Sporadic non-familial ALS participants were recruited at the Pranger ALS Clinic at Michigan Medicine. Disease status was determined by an ALS neurologist based on revised El Escorial World Federation of Neurology criteria (Brooks et al., 2000), electromyography, and clinical and family history data. Age- and sex-matched control participants were recruited via a university-managed recruitment website and were eligible to participate if they: (1) did not have an ALS diagnosis or other neurodegenerative condition, and (2) did not have a family history of ALS in a first- or second-degree blood relative. Adult participants and/or next of kin provided informed consent. This research was approved by the University of Michigan Medical School Institutional Review Board (HUM00028826).

Whole blood samples from participants were drawn and collected in EDTA tubes, and serum was obtained as previously reported (Kanwar et al., 2014). Postmortem CNS tissue homogenates were obtained from approximately 100 mg of frozen frontal cortex and 300 mg of spinal cord as previously reported (Banigan et al., 2013) for participants with available tissue. Tissue homogenates were filtered through a 0.2 μm syringe filter (MilliporeSigma, Burlington, MA, United States). A 1:10 dilution was used to determine protein concentration using a BCA kit (Pierce Biotechnology, Waltham, MA, United States).

The ExoChip device was fabricated as previously reported (Kanwar et al., 2014). Briefly, a polydimethylsiloxane layer was bonded to glass substrate to form a microfluidic chamber with a height of 50 μm. The device was then chemically functionalized with NeutrAvidin (Thermo Fisher Scientific, Waltham, MA, United States) and stored at 4 °C until use. Biotinylated anti-CD63 (Ancell, Bayport, MN, United States) was added before EV isolation to target tetraspanin protein CD63, a common EV marker (Jankovicova et al., 2020).

EVs were isolated by immuno-affinity as previously reported (Kanwar et al., 2014). The saturation capacity of the ExoChip device was first determined using 300 μL of tissue homogenates with total protein concentrations ranging from 0.25 to 1 μg/μL for frontal cortex and 0.5–1.167 μg/μL for spinal cord, or volumes ranging from 200 to 450 μL for serum. This provided the guiding parameters for EV capture to assess EV number, which used sub-saturated conditions to avoid device overload, or to assess EV size, which used saturated conditions to ensure maximal EV capture. Frontal cortex or spinal cord homogenates or serum were infused through the ExoChip at a flow rate of 6 μL/min, followed by a rinse with tris buffered saline at a flow rate of 50 μL/min for 20 min. To measure the number of EVs, 300–350 μL of tissue homogenates with protein concentrations adjusted to 0.3 μg/μL, or 300–350 μL of serum, were processed through the device. To harvest the EV cargo, 300–350 μL of tissue homogenates with protein concentrations adjusted to 1 μg/μL, or 1 mL of serum, was used. The immobilized EVs were stained with membrane dye for on-chip quantification, fixed for scanning electron microscopy (SEM), or lysed for protein and miRNA extraction (see relevant sections below).

To quantify EV number, captured EV membrane-associated lipids were stained by flushing the chambers with the fluorescent carbocyanine dye, Vybrant^TM DiO (Molecular Probes, Eugene, OR, United States) at a flowrate of 20 μL/min for 10 min, followed by incubation with the DiO dye at 37°C for 20 min. Fluorescence intensities were measured at an excitation wavelength of 485 nm and an emission wavelength of 510 nm using a BioTek Synergy Neo multi-purpose plate reader (Winooski, VT, United States). Fluorescence intensities were normalized to the background.

To prepare captured EVs for SEM, standard SEM preparation (Kanwar et al., 2014) was performed with the following modifications. ExoChip immobilized EVs were fixed with 2% paraformaldehyde in phosphate-buffered saline (PBS) for 1 h at room temperature and rinsed with PBS. Next, the samples were serially dehydrated for 10 min in ethanol (30%, 50%, 70%, 95% in distilled water, and 100%) followed by a 10 min incubation in 1:1 ethanol:hexamethyldisilazane (HMDS), transferred to 100% HMDS, and air dried overnight. The samples were then gold- or carbon-coated for SEM imaging at 25,000× or 35,000× magnification with a FEI Nova 200 NanoLab Dual-Beam FIB-SEM microscope under low beam energies (2.0–5.0 kV) at the Michigan Center for Materials Characterization at the University of Michigan.

EV size and morphology were analyzed in SEM micrographs from 3 ALS and 3 control participants using MetaMorph image analysis software (version 7.7.7.0, Molecular Devices, San Jose, CA, United States). Briefly, the threshold feature was used to identify EVs. After all objects were isolated, EVs were verified using a filter for total EV area and shape factor (0 indicated a straight line and 1 indicated a perfect circle). Objects smaller than 2000 nm² and with a shape factor lower than 0.2 were discarded from the data set (Kanwar et al., 2014).

To lyse the captured EVs, the devices were flushed with RIPA buffer (Sigma, St. Louis, MO, United States) supplemented with cOmplete, EDTA-free protease inhibitor tablets (Roche, Basel, Switzerland) at a flowrate of 20 μL/min. The resulting protein lysates were analyzed by standard Western blotting procedures using 4–20% gradient acrylamide gels (Bio-Rad Laboratories, Hercules, CA, United States) and transferred to nitrocellulose membranes (Amersham, St. Louis, MO, United States) per our standard protocol (Kanwar et al., 2014). Immunoblotting was performed using rabbit anti-CD9 [Cell Signaling Technologies (CST), Danvers, MA, United States, cat# 13174] or anti-β-actin (CST, cat#4970) primary antibodies, followed by horseradish peroxidase-linked anti-rabbit IgG secondary antibody (CST, cat#7074). Proteins were visualized using Clarity ECL Western Blotting Substrates (Bio-Rad Laboratories) and a Bio-Rad ChemiDoc imager (Bio-Rad Laboratories).

ExoChip captured EVs were lysed and flushed with Qiazol (QIAGEN, Hilden, Germany) at a flowrate of 20 μL/min. Total RNA was extracted following a standard phenol/chloroform protocol and Norgen columns (Norgen Biotek Corp, Thorold, ON, Canada) as indicated by the manufacturer. The samples were eluted in 50 μL of DNAse- and RNAse-free water and stored at –80°C.

EV miRNAs were profiled with the NanoString nCounter^® Human v3 miRNA Expression Assay Kit (NanoString, Seattle, WA, United States) at the Genomics Shared Resource at the Ohio State University Comprehensive Cancer Center according to the manufacturer and as previously reported (Tank et al., 2018). The nSolver (Wohlfarth et al., 2017) analysis platform was used to assess the quality of the nCounter data and process the data for further analysis. Normalization was done using NanoString’s three ligation control probes. Bioconductor package sva was used to remove batch effects from combining five different assay kits (Leek et al., 2012). Principal component analysis was used to visualize variation in differentially expressed miRNAs (DEmiRNAs) across samples. Probes were excluded from further analysis if their average count across all samples fell below the geometric mean plus one standard deviation (geomean + 1SD) of all negative controls. Differential miRNA expression between control and ALS groups in each tissue was evaluated using a two-tailed t-test, as used in NanoString nSolver v4.0 analysis software. The significance cutoff for DEmiRNAs was a nominal P-value < 0.05.

DEmiRNAs identified in the geomean + 1SD analysis were entered into the pathway analysis. Briefly, DIANA-miRPath v3.0 (Vlachos et al., 2015) was employed to collect predicted gene targets of all significant DEmiRNAs and to characterize associated biological functions and pathways, focusing on the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways. Kyoto Encyclopedia of Genes and Genomes pathways with a false discovery rate (FDR) < 0.05 were considered significantly enriched among the DEmiRNAs from each tissue type (frontal cortex, spinal cord, and serum). To identify the overall theme of these enriched pathways, an association network for each tissue was generated based on the gene-content overlap among the KEGG pathways using richR, our in-house analysis R package,¹ and visualized in Cytoscape (McGregor et al., 2018). P-values were transformed by –log₁₀(P-value) and color-indexed to indicate significance levels; significance ranged from no significance (white) to highest significance (dark red). The inter-relationship among the significant pathways across three tissues was analyzed in Cytoscape, and highly interconnected pathway clusters were detected by MCODE (Bader and Hogue, 2003), a Cytoscape app for network cluster analysis.

Induced pluripotent stem cell (iPSC)-derived neurons (iNeurons) were established from control or sporadic ALS iPSC lines, as previously reported (Zhang et al., 2013; Tank et al., 2018; Weskamp et al., 2020). Briefly, iPSCs were plated on poly-D-lysine (50 μg/L, Sigma, cat# p1149)/laminin (1:100, Sigma, cat# L2020)-coated 6- well plates in iPSC media [E8 media (Gibco, cat# A1517001) supplemented with iROCK Y27632 (Fisher, cat# BDB562822)] at a density of 1 × 10⁵ cells/well. The following day (Day 1), the media was changed to iNeuron media #1 [E8 media supplemented with 1X N2 supplement (Gibco, cat# 17502-048), 1X NEAA supplement (Gibco, cat# 11140-050), 10 ng/mL BDNF (Peprotech, cat# 450-02), 10 ng/mL NT3 (Peprotech, cat# 450-03), 0.2 μg/mL mouse laminin (Sigma, cat# L2020), and 2 mg/mL doxycycline (Sigma, cat# D3447).] On Day 2, the culture media was changed to iNeuron media #2 [1/2 E8, 1/2 DMEM/F12 (Gibco, cat# 11320-033), 1X N2 supplement, 1x NEAA supplement, 10 ng/mL BDNF, 10 ng/mL NT3, 0.2 μg/mL laminin, 2 mg/mL Dox.] On Day 3, the culture media was changed to iNeuron media #3 [Neurobasal-A (Gibco, cat# 12349-015), 1x B27 supplement (Gibco, cat# 17504-044), 1x Glutamax supplement (Gibco, cat# 35050-061), 10 ng/mL BDNF, 10 ng/mL NT3, 0.2 μg/mL mouse laminin, 2 mg/mL doxycycline]. Additional media #3 was added on Day 6 and Day 8. The cells were differentiated for 10 days and maintained at 37°C and 5% CO₂.

Next, EVs were isolated from ALS participant spinal cord tissue by ultracentrifugation. Briefly, ALS spinal cord (200 mg) was homogenized in 2 mL cold PBS supplemented with cOmplete mini, EDTA-free protease inhibitors (Roche, cat# 11836170001). The homogenate was passed 10 times through an 18-gauge needle and then repeated with a 20-gauge needle. The sample was centrifuged at 10,000 g at 4°C for 30 min. The supernatant was passed through a 0.22 μm filter (Millipore, Carrigtowhill, Cork, Ireland) and centrifuged at 100,000 g at 4°C in an Optima L-90K Ultracentrifuge (Beckman Coulter) for 3 h. The pelleted EVs were resuspended in 10 mL of 1X PBS and centrifuged at 100,000 g for 3 h at 4°C. EVs were resuspended in 500 μL 1X PBS with protease inhibitors. EV protein concentration was determined by BCA assay (Pierce Biotechnologies) and nanoparticle tracking analysis using a NanoSight NS300 (Malvern, United Kingdom).

Isolated EVs were incubated with control iNeurons for 8 h by replacing iNeuron media #3 with ∼5.3 × 10⁸ centrifugation extracted spinal cord EV particles resuspended in iNeuron media #3. Cells were then washed with 1X PBS and lysed in QIAzol (Qiagen, cat# 79306) prior to total RNA extraction by phenol/chloroform. The upper layer was passed through a miRNeasy Mini Kit column (Qiagen) and DNase treated using an RNAse-Free DNase Set (Qiagen, cat#79254). cDNA was generated using 1 μg of total RNA and the iScript cDNA synthesis kit (BioRad, cat# 1708840). Quantitative real-time PCR was performed in triplicate using TaqMan sequence specific probes for human N-deacetylase and N-sulfotransferase 4 (NDST4) and YWHAZ, TaqMan^TM 2X gene expression Master Mix (Applied Biosystems/Thermo Fisher Scientific, cat#4369016), and 2 μL of iNeuron cDNA in an Applied Biosystems StepOne^TM RT-PCR system thermocycler. NDST4 was selected by cross-referencing our EV DEmiRNA predicted targets with gene expression data from a previous analysis of ALS spinal cord (Figueroa-Romero et al., 2012). C_T values were used to calculate ΔC_T and ΔΔC_T using YHWAZ as the internal reference. Data were expressed as the mean of the relative quantity of gene expression (2^−ΔΔCT).

Fisher’s exact test and Student’s t-test were used to determine significant demographic differences between ALS cases and controls. Three linear mixed effects models with random participant intercepts were used to determine differences in EV area between ALS participants and controls in the frontal cortex, spinal cord, and serum. The mixed effects models with random participant-specific intercepts identifies differences in EV area between ALS participants and controls, while accounting for the correlation of repeated EV area measurements within participants. The mixed effects models were fit using the lmerTest package in R software (v.3.5.2²) and model parameter estimates were determined using the maximum likelihood method (Kuznetsova et al., 2017). T-tests calculated using Satterthwaite’s degrees of freedom method were evaluated to assess differences in EV size between ALS participants and controls in the three regions (Giesbrecht and Burns, 1985). Statistically significant fold-change differences in gene expression between ALS and control iNeurons were calculated using a Student’s t-test. Statistical analyses were done using Prism (version 7 or 9.1.0; GraphPad Software, La Jolla, CA, United States) and R statistical computing software [v4.0.3 (see text footnote 2)]. Cumulative plots were generated using R software (v.4.0.1). Significance was defined as P-value < 0.05, unless noted otherwise.

A total of 15 ALS and 16 control participants provided frontal cortex, spinal cord, and/or serum (Table 1 and Supplementary Table 1). All three tissues were collected from 11 of the ALS participants. None of the ALS participants exhibited any known familial history, Cu²⁺/Zn²⁺ superoxide dismutase (SOD1) gene mutations, or pathogenic C9orf72 gene expansions. Age distribution and sex were similar between ALS and control participants. Fourteen ALS participants exhibited limb onset disease (93%), while one participant presented with bulbar onset (7%). All ALS participants were Caucasian (100%), while the control group consisted of 15 Caucasians (94%) and one Black/African American (6%).

Using our refined ExoChip coated with CD63 antibody, we captured EVs and their cargo from frozen postmortem frontal cortex and spinal cord and from serum from ALS and control participants (Figure 1A). Saturation experiments indicated that 150 μg of frontal cortex, 300 μg of spinal cord, and 400 μL of serum resulted in device saturation (Figure 1B). These results informed the subsequent EV capture paradigms used to determine EV numbers, which required unsaturated conditions, and for cargo determination, where we used saturated conditions (see below).

Immunoblot analysis indicated that ExoChip-captured EVs were positive for the EV marker CD9 (Figure 2A). Assessment of ExoChip-captured EV size and morphology using SEM revealed spherical EVs from anti-CD63 coated ExoChips, but not from uncoated or PBS control ExoChips, suggesting specific vesicle capture (Figure 2B). The overall diameter of EVs captured from frontal cortex ranged from 5.64 to 447.71 nm for ALS and 7.97–480.15 nm for controls, while EV diameter from spinal cord ranged from 5.64 to 448.17 nm for ALS and 62.46–317.58 nm for controls, and EV diameter from serum ranged from 56.39 to 392.25 μM for ALS and 9.77–636.03 nm for controls. The mean diameter of EVs from frontal cortex was 255.07 ± 5.25 nm for ALS and 245.86 ± 4.45 nm for control; from spinal cord 193.92 ± 3.42 nm for ALS and 141.94 ± 2.59 nm for control; and from serum 187.12 ± 2.74 nm for ALS and 247.33 ± 2.96 nm for control. Size-frequency profiles, cumulative distribution plots, and 3 separate linear mixed effects models with random participant intercepts identified significantly larger EVs in spinal cord from ALS participants versus controls (P = 0.003). This trend was reversed for serum, in which the size of ALS EVs was smaller than controls (P = 0.02) (Figures 2D,E). We did not detect significant differences in EV size between ALS participants and controls in frontal cortex (P = 0.59). Extracellular vesicle number quantification by measuring membrane-associated lipids by fluorescence intensity revealed no significant difference in EV number between ALS and controls across all samples (Figure 2C).

Mature miRNAs contained in ExoChip-captured EVs from frontal cortex, spinal cord, and serum were profiled by NanoString. Principal component analysis indicated that EV-derived miRNAs cluster according to tissue origin (Figure 3A). Using a geomean + 1SD cutoff as negative background and adjusting for batch effect, 33 significant DEmiRNAs (P < 0.05) were identified across the three tissues in ALS versus controls. We observed an overall decrease in EV-derived miRNA levels in ALS frontal cortex and spinal cord compared to controls, while the opposite was observed in serum (Figure 3B and Supplementary Table 2). In ALS frontal cortex, out of 17 dysregulated miRNAs, 3 increased and 14 decreased in level. In spinal cord, 8 miRNAs were dysregulated, with 2 increased and 6 decreased in ALS. In serum, 16 miRNAs were dysregulated, with 11 increased and 5 decreased in ALS. miRNA overlap across all three tissues (Figure 3C) revealed that miR-342-3p was increased and miR-1254 was decreased in ALS. Additionally, miR-4443 dysregulation overlapped between frontal cortex and spinal cord, miR-587 overlapped between frontal cortex and serum, and miR-298 and miR-450a-2-3p overlapped between spinal cord and serum.

To better understand the role of dysregulated EV miRNAs in ALS pathology, we examined the functional KEGG pathways, which correspond to the predicted DEmiRNA targets for each tissue using DIANA-mirPath analysis (Vlachos et al., 2015). A total of 54 KEGG pathways were significantly enriched (FDR < 0.05) (Table 2). Of those pathways, 4 overlapped across all three tissues, and an additional 12 overlapped between two tissues, including axon guidance, long-term potentiation, transforming growth factor-beta (TGF-β)/mitogen-activated protein kinase (MAPK) signaling, and various cancer-associated pathways. The majority of the additional enriched pathways were in frontal cortex, where the hippo signaling pathway, Wnt signaling, and prion diseases were the most significantly enriched. Only 2 pathways were significantly enriched in spinal cord, and 4 pathways in serum.

Connectedness across the identified pathways was then evaluated by the MCODE Cytoscape network analysis app. Connection between nodes (edges) denotes the degree of shared genes between two connected pathways. We identified five pathway clusters with highly interconnected common functional pathways (Figure 4): (1) neurotransmitter systems, such as “glutamatergic synapses” and “GABAergic synapses;” (2) brain function, including “long-term potentiation” and “oxytocin signaling;” (3) intracellular signaling cascades, such as “Ras signaling,” “neurotrophin signaling,” and “ErB signaling;” (4) cancer related, such as “prostate cancer,” “pancreatic cancer,” chronic myeloid leukemia;” and (5) stem cell renewal, including “hippo signaling,” TGF-beta signaling,” and “Wnt signaling.”

To assess potential biological relevance, we compared the predicted target mRNAs of the enriched KEGG pathways in each tissue (Figure 5A and Supplementary Tables 3–5). Transcripts among the most common predicted target genes from the enriched pathways were: serine-threonine kinases, including protein kinase C α/β (PRKCα/β), MAPK1, and AKT3; phosphoinositide 3-kinase (PI3K) subunits, including PIK3CB and PIK3R5; and Ras family GTPases, including NRAS and KRAS.

Since miRNAs regulate mRNA steady-state levels by destabilizing transcripts, we next assessed expression of NDST4, a predicted target of miR-520f-3p (increased in ALS serum EVs, Figure 3B), in an in vitro ALS model. Quantitative real-time PCR of ALS and control participant-derived iNeurons (Weskamp et al., 2020) confirmed that NDST4 gene expression trends to decrease 0.80-fold in ALS versus control iNeurons. This was similar to reduced NDST4 levels in postmortem ALS spinal cord tissue in a previous study (Figueroa-Romero et al., 2012). Additionally, incubating control iNeurons with ALS spinal cord EVs, size verified by NanoSight (Figure 5C), significantly reduced NDST4 expression 0.63-fold (Figure 5B) versus control iNeurons not incubated with ALS EVs.