or cyclin D1 KO EVs. The values represent the mean ± SD from three independent experiments. (*p<0.05; NS, non-significant difference between the two groups, P > 0.05) (P) Quantitative analysis of the percentage of cells containing Pax6 normalized to DAPI stain in differentiated mESCs treated without EVs, with RA-EV or cyclin D1 KO EVs. The values represent the mean ± SD, from two independent experiments. (*p<0.05)

Figure S7
(A) Whole protein profile of RA-EVs and cyclin D1-KO EV was analyzed by MS. The proteome (1339 proteins) overlapped extensively in these two preparations. The list of the EV proteins shown in supplementary file 1.

Figure 8. Model
Neural development includes early-stage neural induction and late-stage neural genesis. During neural genesis, PC12 cells or N2A cells (dark green) respond to NGF or RA to differentiate into neuronal cells (bright green). The content of extracellular vesicles exhibits dynamic changes corresponding to the fate conversion. Cyclin D1 (magenta dots inside the purple EVs) was enriched in EVs from differentiating neuron. Additional cyclin D1 enriched in EVs from the neuronal cells accelerates the commitment of mouse embryonic stem cells (light orange) to neural progenitor cells (mNPC, light green). Exosomal communication between different development stages may contribute to commitment and conversion of mESCs to the neural lineage.
Mouse ESCs (R1) were maintained on N2B27 medium plus 2i (3 μM CHIR99021 and 1 uM PD0325901, Selleckchem, Houston, TX) and LIF (MilliporeSigma, Burlington, MA). Dishes were pre-coated with 0.1% gelatin for 3 h. ESC serum-free monolayer neural progenitor differentiation was performed in N2B27 medium after LIF and 2i withdraw, and ES cells were cultured at a density of 4-5× 10 4 /cm 2 . For ES cell neuronal differentiation, first, cells were cultured in suspension in 5% Knockout™ Serum Replacement medium ( Thermo Scientific, 10828010) for 8 d to form embryonic bodies (EBs) to achieve the neural progenitor status. Then, trypsinized EBs were cultured in Poly-D-lysine (PDL) pre-coated dishes for another 4 days in N2 medium (Bibel et al., 2007;Zhu et al., 2014). During ES 12 d neuronal differentation, the medium was changed each 2 d.
For PC12/N2A EV collection, the medium was harvested from 3 d cultured PC12 or N2A, and medium from differentiated PC12 or N2A was collected each 3 d. For ES EV collection, the medium was harvested at 2 d interval, including 2 d cultured pluripotent ESCs (ES D0-EV), differentiated ESCs derived EBs cultured from 6 to 8 d (ES D8-EV), and EB trypsinized neurons from 10 to 12 d (ES D12-EV).
For EV functional studies, purified EVs from non-differentiated PC12 or N2A cells and from differentiated PC12 or N2A cells were added in ESCs in N2B27 medium for a monolayer neural induction process. The medium together with purified EVs were changed every day. EVs were used at 2~3× 10 9 particle/ml of N2B27 medium. Particle number was quantified by NanoSight NS300. To measure the efficiency of interaction between EVs and recipient cells, the N2B27 monolayer but not the KSR EBs culture system was used for recipient mESC differentiation.

Differential centrifugation and EV purification
For the two-step extracellular vesicular (EV) fractionation (differential velocity centrifugation and linear iodixanol gradient flotation), 420 ml of conditioned medium was harvested from PC12 cells. All subsequent manipulations were performed at 4˚C. Cells and large debris were removed by centrifugation in a Sorvall R6+ centrifuge (Thermo Fisher Scientific) at 1000xg for 15 min followed by 10,000xg for 15 min in 500 ml vessels using a fixed angle FIBERlite F14-6×500 y rotor (Thermo Fisher Scientific). The supernatant was centrifuged at 100,000xg (28,000 RPM) for 1.5 h in a SW-28 rotor (Beckman Coulter) with a sucrose cushion. The cushion consisted of 2 ml of 60% sucrose in buffer C (20mM Tris-HCL pH 7.4, 137mM NaCl). The supernatant was removed carefully, without reaching the cushion, and new conditioned media was added carefully on top without disturbing the cushion, centrifuged again at 100,000xg (28,000 RPM) for 1.5 h in a SW-28 rotor. The interface between the cushion and the conditioned media was collected (~3 ml per tube). About 9 ml interface that came from 3×SW28 tubes (25×89 mm, Beckman Coulter) was loaded in a SW41 tube (14×89 mm, Beckman Coulter) with 0.75ml of a 60% cushion on the bottom, then centrifugated for 15 h in a SW-41 rotor at 160,000xg (36,000 RPM). The combined interface from the first SW28 sucrose cushion should not exceed a sucrose concentration of 21%, as measured by refractometry, for the second centrifugation in the SW41 to be successful. A clear white band that corresponded to the EV fraction (~1ml in each SW41 tube) was collected after centrifugation. For purification of EV sub-populations based on their distinct buoyant density, the cushion-sedimented vesicles were collected and mixed with 60% sucrose until it was a 4ml solution of 40% sucrose (in buffer C). The 4 ml solution was then loaded at the bottom of a SW 41 tube and equal amounts (1.5ml each) of solutions of 25%, 20%,15%,10% and 5% iodixanol (Optiprep) (diluted in buffer C) were layered on top and centrifuged at 160,000xg (36,000RPM) for 15 h. Fractions of 400 ul each from top to bottom were taken for evaluation. For immunoblot analysis, the floated fraction samples were mixed with 2× SDS loading buffer (0.125 M Tris-HCl (pH6.8), 4% SDS, 20% glycerol, 10%-mercaptoethanol, 0.2% bromophenol blue), heated at 95°C for 10 min.
For EV purification without resolution of EV sub-populations, 420 ml of conditioned medium was harvested from PC12 or N2A cells. Cells and large debris were removed by stepwise centrifugation at 1000xg for 15 min, and then 10,000xg for 15 min at 4˚C. The supernatant was centrifuged at~100,000×g (28,000 RPM) for 1.5 h using two SW-28 rotors (Beckman Coulter). The pellet was resuspended with 200 µl phosphate buffered saline (PBS), pH 7.4, and diluted up to~5ml of PBS and followed by centrifugation at~150,000×g (38,500 RPM) in an SW-55 rotor (Beckman Coulter). Washed pellet material was then resuspended in 100 µl PBS as in the first centrifugation step and 900 ul of 60% sucrose (in buffer C) was added and mixed, then 2ml of 40% sucrose (in buffer C) and 1ml of 20% sucrose (in buffer C) were sequentially overlaid and the SW55 tubes (13×51 mm, Beckman Coulter) were centrifuged at~150,000×g (38,500 RPM) for 16 h in an SW-55 rotor. The 20%/40% interface was harvested, washed once with PBS in an SW-55 rotor. The sedimented EV fraction was resuspended in 100 µl PBS for further analysis.

Nanoparticle tracking analysis
Extracellular vesicle sizes and quantities were estimated using the NanoSight NS300 instrument equipped with a 405-nm laser (Malvern Instruments, Malvern, United Kingdom), analyzed in the scatter mode. Silica 100-nm microspheres (Polysciences, Warrington, PA) served as a control to check instrument performance. Vesicles collected as described above were diluted 1000× with filtered PBS (0.02 µm; Whatman). The samples were introduced into the chamber automatically, at a constant flow rate of 50 during 5 repeats of 60-s captures at camera level 13 in scatter mode with Nanosight NTA 3.1 software (Malvern Instruments). The size was estimated at detection threshold 5 using the Nanosight NTA 3.1 software, after which "experiment summary" and "particle data" were exported. Particle numbers in each size category were calculated from the particle data, in which "true" particles with track length >3 was pooled, binned, and counted with Excel (Microsoft).

PKH67 labeling
EVs were labeled with fluorescent dye PKH-67 using the PKH-67 labeling kit (Sigma-Aldrich). Briefly, 5×10 10 EVs were resuspended in 100 µl PBS and mixed with 100 µl of PHK67 dye diluted in diluent C (4 µl of the PKH67 dye solution to 1 ml of Diluent C) for 2 min, followed by continuous mixing for 30s by gentle pipetting. Excess dye was quenched by adding 100 µl 10% BSA in PBS. This mixture was diluted with 4.5 ml of PBS and centrifuged at 150,000xg (38,500RPM) for 1 h to sediment the PKH-67 labeled EVs. The EV pellet was further washed twice with PBS by centrifugation at 150,000xg for 60 min to remove any free dye and the final EV pellet was resuspended in 100 µl PBS and used for uptake studies.

Proteinase K protection assay
The EVs fractionated by differential centrifugation and sucrose flotation were aliquoted into 20 μl of PBS or PBS containing indicated concentrations of proteinase K (proteinase K was dissolved in TBS, pH7.4, 5mM CaCl2, 50% glycerol) on ice for 20 min, and then treatment with or without 1% Triton X-100 on ice for 10 min. The reactions were stopped by sequentially adding PMSF to final concentration of 5 mM and aliquots were mixed with 2×SDS loading buffer followed by at 95°C for 10 min. Samples were processed for SDS-PAGE and evaluated by immunoblot.

RNA preparation and qPCR analysis
Total RNA was extracted from cells using TRIzol reagent (Invitrogen, Carlsbad, CA). RNA (2.5ug) was reverse transcribed by superScript™ III reverse transcriptase (ThermoFisher Scientific, Waltham, MA) according to the manufacturer's instructions. qPCR was performed using TaqMan™ Universal PCR Master Mix, using an ABI-7900 real-time PCR system (ThermoFisher Scientific, Waltham, MA). Primers for qPCR analysis are listed in supplementary file 1.

Negative staining and visualization of EVs by electron microscopy
The EVs collected as described were resuspended in 1% glutaraldehyde, spread onto glow discharged Formvar-coated copper mesh grids (Electron Microscopy Sciences, Hatfield, PA) and stained with 2% uranyl acetate for 2 min. Excess staining solution was removed by blotting with filter paper. Post drying, grids were imaged at 120 kV using a Tecnai 12 Transmission Electron Microscope (FEI, Hillsboro, OR) housed in the Electron Microscopy Laboratory at UC Berkeley.

Coimmunoprecipitation assay
For EV CD9 immunoprecipitation,~5×10 10 EVs collected from the two steps of purification were diluted into 500 μl of PBS and 2 μg of rabbit monoclonal anti-CD9 (Abcam, Cambridge, U.K., ab92726) or rabbit IgG (Fisher Scientific, Hampton, NH) was added and mixed by rotation overnight at 4˚C. Magvigen protein-A/G conjugated magnetic beads (30 μl) (Nvigen, Sunnyvale, CA) were then added to the EV/antibody mixture and mixed by rotation for 2 h at 4˚C. Beads with bound EVs were washed three times in 1 ml PBS and protein was extracted using with 2×SDS loading buffer, the samples were heated at 95°C for 10 min and analyzed by SDS PAGE and immunoblot. Immunoprecipitation of exosomal cyclin D1 was carried out as with the immunoprecipitation of CD9, except that the purified EVs were diluted and washed in co-IP buffer, containing 50 mM Tris·HCl (pH 7.4), 150 mM NaCl, 1 mM EDTA, 1% Nonidet P-40, and 10% glycerol, 1 mM PMSF, 1 mM DTT, 1×proteinase inhibitors (Roche). Anti-cyclin D1 antibody (2 μg) (Abcam, Cambridge, U.K., ab134175) or rabbit IgG (Fisher Scientific, Hampton, NH) was added.
For coimmunoprecipitation of proteins in a cell lysate, the suspended cells were lysed at 4°C with co-IP buffer. Lysates were incubated with required antibodies at 4°C overnight and then, with 30 μl Dynabeads Protein G Thermo Fischer) for another 2 h. The immunocomplexes were centrifuged and washed three times with cold Co-IP buffer and one time with 50 mM Tris·HCl (pH 7.4) in the absence of proteinase inhibitors. The proteins were released from beads by heating to 95°C in SDS sample buffer, and the samples were analyzed by immunoblot. The following antibodies are used for co-IP: anti-Flag (1:1,000; Sigma, St. Louis, MO), anti-HA (1:1,000; Cell Signaling Technology, Danvers, MA).

BrdU assay
A BrdU cell proliferation assay was conducted according to the supplier's instructions (Cell signaling, Danvers, MA). Briefly, the purified EV fraction was added or not to mESC cultures at the onset of the neural induction process, in N2B27 medium for 4 d. After 3d of mESC differentiation, cells were incubated for a further 24h with BrdU. Cells were then fixed and incubated with 1X detection antibody, washed three times with wash buffer and then incubated with HRP-conjugated antibody 0.5 h at RT. Fixed and labeled cells were then incubated with the TMB substrate and reactions terminated with the stop solution. Outcomes were recorded by absorbance at 450 nm.

Plasmid Construction
The plasmids encoding CDK4-Flag, Hsc70-HA and Hsp90-HA were purchased from Sino Biology Inc (Wayne, PA). The cyclin D1-Flag plasmid was generated by PCR insertion of cyclinD1 into the p3XFLAG-CMV™-14 expression vector (Sigma). The XPack-GFP plasmid was generated by inserting GFP from pEGFP-N1 (BD Biosciences Clontech) into the XPack CMV-XP-MCS-EF1α-Puro Cloning Lentivector (System Biosciences, Palo Alto, CA). The cyclin D1-APEX plasmids were constructed by combining the PCR fragment of cyclin D1 from the cyclin D1-Flag plasmid and APEX from pcDNA3 APEX-nes (Addgene) into XPack CMV constructs (System Biosciences). The cyclin D1-GFP plasmids were constructed by combining the PCR fragment of cyclin D1 from the cyclin D1-Flag plasmid and GFP from pEGFP-N1 (BD Biosciences Clontech) into XPack CMV constructs (System Biosciences). For cyclin D1 overexpression, the cyclin D1 was generated by PCR insertion into the XPack CMV constructs (System Biosciences). Hsc70D10N was generated by PCR-based site-directed mutagenesis using the QuikChange II XL Site-Directed Mutagenesis Kit (Agilent Technologies). Cyclin D1 KO was conducted by CRISPR/Cas9 genome editing (Xie et al., 2016). A pX330-based plasmid expressing green fluorescent protein was used to clone the gRNAs targeting cyclin D1. Three CRIPSR guide RNAs targeting the gene were selected using the CRISPR design tool gRNAs targeting exon 1 were selected. Primers for gRNAs are listed in supplementary file 1.

CRISPR interference
N2A cells expressing UCOE-EF1α-dCas9-BFP-KRAB were obtained by lentivirus transfection (Gilbert et al., 2014). Cells were sorted for the BFP signal 3 d post-transduction, and selected cells were expanded by growth for a few generations and then frozen and stored as parental cells (referred to as dCas9). Sequences for gRNAs targeting the promoter of the genes of interest were selected based on data by (Horlbeck et al., 2016). gRNAs were cloned in plasmid pU6-sgRNA EF1Alpha-puro-T2A-BFP (Gilbert et al., 2014)( Addgene, #60955). The top two gRNAs from the V.2 library (Horlbeck et al., 2016) were chosen to transduce the parental dCas9 cells. Posttransduction cells were selected with 2 ug/ml puromycin for 3 d.

APEX reaction in EVs
Biotin-phenol (500 μM) was incubated with purified EVs for 30 min at 37°C in a total mixture volume of <50 μl. The mixture was removed to an SW 41 ultracentrifuge tube and APEX labeling was initiated by addition of 1 mM H2O2. After 1 min, 12 ml quencher solution was added and EVs were sedimented and washed with the quencher solution twice by centrifugation at 110,000xg (31,500RPM) for 1 h. The pellet fraction was suspended in 40 μl PBS, mixed with SDS-loading buffer in preparation for SDS PAGE and blotting Preparation of APEX-labeled proteins for blotting or mass spectrometry Samples in RIPA were briefly sonicated in a bath sonicator (Covaris, S220) then centrifuged at 10,000 x g for 10 min and the supernatant fraction (800 μl) was applied to 40 μl streptavidinagarose beads (Sigma) followed by rotation overnight at 4°C. Streptavidin-agarose beads were washed two times in RIPA lysis buffer, once with 1 M KCl, once with 0.1 M Na2CO3, once with 2 M urea in 10 mM Tris-HCl, pH 7.4. Biotinylated proteins were eluted from the beads by heating the sample in 4× SDS-loading buffer supplemented with 2 mM biotin and 20 mM DTT for 10 min at 95°C. Streptavidin-HRP blotting or mass spectrometry was used to identify the biotinylated proteins. For mass spectrometry, heated samples were electrophoresed in a 4-20% acrylamide Tris-Glycine gradient gel (Life Technologies) for~3 min. The proteins were stained with coomassie and stained bands were excised from the gel with a fresh razor blade. Samples were submitted to the Taplin Mass Spectrometry Facility at Harvard Medical School for in-gel tryptic digestion of proteins followed by liquid chromatography and mass spectrometry analysis according to their standards.

Statistical analysis
Statistical analysis was performed using Prism (GraphPad Software). Groups were compared using Student's t test. The values represent the mean ± SD, from two or three independent experiments. (NS, non-significant difference between the two groups; *p<0.05; **p<0.01)