Five de-identified matched brain tissue and serum samples were obtained from the Bartoli Brain Tumor Laboratory under a protocol approved by the IRB at Columbia University Irving Medical Center. The samples were collected intraoperatively from either epilepsy or glioma brain tumor patients and stored in liquid nitrogen until needed for processing (Table S1). These samples were analyzed for levels of miRNAs and environmental chemicals. Convenient unfiltered serum sample from an HIV-negative, male donor, in his 70s, was obtained from BioIVT (Hicksville, NY). The sample tested negative for several infections tested. This sample was used for metabolomic profiling and characterization of EVs.

Total EVs were isolated from 500 μL of serum using the ExoQuick Ultra kit (System Biosciences, Palo Alto, CA) following the manufacturer’s recommendation. This was performed for each downstream analysis (miRNA sequencing, environmental chemical analysis, and metabolomics) separately. Briefly, samples were centrifuged at 3,000 g for 15 min to remove cellular debris and 134 µL of ExoQuick precipitation solution was added to the supernatant. The mixture was inverted a few times and allowed to incubate on ice for 30min and then centrifuged at 3000 g for 10min at 4°C. After aspirating the supernatant, the EVs in the pellet were resuspended using 200 µL of buffer A. The purification column was pre-washed with buffer B, and then the sample along with 100 µL of buffer B were loaded onto the column. The sample on the column was mixed on a rotating shaker for 5 min, placed in a collection tube, and then centrifuged at 1000 g for 30s to collect the flow through containing purified EVs. Samples were stored at −20°C until needed for downstream analysis and characterization.

NEVs were isolated from 500 µL of serum using the ExoSORT isolation kit (NeuroDex, Natick, MA),^15,36 following manufacturer’s protocol. Like total EV isolations, this was performed for each downstream analysis (miRNA sequencing, environmental chemical analysis, and metabolomics) separately. Briefly, total EVs were precipitated using NeuroDex Total EV isolation reagent, reconstituted in binding buffer and incubated overnight at 4°C with magnetic beads decorated with neuron-specific antibodies.^15,36 After nEV capture, the beads were washed three times on a magnetic separator with NeuroDex Wash Buffer. Beads, with captured nEVs, were air-dried and either stored for environmental chemical and metabolomic analyses or removed from beads using NeuroDex Elution buffer followed by lysis and miRNA extraction for sequencing. Samples were stored at −20°C until needed for downstream analysis and/or characterization.

Tetraspanins are a protein superfamily that organize membrane microdomains and are highly enriched in EVs.³⁷ The tetraspanin profile of nEVs and total EVs was performed using a modified Luminex assay to determine levels of tetraspanins CD63, CD9, and CD81, which are considered pan-EV markers.³⁸ Levels of neuronal proteins synaptophysin (SYP) and Repulsive Guidance Molecule a (RGMa), and a common serum contaminant, human serum albumin (HSA), were measured using enzyme linked immune sorbent assays (ELISA) kits (R&D Systems DuoSet ELISA, cat. No. DY1455, and DY245905) following manufacturer guidelines.

The concentration and size distribution of nEVs and total EVs were determined using the ViewSizer 3000 (Horiba, MA) using methods previously described.³⁹ For all samples, 5 µL of isolated EVs were diluted at 1:1000 with 3 kDa filtered PBS and loaded into a cuvette for analysis. Twenty-five videos (30 frames per second, 300 frames per video) were recorded at 22°C with the following recording parameters: blue laser (445 nm), 210 mW; green laser (520 nm), 12 mW; red laser (635 nm), 8 mW; exposure, 15 ms; camera gain, 30 dB. Samples were processed with the Main Chart in “LogBinSilica” and integrated in the range [50, 1000] nm. The summary of size distribution and particle concentration were extracted. Two technical replicates were run, and their average was reported.

Transmission electron microscopy was used to visualize nEVs and total EVs following methods previously described.⁴⁰ Briefly, 5 μL of EV suspension was placed on a formvar/carbon-coated grid and allowed to settle for 60s. The samples were negative stained and blotted using 1.5% uranyl acetate in water and allowed to air dry at room temperature. Grids were imaged with a JEOL JSM 1400 (JEOL, USA, Ltd, Peabody, MA) transmission electron microscope, operating at 100 kV, and images were captured on a Veleta 2K × 2K CCD camera (Olympus-SIS, Munich, Germany). 10–15 representative images were captured of three randomly selected grid areas per sample at 50,000 and 100,000× lens magnification. Scale bars reflect the camera magnification.

MiRNAs were sequenced using the HTG EdgeSeq miRNA Whole Transcriptome Assay, which quantifies the expression of 2,083 human miRNAs (HTG Molecular Diagnostics, Inc., Tucson, AZ, USA). Brain tissue (5 mg), serum (30 µL), total EVs (30 µL), and nEVs (30 µL), were mixed with 30 µL of HTG biospecimen lysis buffer. The brain tissue was then processed in a bead beater to disrupt the tissue, and the supernatant was used for analysis. Three multi-tissue control lysates were included as internal technical replicates for quality control purposes. All three samples passed quality control criteria.

Brain tissue (weighing between 40 and 170 mg), 300 µL of serum, 300 µL of total EVs in buffer and nEVs on beads were analyzed for levels of 120 chemicals using gas chromatography coupled high-resolution mass spectrometry. These included a wide range of organic pollutants: polychlorinated biphenyls (PCBs), organochlorine pesticides, polyaromatic hydrocarbons (PAHs), dioxins, industrial solvents, etc. These chemicals were chosen based on availability of chemical standards and considered a reasonable number of chemicals that could be mixed. Brain tissue was processed in a bead beater with acetonitrile to extract the chemicals. Briefly, yttria stabilized zirconium oxide beads (∼10 beads, 0.5 mm diameter, Next Advance, Troy, NY) were added to the tissue and samples were placed in a bead beater (Next Advance Bullet Blender Storm, Troy, NY) set at speed 6 m/s for 5min. Serum and total EVs, and nEVs were extracted using a QuEChERS extraction.⁴¹ Briefly, samples were sonicated in 1 mL hexane: acetone: dichloromethane and transferred to a 2 mL QuEChERS containing 150 mg MgSO4 and 50 mg C18. The supernatant was transferred to a new vial and the QuEChERS tube was washed with 0.5 mL hexane: acetone: dichloromethane twice. The extract was evaporated under nitrogen to a final volume of 200–300 µl hexane using an Organomation 30 position Multivap Nitrogen Evaporator (Organomation Associates Inc.) and transferred to a glass autosampler vial. Sample extracts were analyzed using a high-resolution Thermo Q Exactive Orbitrap MS equipped with a Thermo Trace 1300 GC and a TriPlus RSH Autosampler (details are provided in Table S2).

Serum (100 µL), total EVs (in 150 µL buffer), and nEVs (on beads) were used to generate metabolomic profiles using liquid chromatography with high-resolution mass spectrometry. The method has been previously published.⁴² We included buffer B as a blank to filter non-sample signal from total EVs. In brief, the samples were extracted using acetonitrile with a mixture of stable isotopically labeled internal standards.^43,44 The extract was analyzed using two different analytical columns, HILIC and C18, each operated under positive and negative electrospray ionization (four assays total). An acetonitrile gradient with formic acid or 10 mM ammonium acetate adjusted to pH of 9.5 was used for HILIC positive and negative, respectively, an acetonitrile gradient with formic acid or 10 mM ammonium acetate was used for C18 positive and negative, respectively. For each analysis, 10 µL of the sample extract was injected on each column. Mass spectral data were generated on a Thermo Scientific HFX orbitrap mass spectrometer in full scan mode, scanning for mass range 85 to 1250 Da operated at a resolution of 120 000. All raw mass spectral data were extracted using the R packages apLCMS⁴⁵ and xMSanalyzer.⁴⁶ Through the analysis pipeline, four feature tables were generated: HILIC (+ ESI), HILIC (− ESI), C18 (+ ESI), and C18 (−ESI).

All analyses were performed using R (version 3.6.3) unless stated otherwise. The code used for the analysis can be found on the first author’s GitHub account at this link: https://github.com/vrindakalia/molecular_profile_nEVs.

Compared to total EVs, nEVs had detectable, but lower levels of tetraspanins CD9, CD63, and CD81 and higher levels of neuronal proteins SYP and RGMa (Figure 2A and B). nEVs had fewer particles (number of EVs) than total EVs and showed a similar size distribution (Figure 2C) although the median particle size was larger (nEV median = 173.48 nm and total EV = 114.76 nm, Figure 2D). Visual inspection of electron micrographs showed cup shaped vesicles in both total EVs and nEVs isolated (Figure 2E).

The top 30 statistically significant pathways, enriched by the top 100 most expressed miRNA targets in brain tissue, included several neurologically relevant pathways: pathways of neurodegeneration (multiple diseases), amyotrophic lateral sclerosis (ALS), Huntington’s disease (HD), Parkinson’s disease (PD), Alzheimer’s disease (AD), axon guidance, spinocerebellar ataxia, and neurotrophin signaling pathway (Figure 3A and Table S3). Among the enriched pathways, the mean proportion of miRNAs that target each pathway in the brain was most closely related with nEVs, followed by total EVs, and then serum (Table S3).

The correlation between the top 30 most expressed miRNA in brain tissue and their expression in other compartments showed a greater number of positive correlations between brain tissue and nEVs compared to other two compartments (Figure 3B). The expression of miR-221-3p in the brain was perfectly correlated with its expression in nEVs (Spearman’s correlation coefficient (rs) = 1, P = 0.02), positively correlated with its expression in serum (rs = 0.7, P = 0.23) and not correlated with its expression in total EVs (rs = 0.2, P = 0.78). Other miRNA with strong positive correlation (rs ≥ 0.8) between brain tissue and nEVs included: let-7g-5p, miR-29b-3p, miR-103a-3p, let-7b-5p, let-7a-5p, miR-128-3p, miR-24-3p, and miR-125b-5p. The expression of miR-125a-5p in brain tissue was positively correlated with its expression in serum (rs = 0.9, P = 0.08) and modestly correlated with its expression in both nEVs (rs = 0.4, P = 0.52) and total EVs (rs = 0.3, P = 0.68). The correlation coefficients and p-values for all other miRNA analyzed can be found in Table S4.

The miRNAs positively correlated between nEVs and brain tissue were enriched in several neurologically relevant pathways: pathways of neurodegeneration (multiple diseases), ALS, HD, PD, AD, axon guidance, spinocerebellar ataxia, and neurotrophin signaling pathway (Figure 3C). The miRNAs negatively correlated between nEVs and brain tissue were enriched in several neurologically relevant pathways: AD, pathways of neurodegeneration (multiple diseases), ALS, and axon guidance, some signaling pathways, and pathways related to viral infections (Figure 3C and Table S5).

Of the 120 chemicals tested, 12 were not detected in any sample, 60 were detected in at least one brain tissue sample tested, 64 were found in at least one nEV sample tested, 74 were found in at least one total EV sample tested, and 65 were found in at least one serum sample tested (Table S7). Twenty-eight chemicals were found in all four compartments of at least one person and 18 were detected in all four compartments in at least four individuals (Figure 4A). Diethyl phthalate, DEET, and 2,3,7,8-tetrachlorodibenzofuran (2,3,7,8-TCDF) showed the highest absolute concentrations in the brain (Table S8). Concentration of chemicals in brain tissue was consistently the lowest compared to the other three compartments (Table S9).

Looking at all chemicals, as a whole, the correlation between chemicals detected in the brain and their levels in the other three compartments was positive, with the strongest correlation between brain tissue and nEVs (rs = 0.72, P = 7.2e-16), followed by serum (rs = 0.7, P = 5.8e-15) and lowest for total EVs (rs = 0.58, P = 1.5e-09). For individual chemicals, the levels in brain tissue of the pollutants 2,3,7,8-TCDF (rs = 0.8, P = 0.13), polychlorinated biphenyl 49 (PCB 49, rs = 0.8, P = 0.13), and 1,4-dicholorobenzene (1,4-DCB, rs = 0.67, P = 0.22) were positively correlated with their levels in nEVs. The levels of phenanthrene (rs = 0.8, P = 0.13), benz[a]anthracene (rs = 0.7, P = 0.23), and DEET (rs = 0.7, P = 0.23) in the brain were positively correlated with their levels in serum. The correlation between brain tissue and total EVs for diethyl phthalate (rs = 0.8, P = 0.13), 2,3,7,8-TCDF (rs = 0.7, P = 0.23), and safrole (rs = 0.62, P = 0.27) was the strongest positive relationship. None of the individual chemical correlations were statistically significant (Figure S1, Table S10). The correlation of chemical concentrations between brain tissue and all compartments for all 18 chemicals can be found in Table S10. Higher density chemicals were less correlated between serum and brain tissue (rs = −0.51, P = 0.03), but density did not impact the chemical correlations between brain and nEVs or total EVs. Chemical octanol-water partition coefficient, average mass, and biodegradation half-lives were not correlated with correlation coefficients from any of the brain-compartment correlations (Table S11).

Of the annotated metabolites queried through RefMet, 888 (5%) were assigned a class membership. When comparing the super class members across compartments, 15 of the 17 classes were significantly different across all the compartments (Figure 5A). Organic oxygen and organohalogen compounds were not significantly different across all compartments. Post hoc analysis showed that all super classes had highest levels in serum. When comparing total EVs and nEVs, 5 super classes were not significantly different, these were: organic oxygen compounds, organohalogen compounds, prenol lipids, sterol lipids, and sphingolipids. All other classes had higher levels in total EVs compared to nEVs (Table S12).

Fatty acyls were the only main class enriched by the top 10% of most abundant metabolites in all three compartments (FDR < 0.05). Carboxylic acid derivatives and hydroxy acids and derivatives were enriched in both the serum and total EV compartments while steroids and steroid derivatives were enriched in both the total EV and nEV compartments (FDR < 0.05). Classes uniquely enriched to each compartment were imidazopyrimidines in serum, indoles and derivatives in total EVs, and pyrrolidines in nEVs. Other classes unique to each compartment but not statistically significant, included benzene and substituted derivatives, organic sulfuric acid and derivatives, and phenylpropanoic acids in serum; imidazothiozoles in total EVs; and prenol lipids, and thioureas in nEVs (Figure 5B and Table S13).

Testing for the difference in feature abundance between the three compartments revealed 445 metabolites significantly different between serum and nEVs, of which 31 had higher mean levels in nEVs (Figure 5C), and 276 significantly different metabolites between total EVs and nEVs, of which 39 had a higher mean in nEVs (Figure 5D). A total of 36 subclasses were represented by these metabolites with higher levels in nEVs (Figure 5E) and C24 bile acids had the highest number of members (n = 4) followed by primary amides (n = 2), amino sugars (n = 2) and unsaturated fatty acid (n = 2 when comparing nEVs and total EVs). The bile acids higher in both comparisons were: taurocholic acid (level 1), allocholic acid (level 1), and dehydrolithocholic acid (level 1). Chenodeoxycholic acid sulfate (level 3) was higher in nEVs compared to serum and beta-muricholic acid was higher in nEVs compared to total EVs. The primary amides were palmitamide (level 3) and oleamide (level 3); the amino sugars were glucosamine (level 3) and galactosamine (level 1); the unsaturated fatty acids were oleic acid (level 1, higher in both comparisons) and stearidonic acid (level 1, higher only between nEVs and EVs, Figure S3 and Table S14).