Spinal cord injury alters microRNA and CD81+ exosome levels in plasma extracellular nanoparticles with neuroinflammatory potential

Abstract

Extracellular vesicles (EVs) have been implicated mechanistically in the pathobiology of neurodegenerative disorders, including central nervous system injury. However, the role of EVs in spinal cord injury (SCI) has received limited attention to date. Moreover, technical limitations related to EV isolation and characterization methods can lead to misleading or contradictory findings. Here, we examined changes in plasma EVs after mouse SCI at multiple timepoints (1d, 3d, 7d, 14d) using complementary measurement techniques. Plasma EVs isolated by ultracentrifugation (UC) were decreased at 1d post-injury, as shown by nanoparticle tracking analysis (NTA) and paralleled an overall reduction in total plasma extracellular nanoparticles. Western blot (WB) analysis of UC-derived plasma EVs revealed increased expression of the tetraspanin exosome marker, CD81, between 1d and 7d post-injury. To substantiate these findings, we performed interferometric and fluorescence imaging of single, tetraspanin EVs captured directly from plasma with ExoView^®. Consistent with WB, we observed significantly increased plasma CD81+ EV count and cargo at 1d post-injury. The majority of these tetraspanin EVs were smaller than 50nm based on interferometry and were insufficiently resolved by flow cytometry-based detection. At the injury site, there was enhanced expression of EV biogenesis proteins that were also detected in EVs directly isolated from spinal cord tissue by WB. Surface expression of tetraspanins CD9 and CD63 increased in multiple cell types at the injury site; however, astrocyte CD81 expression uniquely decreased, as demonstrated by flow cytometry. UC-isolated plasma EV microRNA cargo was also significantly altered at 1d post-injury, with changes similar to that reported in EVs released by astrocytes after inflammatory stimulation. When injected into the lateral ventricle, plasma EVs from SCI mice increased both pro- and anti-inflammatory gene as well as reactive astrocyte gene expression in brain cortex. These studies provide the first detailed characterization of plasma EV dynamics after SCI and suggest that plasma EVs may be involved in posttraumatic brain inflammation.

1. Introduction

In addition to motor and sensory deficits, spinal cord injury (SCI) causes systemic changes — including pulmonary, immune, and gut dysfunction — that can increase subsequent morbidity and mortality (Benevento & Sipski, 2002; Winslow & Rozovsky, 2003; Sezer et al., 2015; Brommer et al., 2016). Experimental SCI induces chronic neuroinflammation and neurodegeneration in brain regions linked to memory, emotions, and pain regulation (Zhao et al., 2007; Wu, Stoica, et al., 2014; Wu, Zhao, et al., 2014; Luedtke et al., 2014; Maldonado-Bouchard et al., 2016). Importantly, these pathological findings in preclinical models demonstrate that secondary injury processes extend far beyond the injury site and may explain the cognitive deficits and affective changes observed in patients with SCI (Arango-Lasprilla et al., 2011; Shin et al., 2012; Craig et al., 2017; S.-W. Huang et al., 2017). The mechanisms underlying such SCI-mediated brain dysfunction are unclear and require further investigation.

Blood-borne factors play an important role in SCI pathology and contribute to its systemic effects. Whereas plasma cytokine and hormone changes have been elucidated after SCI (Wang et al., 1992; Segal et al., 1997; Bauman & Spungen, 2000), another potential mechanism that contributes changes distant from the lesion site involves the transfer of extracellular vesicles (EVs) through the circulation. EVs can be found in most biological fluids in vivo and may serve as either potential disease biomarkers or mechanisms for disease progression (Quinn et al., 2015; Shah et al., 2018). More recent studies have started to address the role of circulating EVs in long-distance signaling between organ systems, including bidirectional crosstalk between the central nervous system (CNS) and periphery (Ridder et al., 2014; Dickens et al., 2017; Kur et al., 2020). In experimental models of CNS inflammation and injury, EVs have been implicated in the transfer of inflammatory cargo such as cytokines and microRNAs (miRs) that may contribute to the spread of inflammation at a distance (Dickens et al., 2017; Kumar et al., 2017; Kerr et al., 2018). It is well known that local and systemic inflammation has a major impact on SCI progression and outcomes (Sun et al., 2016). However, the role of EVs in the pathobiology of SCI has not been well examined.

There are several challenges in EV research that affect the interpretation and reproducibility of the observations reported, as detailed in recent MISEV2018 guidelines (Théry et al., 2018). Among these are the various EV isolation methods and technologies used for their characterization. Protocols for EV isolation may preferentially select for only certain types of EVs and include different degrees of non-EV contamination (Théry et al., 2018). Technologies for characterization have different capabilities for sizing, counting, phenotyping and visualizing EVs (Rupert et al., 2017). Because of numerous method-dependent factors, multiple working groups have highlighted the urgent need for standardized reporting in EV research as well as the use of complementary approaches for within-study validation (Van Deun et al., 2017; Théry et al., 2018; Welsh et al., 2020). Light scattering methods, including nanoparticle tracking analysis (NTA) and flow cytometry (FC), are still among the most commonly used techniques for EV detection and have a long history in the field (Gardiner et al., 2016; Hartjes et al., 2019). Based on comparative studies with high-resolution microscopy, it is increasingly recognized though that these methods have limited detection capabilities for small EV populations due to the low intrinsic scattering power of biologic materials (Nizamudeen et al., 2018; E. van der Pol et al., 2014). In recent years, new sub-diffraction-limited methods with multiplex phenotyping capability have been developed for high-throughput, single EV analyses, including surface plasmon resonance imaging (Im et al., 2014; Yang et al., 2018) and single particle interferometric reflectance imaging sensor (SP-IRIS) (Daaboul et al., 2016). As these methods rely on affinity-capture, they can directly detect EVs from samples and avoid certain complications and biases associated with sample processing (Van Deun et al., 2017; Théry et al., 2018).

In the present study, we provide the first detailed characterization of plasma EVs in a mouse model of SCI. Multiple techniques were used in parallel to count, size and phenotype EVs, including ExoView^®, a novel detection platform based on SP-IRIS. Significant changes in plasma EV number and biological content were found at 1d post-injury. Moreover, when injected into the brain ventricular system these plasma EVs induced inflammatory effects. This paper presents new information about the temporal dynamics of plasma EVs after SCI and underscores how measurement approaches can impact interpretation of the results of EV studies.

2. Materials and Methods

2.1. Animals and Spinal Cord Injury Model

All experiments were performed using young, adult (10-12 weeks) male C57BL/6 mice from Jackson Laboratories. Mice were housed in a 12:12 light:dark cycle with food and water provided ad libitum. All study procedures were performed according to experimental protocols approved by the University of Maryland School of Medicine Institutional Animal Care and Use Committee (IACUC).

Moderate/severe spinal cord contusion injury was performed using the Infinite Horizon spinal cord impactor (Precision Systems and Instrumentation) as previously detailed (Wu, Zhao, et al., 2014). In brief, mice underwent anesthesia induction at 3% and maintenance at 1.5% with isoflurane throughout the procedure. The spinal column was stabilized with bilateral steel clamps over the T9 and T11 lateral processes and a midline spinal contusion injury performed over the exposed T10 level with a force of 70 kilodynes. Naive animals underwent isoflurane anesthesia for a similar time duration but did not receive laminectomy or spinal cord impact. The bladders of injured mice were manually voided at least twice daily each day post-injury.

2.2. Overview of Experimental Design and Animal Use

To minimize animal usage and maximize information gain, SCI experiments were largely designed to utilize both blood and spinal cord tissue from the same animals. Blood was immediately processed to platelet free plasma (PFP; detailed in 2.3) and used for EV analysis directly from PFP or after ultracentrifugation (UC) isolation, miR profiling of UC-isolated plasma EVs, or intracerebroventricular (ICV) injection experiments of UC-isolated plasma EVs. Spinal cord tissue surrounding the lesion site was isolated and used either fresh for flow cytometry or flash frozen on dry ice prior to tissue EV isolation or whole protein lysate analysis. The specific breakdown by study/cohort is found below.

2.2.1. Study 1

To assess changes in plasma EVs after SCI, blood was collected from an initial cohort of injured animals and time-matched naive controls at 1d (n=4-6/group), 3d (n=5-6/group), 7d (n=5-6/group), and 14d post-injury (n=6/group). As illustrated in Figure 1, EVs were analyzed from separate PFP aliquots of this cohort after UC isolation (detailed in 2.4) or directly from PFP by multiple methods including NTA (detailed in 2.5), Western blot (WB; detailed in 2.7), SP-IRIS (detailed in 2.8), and FC (detailed in 2.9). Frozen spinal cord tissue from this cohort was used for tissue EV isolation (detailed in 2.6).

Figure 1. Schematic diagram of experimental design for plasma EV characterization after SCI.

(Top) Blood was collected from mice by cardiac puncture at various timepoints after SCI and immediately processed to generate PFP aliquots that were stored at −80°C. (Middle) Prior to analysis, PFP was rapidly thawed in a water bath at 37°C. EVs were analyzed after UC isolation or directly from PFP by various techniques. (Bottom) NTA and Western Blot for tetraspanin proteins (i.e. CD81) were performed on UC-isolated EVs. FC and SP-IRIS both used antibody labeling methods to detect tetraspanin-positive EVs (i.e. CD9, C63, CD81) directly from PFP.

A second cohort of animals (n=6/group; 2 groups: naive, 1d SCI) was also included for plasma EV analysis by NTA and WB, but not SP-IRIS or FC. PFP aliquots from this cohort were also used for whole plasma NTA and for WB assessment of non-EV contaminants during UC isolation including lipoprotein markers and albumin (detailed in 2.7). Frozen spinal cord tissue from this cohort was used for whole protein lysate of EV biogenesis-related proteins at 1d post-injury (detailed in 2.7).

2.2.2. Study 2

To analyze miR content of UC-isolated total plasma EVs (detailed in 2.10), blood was collected from 24 mice (n=8/group; 3 groups: naive, 1d SCI, 3d SCI). Spinal cord tissue from 16 of these mice (n=8/group; 2 groups: naïve, 1d SCI) were also collected and used fresh for flow cytometry analysis of cell-specific tetraspanin protein expression at the injury site (detailed in 2.11).

2.2.3. Study 3

To assess the potential inflammatory effects of plasma EVs after SCI on the brain, ICV injections of UC-isolated total plasma EVs were performed (detailed in 2.12).

In the first cohort, blood was collected from 16 mice (n=8/group; 2 groups: naive, 1d SCI) and isolated plasma EVs were injected individually into 16 naive mice (n=8/group; 2 groups: naive EVs, 1d SCI EVs). Total RNA from dissected cortex tissue of injected animals was assessed for expression of inflammatory genes (detailed in 2.13).

In the second cohort, utilizing a different timepoint after SCI, blood was collected from 12 mice (n=6/group; 2 groups: naive, 3d SCI) and isolated plasma EVs were injected individually into 12 naive mice (n=6/group; 2 groups: naive EVs, 3d SCI EVs). Total RNA from dissected cortex tissue of injected animals was assessed for expression of inflammatory genes (detailed in 2.13). Additionally, from the group of injured animals in this cohort, spinal cord tissue was also collected for whole protein lysate of EV biogenesis-related proteins at 3d post-injury (detailed in 2.7).

In the third and final cohort, blood was collected from 10 mice (n=5/group; 2 groups: naive, 1d SCI) and isolated plasma EVs were injected individually into 10 naive mice (n=5/group; 2 groups: naive EVs, 1d SCI EVs) for subsequent brain flow cytometry analysis of cell-specific intracellular cytokine levels (detailed in 2.14).

2.3. Blood Collection and Processing

Blood was collected by syringe from each animal through terminal cardiac puncture under isoflurane anesthesia and immediately placed into precoated EDTA tubes (Cat# 365974, BD Biosciences) and gently inverted ten times for proper mixing (Witwer et al., 2013). Blood was then centrifuged at room temperature (RT) for 500g for 15 min, 2500g for 10 min, and 2500g for 10 min to generate PFP (Lacroix et al., 2013). Blood was kept at RT between collection and centrifugation for no more than 30 minutes to minimize release of platelet extracellular vesicles that occurs under cold temperatures, agitation, and prolonged storage of blood (Ayers et al., 2011; Bæk et al., 2016; Lacroix et al., 2012). PFP was aliquoted into multiple tubes, flash frozen on dry ice, and stored at −80°C. Prior to any analysis, a specific aliquot was thawed in a water bath at 37°C to avoid multiple freeze-thaw cycles, which can affect extracellular vesicle recovery and integrity (Trummer et al., 2009; Jayachandran et al., 2012). A schematic illustration is shown in Figure 1.

2.4. Plasma EV Isolation

EVs were isolated from PFP by a standard ultracentrifugation-based protocol (Kowal et al., 2016; Théry et al., 2006). To isolate “total plasma EVs”, a 100 μL aliquot of PFP was diluted to 4 mL volume in filtered PBS (f-PBS) in a thinwall, polypropylene tube (Cat# 326819, Beckman Coulter). PBS was filtered through a 0.22 μm PVDF filter (Cat# SLGV033RS, Millipore Sigma) prior to all EV isolation and characterization experiments. The sample was then spun at 100,000g for 90 min at 4°C using a SW55Ti swinging bucket rotor in an Optima XE-90 Ultracentrifuge (Beckman Coulter). The supernatant was carefully removed, and the pellet was resuspended in 60 μL f-PBS for downstream analysis by NTA and WB. In a separate procedure using two-step differential ultracentrifugation, a 100 μL aliquot of PFP was diluted as described above and spun first at 10,000g for 40 min at 4°C with the SW55Ti to isolate a “large EV’ pellet (resuspended in 100 μL f-PBS). Then, the supernatant was transferred into a new thinwall, polypropylene tube and spun at 100,000g for 90 min at 4°C with the SW55Ti to isolate a “small EV’ pellet (resuspended in 100 μL f-PBS). A schematic illustration is shown in Figure 1.

2.5. Nanoparticle Tracking Analysis

NTA was performed with the ViewSizer^® 3000 (HORIBA Scientific) fitted with the blue (44 5nm), green (520 nm), and red (635 nm) lasers set at 210 mW, 12 mW, and 8 mW, respectively. Camera gain was set at 30 dB, frame rate at 30 frames per second, and exposure at 15 ms during video recording. EV samples isolated from plasma (or whole PFP directly) were diluted in f-PBS to an appropriate concentration within the linear range of the instrument (between 1x10⁷ – 2x10⁸ particles/mL). The final EV concentration data is presented as “particles/mL plasma” to reflect the original concentration in the plasma, which was calculated by accounting for the dilution factor for NTA, the resuspension volume after isolation, and the original input plasma volume for isolation. Particle counts were integrated from 50nm-2000nm in size, which covers the size range for which particle Brownian motion is relevant for NTA and largely the expected size range of various EV subtypes. The lower size limit was determined by the NTA theoretical limit of detection (LOD) for biological materials, and the upper size limit was determined by the lack of significant Brownian motion for particles greater than a few micrometers (Dragovic et al., 2011). At least 50 videos were recorded to generate enough particle counts (>3000 particles) for an accurate size distribution. Sample was automatically stirred for five seconds with a magnetic stir bar between each video for proper mixing. For assessment of instrument performance, a commercially available bead mixture (Cat# 1493, Apogee Flow Systems) containing six non-fluorescent silica beads and two fluorescent polystyrene beads ranging from 100 nm to 1300 nm in diameter was evaluated as shown in Supplementary Figure 1. Video processing settings for analysis were set as follows: Detection threshold type: Polydisperse; Detection threshold: 2; AutoThreshold: Disabled: Feature radius: 30; Drift correction: 0.

2.6. EV Isolation from Spinal Cord Tissue

After blood collection, mice were perfused with 40 mL normal saline. Approximately 1 cm of tissue around the epicenter of the lesion site (or equivalent length at the same spinal level in naive animals) was carefully dissected and flash frozen on dry ice prior to storage at −80°C. For EV isolation from frozen spinal cord, we followed a modified protocol previously described for human brain EVs (Vella et al., 2017). At the time of isolation, tissue was weighed and carefully cut in half with dissecting scissors. The tissue was then placed in a 50 mL conical tube containing type III collagenase (Cat# LS004176; 40U/mL; Worthington Biochemical) solution in Hibernate™-E medium (Cat# A1247601, Thermo Fisher Scientific) at a ratio of 8μL/mg tissue weight. The sample was placed on an orbital shaker at 37°C. After 15 minutes, the sample was pipetted up and down twice carefully with a 25 mL Stripette and then placed back onto the shaker for another five minutes. Upon completion, a 2x concentrated solution of protease (Cat# 11697498001, Millipore Sigma) and phosphatase (Cat# 4906837001, Millipore Sigma) inhibitors was added to each sample at a volume equal to that previously calculated for the enzyme solution. The sample was then serially centrifuged at 4°C for 5 min at 300g, 10 min at 2000g, and 30 min at 10,000g, transferring the complete supernatant between steps. Then, an equal volume of the remaining supernatant after the 10,000g spin was diluted to 4 mL volume in f-PBS into a thinwall, polypropylene tube and spun at 100,000g for 70 min at 4°C with the SW55Ti. The supernatant was removed carefully with a pipette, and the tissue EV pellet was resuspended in 50 μL f-PBS.

For initial characterization of spinal cord tissue EVs, we followed the protocol above with an additional purification step on a triple sucrose cushion gradient prior to ultracentrifugation as described previously (schematic diagram shown in Supplementary Figure 6A) (Vella et al., 2017). After the 10,000g spin, the supernatant was diluted to 1 mL in f-PBS and transferred on top of a layered sucrose gradient in a thinwall, polypropylene tube containing 1.2 mL of 2.5M sucrose, 1 mL of 1.3M sucrose, and 1 mL of 0.6M sucrose. The sample was then spun at 180,000g for three hours at 4°C in the SW55Ti. 1 mL fractions (designated F1, F2, and F3) were carefully collected from the top of the tube, corresponding with the original sample and gradient placement (F1: sample; F2: 0.6M sucrose; F3: 1.3M sucrose). The density of the fraction was measured by weighing the mass of the fraction volume using a high precision balance ME103TE/00 (Mettler Toledo). These fractions were diluted to 4 mL volume in f-PBS into a thinwall, polypropylene tube and spun at 100,000g for 70 min at 4°C with the SW55Ti. The final pellets were resuspended in f-PBS.

2.7. Protein Extraction and Western Blot Analysis

For injury site analysis of EV biogenesis-related proteins, approximately 1 cm of tissue surrounding the epicenter of the lesion site was homogenized in RIPA buffer (Cat# R0278, Millipore Sigma) supplemented with protease (Cat# P8340, Millipore Sigma) and phosphatase (Cat# P5726, Cat# P0044, Millipore Sigma) inhibitors. Protein concentration was measured by Pierce™ bicinchoninic acid (BCA) protein assay (Cat# 23225, Thermo Fisher Scientific). For detection of proteins in plasma or tissue EV samples, EV samples in PBS were prepared by ultracentrifugation as described above (see 2.4 and 2.6).

For different Western blot experiments, either 30 μg of tissue lysate protein, equal volumes of resuspended plasma EV sample pellets, equal volumes of resuspended tissue EV pellets, or 1 μL of PFP were loaded onto 4-15% Criterion™ TGX Stain-Free™ Precast gels (Cat# 5678083, Bio-Rad) and transferred onto nitrocellulose membranes (Cat# 1704159, Bio-Rad). Membranes were blocked with 5% nonfat milk in PBS containing 0.1% Tween 20 (PBS-T) for one hour at RT, and then incubated in primary antibodies overnight at 4°C. The next day, membranes were washed three times with PBS-T and incubated in species-specific, horseradish peroxidase (HRP)-conjugated secondary antibodies for one hour at RT. Membranes were washed three times with PBS-T prior to the addition of chemiluminescence substrate (Cat# 37071, Thermo Fisher Scientific). Chemiluminescent protein detection was visualized with the ChemiDoc™ MP Imaging System (Bio-Rad), and protein bands were quantified by densitometry analysis by Image Lab™ software Version 6.0.1 (Bio-Rad).

The following primary antibodies were used: anti-CD81 (Cat# 10037S, 1:1000; Cell Signaling Technology), anti-LAMP-1 (Cat# 1D4B, 1:1000; Developmental Studies Hybridoma Bank), anti-Flotillin-1 (Cat# 3253S, 1:1000; Cell Signaling Technology), anti-Alix (Cat# 2171S, 1:1000; Cell Signaling Technology), anti-TSG101 (Cat# ab125011, 1:1000; Abcam), anti-GAPDH (Cat# 919501, 1:1000; BioLegend), and anti-Calnexin (Cat# ab22595, 1:2000; Abcam), anti-ApoB (Cat# sc-393636, 1:100; Santa Cruz Biotechnology), anti-Albumin (Cat# 4929, 1:1000; Cell Signaling Technology), and anti-ApoA1 (Cat# PA5-88109, 1:2000; Invitrogen). The following HRP-conjugated secondary antibodies were used: Goat anti-Rabbit (Cat# 111-035-003, 1:3000; Jackson ImmunoResearch Laboratories), Goat anti-Mouse (Cat# 115-035-003, 1:3000; Jackson ImmunoResearch Laboratories), and Goat anti-Rat (Cat# 112-035-003, 1:3000; Jackson ImmunoResearch Laboratories).

2.8. Single Particle Interferometric Reflectance Imaging Sensor with ExoView^®

PFP for each animal was diluted 1:30 in separate tubes with PBS containing 0.05% Tween 20 (PBS-T). 35 μL of this sample was carefully pipetted onto the silicon chip coated with individual antibody spots against mouse CD9, CD63, and CD81 as well as negative isotype controls. After overnight incubation in a 24-well plate, chips were washed three times on an orbital shaker with PBS-T. Then, the chips were incubated for one hour at RT with a cocktail of fluorescent antibodies (anti-CD9-AF647; anti-CD63-AF488; anti-CD81-AF555) diluted in BSA (5% final concentration in well). Chips were washed once in PBS-T, three times in PBS, and once in deionized water. Chips were carefully removed from the 24-well plate, washed further in deionized water and removed for drying. Image and data acquisition for each chip was performed with the ExoView^® R100 (NanoView Biosciences). Data analysis was performed with NanoViewer 2.9 and ExoViewer 3 (NanoView Biosciences). Antibodies were purchased from BioLegend (CD9: Cat# 124802, Clone: MZ3; CD63: Cat# 143902, Clone: NVG-2; CD81: Cat# 104902, Clone: Eat-2). Fluorescent conjugation of antibodies was performed at NanoView Biosciences. A schematic illustration is shown in Figure 4A.

2.9. EV Flow Cytometry

FC for tetraspanin-positive EVs was performed with the Cytek^® Aurora at the University of Maryland Flow Core Shared Facility. Instrument performance for small particle detection was evaluated with a commercially available bead mixture (Cat# 1493, Apogee Flow Systems) containing eight bead populations ranging from 100 nm to 1300 nm in diameter. Compared across all channels, the R2 (APC/AF647) detector was chosen for fluorescence (FL) trigger detection of antibody stained EVs since it provided the least background signal in control, unstained samples. Threshold was set to the minimum (500) and R2 detector gain was increased until 10-20 events/second were detected in antibody only samples (to establish antibody aggregate noise). Flow rate was kept at the lowest setting (approximately 12-15 μL/min). A FL trigger detection was used rather than a side scatter (SSC) trigger detection based on prior published reports (Nolan & Stoner, 2013; Arraud et al., 2015) and our own preliminary testing that found a high number of stained EVs were undetected by SSC triggering, likely due to their small size (see Supplementary Figure 5A-B).

5 μL of PFP was diluted into f-PBS to a total volume of 100 μL with either anti-CD9-AF647 (Cat# 124810, 1:200, Clone: MZ3; BioLegend), anti-CD63-AF647 (Cat# FAB5417R, 1:20, Clone: 446703; R&D Systems), or anti-CD81-AF647 (Cat# FAB4865R, 1:20, Clone: 431301; R&D Systems) and stained for one hour at room temperature. The reaction was stopped by adding 1 mL of f-PBS to each tube and samples were acquired on the Cytek^® Aurora for an equal amount of time. The volume of sample taken from each tube was automatically recorded by the software and used to normalize calculations. The sample dilution was tested beforehand to ensure detection of single events and avoid “swarm-based” detection of multiple particles that are considered a single event, as previously demonstrated in FC analysis of concentrated nanoparticle samples (Nolan & Stoner, 2013; E. V. D. Pol et al., 2012). Data was analyzed using FCS Express Version 6.0 (De Novo™ Software).

2.10. FirePlex^® microRNA assay and analysis

A miR assay using FirePlex^® technology (Abcam) was performed to assess the profile of 65 miRs after SCI in total plasma EVs. The selected miRs were part of the Neurology Panel V2 (ab218371) that have published association in plasma/serum with neurological disease. The FirePlex^® assay uses a hybridization technique to detect miRs on three-dimensional hydrogel particles that are analyzed by fluorescent intensity readout on a flow cytometer (Tackett & Diwan, 2017). Total plasma EVs were isolated by ultracentrifugation from 360 μL of PFP by the protocol described above (see 2.4) and resuspended in 40 μL of f-PBS. Three groups were sent for analysis: Naive control, 1d SCI, and 3d SCI. miRs were detected directly from these equal sample volumes without the need for RNA isolation (Tackett & Diwan, 2017).

For miR data analysis, fluorescence intensity readout values were normalized using a geNorm-like algorithm in the Firefly Analysis Workbench (Chen et al., 2013). Partial least squares discriminant analysis (PLSDA) of normalized data was performed in the R language using RStudio Version 1.2.5033 with mixOmics v6.8.2 (Rohart et al., 2017). Individual miR fluorescent intensity data were statistically analyzed and plotted in boxplots with Prism Version 8.4.2 (GraphPad). Heatmap with hierarchical clustering of individual samples (columns) and differentially expressed miRs (rows) was generated with “pheatmap()” function in RStudio.

2.11. Spinal Cord Flow Cytometry

For FC analysis of spinal cord cells after SCI, anesthetized mice were perfused with 40 mL of cold PBS and ~1.5 cm of spinal cord tissue surrounding the epicenter of the lesion site was extracted. The tissue was mechanically digested through a 70-μm filter in complete Roswell Park Memorial Institute (RPMI) 1640 medium (Cat# 22400105, Invitrogen) followed by enzymatic digestion in collagenase/dispase (Cat# 10269638001, 1mg/mL; Roche Diagnostics), papain (Cat# LS003119, 5U/mL; Worthington Biochemical), 0.5M EDTA (Cat# 15575020, 1:1000; Invitrogen), and DNAse I (Cat# 10104159001, 10 mg/mL; Roche Diagnostics) on a shaking incubator (200 rpm) for 1 h at 37°C. Cells were washed twice with RPMI, filtered through a 70-μm filter, and RPMI was added to a final volume of 5 mL per sample. After each wash, cells were spun down at 1500 rpm for 5 min at 4°C and samples were kept on ice unless noted elsewhere. Cells were then transferred to FACS tubes and washed with FACS buffer. Cells were incubated with Fc block (Cat#101320, Clone: 93; BioLegend) for 10 min on ice, and then stained with the following surface antigens at room temperature for 15 min: CD45-PerCP-Cy5.5 (Cat# 103132, Clone: 30_F11; BioLegend), CD11b-APC/Fire™750 (Cat# 101262, Clone: M1/70; BioLegend), Ly6C-AF700 (Cat# 128024, Clone: HK1.4; BioLegend), CD200-PE (Cat# 123808, Clone: OX-2; BioLegend), CD9-PE-Cy7 (Cat# 124816, Clone: MZ3; BioLegend), CD63-PE-Cy7 (Cat# 143910, Clone: NVG-2; BioLegend), or CD81-PE-Cy7 (Cat# 104914, Clone: Eat-2; BioLegend). Cells were then washed with FACS buffer and then fixed and permeabilized for intracellular staining with BD Cytofix/Cytoperm™ (Cat#51-2090KZ; BD Biosciences) for 20 min at 4°C. Cells were washed with BD Perm/Wash™ (Cat# 51-2091KZ; BD Biosciences) and then stained for intracellular detection of GFAP-AF647 (Cat# 644706, Clone: 2E1.E9; BioLegend) or TUBB3(Tubulin Beta 3)-AF647 (Cat# 657406, Clone: AA10; BioLegend) for 30 min at 4°C. Cells were washed with Perm/Wash™ and then fixed in 2% paraformaldehyde for 10 min. Cells were washed in FACS Buffer and then resuspended and stored in 500 μL of FACS buffer at 4°C.

Prior to analysis, samples were stained with CytoPhase Violet (Cat# 425701, 1:500; BioLegend) for 15 min at 37°C to stain nucleated cells. Data was then acquired on a BD LSRFortessa cytometer using FACSDiva 6.0 (BD Biosciences) and analyzed using FCS Express Version 6.0 (De Novo™ Software). To maximize cell count for analysis, samples were run to completion at a flow-rate of 10,000-15,000 events/second for a total of about 5-10 million events. Nucleated cells were identified by CytoPhase staining, and singlets were gated on FSC-H vs. FSC-W. Resident microglia were identified as CD45+CD11b+Ly6C− and differentiated from infiltrating leukocytes identified as CD45+CD11b+Ly6C+; astrocytes were identified as CD45-CD11b-GFAP+ and neurons were identified as CD45-CD11b-TUBB3+CD200+ (see Figure 7A). Cell-type matched fluorescence minus one (FMO) controls were used to identify positive antibody staining.

2.12. Intracerebroventricular Injection

A small craniotomy was performed on the right skull to expose the site for injection centered at stereotactic coordinates of 0.2 mm anteroposterior and 1.0 mm mediolateral relative to bregma. Total plasma EVs were isolated from individual naive and SCI mice (at either 1d or 3d post-injury) by UC from 350 μL of PFP (see 2.4 above) and resuspended in 10 μL of f-PBS. In individual mice, 5 μL of these EV samples were injected into the right lateral ventricle at 2.5 mm depth below the pia mater at the craniotomy site using a 33-gauge needle attached to a 10 μL Hamilton syringe as described before (Kumar et al., 2017). In one set of experiments using 1d and 3d SCI plasma EVs, anesthetized animals were perfused 24 hours later and cortical tissue ipsilateral to the injection site was dissected for RNA isolation (see 2.13 below). In a subsequent set of experiments using 1d SCI plasma EVs, anesthetized animals were perfused 24 hours later and the brain hemisphere ipsilateral to the injection site was processed for flow cytometry analysis (see 2.14 below).

2.13. RNA Isolation and Quantitative RT-PCR

Total RNA was extracted from flash frozen cerebral cortex after ICV injection using the miRNeasy Kit (Cat# 217004, Qiagen) following manufacturer’s instructions. Complementary DNA (cDNA) was synthesized from RNA with the Verso™ cDNA RT kit (Cat# AB1453B, Thermo Scientific) following manufacturer’s instructions. Quantitative real-time PCR for target mRNAs was performed with QuantStudio™ 5 Real-time PCR System (Applied Biosystems) using TaqMan^® Gene Expression assays for the following mouse genes: Tnf (Mm00443258_m1), Nos2 (Mm00440502_m1), Il6 (Mm00446190_m1), Il1b (Mm00434228_m1), Il4ra (Mm01275139_m1), Arg1 (Mm00475988_m1), Chil3 (Mm00657889_mH), Il1a (Mm00439620_m1), Tgfb1 (Mm01178820_m1), Nlrp3 (Mm00840904_m1), Casp1 (Mm00438023_m1), Gfap (Mm01253033_m1), Lcn2 (Mm01324470_m1), Cd44 (Mm01277161_m1), Gbp2 (Mm00494576_g1), Osmr (Mm01307326_m1), Vim (Mm01333430_m1), Cd14 (Mm01158466_g1), and Gapdh (Mm99999915_g1). Samples were run in duplicate. Relative quantity of mRNA was calculated based on the comparative Ct method after normalization to Gapdh.

2.14. Brain Flow Cytometry

For flow cytometry analysis of brain cells after ICV injection of plasma EVs, anesthetized mice were perfused with 40 mL of cold PBS and the ipsilateral (i.e. injection side) hemisphere was extracted. The olfactory bulb and cerebellum were removed, and the remaining tissue was mechanically and enzymatically digested as described above for spinal cord tissue (see 2.11). After enzyme treatment, cells were washed, filtered and transferred to FACS tubes (see 2.11). Cells were incubated with Fc block (Cat#101320, Clone: 93; BioLegend) for 10 min on ice, and then stained with the following surface antigens at room temperature for 15 min: CD45-PerCP-Cy5.5 (Cat# 103132, Clone: 30_F11; BioLegend)/CD45-eF450(Cat# 48-0451-82, Clone: 30-F11; eBioscience), CD11b-APC/Fire™750 (Cat# 101262, Clone: M1/70; BioLegend). Cells were then washed with FACS buffer and then fixed and permeabilized for intracellular staining with BD Cytofix/Cytoperm™ (Cat#51-2090KZ; BD Biosciences) for 20 min at 4°C. Cells were washed with BD Perm/Wash™ (Cat# 51-2091KZ; BD Biosciences) and then stained for intracellular detection of GFAP-AF647 (Cat# 644706, Clone: 2E1.E9; BioLegend), S100B-AF700 (Cat# NBP2-54399, Clone: SPM354; Novus Biologicals), IL-1β-PerCP-eF710 (Cat# 46-7114-82, Clone: NJTEN3; Invitrogen), TNFα PE-Cy7 (Cat# 506324, Clone: MP6-XT22; BioLegend), and either IL-6-PE (Cat# 504504, Clone: MP5-20F3; BioLegend) or IL-1α-PE (Cat# 503203, Clone: ALF-161; BioLegend) overnight on ice at 4°C. Cells were washed the next day with Perm/Wash™ and then fixed in 2% paraformaldehyde for 10 min. Cells were washed in FACS Buffer and then resuspended and stored in 500 μL of FACS buffer at 4°C.

Prior to analysis, samples were stained with CytoPhase Violet (Cat# 425701, 1:500; BioLegend) for 15 min at 37°C to stain nucleated cells. Data was then acquired on a BD LSRFortessa cytometer using FACSDiva 6.0 (BD Biosciences) and analyzed using FCS Express Version 6.0 (De Novo™ Software). To maximize cell count for analysis, samples were run to completion at a flow-rate of 10,000-15,000 events/second for a total of about 5-10 million events. Nucleated cells were identified by CytoPhase staining, and singlets were gated on FSC-H vs. FSC-W. Resident brain microglia were identified as the CD45^intCD11b+ population, and astrocytes were identified as CD45-CD11b-GFAP+S100B+ (see Figure 9C). Cell-type matched fluorescence minus one (FMO) controls were used to identify positive antibody staining.

2.15. Statistical Analysis

All statistical analysis was performed in Prism Version 8.4.2 (GraphPad). Data normality was assessed using the Shapiro-Wilk test. Quantitative data are plotted as mean ± standard error of the mean (S.E.M.) and individual data points are presented for each graph (exceptions noted in relevant figure legends). Statistical analysis in each assay was detailed in figure legends. For NTA, plasma EV Western Blot (WB) densitometry, and quantitative RT-PCR data, Mann Whitney U test was performed between two groups. For all other two group comparisons, a two-tailed, unpaired Student’s t test was used. For miR assay data, a one-way ANOVA followed by Dunnett’s multiple comparisons test was used to compare each SCI group to the Naive control group. For comparison of EV and non-EV markers by WB densitometry in one- or two-step UC, a one-way ANOVA followed by Tukey’s multiple comparisons test was used. For comparison of tetraspanin expression in two subpopulations of neurons within injury by flow cytometry, a two-way ANOVA followed by Tukey’s multiple comparisons test was used. A p-value < 0.05 was considered statistically significant.

3. Results

We used multiple characterization methods to analyze changes in plasma EVs after mouse SCI in a time course as illustrated in Figure 1. To avoid batch effects related to EV storage, SCI animals had matched naive controls at each timepoint for comparison. Plasma EVs were analyzed after UC isolation or directly from PFP.

3.1. SCI results in decreased plasma EV count by NTA but increased CD81 expression

We isolated a “total plasma EV” pellet by UC of diluted PFP at 100,000g, which was assessed by nanoparticle tracking analysis (NTA) for particle count and size distribution with ViewSizer^® 3000. Unlike conventional, single-laser NTA, ViewSizer^® is designed to illuminate samples with three controllable lasers simultaneously, which allows for more accurate counting and sizing of polydisperse samples (Maguire et al., 2018). We confirmed that ViewSizer^® could resolve multiple peaks in a complex, bead mixture consisting of six silica beads and two fluorescent polystyrene beads ranging in size from 100 nm to 1300 nm (Supplementary Figure 1A-C). The illumination of nanoparticles within a sample by NTA is dependent on the laser wavelength used for detection based on Mie and Rayleigh scattering approximation, which states that scattering intensity is inversely related to the fourth power of the wavelength (F.R.S, 1899; Maguire et al., 2018). Thus, a shift to lower wavelength light can dramatically improve scattering detection. We were able to illustrate this principle clearly with ViewSizer^® using a single laser at a time to analyze the same sample of total plasma EVs. A blue (445nm) laser was essential to visualize the majority of small particles within the sample, which were not well detected when using the green (520nm) or red (635nm) laser alone (Supplementary Figure 2A). This technical limitation led to an underestimation of the particle count and a right-shift in the size distribution without the blue laser (Supplementary Figure 2B). Compared to the NanoSight LM10, a conventional NTA instrument fitted with a single red (642nm) laser, the particle size distribution of total plasma EVs was similar to ViewSizer^® when using the red laser alone, indicating this phenomenon was not an instrument-specific effect (Supplementary Figure 2C).

With the appropriate NTA settings established, we found a robust decrease in total plasma EV count at 1d post-SCI, whereas there was no statistically significant difference at 3d, 7d or 14d (Figure 2A). Particle counts were integrated from 50nm-2000nm in size, which covers the size range detectable by NTA (Dragovic et al., 2011). The overall particle size distribution (PSD) profile was similar between naive and SCI animals at all time points assessed with the majority (>90%) of particles below 300nm in size and a median size of ~100nm (Figure 2B). The peak of the PSD was broad and extended down to the lower limit of detection (LOD) at 50nm. We also isolated EVs by a two-step, sequential UC protocol to differentiate “large EVs” (lEVs; pelleting at 10,000g) from “small EVs” (sEVs; pelleting at 100,000g) as commonly described in the literature (Théry et al., 2006). Similar to one-step UC, there was a significant decrease in particle count in both the lEV and sEV pellet at 1d post-SCI (Figure 2C-D). Interestingly, NTA of diluted PFP revealed an overall decrease in plasma extracellular nanoparticle (ENP) count at 1d post-SCI and similar PSD profiles between groups (Figure 2E). Multiple independent cohorts supported this consistent drop in UC-isolated plasma EV (Supplementary Figure 3A-B) and overall plasma ENP count (Supplementary Figure 3C) at 1d post-SCI.

Figure 2. NTA reveals an acute decrease in UC-isolated plasma EVs after SCI.

(A) NTA particle concentration for total plasma EVs isolated by UC at 100,000g directly. There was a significant decrease in particle count at 1d post-injury. (B) NTA particle size distribution (left) and cumulative frequency percent distribution (right) for total plasma EVs. The median size is indicated by the dashed line at 50% cumulative frequency. (C) NTA particle concentration for large plasma EVs isolated at 10,000g. There was a significant decrease at 1d post-injury. (D) NTA particle concentration for small plasma EVs isolated at 100,000g after a previous spin at 10,000g. There was a significant decrease at 1d post-injury. (E) NTA particle concentration of total plasma at 1d post-injury (left) with particle size distribution (top right) and cumulative frequency plots (bottom right). There was a significant decrease in total plasma extracellular nanoparticle (ENP) count at 1d post-injury. *p<0.05, **p<0.01 vs naive control by Mann-Whitney U test (A, C-D, E). n = 5-6/group.

As EV isolation procedures can also contain non-EV components detectable by NTA, we then analyzed total plasma EV pellets by Western Blot (WB) for tetraspanin proteins that are specifically present in EVs. If EVs were decreased at 1d post-injury, as our NTA results suggested, we expected to find an associated decrease in EV protein markers. Instead, we found a striking increase in CD81 expression at 1d, 3d and 7d post-injury by equal volume loading (Figure 3A-B, Supplementary Figure 3D). We estimated the amount of CD81 per plasma EV by taking the ratio of the WB densitometry and NTA count measured independently from the same samples (NTA data shown in Figure 2A), and this ratio was markedly increased at 1d and 3d post-injury (Figure 3C). CD81 was present in sEVs predominantly compared to lEVs, and its quantification by WB was not affected by either one-step or two-step UC (Supplementary Figure 3E-G).

Figure 3. CD81 expression in total plasma EVs is increased after SCI.

(A) Representative Western blot (WB) images for detection of CD81 in total plasma EVs. Equal volumes of the sample were loaded onto the gel that were isolated from the same volume of PFP. (B) Densitometry quantification for WB images demonstrating increased CD81 expression in total plasma EVs at 1d, 3d, and 7d post-injury. Naive and SCI samples at 1d were combined from two independent animal experiments run (both n = 3-6/group) on separate gels (additional image shown in Supplementary Figure 3D). n = 5-12/group. (C) The ratio of CD81 per plasma EV was calculated by taking the ratio of the WB densitometry to the NTA particle count from the same sample. NTA data for the WB images presented in Figure 3A were shown previously in Figure 2A. n = 5-6/group. *p<0.05, **p<0.01 vs naive control by Mann-Whitney U test (B-C).

Previous reports suggest that lipoproteins are the predominant ENP in plasma and may outnumber EVs with commonly used isolation methods, including UC (Dragovic et al., 2011; Sódar et al., 2016). ApoB-carrying lipoproteins -- especially chylomicron and very low density lipoproteins (VLDL) -- can overlap in the size range with EVs, while other contaminants may be present in the EV sample but are below the theoretical LOD for NTA and would not be significantly quantified including low-density lipoproteins (LDL), intermediate-density lipoproteins (IDL), high-density lipoproteins (HDL), and albumin (Mørk et al., 2017). With our UC protocol, we found expression of ApoB100/ApoB48 (chylomicron/VLDL/IDL/LDL markers), ApoA1 (HDL marker), and albumin in the total plasma EV pellet as well as the lEV pellet and sEV pellet (Supplementary Figure 3E-G). ApoB expression is directly related to ApoB+ lipoprotein count as there is only one copy of ApoB per particle (Elovson et al., 1988). Considering the parallel decrease in overall plasma ENP and UC-isolated EV counts at 1d post-injury, we examined expression levels of ApoB in PFP as well as the lEV and sEV pellets. Surprisingly, we found elevated ApoB in plasma of SCI animals at 1d-post injury (Supplementary Figure 4A-B), and this relative difference was even greater in both the lEV and sEV pellets (Supplementary Figure 4C-F).

3.2. Interferometric imaging with ExoView^® reveals an abundance of tetraspanin proteins in small plasma EVs

Our results in Figures 2 and 3 show an intriguing reduction in the overall particle count by NTA with an increase in CD81 content in total plasma EVs at 1d post-injury. Considering the heterogeneity of EV subtypes and their marker expression, this increase in CD81 relative to the particle count could potentially reflect an increase in the number of CD81+ EVs and/or an increase in the amount of CD81 per EV. To address this issue without concern for measuring non-EV contaminants, we used a new interferometric imaging platform, ExoView^®, to count, size and phenotype single EVs in a multiplex manner. This technology is based on the single particle interferometric reflectance imaging sensor (SP-IRIS) technique, which has previously been used to detect viruses and EVs down to 50nm in size (Daaboul et al., 2014, 2016). Diluted PFP was incubated overnight on a silicon chip containing unique antibody capture spots for tetraspanin proteins CD9, CD63 and CD81 (Figure 4A). Captured EVs were then stained with fluorescent antibodies against these three tetraspanins, and spot images were acquired in both label-free interferometry (IFM) and fluorescence (FL) modes (Figure 4A). We were able to capture EVs bearing CD9 and CD81, but not CD63. The size distribution profile of CD9+ and CD81+ EVs peaked sharply at the limit of detection for IFM around 50 nm in size (Figure 4B-C).The addition of fluorescent-based detection was critical for accurate quantification as we found that a large majority of EVs (>90%) were undetected by IFM, indicating that the peak distribution is actually below 50 nm in size (Figure 4D-E). We found that CD9 and CD81 were colocalized on a significant number of EVs, whereas CD63 did not reliably colocalize with either of these tetraspanin markers (Figure 4F-H). More than 50% of CD9 captured EVs also expressed CD81, while less than 25% of CD81 captured EVs expressed CD9 (Figure 4H).

Figure 4. Detection of tetraspanin proteins in plasma EVs captured on a chip with ExoView^®.

(A) Schematic diagram illustrating the protocol for tetraspanin-positive plasma EV detection via SP-IRIS (details in 2.8). CD63+ EVs were neither captured nor reliably stained with fluorescent antibodies and thus excluded from further analysis. (B) Label-free interferometry (IFM) images of a representative anti-CD9 capture spot prescan (left) and postscan (middle) after overnight incubation of mouse PFP. White circles on the high magnification postscan image (right) identify detectable EVs. (C) Representative size distribution profile of tetraspanin-positive EVs by label-free IFM. EVs were primarily between 50-59 nm in size, near the IFM detection limit. (D) Representative images for EV detection by label-free IFM (left) and by fluorescence (FL; middle) on the same anti-CD9 capture spot. White circles indicate detectable EVs by either IFM alone (left) or combined with FL (right). The vast majority of EVs were only detected by FL and not IFM, indicating their size was below 50 nm. (E) Representative particle counts on different capture spots from a single chip using IFM, FL green (anti-CD81-AF555) and FL red channels (anti-CD9-AF647). Low counts on the control Hamster IgG and Rat IgG antibody spots indicate the specificity of detection. Data are presented with mean ± S.D. for three capture spots for each antibody on a single chip. (F) Representative low (left) and high (right) magnification FL images on different capture spots demonstrating colocalization of CD9 and CD81 on single EVs. (G) Representative scatter plot of particle diameter vs. FL intensity for EVs detected in both IFM (>50nm) and FL modes on different capture spots. Background signal thresholds for each FL channel are indicated by the colored dashed lines. (H) Representative pie charts from a single chip that show the percent of total particles detected with either CD9-AF647 (red), CD81-AF555 (green), or both (yellow) on different capture spots. More than 50 percent of CD9 captured EVs also contained CD81.

FC has been the predominant antibody-based, single EV analysis method to date, but this technique is limited to medium/large EVs since small EVs evade event detection by most cytometers due to their low scattering signal (E. van der Pol et al., 2014; Welsh et al., 2018). Using the Cytek^® Aurora, a high-resolution cytometer with detection capabilities down to 100 nm polystyrene beads (Supplementary Figure 1A), we were unable to detect CD63+ or CD81+ EVs in mouse plasma (Supplementary Figure 5A-D). We did detect some CD9+ EVs, but most of the events were below the scatter triggering threshold and did not clearly separate from background fluorescent antibody signal (Supplementary Figure 5A-D). Thus, SP-IRIS with ExoView^® represents a more sensitive antibody-based EV detection method that can quantify a highly abundant population of EVs below 50 nm in size, which may not be sufficiently resolved by current FC or NTA technology.

3.3. ExoView^® confirms increased plasma CD81+ EV count and cargo after SCI

To compare changes in tetraspanin-positive plasma EVs after SCI, we analyzed aliquots of PFP with ExoView^® from the cohort of animals previously analyzed by NTA and WB. We found a statistically significant increase in the number of CD81 captured EVs at 1d post-injury, with a non-significant trend at other timepoints (Figure 5A-B), supporting our prior WB results (Figure 3). We did not observe any differences in the overall count on the CD9 capture spot at any timepoint (Figure 5A), which was further supported by FC quantification of CD9+ EVs (Supplementary Figure 5E). There was no remarkable difference in EV size or colocalization distribution with injury at any timepoint (data not shown). Since CD81 colocalized significantly with CD9 captured EVs (Figure 4H), we analyzed the fluorescent intensity distribution of CD81 on the CD9 capture spot as a measure of CD81 cargo per EV. There was a statistically significant increase in the median value at 1d post-injury, with a non-significant increase at 7d (Figure 5C). Collectively, our data using both isolation-based and purification-free strategies for EV analysis strongly support the conclusion that CD81+ EV count and cargo are increased in plasma at 1d post-SCI and may remain elevated longer.

Figure 5. ExoView^® demonstrates an acute increase in CD81+ EV count and cargo in plasma after SCI.

(A) Quantification of total EVs captured on the anti-CD81 (left) and anti-CD9 (right) spots. CD81+ EVs were significantly increased at 1d post-injury with non-significant trends for an increase at later timepoints. There was no difference in CD9+ EV count. (B) Representative low (left) and high (right) magnification images of the anti-CD81 capture spot showing increased CD81+ EV count 1d post-SCI (bottom) relative to naive (top). (C) Plot of the first quartile, median, and third quartile of CD81 fluorescent intensity distribution on the anti-CD9 capture spot at 1d (left) and 7d (right) post-SCI. The amount of CD81 per EV was increased at 1d post-injury while there was a non-significant trend at 7d. *p<0.05 vs naive control by two-tailed, Student’s unpaired t-test (A, C). n = 4-6/group.

3.4. SCI alters expression of EV proteins at the injury site while modifying surface tetraspanin levels in a cell-specific manner

We hypothesized that alterations in plasma EV count and cargo was due in part to direct release of EVs from cells at the injury site. To address this question, we first analyzed whole tissue lysates surrounding the injury site by WB for proteins related to EV biogenesis. Although we did not observe a change in CD81 expression, other EV-related proteins such as TSG101, LAMP-1 and Flotillin-1 were increased at either 1d or 3d post-injury (Figure 6A-D). We confirmed the presence of these proteins in EVs isolated from spinal cord tissue using a protocol recently described for human brain (Vella et al., 2017) (Supplementary Figure 6A). Tissue was treated with type III collagenase in solution to release EVs from the extracellular spaces, followed by sequential centrifugation up to 10,000g to remove cells and large debris. The remaining supernatant was overlaid onto a triple sucrose cushion for density-gradient ultracentrifugation. At the end of the spin, separate fractions (F1, F2, F3) were collected and diluted in PBS to generate final pellets by ultracentrifugation at 100,000g. NTA demonstrated the presence of particles consistent with an EV size distribution in all three fractions (Supplementary Figure 6B). But, EV-marker proteins were found to equilibrate almost uniquely in F2 (0.6M sucrose) at a buoyant density (~1.07-1.08g/mL) similar to tissue EVs derived from human (Vella et al., 2017), macaque (Y. Huang et al., 2020), and mouse brain (Perez-Gonzalez et al., 2012) (Supplementary Figure 6C). For higher throughput, we used the same procedure without the sucrose gradient step to compare EV-marker expression by WB in tissue EVs after SCI and observed a striking increase in LAMP-1 at 7d and 14d post-injury (Figure 6E-F).

Figure 6. EV biogenesis-related proteins are increased at the injury site and found in EVs isolated directly from spinal cord tissue.

(A) Representative Western blot (WB) images for detection of proteins related to EV biogenesis at the injury site 1d post-SCI. (B) Densitometry quantification for WB images at 1d post-injury. One SCI sample was excluded as an outlier by ROUT statistical test, likely due to low GAPDH signal. TSG101 expression was significantly increased after SCI. (C) Representative WB images for detection of proteins related to EV biogenesis at the injury site 3d post-SCI. (D) Densitometry quantification for WB images at 3d post-injury. LAMP-1 and Flotillin-1 expression were significantly increased after SCI. (E) Representative WB images for detection of proteins in EVs isolated directly from tissue at multiple timepoints after SCI. (F) Densitometry quantification of tissue EVs at multiple timepoints after SCI. LAMP-1 expression was increased in tissue EVs at 7d and 14d post-injury. *p<0.05, **p<0.01, ***p<0.001 vs naive control by two-tailed, Student’s unpaired t-test (B, D, F). n = 4-6/group.

To gain more insight into which cells at the injury site may serve as a source of altered plasma EVs after SCI, we analyzed the surface expression of tetraspanin proteins (i.e. CD9, CD63, CD81) on cells isolated from the injured area by flow cytometry (Figure 7A). Surface expression of both CD9 and CD63, as measured by mean fluorescent intensity (MFI), increased at 1d post-injury on astrocytes, neurons, and microglia (Figure 7B-D). In direct contrast, changes in surface CD81 expression were cell type dependent. CD81 was unchanged in neurons but increased in microglia and decreased in astrocytes at 1d post-injury (Figure 7B-D). This mix of upregulation and downregulation of CD81 in different cell types may explain why we did not observe a change in whole tissue CD81 at the injury site by WB at the same timepoint (Figure 6A-B). Astrocyte CD81 and microglia CD63 showed the most marked population shifts among the cell types and tetraspanins analyzed (Figure 7B, D, E, G). The mean percent of CD81+ astrocytes decreased from 95% to 68% while the mean percent of CD63+ microglia increased from 13% to 67% with injury (Figure 7H). Within neurons, we observed two distinct populations of cells based on high or low expression of CD200 (Figure 7F), a membrane glycoprotein that inhibits microglial activation through ligand-receptor signaling (Hoek et al., 2000). While CD63 and CD9 increased with SCI in both neuronal populations, there was a statistically significant higher expression of CD63 (and non-significant increase for CD9) in the CD200^lo relative to the CD200^hi (Figure 7I), suggesting a correlation with loss of inflammatory inhibition and tetraspanin expression. Together, these data demonstrate multifaceted changes in cell surface tetraspanins after injury that may be associated with inflammatory activation/signaling and EV release.

Figure 7. Surface CD81 expression decreases on astrocytes at the injury site while CD9 and CD63 increases in multiple CNS cell types.

(A) Representative gating strategy to identify surface expression of tetraspanin proteins CD9, CD63, and CD81 in different cell types by flow cytometry (details in 2.11). (B-D) Mean fluorescent intensity (MFI) quantification (left) of tetraspanin proteins in astrocytes (B), neurons (C), and microglia (D) with representative histograms (right). CD9 and CD63 MFI increased in all three cell types, while CD81 MFI decreased in astrocytes, increased in microglia, and was unchanged in neurons. (E-G) Representative contour maps of tetraspanin expression in astrocytes (E), neurons (F), and microglia (G). Arrows indicate direction of change in fluorescent intensity after SCI. Note that neurons separated into two populations based on CD200 expression (high: CD200^hi; and low: CD200^lo), which was more evident after injury. (H) Plots showing the percentage of cells positive for individual surface tetraspanins within each cell type. The percent of CD63+ microglia dramatically increased with injury while the percent of CD81+ astrocytes dramatically decreased. (I) MFI quantification of CD9 and CD63 in the CD200^hi and CD200^lo neuronal populations after injury. There was statistically significant higher CD63 (and trend for higher CD9) MFI in the CD200^lo population compared to the CD200^hi population within SCI. For all flow cytometry histograms, fluorescence minus one (FMO) controls are shown in blue (B-D). **p<0.01, ***p<0.001, ****p<0.0001 vs naive control by two-tailed, Student’s unpaired t-test (B-D, H) and using two-way ANOVA followed by Tukey’s multiple comparisons test (I). n = 8/group.

3.5. UC-isolated plasma EVs after SCI have altered miR content and can promote inflammatory changes in the brain

MicroRNAs (miRs) can be transported in circulation by EVs (Valadi et al., 2007) -- as well as by other plasma ENPs including lipoproteins (Vickers et al., 2011) and ribonucleoprotein complexes (Arroyo et al., 2011) -- and exert powerful transcriptional regulatory effects when delivered to recipient cells. Based on our EV characterization data, we chose the 1d and 3d timepoints for cargo analysis via Abcam’s miR Fireplex^® assay. Total plasma EVs were isolated by UC at 100,000g for analysis. We selected a focused Neurology panel that assayed miRs with published evidence for association in plasma/serum with neurological disease. Partial least squares discriminant analysis (PLSDA) of miR data showed a separation of naive, 1d SCI and 3d SCI animals along the first two principle component axes (Figure 8A). The first principle component (PC1) effectively separated naive and 1d SCI animals, indicating the greatest variation in the data was between these two groups (Figure 8A-B). In accordance, we found more differentially expressed (DE) miRs at 1d post-injury (Figure 8C). Out of the 65 miRs tested, 18 miRs were increased and five miRs were decreased in EVs at this timepoint relative to naive controls (Figure 8C-E). Only six miRs were DE at 3d post-injury relative to naive controls, and five of them were already altered at 1d post-injury with similar magnitude and direction of change including miR-93-5p, miR-486-5p, miR-21-5p, miR-29b-3p, and miR-206 (Figure 8C-E).

Figure 8. Neurology-related miRs are altered in UC-isolated total plasma EVs acutely after SCI.

The Fireplex^® Neurology miR assay was used to assess changes in 65 miRs in total plasma EVs isolated by UC at 100,000g. (A) Partial least squares discriminant analysis (PLSDA) of miR data separated individual animals in each group (Naive: blue circle; SCI 1d: orange triangle; SCI 3d: gray “+” sign) based on the first two principle component axes. The first principle component (PC1) represented the greatest variation in the data (27%) and efficiently separated naive animals from 1d SCI animals. (B) Graph of PLSDA loading values for miRs with the greatest influence (absolute value > 0.1) in separating animals along PC1. miRs with negative loading scores were associated with increased expression while miRs with positive loading scores were associated with decreased expression in EVs 1d post-SCI. (C) Venn diagram (top) showing the number of differentially expressed (DE) miRs at 1d and 3d post-injury relative to naive control animals. Table (bottom) displays these DE miRs, their linear fold change based on mean fluorescent intensity (MFI) of assay data and adjusted (adj.) p-value of significance. (D) Box plots of individual fluorescent intensity assay data for miRs that are DE relative to naive controls at 1d or 3d post-SCI. (E) Heatmap of DE miRs in individual samples with hierarchical clustering. Heatmap color is based on z-score scaling of miR data across samples. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001 vs naive control using one-way ANOVA followed by Dunnett’s multiple comparisons test (C-D). n = 8/group.

Interestingly, a number of miR changes reflected alterations in astrocyte EV cargo after pro-inflammatory stimulation in vitro with either TNFα or IL-1β (Chaudhuri et al., 2018) including upregulation of miR-145-5p, miR-214-3p, miR-24-3p, miR-206, and miR-34c-5p and downregulation of miR-451a, miR-21-5p, and miR-29b-3p. Astrocyte EVs carried in the blood have been shown to stimulate acute cytokine and chemokine gene responses in the liver, lung and spleen in response to central nervous system inflammation (Dickens et al., 2017). Whether remote inflammation in the brain after SCI may be mediated in a similar fashion through circulating EVs has not been previously examined. As a first step in addressing this question, we performed an ICV injection of UC-isolated total plasma EVs from either naive or SCI mice and collected the ipsilateral cortex for molecular analysis. Consistent with the extent of differential miR changes, we found that plasma EVs from 1d SCI mice (Figure 9A), but not 3d SCI mice (Figure 9B), were able to induce a robust inflammatory gene response in the cortex compared to plasma EVs from naive mice. Both pro-inflammatory (Tnf, Nos2, Il1a, Il1b, Nlrp3) and anti-inflammatory gene (Arg1, Chil3, Tgfb1) mRNA increased with SCI plasma EVs (Figure 9A). To determine which glial cell types may generate inflammatory cytokines in response to plasma EVs, we performed flow cytometry analysis of the ipsilateral hemisphere in a separate ICV injection cohort (Figure 9C). We found significant changes in astrocyte intracellular cytokine levels with SCI plasma EVs, including increased IL-1β and IL-1α and decreased IL-6 (Figure 9D), whereas the same cytokines were not affected in microglia (Figure 9E). Furthermore, mRNA expression of multiple reactive astrocyte genes (Gfap, Lcn2, Cd14) increased in the cortex with SCI plasma EVs (Figure 9F). Overall, these data suggest that acute alterations in plasma EVs after SCI have the potential to trigger an inflammatory response in remote regions targeted through the circulation and that astrocytes may play an important role in both the source and response to circulating EVs after injury.

Figure 9. Intracerebroventricular injection of UC-isolated total plasma EVs from SCI animals promotes inflammation in the brain.

(A) qPCR data of inflammatory genes in cortical tissue after ICV injection of plasma EVs from either 1d SCI or naive control mice. There was increased mRNA expression for both pro-inflammatory and anti-inflammatory genes. n = 8/group. (B) qPCR data of inflammatory genes in cortical tissue after ICV injection of plasma EVs from either 3d SCI or naive control mice. There was no significant difference in inflammatory genes tested. n = 6/group. (C) Representative flow cytometry gating strategy to analyze intracellular cytokine changes in different brain cell types after ICV injection of plasma EVs (details in 2.14). (D) Mean fluorescent intensity (MFI) quantification of intracellular cytokine levels in brain astrocytes with representative histograms (right). IL-1β and IL-1α expression increased, while IL-6 expression decreased after injection of 1d SCI plasma EVs. n = 5/group. (E) MFI quantification of intracellular cytokine levels in brain microglia after injection of 1d SCI plasma EVs. There was no significant difference in the cytokines tested. n = 5/group. (F) qPCR data of reactive astrocyte gene markers in cortical tissue after ICV injection from the same cohort of animals in Figure 9A. Multiple genes had increased mRNA expression with SCI plasma EVs including Gfap, Lcn2, and CD14. *p<0.05, **p<0.01, ***p<0.001 vs naive control by Mann-Whitney U test (A-B, F) and by two-tailed, Student’s unpaired t-test (D-E).

4. Discussion

In this study, we demonstrate that SCI alters plasma EV count and biological content, which may be associated with remote inflammatory effects in the brain. These experiments are the first to characterize plasma EVs in a mouse SCI model, including temporal profiling of altered cargo content. The most significant changes occurred at 1d post-injury, the earliest timepoint examined. Two recent reports analyzed miR changes in serum EVs after rat SCI for their biomarker potential, but neither study included a characterization of EVs or an assessment of their potential biological effects (Ding et al., 2019, 2020).

Our approach highlights the importance of analyzing EVs with multiple techniques and suggests critical evaluation of previous EV studies in relation to the methods utilized (Hartjes et al., 2019). To analyze plasma EV changes after mouse SCI, we used four different characterization methods. NTA demonstrated a decreased count in total plasma EVs isolated by UC early after injury. In contrast to the decreased particle count, expression of the EV-specific tetraspanin protein CD81 within the sample increased, as demonstrated by WB analysis. We addressed these seemingly contradictory results using ExoView^®, a new technology based on SP-IRIS (Daaboul et al., 2016). EVs were affinity captured directly from plasma based on expression of tetraspanin proteins CD9 and CD81. CD81+ EV count and cargo were increased at 1d post-injury, with the peak size distribution of tetraspanin-positive EVs below 50 nm. EVs in this size range have been consistently reported in electron microscopy and high-resolution microscopy data, but may not be completely detected by most traditionally used analytical techniques, including NTA (E. van der Pol et al., 2014; Welsh et al., 2018). We sought to confirm our SP-IRIS results on tetraspanin EVs with FC, but detection of antibody-labeled EVs was hindered by a poor separation of fluorescence from background noise and poor overall scattering signal due to their small size (<50 nm) (Welsh et al., 2018). Nevertheless, the combination of bulk (WB) and single EV characterization (SP-IRIS) using isolation-dependent and isolation-free methods, respectively, strengthens our conclusions.

UC has been the gold-standard for EV isolation, but recent studies have highlighted that UC also pellets non-EV components that fall into a similar size range and mimic EV detection by standard analytical techniques. These non-EV components may include protein aggregates (György et al., 2011), lipoproteins (Sódar et al., 2016), or even recently discovered exomeres (Zhang et al., 2018). As UC-based isolation contains a heterogeneous nanoparticle mixture, the nature of the decreased particle count that we observed is unclear. This decrease could reflect alterations in an EV subtype that we were unable to analyze in this study (such as CD63+ EVs or Annexin V+ EVs) or a non-EV component. Lipoproteins are highly abundant extracellular nanoparticles (ENPs) in blood plasma and may significantly outnumber EVs during isolation by commonly used methods, including UC (Dragovic et al., 2011; Sódar et al., 2016). As we observed decreased plasma ENP count at 1d post-injury, we quantified levels of ApoB, a specific chylomicron/low-density lipoprotein family marker, in whole plasma and UC-isolated EV samples by WB. Contrary to our predictions, ApoB expression (directly correlated to ApoB+ particle count) increased after SCI in both plasma and the EV sample pellets, with greater relative difference in the latter. There are a few potential explanations that require further investigation. While CD81+ EVs and ApoB+ lipoproteins increase in response to injury, the overall size distribution of plasma ENPs may shift below the NTA detection limit as most lipoproteins and EVs are less than 50 nm. It is also possible that EVs may aggregate with each other (Rogers et al., 2020) or with lipoproteins (Sódar et al., 2016) in circulation, leading to a net decrease in count. Alternatively, this decrease could reflect a change in another plasma ENP as yet unidentified.

Tetraspanin proteins CD9, CD63, and CD81 are among the most commonly cited EV markers by WB characterization (Kowal et al., 2016; Théry et al., 2018). Jeppesen et al. showed that tetraspanin-positive EVs can be separated from other EV subtypes through immunoaffinity capture methods and contain a more restricted repertoire of biological cargo than previously recognized (Jeppesen et al., 2019). Based on recent work in HEK293 cells using SP-IRIS, the relative colocalization of tetraspanin markers within single EVs appears to depend on their subcellular distribution (plasma membrane vs. endosomal) and, therefore their site of biogenesis and release (Fordjour et al., 2019). Our SP-IRIS results also demonstrate considerable heterogeneity in tetraspanin expression within mouse plasma EVs. Similar to in vitro studies (Fordjour et al., 2019; Jeppesen et al., 2019; Kowal et al., 2016), there was significant colocalization of CD9 and CD81 within single EVs, but there were also EVs that contained only one of these proteins. We were unable to capture or detect CD63+ EVs in mouse plasma by any method in this study. Although this could reflect technical limitations with antibody binding, we have observed CD63 staining with this antibody using SP-IRIS in cell culture EVs (unpublished results). Mouse CD63+ plasma EVs may represent a subset distinct from those that are CD9+ and/or CD81+.

We also provide evidence that changes in plasma EVs may reflect molecular and cellular changes at the injury site after SCI. There was increased expression of proteins related to EV biogenesis at the whole tissue and single cell levels. Using flow cytometry, we found increased surface CD9 and CD63 at the injury site on astrocytes, neurons, and microglia at 1d. In contrast, injury-related changes in surface CD81 were cell-type dependent: decreased in astrocytes, increased in microglia, and unchanged in neurons. The decrease in CD81 expression in astrocytes may be connected to the increased CD81+ EVs in the blood after SCI, as astrocytes are an important component of the blood-brain-barrier interface. EVs from astrocytes can shed directly from the plasma membrane, which may explain the decline in surface CD81 (Dickens et al., 2017). Previous in vitro studies have suggested an association between decreased cellular surface tetraspanins and increased levels in their EVs (Crescitelli et al., 2013) and that the secretion of tetraspanin-positive EVs is ~5-fold greater at the plasma membrane than from the endosome (Fordjour et al., 2019). We cannot rule out the possibility that increased CD81 in microglia at the injury site may also reflect alterations in plasma EVs. Our results are consistent with a previous study showing increased histological staining of CD81 in microglia/macrophages at the injury site after SCI in rats (Dijkstra et al., 2000). Intraspinal administration of an anti-CD81 antibody improved behavioral and histological outcomes after SCI, but the authors did not speculate on a role for EVs in SCI pathology (Dijkstra et al., 2006). Identification of cell-specific markers on tetraspanin-captured EVs may help clarify whether increased CD81+ plasma EVs after SCI originate from specific cells in the CNS.

Astrocyte EVs can travel to immune organs in response to IL-1β injection in the brain to stimulate acute cytokine responses regulating leukocyte infiltration to the site (Dickens et al., 2017). Inflammation and peripheral leukocyte infiltration are pathological features of trauma in both the spinal cord and brain (Das et al., 2012; Trivedi et al., 2006), and astrocyte EVs may contribute to these pathophysiological changes at the injury site. Our data suggest that the brain may be an additional target for inflammatory plasma EVs after CNS trauma. Multiple miR changes observed in UC-isolated total plasma EVs 1d after SCI overlapped with alterations in astrocyte EV cargo within two hours of pro-inflammatory stimulation in vitro with either TNFα or IL-1β (Chaudhuri et al., 2018). When injected into the cerebroventricular system, UC-isolated plasma EVs from SCI animals at 1d post-injury (but not 3d) induced an inflammatory response in the cortex, as shown by increased mRNA for both pro- and anti-inflammatory genes as well as reactive astrocyte genes. Intracellular inflammatory cytokine levels were significantly altered in astrocytes at 24h post-injection, whereas microglial cytokines were largely unchanged. A limitation of our study is that we cannot ascribe the miR cargo and inflammatory effects observed here to EVs specifically as isolation by UC also contains non-EV components that can deliver functional RNA cargo.

Our laboratory has previously shown that chronic inflammatory activation of microglia in the brain is associated with neurodegeneration in multiple brain regions after rodent SCI (Wu, Stoica, et al., 2014; Wu, Zhao, et al., 2014; Wu et al., 2016; Li et al., 2020). Although we did not observe any changes in microglial pro-inflammatory cytokines specifically after ICV injection, it is possible that microglial activation may be more delayed and occur downstream in response to astrocyte activation or neuronal cell death. Recently, Kur et al. reported that microglia are a major recipient of plasma EVs within two days after intraperitoneal injection of lipopolysaccharide (LPS) to induce peripheral inflammation (Kur et al., 2020), supporting the idea that microglia can participate in direct blood-to-brain EV signaling.

EVs also circulate through CSF, which could provide an alternative route for the transport of cargo promoting brain inflammation after SCI. Increased levels of inflammasome proteins have been observed in CSF EVs from human SCI patients (de Rivero Vaccari et al., 2016). Serum EVs can carry inflammasome proteins after TBI and may contribute to TBI-induced lung inflammation and pathology (Kerr et al., 2018). Our ICV injection data also support an association between EVs and the inflammasome as we observed increased Nlrp3 and Il1b mRNA expression in the cortex with plasma EVs from SCI animals. Systemic LPS administration has been shown to increase CSF EV count, particularly derived from choroid plexus epithelial cells, and pro-inflammatory miR content, which can lead to subsequent brain inflammation (Balusu et al., 2016).

We also provide the first demonstration of EVs isolated directly from mouse spinal cord tissue. Evaluation of tissue-derived EVs is an important step towards confirming the physiological relevance of EV functions that are largely proposed from in vitro studies (Gardiner et al., 2016). Using a method initially described for human brain tissue (Vella et al., 2017), we similarly found that the majority of EV proteins were located almost uniquely in the 0.6M sucrose fraction at a density of ~1.07-1.08 g/mL. We also analyzed protein content after SCI in tissue EVs isolated by a simpler protocol that bypassed the sucrose gradient. Expression of LAMP-1, an established EV and cellular lysosome marker, was increased in tissue EVs at 7d and 14d post-injury. This alteration in LAMP-1 may reflect an important intersection between lysosomal-autophagy dysfunction, which is known to occur after CNS trauma (Li et al., 2019; Liu et al., 2015; Wu & Lipinski, 2019), and EV secretion pathways. Autophagy-related EVs carrying classical markers LC3 and p62 have been recently described as a distinct population from tetraspanin-positive EVs (Jeppesen et al., 2019).

In conclusion, this work supports the hypothesis that EVs released after spinal cord trauma may contribute to the pathophysiology of SCI at both the systemic and local levels. We identified increased tetraspanin CD81+ plasma EVs, which may reflect alterations in tetraspanin expression at the injury site. UC-isolated plasma EVs after SCI displayed alterations in inflammatory-associated/neurology-related miR cargo and promoted inflammation in the brain after ICV injection. Together, our data are consistent with the hypothesis that SCI-induced inflammatory brain changes may be mediated, in part, by circulating plasma EVs early after injury.

Highlights.

Total plasma EVs isolated by ultracentrifugation (UC) decrease 1d post-SCI.

Tetraspanin CD81+ plasma EV count and cargo increase 1d post-SCI.

Surface CD81 decreases in astrocytes at the injury site while CD9 and CD63 increase.

SCI acutely modifies CNS-related microRNA content in UC-isolated plasma EVs.

UC-isolated plasma EVs after SCI promote brain inflammation related to astrocytes.