Ethics statement

This study was conducted in compliance with recommendations of the ARRIVE Guidelines (Animal Research: Reporting of In Vivo Experiments) for reporting in vivo experiments with research animals [30]. All experimental protocols in this study were approved by the Institutional Animal Care and Use Committee of The Ohio State University and followed the National Institutes of Health Guide for the Care and Use of Laboratory Animals. All experiments were performed in a manner that minimized animal number used, pain, and discomfort.

Animals

For all in vivo experiments, we utilized 3-month-old wild-type FVB/NJ mice (Jackson Laboratories; Stock #001800). Four mice were used for the in vivo EV study. For in vitro studies, E16 embryos were dissected from pregnant FVB/NJ female dams and dissociated and incubated as described. Mice were housed on a 12-h light/dark cycle with access to food and water ab libitum.

Mouse forebrain dissociation

Three-month old FVB/NJ mice were deeply anesthetized with isoflurane until non-responsive to toe pinch. Brains were removed quickly after cervical dislocation and placed in chilled artificial cerebrospinal fluid. The brains were placed and sliced in a brain matrix (Braintree scientific 1 mm, coronal). The forebrain sections were visually identified as those containing hippocampus (used to enrich cells from the cortex and hippocampus, while also including striatum and other subcortical structures, but excluding cerebellum) and were transferred to a 6-well plate with 75 U/ml of collagenase type 3 (Worthington Biochemical Corporation, Cat# NC9405360) in serum-free DMEM (Corning^™, Cat# 10013CV). The tissue sections were subsequently finely chopped, and the tissue was incubated at 37°C for a total of 20 minutes with gentle shaking every 5 minutes. The tissue was further dissociated by pipetting. The dissociated tissue was spun at 300 × g for 5 min at 4°C to remove the undigested tissue as a pellet. The protein concentration of the supernatant was determined by DC protein assay kit II (Bio-Rad, Cat# 5000112) on an iMark Microplate Absorbance Reader (Bio-Rad, Cat# 1681130) to ensure an equal amount of starting tissue was used (3 mg) for each sample. The supernatant was transferred to a fresh tube and spun at 2,000 × g for 10 min at 4°C to remove the cellular fraction. Next, the supernatant was spun at 10,000 × g for 30 min at 4°C to remove cell debris and large intracellular vesicles. Finally, to concentrate the samples, the supernatant was spun at 100,000 × g for 70 min in a Bechman (model Optima L-90K) ultracentrifuge and the pellets (P4, raw EVs) were resuspended in 500 μl 1× PBS for use as the input for SEC EV isolation. A schematic diagram of the procedure is presented in S1 Fig.

Mouse primary dissociated forebrain cell cultures

Sixteen-day old embryos (E16) were harvested from pregnant dams. The pregnant dams were deeply anesthetized with isoflurane until unresponsive to toe pinch before cervical dislocation and subsequent cesarean section removal of the embryos. The forebrain (defined as the whole brain including cortex, hippocampus, striatum, and additional subcortical regions, but with the cerebellum removed) of each embryo was dissected in 1× PBS on ice. The brain tissue was mechanically dissociated in 1 mL of DMEM (Corning^™, Cat# 10013CV) containing papain (20 U/ml, Sigma Aldrich, Cat# P3125) and was incubated for 30 min at 37°C. The dissociated cells were then separated from the remaining undigested tissue by a 70 μm cell strainer (MTC Bio, Cat# C4070). Dissociated cells were added to 14 mL of plating media (DMEM supplemented with 5% fetal bovine serum [FBS, Alkali Scientific, Cat# FB72]) and plated into T75 flasks (VWR, Cat# 10062–860). Embryos from 2 dams were combined to generate 4 flasks of cells (each flask representing 1 replicate). At the end of day in vitro (DIV) 1, the plating media was fully replaced with fresh maintenance media (DMEM supplemented with 0.5% FBS, and 1× B27 [Invitrogen, Cat# 17504044]). The percentage of supplemented FBS was reduced by 90% to prevent glial proliferation. Beginning on DIV2 and continuing on every third day during DIV 2–14, 50% of the medium was replaced with fresh maintenance media also containing 100 μg/ml penicillin-streptomycin (Gibco, Cat# 15140122). On DIV 15, the maintenance media was fully removed, the cells were gently washed with 1× PBS three times to remove accumulated EVs and cellular debris. Extracellular vesicle-free, FBS-free DMEM was added to the flasks. On DIV 16, after allowing for EV accumulation in the DMEM for 24 hr, the media was collected and spun down at 1,000 × g for 10 min to remove the large debris and dead cells. The supernatant was collected and further spun at 10,000 × g for 30 min to remove large intracellular vesicles and additional cellular debris. The supernatant was then concentrated (and initially purified) down to 500 μl using UF (Amicon Ultra-15, 100 kDa membrane, Cat# UFC910024) and used as input for EV isolation via SEC. The 100 kDa membrane is predicted to trap vesicles ≥10 nm for collection. A schematic diagram of the procedure is presented in S1 Fig.

Izon SEC column EV isolation and UF concentration

The raw EVs were isolated from both dissociated in vivo adult brain and embryonic in vitro neuron-astrocyte culture medium using a qEV SEC column (Izon) following the manufacture’s manual. The qEVOriginal / 35 nm column (Izon, Cat# ICO-35), designed for separating EVs between 35–350 nm from a 500 μL biological sample, was used for this study. Briefly, qEVOriginal columns were flushed with 1× PBS before the samples were loaded. All the samples were concentrated into 500 μl in 1× PBS as described above before being loaded onto the column. The qEV Automatic Fraction Collector (AFC, Izon, Cat# AFC-V2) was programmed for automatic isolation of the EVs. A total of 4 fractions (corresponding to fraction 7–10; fractions 1–6 correspond to the void volume that is deplete of EVs) were collected for each sample. Fraction 9 and fraction 10, which contained the small EVs enriched in sizes ranging between 50–150 nm, were pooled and concentrated (as well as further purified) by UF (Amicon Ultra-4, 100 kDa membrane, Cat# UFC810096) to a final volume of 50 μl. The 100 kDa membrane for UF was selected as the pore size is predicted to isolate EVs of 10–20 nm and larger, but deplete smaller, empty vesicles and free protein under 100 kDa. A schematic diagram of the procedure is presented in S1 Fig. The obtained samples were subjected to Western blot analysis to assess markers of EVs, including Tsg 101, Alix and CD81. Results are presented in S2 Fig.

Nanoparticle tracking analysis

The concentration and size of the vesicles were characterized by Nanoparticle Tracking Analysis (NTA; Malvern Pharmaceuticals, NanoSight NS300) using the manufacturer’s instructions. A small portion of the isolated EVs were diluted 1:1000 with 1× PBS to achieve 15 to 50 particles per frame for tracking. The sample was injected via syringe and slowly pumped into the chamber for analysis. A laser passed through the chamber and the particles in suspension in the path of the beam scattered the light which could be visualized via 20× magnification. Each sample was analyzed three times for 30s in the chamber to generate an aggregate average.

Sample clean-up and protein digestion via S-Trap

After NTA analysis, the remaining samples were denatured by the addition of 5× SDS loading buffer (250 mM Tris—HCl pH 6.8, 50% glycerol, 10% SDS, 500 mM DTT, 0.5% bromophenol blue, 5% β-mercaptoethanol) to a final concentration of 1× and heated at 95°C for 5 min. Up to 50 ug of the samples were used for trypsin digestion. The samples were purified of macromolecules other than proteins using Suspension Trapping (S-Trap) micro columns (≤ 100 μg) per the manufacturer’s instructions (Protifi, Cat# K02MICRO10). Briefly, 5 μL of 50 mM ammonium bicarbonate containing 5 μg/μL DTT was added, and the sample was incubated at 65°C for 15 min followed by addition of 5 μL of 50 mM ammonium bicarbonate containing 15 μg/μL iodoacetamide (incubated at RT for 15 min in the dark). Samples were then acidified by adding 12% phosphoric acid (1:10 v/v acid to sample). For every 25 μL of sample, 165 μL of Triethylammonium bicarbonate (TEAB 1M)/MeOH (10:90 v/v) was added and then loaded to the S-Trap spin column for further washing. Samples were centrifuged at 4,000 × g for 3 min (4°C) to remove supernatant. Then, 150 μL of TEAB/MeOH (10:90 v/v) was added to the spin column as wash solution and the trap was washed 3–6 times dependent upon the initial loading volume. After the final wash, sequencing grade trypsin dissolved in 50 mM TEAB was added and digested O/N at 37 °C. The following day, peptides were eluted by adding 40 μL of 50 mM TEAB, 0.1% formic acid (FA) and 0.1% FA in acetonitrile (50:50), sequentially. The sample was pooled together and dried in a vacufuge and resuspended in 20 μL of 50 mM acetic acid. Peptide concentration was determined by nanodrop (A280nm). The samples were then submitted for mass spectrometry analysis.

Mass spectrometry

Experimental procedures were performed similar as previously indicated with modifications [31]. Capillary-liquid chromatography-nanospray tandem mass spectrometry (Capillary-LC/MS/MS) was performed on an Orbitrap Fusion mass spectrometer (Thermo Scientific, Cat# FETD2-10002) equipped with an EASY-Spray^™ Source operated in positive ion mode. Samples were separated on an EASY-Spray^™ nano column (PepMap^™ RSLC, C18 3μ 100A, 75μm X150mm Thermo Scientific) using a 2D rapid separation liquid chromatography HPLC system from Thermo Scientific. Each sample was injected into the μ-Precolumn Cartridge (Thermo Scientific) and desalted with 0.1% FA in water for 5 minutes. The injector port was then switched to inject, and the peptides were eluted off of the trap onto the column. Mobile phase A was 0.1% FA in water while acetonitrile (with 0.1% FA) was used as mobile phase B. Flow rate was set at 300nL/min. Mobile phase B was increased from 2% to 16% in 105 min and then increased from 16–25% in 10 min and again from 25–85% in 1 min and then kept at 95% for another 4 min before being brought back quickly to 2% in 1 min. The column was equilibrated at 2% of mobile phase B (or 98% A) for 15 min before the next sample injection.

MS/MS data was acquired with a spray voltage of 1.6 kV and a capillary temperature of 305°C was used. The scan sequence of the mass spectrometer was based on the preview mode data dependent TopSpeed^™ method. The analysis was programmed for a full scan recorded between m/z 375–1500 and a MS/MS scan to generate product ion spectra to determine amino acid sequence in consecutive scans starting from the most abundant peaks in the spectrum in the next 3 seconds. To achieve high mass accuracy MS determination, the full scan was performed at Fourier transformation (FT) mode and the resolution was set at 120,000 with internal mass calibration. The automatic gain control (AGC) target ion number for FT full scan was set at 4 × 105 ions, maximum ion injection time was set at 50 ms and micro scan number was set at 1. Multistage mass spectrometry (MSn) was performed using higher energy collisional dissociation (HCD) in ion trap mode to ensure the highest signal intensity of MSn spectra. The HCD collision energy was set at 32%. The AGC target ion number for ion trap MSn scan was set at 30,000 ions, maximum ion injection time was set at 35 ms and micro scan number was set at 1. Dynamic exclusion is enabled with a repeat count of 1 within 60s and a low mass width and high mass width of 10ppm.

Data were searched using Mascot Daemon by Matrix Science version 2.7.0 (Boston, MA) via Proteome Discoverer (version 2.4, Thermo Scientific) and the database searched against the most recent Uniprot databases. The mass accuracy of the precursor ions was set to 10 ppm, accidental pick of one ¹³C peaks was also included into the search. The fragment mass tolerance was set to 0.5 Da. Carbamidomethylation (Cys) is used as a fixed modification and considered variable modifications were oxidation (Met) and deamidation (N and Q). Four missed cleavages for the enzyme were permitted. A decoy database was also searched to determine the false discovery rate (FDR) and peptides were filtered at 1% FDR. Proteins identified with at least two unique peptides were considered as reliable identification. Any modified peptides are manually checked for validation. Relative quantitation was done using the label free quantitation (LFQ) mass spectral peak intensity approach. Peptide precursor (MS1) intensities from the same protein were generated from ProteomeDiscoverer MASCOT search and summed up for quantitation comparison.

Statistical analysis

The proteomic data for in vivo EVs and in vitro EVs were first merged and aligned (n = 4/group). Proteins that did not have values for at least 2 replicate samples for either in vivo or in vitro were excluded (4 proteins total were excluded). The proteins were analyzed using the Database for Annotation, Visualization and Integrated Discovery (DAVID) [32] to determine the cellular component (CC), biological process (BP), and molecular function (MF) of the proteins of the total proteome of both the in vivo and in vitro EVs. DAVID was also used to examine biological pathways enriched from proteins identified in both in vivo and in vitro samples through conducting a Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis. The Benjamini—Hochberg method (p-value converted to -log₁₀ for data presentation) was used to determine if a category within a gene ontology (GO) group or KEGG pathway is significantly enriched. For relative enrichment of individual proteins in vivo or in vitro, the LFQ for each protein was converted to a Z-score for comparison between sample types. Proteins enriched in vivo, enriched in vitro, or not enriched were analyzed by DAVID for CC, BP, MF (p < 0.05 presented) and KEGG pathway analysis (p < 0.05 for Benjamini-Hochberg test presented). Further, a list for each of the 5 categories from the MISEV2023 guidelines [28] were generated as well as the Vesiclepedia Top 100 proteins [29]. For the MISEV2023 categories, the HUGO Gene Nomenclature Committee website [33] was used to identify all proteins in the same family for those marked as having multiple proteins. All proteins from both sample types were cross-referenced with these lists and heatmaps (generated online by Heatmapper [34]) were generated from the replicate LFQ Z-scores for each sample type. The proteins from each categorical list were further analyzed for significant enrichment to a sample type (in vivo or in vitro) using a Students T-test comparing the Z-scores from the 4 replicates for each of the in vivo and in vitro EV sample types. The statistical values presented are indicated in each figure but represent at least p < 0.05.

Biophysical characterization of EVs isolated from mouse intact brain and dissociated culture

A number of studies have reported the procedures for the isolation of EVs from brain tissues [2, 35]. The protocol generally starts with a gentle dissociation of the tissue to free EVs from the extracellular space followed by sequential centrifugations (in a series of increasing speed) to remove cells, cellular debris, microvesicles, non-EV vesicles, and free protein. The EVs are then isolated using one of various approaches including, density/gradient centrifugation, UF, immunoprecipitation, or SEC to obtain exosome-enriched EVs [36–38]. In the current study, we combined SEC with UF to isolate EVs. EVs isolated from the forebrain of the mouse were either collected acutely or after cells from the region were dissociated and grown in culture for 15 days. Acutely isolated EVs from the forebrain (hereinafter referred to as in vivo EVs) showed an approximate normal distribution between 50–250 nm in diameter with a major peak of abundant EVs of around 114 nm (Fig 1A). To compare in vivo EVs with the forebrain EVs released under cultured conditions, we prepared primary cell cultures from the forebrain of 16-day-old mouse embryos. Cultures were grown for 15 DIV before beginning EV collection for 24 hr. The EVs (hereinafter referred to as in vitro EVs) isolated from the culture media were also characterized by NTA. Results showed a more heterogeneous population of EVs, primarily sized between 20–250 nm in diameter, with a major peak at 133 nm and three smaller peaks at 24, 57, and 193 nm (Fig 1B). Comparatively, in vitro EVs yielded, on average, larger EVs but these EVs were reduced ~50% in concentration compared to in vivo EV isolates (Fig 1A and 1B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Characterization of EVs isolated via SEC and UF from mouse forebrain and primary culture.

EVs were collected from either (1) acutely dissociated mouse forebrain, in vivo EVs, or (2) primary cells dissociated from embryonic mouse forebrain, in vitro EVs. (A, B) We first characterized the biophysical properties of both EV groups using NTA. (A) In vivo EVs yielded high density EVs with an average diameter of 131.5 nm and a peak at 114.5 nm while (B) in vitro EVs yielded lower concentrations as well as slightly larger (mean: 163.4 nm; primary peak: 142.0 nm) EVs. For both EV groups, the majority of EVs assessed fell within the typical parameters for small, exosome-like EVs (30–200 nm). (C-E) The protein content of each EV group was assessed using GO analysis of mass spectrometry data. A total of 3,204 and 3,583 proteins were identified in in vivo and in vitro EVs, respectively, and used for analysis. (C) Principle CC analysis of both in vivo and in vitro EVs yielded very similar term enrichment, specifically with the top terms cytoplasm and synapse followed by cytoplasmic vesicle, endosome, and Golgi apparatus consistently being the top 5 most significantly enriched terms. (D) Consistently, principle BP analysis shows that the same top 4 terms (in the same order) are enriched for both in vivo and in vitro including: protein transport, transport, mRNA processing, and mRNA splicing. (E) Principle MF analysis, again, suggests comparable term enrichment between the EV groups with the top terms being RNA-binding and ribonucleoprotein, followed by actin-binding, hydrolase, and chaperone. (F) KEGG pathway enrichment analysis indicates similar pathway enrichment between in vivo and in vitro EVs, with the most enriched pathways being endocytosis, salmonella infection, proteasome, and focal adhesion. Data displayed is the Benjamini-Hochberg p-value for term enrichment converted to -log₁₀. Data shown in C-E is p < 0.001. Data shown in F is the Benjamini-Hochberg p < 1e^-9. Values denoted in F with “a” were found to be present, but with a Benjamini-Hochberg p ≥ 1e^-9. n = 4 replicates per EV group.

https://doi.org/10.1371/journal.pone.0309716.g001

Proteome characterization of in vivo and in vitro EVs

Next, we sought to determine how the proteome of the in vivo and in vitro compared. The in vivo and in vitro EVs samples were then subjected to proteomic analysis by mass spectrometry to determine protein composition. A total of 3,204 proteins were identified across all in vivo EV samples while a total of 3,583 proteins were identified across all in vitro EVs samples (n = 4/group). A protein was considered expressed in either group of EVs if at least 2 replicates indicated expression of the protein. GO analysis was carried out on the proteins to determine the EV composition as indicated by CC, BP, and MF. Note that only terms resulting in p < 0.001 are displayed in (Fig 1C–1E). For CC GO enrichment (Fig 1C), a total of 29 terms were significantly enriched (p < 0.05 for Benjamini-Hochberg test) in in vivo EVs. The most notably enriched terms (p < 1e^-30) include: cytoplasm (45.3% of 3,204), synapse (6.8%), cytoplasmic vesicles (7.8%), endosome (7.1%), and Golgi apparatus (9.1%). Similarly, a total of 28 CC GO terms (Fig 1C) were significantly enriched (p < 0.05) for in vitro EVs. The most notably enriched terms (p < 1e^-30) were, again, for cytoplasm (45.3% of 3,583), synapse (6.3%), endosome (7.0%), Golgi apparatus (8.8%), and cytoplasmic vesicles (7.3%). For BP GO enrichment, a total of 45 terms (Fig 1D) were significantly enriched (p < 0.05) in in vivo EVs. The most notably enriched terms (p < 1e^-15) included: protein transport (8.2%), transport (17.1%), mRNA processing (4.8%), mRNA splicing (4.1%), endoplasmic reticulum—Golgi transport (1.7%), and protein biosynthesis (2.2%). Similarly, again, a total of 49 BP GO terms (Fig 1D) were significantly enriched (p < 0.05) for in vitro EVs. The most notably enriched terms (p < 1e^-15) were for protein transport (8.1%), transport (16.6%), mRNA processing (4.6%), mRNA splicing (3.8%), protein biosynthesis (2.0%), and endoplasmic reticulum—Golgi transport (1.7%). Finally, for MF GO enrichment, a total of 36 terms (Fig 1E) were significantly enriched (p < 0.05) in in vivo EVs. The most notably enriched terms (p < 1e^-10) were for RNA-binding (7.9%), ribonucleoprotein (4.1%), actin-binding (3.1%), hydrolase (12.6%), chaperone (2.5%), ribosomal protein (2.5%), and ligase (1.9%). Again, a similar number (38 total) of MF GO terms (Fig 1E) were significantly enriched (p < 0.05) in in vitro EVs. The most notably enriched terms (p < 1e^-10) were for RNA-binding (7.8%), ribonucleoprotein (3.9%), hydrolase (12.7%) actin-binding (2.9%), chaperone (2.3%), ligase (1.7%), and ribosomal protein (2.2%).

Further, 58 proteins were identified in our total sample proteome (containing both in vivo and in vitro samples) that corresponded with the term “extracellular exosome” (GO:0070062). This term contains a list of 154 genes, resulting in an approximate 38% overlap of our identified proteins with the GO term’s total list of proteins for “extracellular exosome”. When looking at the total proteome for in vivo, 57 proteins corresponded with the CC GO term “extracellular exosome” (p = 1e^-13; 1.8%), while the total proteome of in vitro had 58 proteins corresponding to this term (p = 2.7e^-12; 1.6%).

Additionally, KEGG pathway enrichment analysis was carried out on the identified proteins to determine molecular biological functions. A total of 128 pathways were found to be significantly enriched in the in vivo samples (p < 0.05, Benjamini-Hochberg) (Fig 1F). Note that only terms resulting in Benjamini-Hochberg p < 1e^-9 are displayed in Fig 1F. The most notably enriched terms (p < 1e^-13) included: endocytosis (3.6%), salmonella infection (3.2%), proteosome (1%), carbon metabolism (1.9%), and pathways of neurodegeneration in multiple diseases (4.9%). Similarly, a total of 129 terms were found to be significantly enriched in the in vitro samples (p < 0.05, Benjamini-Hochberg) (Fig 1F). The most notably enriched terms (p < 1e^-13) included: endocytosis (3.5%), salmonella infection (3.1%), proteasome (1%), focal adhesion (2.5%), and carbon metabolism (1.8%).

Overall, the data indicates that the current combined method of SEC and UF isolates EVs released from in vitro primary dissociated brain cell culture medium that have a remarkably similar proteome as those released from in vivo EVs based on GO analysis and KEGG pathway analysis.

In vivo and in vitro EVs proteome content comparison

Next, we compared protein composition between in vivo EVs (3,204 proteins) and in vitro EVs (3,583 proteins). Overlap between in vivo and in vitro EVs are most likely to represent actual EV proteins. There were 3,147 proteins overlapped between in vivo EVs and in vitro EVs, representing 86.4% total overlap, with only 57 proteins found exclusively in the in vivo EVs (1.6%), and 436 proteins found exclusively in the in vitro EVs (12.0%; Fig 2A). CC GO enrichment analysis revealed that the 3,147 overlapping proteins were significantly enriched for 26 CC GO terms (p-value < 0.05). These top CC GO terms included the following (p < 0.001): cytoplasm (45.8% of all the overlapped 3,147 proteins), synapse (6.7%), endosome (7.1%), cytoplasmic vesicle (7.8%), Golgi apparatus (9.0%), cell projection (10.5%), endoplasmic reticulum (10.7%), protease (1.3%), spliceosome (2.3%), cytoskeleton (10.6%), nucleus (31.5%), coated pit (0.9%), lysosome (3.3%), microtubule (2.5%), synaptosome (0.8%), cell junction (3.2%), and ECM (2.1%; Fig 2B). This is consistent with the CC GO term enrichment for each EV proteome alone (Fig 1C). The 57 proteins found exclusively in the in vivo EVs were significantly enriched for 9 GO cellular component terms (p < 0.05). Of these CC GO terms (Fig 2C), endoplasmic reticulum was most robustly enriched (36.4% of 57 proteins; p < 0.001). The 436 proteins found exclusively in the in vitro EVs were significantly enriched for 12 CC GO terms (p < 0.05). The most enriched CC GO terms (Fig 2D) included: cytoplasm (41.2% of 436 proteins), ECM (4.4%), and nucleus (33.7%).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Comparison of protein composition between in vivo EVs and in vitro EVs.

The proteomes for in vivo and in vitro EVs were overlaid to identify which proteins were expressed in both or were unique to one EV group. (A) A total of 3,204 and 3,583 proteins were identified in vivo and in vitro respectively. Of these proteins, 3,147 were expressed in both EV groups while only 57 proteins were exclusive to in vivo and 436 were exclusively identified in vitro. Data displayed is CC analysis of GO terms expressed (B) in both in vivo and in vitro, (C) only in vivo, and (D) only in vitro. (E-G) Data displayed is KEGG pathway enrichment analysis: (E) proteins found in both in vivo and in vitro EVs (p < 1e^-6), (F) proteins found only in vivo EVs (p < 0.05), and (G) proteins found only in vitro EVs (p < 0.05). Data displayed for CC GO terms and KEGG pathways are the Benjamini-Hochberg p-value (B-E) and normal p-value (F, G) for term/pathway enrichment converted to -log10. All CC GO data shown is p < 0.05. KEGG pathway data shown is p < 0.05 (F, G) and p < 1e^-7 (E). n = 4 replicates per EV group.

https://doi.org/10.1371/journal.pone.0309716.g002

KEGG pathway analysis revealed that the 3,147 overlapping proteins were significantly enriched for 112 KEGG pathway terms (p < 0.05, Benjamini-Hochberg). The most notably enriched KEGG pathway terms (p < 1e^-13) included: endocytosis (3.6%), salmonella infection (3.2%), proteosome (1.1%), carbon metabolism (1.9%), and pathways of neurodegeneration in multiple diseases (5%) (Fig 2E). The 57 proteins found exclusively in the in vivo EVs were significantly enriched for 4 KEGG pathway terms (p < 0.05) (Fig 2F). Of these KEGG pathway terms, metabolic pathways were the most robustly enriched (26.3%, p < 0.001), along with sphingolipid metabolism (5.3%), PPAR signaling pathway (5.3%), and glycerophospholipid metabolism (5.3%). The 436 proteins found exclusively in the in vitro EVs were significantly enriched for 28 KEGG pathway terms (Fig 2G; p < 0.05). The most enriched terms included: animal autophagy (2.9%), proteoglycans in cancer (3.1%), adherens junction (2%), chemokine signaling pathway (2.9%), and cytoskeleton in muscle cells (3.1%).

As the vast majority of proteins between the in vivo and in vitro EV groups were so similar, it is conceivable that all of these common proteins are associated with EVs and/or are representative of common contaminants associated with EVs. The limited number of proteins exclusively identified in in vivo EVs makes it difficult to determine what this small population of proteins represents but may be indicative of contaminants unique or selectively increased in in vivo EV isolation. The proteins exclusively identified in the in vitro EVs could be associated with the EVs that carry nuclear content or from ECM turnover as cells were not damaged during EV collection.

In vivo and in vitro EVs proteomes have different relative expression

While the overall proteome content of both in vivo and in vitro EVs is mostly conserved, the relative abundance of individual proteins between the two sample types is quite disparate. To compare the enrichment of each protein between sample types, the individual LFQ values for each replicate were converted to a Z-score to describe the relative abundance of each protein in relation to the overall sample. The values of each replicate were then compared via a Student’s T-test (two-tailed) to determine if each protein is significantly enriched in either sample type (in vivo compared to in vitro; p < 0.05). Of the total 3,640 proteins identified when considering both in vivo and in vitro EVs, only 417 (11.5%) of the proteins were not significantly enriched (p ≥ 0.05, t-test of Z-scores abundancy values) in either sample type (Fig 3A). The composition of this category of proteins was assessed by CC GO analysis with 21 terms significantly enriched (p<0.05, Benjamini-Hochberg). The following terms represent the most robustly enriched CC GO terms (p < 0.001, Benjamini-Hochberg): endoplasmic reticulum (19.9% of 417 proteins), endosome (9.2%), Golgi apparatus (11.1%), coated pit (2.5%), synaptosome (2.3%), cytoplasmic vesicle (8.3%), synapse (6.7%), membrane (55.7%), cell membrane (26.1%), and intermediate filament (2.1%) (Fig 3B). Similar to the exclusive protein expression pattern (Fig 2A), only a small percentage of proteins were significantly enriched or exclusively expressed in the in vivo EVs (268 proteins; 7.4%; p < 0.05, t-test of abundancy Z-score abundancy value) while in vitro EVs (2,955 proteins) yielded a staggering 81.2% of enriched (2,519 proteins) or exclusive (436 proteins) protein expression (Fig 3A). For proteins enriched in vivo, there were 9 CC GO terms significantly enriched (p < 0.05, Benjamini-Hochberg) with the following terms representing the most robustly enriched (p < 0.001, Benjamini-Hochberg): endoplasmic reticulum (23.3% of 268 proteins), cytoplasmic vesicle (13.5%), cytoplasm (44.4%), and synapse (10.2%) (Fig 3C). For those proteins enriched in vitro there were 28 CC GO terms significantly enriched (p < 0.05, Benjamini-Hochberg) with the following terms representing the most robustly enriched (p < 0.001, Benjamini-Hochberg): cytoplasm (45.1% of 2,955 proteins), endosome (6.8%), nucleus (33.0%), synapse (5.7%), proteasome (1.3%), Golgi apparatus (8.3%), cell projection (9.7%), cytoplasmic vesicle (6.4%), cytoskeleton (10.1%), spliceosome (2.0%), lysosome (3.4%), basement membrane (0.7%), ECM (2.3%), microtubule (2.4%), nuclear pore complex (0.6%), and cell junction (2.9%) (Fig 3D).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Proteome enrichment of in vivo and in vitro EVs.

(A-D) As the overlap for both proteomes was largely conserved, the proteomes for in vivo and in vitro were compared for relative enrichment of proteins. Relative enrichment was determined by converting the LFQ mass spectrometry values to a Z-score for comparison between replicates and between groups. A two-tailed Student’s t-test was used to determine if a protein was enriched into either in vivo or in vitro EVs (p < 0.05) or was not enriched in either group (p ≥ 0.05). Proteins that were unique to either EV type were included for analysis. (A) Overall, only 417 proteins were similarly expressed in in vivo and in vitro EVs (p < 0.05). In vitro EVs were significantly enriched for most of the proteins analyzed (2,955, p < 0.05), while the in vivo EVs, again, only exhibited a few significantly enriched proteins (268, p < 0.05). Data displayed is CC analysis of GO terms that were (B) not enriched either in vivo or in vitro EVs, (C) enriched specifically in in vivo EVs, and (D) enriched specifically in vitro EVs. (E-G) Both enriched and not enriched proteins were further analyzed for KEGG pathway enrichment analysis. Proteins that were unique for either in vivo or in vitro EVs were included for analysis. Data displayed is KEGG pathway enrichment analysis of proteins that were (E) not enriched in either in vivo or in vitro EVs (p < 0.05), (F) specifically enriched in in vivo EVs (p < 0.05), and (G) specifically enriched in in vitro EVs (p < 1e^-7). Data displayed for CC GO terms and KEGG pathways are the Benjamini-Hochberg p-value for term/pathway enrichment converted to -log₁₀. All CC GO data shown is p < 0.05. KEGG pathway data shown is p < 0.05 (E, F) and p < 1e^-7 (G). n = 4 replicates per EV group.

https://doi.org/10.1371/journal.pone.0309716.g003

In addition to the GO CC analysis, a KEGG pathway analysis was performed to assess these categories of proteins. There were 11 KEGG pathway terms (Fig 3E) that were not significantly enriched (p < 0.05, Benjamini-Hochberg) which included: endocytosis (5.5%), endocrine and other factor-regulated calcium reabsorption (2.5%), Huntington disease (4.8%), protein processing in endoplasmic reticulum (3.5%), cholesterol metabolism (1.8%), synaptic vesicle cycle (2.1%), phagosome (3.2%), cytoskeleton in muscle cells (3.7%), pathways of neurodegeneration in multiple diseases (5.8%), cGMP-PKG signaling pathway (3%), and long-term potentiation (1.8%). For proteins enriched in vivo EVs, there were 19 KEGG pathway terms that were significantly enriched (p < 0.05, Benjamini-Hochberg) with the following terms representing the most robustly enriched (p < 0.001, Benjamini-Hochberg): ribosome (25.5%), coronavirus disease (27.3%), synaptic vesicle cycle (6.5%), collecting duct acid secretion (3.6%), and pancreatic secretion (4.4%) (Fig 3F). For the proteins enriched in vitro, there were 97 KEGG pathway terms significantly enriched (p < 0.05, Benjamini-Hochberg) with the following terms representing the most robustly enriched (p < 1e^-10, Benjamini-Hochberg): proteasome (1.2%), focal adhesion (2.8%), salmonella infection (3.2%), carbon metabolism (1.9%), and endocytosis (3.3%) (Fig 3G).

Comparison of in vivo and in vitro EV proteins using MISEV2023 EV content categories and Vesiclepedia top 100 list

Beyond looking exclusively at GO analysis and KEGG analysis to determine the composition and relative enrichment of protein categories in the EV proteomes, we further sought to compare each proteome with known EV and common EV contaminating proteins. To accomplish this, we utilized two categorical lists: (A) MISEV2023 established by the Executive Committee of the International Society for Extracellular Vesicles [28] and (B) the Vesiclepedia Top 100 proteins identified in EV studies [29]. First we created categorical lists of all proteins indicated in the MISEV2023 recommendations (including family homologues as indicated) for the following 5 categories: category 1, transmembrane or GPI-anchored proteins associated with plasma membrane and/or endosomes; category 2, cytosolic proteins recovered in EVs; category 3, major components of non-EV co-isolated structures; category 4, transmembrane, lipid bound and soluble proteins associated to intracellular compartments other than plasma membrane and/or endosomes; and category 5, secreted proteins recovered with EVs [28]. We then used the proteins from the 5 categorical lists and identified which proteins were expressed in our EV samples. Of those proteins identified, we next compared the relative abundance of each protein identified between in vivo and in vitro EVs. Results are presented in a heatmap for each category (Figs 4–8), which demonstrates the relative enrichment of each protein in the in vivo and in vitro EVs represented by the Z-score (converted from LFQ value as described above). A high Z-score (relative enrichment) is represented in red color indicating higher protein abundance versus the mean of all proteins in that individual sample, and a low Z-score (relative depletion) is represented by blue color indicating lower protein abundance versus the mean of all proteins in that independent sample. Further, the statistical p-values for average sample type enrichment of each protein, based on Student’s T-test of Z-score values, are indicated at the end of the protein name. Note the same strategy was used for analysis of the Vesiclepedia Top 100 proteins.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Enrichment of in vivo and in vitro EV proteins using MISEV2023 EV content category 1.

Proteins identified in either or both EV groups were organized based on the MISEV2023 EV content categories for analysis. Relative enrichment for each protein in each category was determined by converting the LFQ mass spectrometry values to a Z-score for comparison between replicates and between groups. The results of the enrichment analysis are displayed as a heatmap. EV category 1 proteins are enriched in vitro. Heatmaps display the row Z-score values with red representing relative enrichment, blue indicating relative depletion, and white indicating no enrichment or depletion for either in vivo or in vitro EVs. A two-tailed Student’s t-test was used to determine if a protein was significantly enriched in either in vivo or in vitro EVs (p < 0.05) or was not enriched in either group (p ≥ 0.05). *p < 0.05, **p < 0.01, ***p < 0.001. n = 4 replicates per EV group.

https://doi.org/10.1371/journal.pone.0309716.g004

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Enrichment of in vivo and in vitro EV proteins using MISEV2023 EV content category 2.

Proteins identified were analyzed as described in the Fig 4 figure legend with EV category 2 proteins being enriched in vitro.

https://doi.org/10.1371/journal.pone.0309716.g005

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Enrichment of in vivo and in vitro EV proteins using MISEV2023 EV content category 3.

Proteins identified were analyzed as described in the Fig 4 figure legend with EV category 3 proteins, considered to be likely co-elution contaminants, primarily enriched in vivo.

https://doi.org/10.1371/journal.pone.0309716.g006

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Enrichment of in vivo and in vitro EV proteins using MISEV2023 EV content category 4.

Proteins identified were analyzed as described in the Fig 4 figure legend with EV category 4 proteins that may suggest EV subtypes were mostly enriched in vitro.

https://doi.org/10.1371/journal.pone.0309716.g007

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. Enrichment of in vivo and in vitro EV proteins using MISEV2023 EV content category 5.

Proteins identified were analyzed as described in the Fig 4 figure legend with EV category 5 proteins that may be used to assess possible EV function were mostly enriched in vitro.

https://doi.org/10.1371/journal.pone.0309716.g008

Fig 4 displays category 1 transmembrane and membrane bound proteins that are expected to be enriched in EVs. There were 52 proteins found to belong to category 1 expressed in the current EV data set. Overall, the majority of these proteins were significantly more abundant in the in vitro EVs as compared to the in vivo EVs (Fig 4). Notably, this included the tetraspanins Cd63 and Cd47; integrin proteins Itga1/3/5-7/m/v and Itgb1/2/5/8; and heterotrimeric G proteins Gnas, Gnai1/3, Gna11/l/z. Interestingly, the common, canonical EV proteins Cd9, Cd81, Cd82 and Bsg did not show significant enrichment in either EV type indicating equivalent relative expression.

Fig 5 displays category 2 cytosolic proteins that are expected to be enriched in EVs. There were 42 proteins found to belong to category 2. Consistent with category 1 proteins, category 2 proteins continued to display significant enrichment in the in vitro EVs relative to the in vivo EVs. Importantly, these proteins included the canonical EV proteins: endosomal sorting proteins Tsg101, Vps4a, Arrdc1, and Chmpa1a/1b2/2b/3; Flot1; Anxa1/7; heat shock proteins, Hspa8 and Hsp90ab1; and structural proteins Tubb1a/2a/3 among others.

Fig 6 displays category 3 co-eluted proteins that are expected to be contaminants or underrepresented in EVs, e.g., lipoproteins and ribosomal proteins. There were 79 category 3 likely contaminant proteins identified which were almost exclusively enriched in in vivo EV samples. Specifically, many components of the 40S small (Rps* proteins) and the 60S large (Rpl* proteins) subunits of the ribosome complex were robustly enriched in vivo. In regard to ribosomal proteins, only Rps12 and Rps21 were preferentially enriched in vitro. Of note, the plasma protein Alb was significantly enriched in vitro suggesting some contamination remaining from the serum containing media used for maintaining the cells in vitro before switching to serum free media for EV collection.

Fig 7 displays category 4 proteins that may be used to show EV subtype or cellular origin. A total of 13 proteins from this category were identified with 8 showing significant enrichment. However, the enrichment pattern was mixed between in vivo and in vitro (although in vitro still exhibited more proteins significantly enriched than in vivo). Both in vivo and in vitro EVs were enriched for nuclear proteins which suggests that these EVs may have a nuclear origin (H1.0 and H1.5/Lmna/Actn1/Actn4 respectively) or are carrying nuclear originating material. However, the endoplasmic reticulum protein calnexin (Canx) showed significant enrichment in vivo while the Golgi apparatus protein golgin A2 (GM130; Golga2) showed significant enrichment in in vitro suggesting additional origins for both EV types. Overall, the exact intracellular origin of the EVs from within the cell was indeterminant and not the focus of the current study as all EVs were collected from cells originating from the cells in the brain.

Fig 8 displays category 5 proteins which may represent functional signaling components of EVs. While only 9 proteins from this category were identified, 8 of these proteins were selectively enriched in the in vitro EVs. The proteins include those associated with growth factors (Pdgfrb and Mfge8), collagen production or the ECM (Col6a1, Col6a2, Col26a1, and Fn1), and pro-inflammatory cytokine release (Il34). Overall, the potential specific function of the EVs isolated here could not be determined and was not the focus of the current study.

In addition to the MISEV2023 method of categorizing EV proteins, we further compared the proteins identified in the present study (both in vivo and in vitro EVs) with the Vesiclepedia top 100 listed proteins [29], which are proteins most often reported in EV studies. Of the Top 100 listed proteins, we were able to identify 97 in our dataset (only Hist1h4a, Hla-a, and Tubb4b were not identified). In accordance with the results found for MISEV2023 category 1 and 2 proteins, most of the Top 100 commonly found EV proteins were either preferentially or significantly enriched in the in vitro EVs (Fig 9). Of all 100 proteins, only Atp1a1 and Anxa5 were significantly enriched in the in vivo EVs. This result further indicates that in vitro EVs are selectively enriched for EV proteins or are selectively depleted of non-EV contaminants relative to in vivo EVs isolated in a similar manner from a similar source.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 9. Enrichment of in vivo and in vitro EV proteins using Vesiclepedia top 100 list.

The top 100 most often identified EV proteins were extracted from the Vesiclepedia database [29] and compared for relative enrichment between in vivo and in vitro EVs. Of these proteins, 97/100 were found to be expressed in the overlapping experimental EV proteome and were used for subsequent analysis. Overall, the in vitro EVs showed the most significant enrichment of the proteins assessed with 41/97 showing significant enrichment while only 2/97 were significantly enriched in vivo. However, it is important to note that 54/97 showed no significant enrichment and that all 97 proteins listed were identified both in vivo and in vitro. This data is represented in the format of a heatmap where red represents relative enrichment, blue indicates relative depletion, and white indicates no enrichment or depletion for either in vivo or in vitro EVs (based on row Z-scores). A two-tailed Student’s t-test was used to determine if a protein was significantly enriched in either in vivo or in vitro EVs (p < 0.05) or was not enriched in either group (p ≥ 0.05). *p < 0.05, **p < 0.01, ***p < 0.001. n = 4 replicates per EV group.

https://doi.org/10.1371/journal.pone.0309716.g009

Characterizing the protein composition of EVs is important to study the functional impact of EVs [28, 37, 39, 40]. Isolating EVs from cell culture medium is a common approach to identify EV proteins considering many of the EV proteins were originally identified in culture systems. However, the extent to which culture conditions accurately reflect the physiological conditions of tissues with regard to EV protein composition remains unclear. Here, we present an in-depth analysis of the protein composition of brain EVs isolated from in vivo mouse forebrains and in vitro primary cells dissociated from the mouse forebrain to address this knowledge gap. We first characterized the size of isolated vesicles and found that the most prevalent size of vesicles from each source were similar, but not identical (131.5 vs 163.4, in vivo to in vitro). Further, most of the proteins (~86%), including 97 of the top 100 most commonly identified EV proteins, were expressed in both EV groups. However, a notable subset of proteins was exclusively identified in vitro and were composed mainly of cytoplasmic, ECM, and nuclear proteins. Contrastingly, very few proteins were found to be exclusively identified in vivo; however, those that were identified were enriched with endoplasmic reticulum proteins suggesting a source of contamination in in vivo EVs. Importantly, while the total proteome for both pools of EV were remarkably similar in terms of proteins expressed, the relative abundance of each protein between in vivo and in vitro EVs was vastly different. The in vitro EVs indicated a clear enrichment of canonical EV proteins. When the relative expression of all proteins was categorized into MISEV2023 category 1 and 2 (Figs 4 and 5), as well as the Vesiclepedia Top 100 proteins (Fig 9), in vitro EVs consistently showed enrichment for these proteins as compared to in vivo EVs. Further, in vitro EVs showed very robust depletion of MISEV2023 category 3 proteins, which represent co-eluting, non-EV contaminants, specifically indicating contamination from ribosomes in the in vivo EVs (Fig 6). Overall, the similarity of total proteome of both in vivo and in vitro EVs suggest that both EVs may be similar in function, but that they are not identical, especially as there is a clear difference in relative expression of canonical EV proteins, as well as probable sources of contamination between the 2 types of brain EVs. This should be considered when interpreting results between EVs collected from acute dissociation of intact tissues versus cell culture systems.

To isolate EVs from the brain in vivo, it is necessary to disrupt the tissue to facilitate release of EVs from the ECM that can insnare EVs released from the cells of origin. Our approach, commonly used in the field, involved gentle mechanical disruption to break the tissue into small pieces, followed by collagenase treatment. Collagenase is an enzyme commonly used to facilitate the dissociation of cells and EVs in larger tissues by cleaving collagen peptide bonds breaking down the ECM [2]. However, this gentle mechanical disruption can still cause cellular damage resulting in contamination of plasma membrane and intracellular organelles, which may explain why a large number of ribosomal contaminating proteins were enriched in vivo and not in vitro.

We used SEC combined with UF to enrich EVs isolated from brain tissue and cell culture medium. SEC is a gentle isolation method that reduces the risk of damage to the isolated EVs. It is important to note that each EV isolation method can affect the composition of the isolated EVs and co-isolated entities, which may vary depending on the sample type. Veerman et al., conducted a comparison of several EV isolation methods in cell culture and plasma samples [37]. In cell culture, they found that SEC yielded the highest number of EV associated MISEV2023 category 1 transmembrane/GPI proteins, but also co-isolated a greater amount of MISEV2023 category 3 non-EV proteins, especially lipoproteins. In human plasma, SEC isolated the fewest EV proteins but had a high presence of tetraspanins. These findings indicate that the efficiency of EV isolation can vary depending on the sample type as well as the isolation method, which may explain the differences in EV protein enrichment observed in vitro versus in vivo EVs in the current study.

While SEC is an increasingly common and viable method to rapidly isolate EVs from various sources, it is not, however, able to efficiently isolate EVs from other similarly sized particles, which can lead to contamination and interfere with downstream analyses [11]. For instance, EVs enriched using SEC may still contain significant amounts of other lipoprotein particles, especially when plasma or serum is the source, such as chylomicrons (100–600 nm) and very low-density lipoproteins (30–80 nm), as well as microvesicles (100–1000 nm) from other cellular compartments that are in the non-exclusion size range of the respective column [41–43]. In our study, we identified an enrichment of proteins associated with the nucleus and ER in the in vitro and in vivo EV samples suggesting that vesicles from these cellular compartments were co-isolated with the EVs. It is yet to be definitively determined whether these represent bona fide EV proteins or are sources of non-EV contaminants. Further, in the current study, UF was used not only to concentrate the large volumes produced by SEC, but to also deplete contaminating free proteins and very small vesicles (<10 nm). However, this method is ineffective at reducing contamination from similarly sized particles (>30 nm) that will co-elute using SEC. In the future, attempts should be made to further purify EVs from potential contaminating sources by adding an additional step to the SEC with UF (and likely any other) method used. For example, adding an immunoprecipitation step using antibodies that recognize the extracellular-facing EV-associated tetraspanin proteins CD9, CD63, and CD81 following SEC may substantially enrich EVs by further depleting contaminating proteins and similar sized particles. Any single-step (and possibly two-step) method is unlikely to adequately purify EVs for analysis of low expression proteins that may be masked by abundant contaminants.

Intracellular proteins with membrane binding ability are often co-isolates of EVs. An example of this is the Rab GTPase family which is a small GTP-binding protein that plays an important role in intracellular vesicle and EV trafficking [44]. Rab GTPases are involved in several steps of EV biogenesis and release, including trafficking of EVs through the endosomal system, release of EVs from the plasma membrane, and fusion of EVs with target cells [45]. Specific Rab GTPases are involved in EV biogenesis and trafficking, and the expression and activity levels of these Rab GTPases can alter the quantity and quality of EVs released. For example, Rab27a and Rab27b have been shown to be involved in the secretion of EVs from different cell types [46]. Interestingly, we only identified Rab27b in the current study of brain derived EVs. Rab11 has also been implicated in the trafficking of EVs from the recycling endosome to the plasma membrane for release [47]. Additionally, Rab35 has been shown to be involved in the biogenesis of EVs derived from multivesicular bodies (MVBs) [48]. In the current study, we detected a total of 29 Rab GTPase proteins common to both EV types. Most of these proteins were selectively enriched in in vitro EVs including: Rab- 3a, 3c,4a, 4b, 5b, 5c, 6a, 6b, 7a, 8a, 8b, 9a, 9b, 21, 22a, 23, 24, and 27b. Further, 7 Rab GTPase proteins were only identified in in vitro including: Rab- ac1, l6, 3 12, 32, 33b, and 40c. Not all Rab GTPases were enriched in either group, including: Rab- 1a, 1b, 2a, 5a, 10, 11b, 14, 18, and 35. Interestingly, no Rab GTPases were selectively enriched in vivo. Previous studies demonstrated that Rab- 4a, 5b, 5c, 7a, 22a, and 27b (expressed in vivo and in vitro, but enriched in vitro EVs) were enriched in small EVs (defined as vesicles enriched at 118,000 × g) compared to large EVs (defined as vesicles enriched at 16,500 × g) [48]. Both Rab1b and 18 (not enriched in vivo or in vitro in the current study) were found to be enriched in large EVs compared to small EVs. This suggests that EVs isolated from both sources likely have some large EV contamination but that the in vitro EVs are more representative of small EVs even though in vitro EVs were slightly larger in diameter on average.