Eight aged female rhesus monkeys (Macaca mulatta) (18–23 years, equivalent to approximately 54–69-year-old humans) were used in this study and were randomly assigned to MSC-EV treated versus vehicle treated groups (n = 4 for each group). The experimental timeline was based on our previous work (Moore et al., 2019) and is summarized in Figure 1A. Briefly, the monkeys were trained for 4 weeks on a task of fine motor function of the hand prior to surgical induction of a lesion in the hand representation of the primary motor cortex (M1). Monkeys were administered either MSC-EVs or vehicle intravenously at 24 h and then again 2 weeks after surgery and then began retesting on the same fine motor task for 4 weeks. Cerebrospinal fluid (CSF) and blood (for plasma) were drawn at multiple time points across the recovery period, and monkeys were then euthanized at 6 weeks post injury for harvesting of brain tissue (Figure 1B). Data from brain tissue were also compared to our archived data from our previous studies (Moore et al., 2019; Go et al., 2020; Zhou et al., 2023) with a cohort of aged female monkeys (16–26 years) from which brains were harvested 16-weeks post injury (MSC-EV n = 5, Vehicle n = 5).

All monkeys were obtained from national primate research facilities or private vendors and had known birth dates and complete health records. Only female monkeys were available at the time of this study. Monkeys received medical examinations and magnetic resonance imaging to ensure there were no occult health problems or neurological damage. Monkeys were housed in the Animal Science Center of Boston University Chobanian and Avedisian School of Medicine which is AAALAC accredited. All procedures were approved by the Boston University Institutional Animal Care and Use Committee (2018–00053).

Testing of the fine motor function of the hand is fully described in Moore et al. (2019). In brief, monkeys were trained on a task of fine motor function, the Hand Dexterity Task (HDT), using a testing apparatus that quantifies the response latency and video records responses from each hand (Pessina et al., 2019). The HDT is a modified version of a Kluver board (Klüver, 1935) and requires precise control of the digits, particularly opposition of the thumb and index finger, to retrieve a small, visible food reward from two different size round wells in a plexiglas tray.

All surgical procedures are described in detail in Moore et al. (2019) and briefly summarized as follows: To create reproducible cortical injury and motor deficits, the precentral gyrus was systematically mapped using electrical stimulation delivered through a small monopolar silver ball electrode placed on the surface of the pia to evoke movements. During each stimulation, a trained observer noted muscle movements (e.g., distinct movement or twitches of muscle) in specific areas of the hand, forearm or arm, both visually and by palpation. Responses were recorded on a calibrated photograph creating a cortical surface map of the hand area that was used to guide placement of the lesion.

Using the map described above, the cortical injury was induced by inserting a small glass suction pipette under the pia to bluntly transect the small penetrating arterioles as they enter the underlying cortex. Since the hand representation is known to extend down the rostral bank of the central sulcus, the sulcus was opened down to the fundus along the length of the gyral hand representation by microdissection with a small glass pipette and the pia was dissected with the glass pipette all the way down to the fundus of the sulcus.

Bone marrow was harvested from the head of the humerus of a young female monkey (~5 years of age) in our colony, as described in detail in Moore et al. (2019). Briefly, the monkey was sedated with ketamine (10 mg/kg, IM) and anesthetized with sodium pentobarbital (15–25 mg/kg IV). After sterilizing the field, bone marrow was aspirated from the humerus using a bone biopsy needle. After biopsy, monkeys were treated with buprenex (0.01–0.03 mg/kg). The collected bone marrow was shipped on ice, the same day to the Henry Ford Health. Upon arrival, the bone marrow sample was centrifuged (4,000xg for 15 min) to separate cells. The MSCs were isolated and cultured. MSC-EVs were extracted from the culture media, and purified via multi-step centrifugation and filtration, as described (Moore et al., 2019). The size distribution and concentration of the MSC-EVs was measured by qNano (Izon, Cambridge, MA), which confirmed a distribution of EVs ranging from 60 to 300 nm, with a prominent peak observed at 157 nm. Expression of EV markers (CD9, CD63) were confirmed via western blot (Supplementary Figure 1; Théry et al., 2018; Welsh et al., 2024). Each treated monkey received 4 × 10¹¹ particles/kg in 10 mL of PBS and control monkeys received 10 mL of PBS. Both MSC-EV and Vehicle doses were administered intravenously 24 h and again 2 weeks after cortical injury (Figure 1A). Dosing concentration and intervals were identical to our previous work in Moore et al. (2019), which were based on studies in rodents (Xin et al., 2013a; Xin et al., 2013b).

Post-operative testing on the HDT began 2 weeks after surgery and continued for 4 weeks (Figure 1A). Testing occurred three times per week following injury to mimic rehabilitation measures used in human stroke patients.

We developed a Non-Human Primate Grasp Assessment Scale (GRAS) (Moore et al., 2016; Pessina et al., 2019) to detect and quantify significant impairments in fine motor function of the hand and to document recovery of function of individual digits and the precise finger-thumb pinch used by monkeys to retrieve food morsels. Using the video-recorded responses, this scale allows us to distinguish between compensatory grasp patterns and a return to pre-injury grasp patterns. A maximum score of 8 is normal grasp function (Pessina et al., 2019).

Cerebrospinal fluid (CSF) and blood (for plasma) were drawn at multiple time points across the study period (Pre-surgery/baseline, 24 h, 2 weeks, 4 weeks, and at the terminal timepoint, approximately 6 weeks post-injury, Figure 1A). At the timepoints that coincide with MSC-EV or vehicle administration (24 h, 2 weeks), blood and CSF were collected immediately prior to treatment administration. Briefly, monkeys were sedated with Ketamine (10 mg/kg IM), CSF was drawn from the cisterna magna and transferred into a K3 EDTA tube and blood was drawn from the femoral vein into a K3 EDTA tube with a Vacutube and both were immediately placed in 4°C for temporary storage (less than 2 h). Plasma was centrifuged at 3,000 RPM for 15 min at 4°C and then stored in 250 μL aliquots at −80°C. CSF was centrifuged at 6,000 RPM for 1 min and stored in 100 μL aliquots and frozen at −80°C until processing.

Following post-operative motor testing, at 6-weeks post-injury (Figure 1A), brain tissue was harvested in a terminal procedure as follows: Monkeys were sedated with Ketamine (10 mg/kg IM) then anesthetized with either sodium pentobarbital (25 mg/kg IV to effect) or propofol (2.5 mg/kg) followed by exsanguination during transcardial perfusion of the brain. The perfusion began with cold Krebs-Heinsleit buffer (4°C, pH 7.4) during which, small tissue biopsies from pre-motor cortex were collected for separate studies, and perfusate was switched to 8 L of 4% paraformaldehyde (30°C, pH 7.4) to fix the remainder of the brain. Brains were blocked in situ, in the coronal plane, removed from the skull, and drop fixed with 4% paraformaldehyde for 24 h (4°C, pH 7.4), then cryoprotected in a solution of 0.1 M phosphate buffer, 10% glycerol, and 2% DMSO followed by buffer with 2% DMSO and 20% glycerol. Brains were then flash-frozen in −75°C 2-methylbutane and stored at −80°C until cut on a microtome in the coronal plane into series (8 series of 30-μm sections, and one 60-μm section series). Sections were stored at −80°C in a cryoprotectant of 15% glycerol in buffer (Estrada et al., 2017).

To assess levels of inflammatory proteins in plasma and CSF, the Olink^® PEA Inflammation panel (Olink, Uppsala, Sweden) was used on the samples collected across recovery. Samples from baseline, 24 h, 2 weeks, 4 weeks, and 6 weeks were sent to the Olink^® analysis facility in Waltham, MA. This multiplex protein measurement assay (92 inflammatory proteins) used two antibodies for every protein of interest. Each antibody was tagged with a single stranded DNA fragment. When two paired antibodies properly bind the targeted protein of interest, the DNA fragments hybridized. The hybridized DNA fragments were extended to create barcoded DNA fragments that were specific to the target protein. The number of coded fragments in the DNA library was specific to the relative concentration of the proteins (Assarsson et al., 2014). Positive control samples were run to determine efficacy and negative control samples were run to determine the limit of detection (Supplementary Table S1). Multiplex qPCR was used to quantify the levels of target proteins, and a normalized protein concentration (NPX) was reported in log2 format. For this study, all NPX log2 values were converted to base values and normalized to baseline protein measurements from their own baseline sample. Proteins that did not reach our criterion (more than 25% of samples did not reach detectable levels, or any monkey baseline values did not reach detectable levels) were excluded from analyses so that in plasma, 64 of 92 proteins reached criterion, and in CSF 39 of 92 proteins reached criterion (Supplementary Tables S2, S9). A similar rate of protein detection with the Olink^® PEA was reported for the CSF from a Rhesus monkey model of SARS-Cov-2 infection (Verma et al., 2021). We then further analyzed 36 of these proteins in plasma and 23 in CSF that are known to be implicated in stroke and cortical injury by normalizing the values against baseline and comparing across treatment groups and timepoints (Supplementary Tables S4, S11).

Immunofluorescence was performed on 30 μm brain tissue coronal sections stored at −80°C from the same cohort used in the biomarker analysis (4 MSC-EV and 4 vehicle), based on our previous work (Go et al., 2020; Zhou et al., 2023). Serial coronal sections were selected (n = 2 sections per case) at the level of the lesion in M1. Free-floating sections were washed in 0.01 M PBS (3 × 10 min) and then incubated in 50 mM glycine for 1 h (at 25°C, room temperature, while rocking). Sections were washed (3×10 min, PBS) and then incubated in 10 mM sodium citrate buffer (pH = 8.5) for 20 min in water bath at 75°C for antigen retrieval. Sections were washed then placed in pre-block solution (5% Bovine Serum Albumin, 5% Normal Donkey Serum, 0.2% Triton-X in 0.01 M Phosphate buffered saline) for 1 h (at 25°C, room temperature, while rocking). All antibodies were diluted in an antibody diluent solution containing 0.2% acetylated Bovine Serum Albumin, 1% Normal Donkey Serum, 0.1% Triton-X in 0.1 M Phosphate buffer. Sections were incubated in primary antibody against Iba1 and P2RY12 for enhancement of microglial processes (1:500 Rabbit anti-Iba1, Wako Cat# 019–19741; 1:250 Rabbit anti-P2RY12, Abcam Cat# ab1030066), MHCII (1:100 Mouse anti-LN3, MPBiosystems, Cat# 08634801), first in a variable wattage microwave (Biowave, Ted Pella, Inc., Redding, CA, United States) for 20 min (at 150 Watts and 25°C), then followed by an incubation at 4°C (while rocking) for approximately 65 h. Sections were washed (3× 10 min PBS), then incubated in secondary antibodies, Donkey anti-Mouse IgG conjugated to Alexa 568 fluorescence probe (diluted 1:200, Jackson ImmunoResearch, Cat# 703–605-155) and biotinylated donkey anti-rabbit (diluted 1:200, Jackson ImmunoResearch, Cat#711–065-152), first in the microwave (Biowave) for 20 min (at 150 Watts and 25°C), followed by incubation at 4°C (while rocking) for 20 h. For 4 h at room temperature, slices were incubated in streptavidin conjugated with Alexa 405 (diluted 1:200, Invitrogen Cat# S32351). Sections were then washed (3× 10 min in 0.1 M PB) and mounted on gelatin subbed slides. Slides were coverslipped with Prolong Gold anti-fade (Invitrogen, Cat# P36930), cured in the dark for 48 h at room temperature, and then stored at 4°C, until imaged.

Sections were imaged on a Zeiss LSM 710 confocal microscope (Zeiss, Jena, Germany) at four wavelengths (405 nm, 488 nm, 561 nm, and 633 nm), using a 40x oil immersion lens (Zeiss EC Plan-Neofluar 40x/1.3 Oil DIC) at a voxel resolution of 0.21 × 0.21 × 0.5 μm. For imaging the perilesional gray matter, the surface of the lesion was identified. A total of 4 images were acquired for each region of interest in the lesion section as follows: The first image was acquired 200 μm away from the lesion surface towards the white matter. Subsequent images throughout the grey matter depth were then acquired every 400 μm, perpendicular to the lesion surface. For the white matter, the first image was acquired 200 μm away from the gray/white interface below the lesion core. Subsequent images throughout the white matter depth were then acquired every 400 μm, perpendicular to the lesion surface (Figure 1B).

All images were processed and analyzed as described in our previous studies (Go et al., 2020; Zhou et al., 2023; Medalla and Luebke, 2015). First, confocal image stacks were deconvolved using AutoQuantX3 software (MediaCybernetics, Bethesda, MD, United States) to enhance signal-to-noise ratio. Images were assessed for microglia cell counts using Neurolucida software (MBF Bioscience, Williston, VT, United States), employing stereological 3D cell counting methods. Microglia were first identified with staining for Iba1 and then categorized into the following groups according to morphology and expression (double labeling) of MHCII: Ramified/MHCII-, Ramified/MHCII+, Hypertrophic/Ameboid/MHCII-, and Hypertrophic/Ameboid/MHCII+, as previously performed in Go et al. (2020). Microglia were classified as “ramified” versus “hypertrophic/ameboid” morphology, based on criteria described in Karperien et al. (2013). Briefly, ramified morphology is characterized by a small circular soma (about 5–10 μm in diameter), and long thin processes that extend to about 2-4x the diameter of the soma. Hypertrophic morphology is characterized by an elliptical, elongated or irregular shaped soma with large aspect ratio and/or major diameters (about > 10 μm), and thick primary processes. Amoeboid morphology is characterized by large soma that either lack processes or have thick and short processes (with widths >0.5x the soma diameter and lengths about < 1x the diameter of the soma) (See results). Microglia were counted in each z-stack using 3D counting rules with inclusion and exclusion borders, and values were expressed as density (number of cells/total volume of tissue per field or area) or proportion (total number of cells counted in each type/ total cells counted per field or area), as described previously (Tsolias and Medalla, 2021; Tsolias et al., 2024; Go et al., 2020; Zhou et al., 2023).

As summarized in Figure 1A, aged female monkeys were assessed for the effects of MSC-EV treatment on the recovery of fine motor hand function, 4–6 weeks after induction of cortical injury to the hand representation of M1 (Figures 1A,B). The aged female monkeys in the current study exhibited patterns of functional recovery after injury and treatment, consistent with our previous study (Moore et al., 2019). Specifically, between-group comparisons revealed that MSC-EV treated monkeys took fewer mean number of days to recover pre-operative grasp function or reach a plateau in recovery compared to vehicle monkeys (Student’s t-test, p = 0.031, Figure 1C). In contrast, vehicle monkeys exhibited persistent compensatory grasp patterns and reached a recovery plateau, not returning to pre-operative grasp function (Figure 1C). Similarly, the mean grasp pattern in the first week of post-operative testing (3 weeks post injury) in the current cohort revealed a similar trend to our previous cohort, with higher grasp rating in MSC-EVs compared to vehicle monkeys but did not reach statistical significance (Mann–Whitney Test, p = 0.086, Figure 1D). Notably, 3 of the 4 MSC-EV treated animals in the current cohort demonstrated a complete return to preoperative grasp function (a score of 8 on the GRAS) in the 1st week of testing, while none of the vehicle control monkeys showed this level of recovery (Figure 1D). Across the whole 6 week recovery period, we did not find a significant difference in the mean grasp rating (Mann–Whitney Test, p = 0.20, Figure 1E); However, pooling the data from the current and previous cohort (Moore et al., 2019) revealed consistent significant between-group differences across all measures of functional recovery (Moore et al., 2019). Specifically, we found that when considering this larger cohort of monkeys, significant between-group differences (p < 0.05) were found in the mean number of days to recover, as well as post-operative grasp rating in both the first week of post-operative testing, and the mean across the first 4 weeks of the recovery period (Supplementary Figure 2).

To understand the biological effects of lesion and MSC-EV treatment across recovery, we collected blood samples at the baseline, 24-h, 2-week, 4-week, and 6-week time points across recovery. Using the Olink^® PEA, levels of inflammatory proteins, cytokines, chemokines, neurotrophins, and growth factors in plasma were quantified. First, we assessed the effect of lesion in both the MSC-EV and vehicle treated groups, comparing raw NPX expression values of each biomarker between groups and timepoints (two-way ANOVA, treatment x timepoint, with Fisher’s LSD post-hoc; Figure 2; Supplementary Table S3). Largely, we found that the groups were similar in inflammatory plasma profiles at baseline and at 24 h post-injury, prior to treatment, with only a small subset of inflammatory biomarkers demonstrating between group differences at the baseline (CCL20, uPA, CXCL10, Supplementary Table S3) and at 24-h (CCL20, CASP-8, IL10-Rβ, CD40, Supplementary Table S3).

We then assessed the changes of plasma biomarkers across recovery, comparing raw expression values between-timepoints, within each group (Figure 2; Table 1; Supplementary Table S3). Significant differences in inflammatory plasma proteins were found between the 24-h post-injury timepoint and other timepoints, in both vehicle and MSC-EV treated animals (Supplementary Table S3). Specifically, we found a high number of differentially expressed proteins (DEPs) at 24 h post-injury relative to baseline, indicating a significant effect of injury on the inflammatory profile of plasma, which was consistent between the treatment groups (Figure 2A; Supplementary Table S3). Overall, the majority of DEPs at 24 h post-injury vs. baseline were canonical pro-inflammatory proteins (Figure 2A; Table 1). At 24 h post-injury, but prior to MSC-EV or vehicle treatment, we found a significant upregulation of a subset proteins compared to baseline in both groups, including IL-6, VEGF⍺, PD-L1 (Figure 2B; Supplementary Table S3). Also at 24 h, we found that there was a significant downregulation of CXCL11, MCP-4 and TNFβ (Figure 2C; Table 1; Supplementary Table S3), at 24 h as compared to baseline. These DEPs all have important roles in the post-injury inflammatory response, VEGF⍺ is a growth factor that has context-dependent beneficial effects of angiogenesis and detrimental effects on the blood brain barrier (Li et al., 2016; Geiseler and Morland, 2018), while IL-6 is a cytokine with a dual role as a pro-inflammatory molecule related to more severe stroke (Mosarrezaii et al., 2020; Zhu et al., 2022) and possesses neurotrophic properties (Zhu et al., 2022). MCP-4 is a chemoattractant protein that targets multiple inflammatory cells (Lamkhioued et al., 2000).

In addition to DEPs between 24-h post-injury and baseline, we found a subset of proteins that were differentially expressed at 2 weeks, 4 weeks, and 6 weeks post-injury compared to 24-h timepoint in both groups (Supplementary Table S3). Interestingly, we observed a divergence in the temporal patterns of recovery-associated plasma profiles of the two groups at 2 weeks, when significant DEPs between treatment groups were found (Table 1). Further, in the vehicle group, but not the MSC-EV group, we found significant differential expression of inflammatory markers at 2 weeks relative to baseline (Table 1; Supplementary Table S3). These differences from baseline at 2 weeks were not observed to the same degree in the MSC-EV group. These data indicate a treatment effect on the time course of plasma inflammatory profiles, consistent with a sustained inflammatory environment in the vehicle control animals, and an earlier shift towards a homeostatic environment in the MSC-EV animals.

To isolate the effect of MSC-EV treatment at each post-injury timepoint on inflammatory proteins, we normalized values at each timepoint to baseline values to account for inter-subject variability. Two-way repeated measures ANOVA with treatment x timepoint as independent variables was employed on post-treatment timepoints (2 weeks, 4 weeks, 6 weeks) (Supplementary Table S5A). All p-values reported were adjusted for multiple comparisons, and the effect sizes were reported as Cohen’s d in standard deviational units (Table 2; Supplementary Table S5B). There were significant effects of treatment for multiple inflammatory plasma proteins at all post-treatment timepoints, but mainly at 2 weeks post-injury. Overall, the data revealed a significant downregulation of a subset of both pro- and anti-inflammatory proteins with treatment (Table 2; Figure 3). However, the majority (67%) of the proteins downregulated across all timepoints were canonically pro-inflammatory (Figure 3A). The remainder of the downregulated proteins, which were canonically anti-inflammatory, were downregulated mostly at 2 weeks post injury (Figure 3A). At 2 weeks post-injury, the numbers of downregulated pro- and anti-inflammatory proteins were the highest relative to other timepoints. At 4 weeks and 6 weeks post-injury, a relatively high number of downregulated pro-inflammatory proteins persisted, while the number of downregulated anti-inflammatory proteins declined (Figure 3A).

Between-group post-hoc comparisons at each timepoint (Tukey’s HSD) revealed that plasma from MSC-EV treated monkeys exhibited significantly lower levels of a subset of pro-inflammatory markers throughout the recovery period (Table 2 for downregulated proteins, Cohen’s d effect size, p-value; Supplementary Table S5B for full dataset). The following pro-inflammatory markers were downregulated significantly (p < 0.05) or approached significance (p < 0.10) (Table 2) in MSC-EV vs. Vehicle group only at 2 weeks post injury: CCL3, a chemoattractant protein that contributes to secondary damage in spinal cord injury (Pelisch et al., 2020). TNFβ (also known as lymphotoxin alpha) and TNFRSF9 (also known as CD137), which activate downstream inflammatory pathways, increasing inflammation (Stahel et al., 2000; He et al., 2018) and CCL19, a protein involved in arteriole growth following ischemia (Nossent et al., 2017) that is correlated with worse outcome following stroke (Che et al., 2023).

A substantial subset of pro-inflammatory markers relevant to brain injury was downregulated in MSC-EV vs. vehicle monkeys, at multiple timepoints across the 2–6 week post-injury recovery time window. In particular, we found CCL11, MCP1, and TNF⍺, which are robust pro-inflammatory mediators that impair recovery following cortical injury while driving secondary damage (Lieschke et al., 2019; Ho et al., 2012; Shohami et al., 1999; Longhi et al., 2013), downregulated in the MSC-EV vs. vehicle monkeys at all three timepoints. Similarly, IL-18 was downregulated in MSC-EV vs. vehicle monkeys at 2 weeks and 4 weeks, and CXCL10 at 2 weeks and 6 weeks post-injury. Both proteins are pro-inflammatory cytokines that can induce inflammatory cell activity and are higher in humans with worse outcome following stroke (Hao et al., 2019; Landreneau et al., 2018). In addition, IL-6 and CCL20, which are pro-inflammatory mediators linked to microglial activation and microglial-mediated neurodegeneration in mouse models of cortical injury (Liao et al., 2020; Mosarrezaii et al., 2020; Zhu et al., 2022), were downregulated in MSC-EV vs. vehicle at 2 weeks and 6 weeks post-injury. A number of pro-inflammatory proteins were downregulated in MSC-EV vs. vehicle monkeys only at the 4-week timepoint. These include PD-L1 and TRAIL, two pro-inflammatory proteins whose downregulation has beneficial effects in models of stroke (Bodhankar et al., 2015; Hoffmann et al., 2009) and CCL23, a chemokine that induces further expression of inflammatory proteins in a feed-forward mechanism (Simats et al., 2018). Representative plots of pro-inflammatory proteins MCP1 and IL-6 across recovery are presented (Figure 3B).

Compared to the pro-inflammatory markers, there were fewer anti-inflammatory markers downregulated with MSC-EV treatment. Further, most of these downregulated anti-inflammatory markers in the MSC-EV treated group were significantly different only at 2 weeks post-injury, early in the recovery timeline and before the second MSC-EV treatment. There was a smaller subset of anti-inflammatory proteins that were also downregulated in MSC-EVs vs. vehicle group at 4 weeks post-injury (Table 2). Specifically, at 2 weeks, both TGF⍺ and LIF-R, two proteins involved in post-injury neuroprotection (Dai et al., 2020; Davis and Pennypacker, 2018), had lower levels in MSC-EVs vs. vehicle monkeys, along with urokinase-type plasminogen activator (uPA), a protein involved in axonal repair mechanisms (Merino et al., 2017). In addition, the following markers were downregulated in MSC-EV vs. vehicle monkeys both at 2 weeks and 4 weeks post injury: CX3CL1 and IL10Rβ, anti-inflammatory proteins that are involved in neuroprotection and repair following cortical injury (Lauro et al., 2019; Pawelec et al., 2020; Garcia et al., 2017) and VEGF⍺ (Li et al., 2016; Geiseler and Morland, 2018). Only one anti-inflammatory protein was downregulated at all timepoints post-injury: NT3, a growth factor associated with neuroprotection (Lin et al., 2021). Representative plots of anti-inflammatory proteins CX3CL1 and NT3 across recovery are presented (Figure 3C).

Overall, these results suggest that MSC-EV treatment is associated with a general suppression of the immune response, mainly 2-weeks following cortical injury, by modulating the levels of pro- and anti-inflammatory plasma biomarkers. Further, while MSC-EV-associated downregulation of anti-inflammatory proteins was predominant at 2 weeks post-injury, the downregulation of pro-inflammatory proteins persisted across the 2 to 6-week post-injury recovery period. These data demonstrate that as recovery progresses, the effect of MSC-EVs shifts from an acute suppression of overall inflammatory markers to a more selective suppression of pro-inflammatory plasma markers during more chronic recovery timepoints.

To further investigate the multivariate effects of treatment and timepoint on inflammatory protein expression profiles, we performed principal component analyses and dimensional reduction of our full 64-marker Olink dataset (Figure 4, Supplementary Table S6). The analyses revealed that 18 Principal Components (PC) explained 95% of the variance in the dataset. The first two PCs explained 20% and 15% of the variance, respectively. The variables that contributed most to the variance of the top 9 PCs include many of the differentially expressed markers we assessed (Figures 4A,B, Supplementary Table S6). Plotting principal components 1 and 2 shows some segregation of data points from the 24-h timepoint from both groups, and 2-week timepoint from the vehicle group, clustering away from the rest of the datapoints (Figure 4A). This was confirmed with non-parametric HCA, assessing the relative (dis)similarities across cases and timepoints based on Olink inflammatory biomarker expression profiles (Figure 4B). We found that 24-h timepoint initially clustered separately and was the most different (largest distance) from the rest of the subgroups, demonstrating the early effects of the lesion causing the greatest disruption to the inflammatory environment. The next node of clustering segregated 2-week Vehicle from the rest of the data. Interestingly, the next clustering node subclustered Vehicle 4-week and MSC-EV 2-week timepoints, distinct from the other groups. The remaining cluster consist of baseline and 6-week data from both groups and 4-week data from the MSC-EV group, which had high relative similarity to each other (small distances). These findings are indicative of the MSC-EV treatment accelerating the shifts in peripheral inflammatory plasma profiles towards a more homeostatic, baseline state, while the profile of vehicle control monkeys remain in a chronic state of inflammation (more dissimilar from baseline values), for a more prolonged period following injury.

To understand the functional relevance of the inflammatory biomarkers affected by treatment, we performed functional annotation (Gene Ontology and KEGG pathway) analyses on the subset of DEPs (all of which were downregulated at 2-weeks post injury) in MSC-EV versus vehicle group (DAVID v6.8) (Huang et al., 2009a,b) (Proteins presented in Table 2 and pathway analyses in Figure 4C and Supplementary Table S7). Analysis of these proteins yielded a list of functional terms associated with an upregulation of immune response (Figure 4C, Supplementary Table S7). Some of the highest significantly enriched pathways include: positive regulation of inflammatory response to antigen, lymphocyte chemotaxis, monocyte chemotaxis, eosinophil chemotaxis, positive regulation of glial cell proliferation, negative regulation of neurogenesis, and chemokine mediated signaling pathway (Figure 4C, Supplementary Table S7). Thus, these analyses show that MSC-EV treatment after cortical injury is associated with a downregulated peripheral immune response and inflammatory cascades. Interestingly, some functions related to negative modulation of plasticity and repair (negative regulation of neurogenesis) were also associated with the MSC-EV mediated protein downregulation (Figure 4C) and thus decreases in these proteins could be beneficial for neurite growth and recovery. In summary, these functional annotation analyses suggest that the downregulation of inflammatory proteins with MSC-EV treatment following cortical injury can lead to dampening of the peripheral immune response as well as direct suppression of the negative modulation (hence promotion) of plasticity and repair.

We then assessed how the inflammatory changes observed in plasma biomarkers (specifically at the 2-week timepoint) correlate with measures of recovery using linear regression based on Pearson’s correlation. We found that lower plasma levels of inflammatory biomarkers at 2 weeks post-injury were significantly correlated (p < 0.05) with enhanced recovery of grasp function post-injury (fewer days to return to pre-operative grasp patterns; Figure 5). Figure 5 shows individual scatter plots with linear regression of a subset of significantly correlated biomarkers and functional measures. The number of days to return to grasp was positively correlated with plasma levels of the following biomarkers at 14-days post-injury: pro-inflammatory MCP1 (R² = 0.662, p = 0.014, Figure 5A), CCL19 (R² = 0.764, p = 0.005, Figure 5B), CCL11 (R² = 0.793, p = 0.003, data not shown), and IL12β (R² = 0.684, p = 0.0113, data not shown); anti-inflammatory TGF⍺ (R² = 0.831, p = 0.002, Figure 5C).

Further, we assessed the correlation between inflammatory markers at the 2-week timepoint, using a Pearson’s correlation matrix. We found significant positive correlations (p < 0.05, Supplementary Table S8) of pro-inflammatory biomarkers with both other pro- and anti-inflammatory markers (Figure 5D). For instance, the major proinflammatory signal, TNF⍺ was significantly positively correlated with MCP1, IL6 and TNFβ. MCP1 is additionally correlated with IL12β, CCL11, CCL19 (Figure 5D, P-values presented in Supplementary Table S8). Similarly, anti-inflammatory CX3CL1 showed significant correlation with anti-inflammatory factor NT3 (p < 0.05). Our data also showed a strong positive correlation between several pro- and anti-inflammatory markers. For example, canonically anti-inflammatory marker CX3CL1 was significantly positively correlated with pro-inflammatory markers (MCP1, TNF⍺, IL6, TNFβ; Figure 5D). These correlations suggest a complex balance between plasma inflammatory biomarkers after injury. In summary, these data further support how MSC-EV treatment may lead to a general decrease in the expression of plasma inflammatory biomarkers, which can alter the time course of complex pro- and anti-inflammatory signaling to support functional recovery after cortical injury.

Our previous studies have demonstrated the efficacy of MSC-EVs in ameliorating the consequences of cortical injury in brain tissue, at 16 weeks post-injury (Moore et al., 2019; Go et al., 2020; Medalla et al., 2020; Go et al., 2021; Calderazzo et al., 2022; Zhou et al., 2023), a later recovery timepoint than that of the current study. Since the treatment was given via IV infusion, whether the effect of MSC-EV treatment on brain tissue is direct or via peripheral mechanisms that are reflected in changes in blood and CSF remains unclear. Thus, we analyzed inflammatory profiles of CSF samples collected in parallel to plasma samples. Analysis of CSF samples from monkeys in this study using the Olink^® PEA yielded results for 39 proteins, 23 of which were of interest in cortical injury (Supplementary Table S9). In contrast to plasma, two-way ANOVA comparisons of raw expression values, by treatment x timepoint, revealed largely no significant differences for the majority of markers, except for in a few proteins (Timepoint main effect, CXCL10 p = 0.001, CXCL11 p = 0.004, and CCL11 p = 0.024, Group main effect, CXCL11 p = 0.049, CD40 p = 0.013 representative examples Figure 6A; Supplementary Table S10). Similarly, comparisons of normalized values (normalized to baseline) of these CSF biomarkers at post-treatment timepoints (2-, 4- and 6-weeks post-injury; two-way repeated measures ANOVA, treatment x timepoint) revealed no significant effects of treatment or timepoint (Figures 6B,C; Supplementary Table S12). Unlike the pattern seen with plasma biomarkers, there were no correlations between levels of inflammatory proteins in CSF and functional recovery (representative example, Figure 6D).

Our previous analyses of brain tissue harvested at 16 weeks post-injury indicate effects of MSC-EVs on the phenotypes of microglia, the major immune cells of the brain (Go et al., 2020; Zhou et al., 2023). Thus, we investigated whether these microglia differences are evident at an earlier timepoint, 6-weeks post injury, and if any changes are related to peripheral inflammatory changes and functional recovery. We analyzed perilesional cortex from the brain tissue harvested 6 weeks post-injury and quantitatively assessed injury and treatment-related changes in microglial morphology (labeled with Iba1) and expression of MHCII, a marker of immune-activated, antigen presenting microglia when colocalized with Iba1 (Schetters et al., 2017). Images from perilesional motor cortex and sublesional white matter were acquired radially beginning 200 μm from the lesion and each subsequent field 400 μm away from the previous field, towards the white matter (Imaging methodology presented in Figure 1B). Representative images from both perilesional gray and sublesional white matter are presented for both MSC-EV treated and vehicle control monkeys (Figures 7A–D).

The density and proportion of distinct microglia phenotypes based on morphology and expression of MHCII (Schetters et al., 2017) were quantified using 3D counting methods, as described previously (Tsolias and Medalla, 2021; Tsolias et al., 2024). Microglia were classified based on MHCII expression as well as their “ramified” versus “hypertrophic/ameboid” morphology, as defined in Karperien et al. (2013) (Figure 7E). Ramified microglia have long thin processes and are thought to be in a surveilling, homeostatic state. In contrast, hypertrophic and amoeboid microglia have larger cell bodies and thicker processes, and are thought to be in an immune activated or phagocytic state (Karperien et al., 2013; Kumar et al., 2016; Anttila et al., 2017; Woodburn et al., 2021). With these criteria, four distinct microglia subtypes were identified and counted: Ramified/MHCII-, Ramified/MHCII+, Hypertrophic/Ameboid/MHCII-, and Hypertrophic/Ameboid/MHCII+. When assessing the total density of microglial phenotype in perilesional gray and sublesional white, there were no significant between group differences in either region of interest (Figures 7F,G, Student’s t-test, p > 0.05). However, in perilesional gray matter specifically, we found a significant interaction between field and group: in field 1 (closest to the lesion) and field 4 (farthest from the lesion) of perilesional gray, there is a significantly greater density of ramified/MHCII- and total ramified microglia in MSC-EV compared to vehicle group (Figure 7H: field 1, p = 0.034; field 4, p = 0.044; Figure 7I: field 1, p = 0.027; field 4, p = 0.017).

When comparing the overall percentages of each microglial phenotype, we found a significant effect of treatment. We found that compared to vehicle monkeys, the ratio of total ramified:hypertrophic/ameboid (Ram: Hyp/Am) microglia were significantly greater in the perilesional gray matter of MSC-EV treated monkeys (Figure 7J, Vehicle = 0.15 Ram: Hyp/Am, MSC-EV = 0.27 Ram: Hyp/Am, p = 0.00097). This pattern was also evident in the subset of MHCII+ microglia, which exhibited a significantly higher ratio of Ram: Hyp/Am in MSC-EV compared to vehicle perilesional gray matter (Figure 7K, p = 0.041). In the perilesional gray matter of MSC-EV brains, there was a higher percentage of the microglia that are ramified/MHCII- (Figure 7L, Vehicle 9.7%, MSC-EV 15.4%, p = 0.038). These between-group differences were only seen in the perilesional gray matter and did not extend to the sublesional white matter. The higher proportion of total ramified microglia, especially the ramified/MHCII- subpopulation, can indicate a more homeostatic/anti-inflammatory micro-environment in perilesional gray matter of MSC-EV relative to vehicle monkeys (Anttila et al., 2017; Woodburn et al., 2021).

We then assessed whether there are longitudinal differences in microglial morphology across recovery, by directly comparing the data from the current cohort at 6-weeks post-lesion (Figure 7L), to our archived data from the 16-week survival cohort from our previous studies (Go et al., 2020) (Figure 7M). The archived dataset consist of aged female monkeys of the same age range (16-26yo) as the current 6-week cohort (Go et al., 2020). Two-way ANOVA (treatment x timepoint) revealed a significant main effect of time on the proportion of Hypertrophic/Ameboid/MHCII- (p = 0.0017) and an interactive effect of treatment x timepoint on ramified/MHCII+ microglia (p < 0.0001). For both groups the proportion of Hypertrophic/Ameboid/MHCII- was significantly lower in the 16-week (Vehicle 16.0%, MSC-EV 21.1%) compared to the 6-week (Vehicle 66.6%, MSC-EV 63.6%) (Figures 7L,M, Vehicle, p = 0.035; MSC-EV, p = 0.002). In the MSC-EV group, but not vehicle group, the proportion of ramified/MHCII+ microglia was higher at 16-weeks compared to 6-weeks, increasing from 4.9% at 6-weeks to 34.4% at 16-weeks (Figures 7L,M, p = 0.027). These data suggest that in the MSC-EV group the subset of immune activated MHCII+ microglia at 6 weeks may be reverting to a homeostatic, ramified morphological state by 16-weeks post injury.

We assessed whether the density of microglia phenotypes correlated with functional outcome measures. We found a significant negative correlation between ramified/MHCII- microglia and the days to return to pre-operative grasp (Figure 8A). In line with our previous study (Go et al., 2020), this data suggests that a more anti-inflammatory, homeostatic environment in the perilesional gray matter is correlated with enhanced recovery of function.

Further, we assessed how these microglial measures of brain neuroinflammation may be related to peripheral changes in plasma inflammatory biomarkers. Interestingly, we found that microglial expression at 6 weeks post injury were significantly correlated with a subset of plasma inflammatory biomarkers at 2 weeks post-injury. Biomarkers in the plasma such as CCL19 (Figure 8B), TNFβ (Figure 8C), CCL11 and TGFα (data not shown) are negatively correlated with the density of homeostatic ramified/MHCII- microglia. These data indicate that higher levels of inflammatory markers in the plasma are correlated with lower density of microglia demonstrating a more homeostatic phenotype. Further, major inflammatory cytokine TNFα (Figure 8D), along with CXCL10 and NT3 (data not shown), were positively correlated with the density of Hypertrophic/Ameboid/MHCII+ microglia. These data show that higher levels of inflammatory plasma proteins are correlated with more pro-inflammatory microglial phenotypes.

The initial acute pro-inflammatory response after cortical injury involves local cytokine and chemokine signaling, reactive oxygen species, and infiltration of peripheral immune cells across the disrupted blood brain barrier (Wang et al., 2007). Pro-inflammatory signaling cascades activate microglia (Gottlieb et al., 2022; Wang et al., 2022) and recruit peripheral immune cells to the brain to contain and clear damage (Kim et al., 2014; Zhu et al., 2022; Lieschke et al., 2019; Simats et al., 2018). This is followed by an anti-inflammatory response that can promote repair of damaged cortical tissue (Jassam et al., 2017; Jin and Yamashita, 2016). Failure to resolve the pro-inflammatory environment can cause further secondary damage and cell death (Postolache et al., 2020). Our current and previous work, as well as studies in rodents, have shown that treatment with MSC-EVs results in an overall suppression of the inflammatory response across the acute and chronic stages, which in turn can mitigate secondary damage and facilitate downstream repair, plasticity and functional recovery (Xin et al., 2013a; Xin et al., 2013b; Xin et al., 2021; Kodali et al., 2023; Wang et al., 2022).

The current study showed that the majority of treatment related DEPs in plasma were found during the acute stage, at 2 weeks post-injury, where both pro- and anti-inflammatory plasma biomarkers are downregulated with MSC-EV treatment. However, the downregulation of anti-inflammatory markers in MSC-EV treated monkeys was specific to the 2-week timepoint but did not persist at 4- and 6-weeks post-injury. As recovery progressed, at 4- and 6-weeks post-injury, MSC-EV treated monkeys showed a sustained downregulation of primarily pro-inflammatory plasma proteins. Functional annotation analyses revealed that these downregulated inflammatory proteins in plasma are implicated in many biological processes associated with functional recovery.

In the early stages of recovery, an acute pro-inflammatory response can be protective after injury, by clearing and containing damage (Anrather and Iadecola, 2016). However, sustained inflammation can promote chronic secondary damage (Schimmel et al., 2017; Kokiko-Cochran and Godbout, 2018) and severe functional impairment (Hou et al., 2021). Specifically, our data showed that the majority of downregulated proteins, mostly found at 2-weeks post-injury, were related to positive modulation of the humoral immune response and the chemotaxis of peripheral immune cells. In addition, our data suggest that MSC-EVs may suppress neurotoxic effects of peripheral immune cells, which have been described to drive secondary inflammation and impair recovery (Zhang Z. et al., 2022). For instance, CCL11, CCL23, and IL-6 have all been shown to attract peripheral immune cells such as macrophages, B-cells, and neutrophils to the site of injury (Zhu et al., 2022; Lieschke et al., 2019; Simats et al., 2018), and all of these proteins are downregulated with MSC-EV treatment. The concurrent downregulation of anti-inflammatory proteins suggests that MSC-EV mediated suppression of acute inflammatory signaling maybe due to an overall lower level of acute damage (Cicchese et al., 2018). This points to a potential neuroprotective mechanism (Kodali et al., 2023; Wang et al., 2022) or facilitation of debris clearance pathways within the very acute stages (Mei et al., 2010; Yin et al., 2023), earlier than 2-weeks post injury. Assessment of biomarkers at even more acute timepoints (e.g., the first days following treatment) would be important to address in future studies.

As the major immune cells of the brain, microglia play a critical role in recovery after cortical injury. After an insult, microglia undergo a transition from the ramified “homeostatic” surveying state to an “inflammatory” reactive state, characterized by short processes and an enlarged cell body, associated with phagocytic and pro-inflammatory functions (Kumar et al., 2016; Anttila et al., 2017). When damage associated markers are released following injury, microglia enter this immune/inflammatory reactive state, where they can release pro-inflammatory cytokines such as TNFα, interleukins, and reactive oxygen species (Lull and Block, 2010). These reactive microglia proliferate and gather in the damaged areas and can phagocytose debris and damaged cells (Jia et al., 2021). Cytokines released from microglia can attract peripheral immune cells to the lesion, which can further contribute to the pro-inflammatory environment (Berchtold et al., 2020). While clearance of damaged tissue is necessary for recovery, sustained activation of microglia and infiltrating peripheral immune cells can activate apoptotic pathways that further secondary neurodegeneration (Schimmel et al., 2017; Galgano et al., 2017). The role of microglia in recovery from cortical injury and in promoting secondary damage underscores the necessity for a therapy that can target these cells.

In the current study, we found that MSC-EV treatment reduced the expression of inflammatory microglial phenotypes identified based on morphology and MHCII expression. We found that at 6-weeks post injury, MSC-EVs were associated with a higher proportion of ramified microglia, specifically the Ram/MHCII- subtype, which are associated with homeostatic functions (Anttila et al., 2017; Woodburn et al., 2021), in perilesional grey matter. These data suggest that the transition to a less inflammatory environment is occurring by 6-weeks post-injury, leading to the more resolved inflammatory micro-environment seen at the chronic 16-week recovery time point (Go et al., 2020; Zhou et al., 2023). Comparisons of these longitudinal changes in microglial phenotypes across recovery from our current and previous data (Go et al., 2020) revealed that in both treatment groups, there was a transitional decrease in the proportion of inflammatory hypertrophic/amoeboid microglia from 6 to 16-weeks post-injury. However, in the MSC-EV treated group, there was a specific increase in the proportion of ramified, MHCII+ immune stimulated/antigen presenting microglia. These data support the idea that MSC-EV treatment may decrease inflammatory hypertrophic/ameboid MHCII+ microglia via signaling them to shift back to ramified, homeostatic states over time, resulting in more ramified MHCII+ microglia at 16 weeks (Go et al., 2020; Zhou et al., 2023). These data are consistent with rodent in vivo and in vitro models of injury and neurodegenerative disease showing that MSC-EVs alter the phenotypes of microglia (Hao et al., 2022; Go et al., 2020; Zhou et al., 2023; Mukai et al., 2021) and reduce pro-inflammatory microglial markers (Garcia-Contreras and Thakor, 2021).

Taken together, our current and previous studies in monkeys (Go et al., 2020; Zhou et al., 2023) provide evidence that MSC-EV treatment following cortical injury leads to a decrease in inflammatory microglia starting at 6 weeks post-injury, with a more pronounced shift towards anti-inflammatory phenotypes at 16 weeks post injury. This increase in homeostatic microglia correlated with a more rapid recovery of pre-injury grasp, suggesting a role of this microglial phenotypic shift in functional recovery.

Our data from the current study and in previous work (Go et al., 2020; Zhou et al., 2023) indicate that MSC-EVs modulate both the peripheral and central inflammatory responses after injury, which may underlie enhancements in functional recovery. However, the direct versus indirect targets of MSC-EVs and the relationship between peripheral and central inflammatory responses in our model remain open questions. Levels of CSF inflammatory biomarkers have been shown to predict cortical injury outcome in humans (Naik et al., 2023). However, the current study did not reveal significant effects of MSC-EV treatment on CSF biomarkers. Considering that cytokine expression can peak at 3 days post-injury, and then return to baseline levels with time (Doll et al., 2014; Nayak et al., 2012), it is possible that we are not capturing the earliest injury- and treatment-related changes in CSF biomarkers. The fact that we did not see MSC-EV effects on CSF inflammatory markers, but did observe robust changes in plasma and in brain tissue, underscore the complex relationship between peripheral and central neuro-immune signaling reported in the literature (Zang et al., 2022). Further, understanding the biodistribution of the MSC-EVs would be important to assess in future studies to understand the complex peripheral-central blood–brain relationships.

Our data revealed significant correlations between plasma biomarkers at 2 weeks post-injury and microglial phenotypes at 6-weeks post injury. Specifically, these correlations indicate that higher levels of inflammatory markers in the plasma are correlated with lower density of microglia demonstrating a more homeostatic phenotype. Conversely, positive correlations were found between multiple inflammatory proteins (TNFα, CXCL10, NT3) in plasma 2-weeks post injury and the density of inflammatory (Hyp/Am/MHCII+) microglia subtype in brain tissue. These data suggest that the effects of MSC-EVs on inflammatory signaling early in recovery (2 weeks) can affect the brain microenvironment at later recovery timepoints (6 weeks). To build upon the findings here, direct assessments of plasma and brain extracellular fluids (e.g., through microdialysis) at earlier timepoints across recovery, together with parallel proteomic and transcriptomic profiling of microglia, could further elucidate these neuroimmune signaling mechanisms (Thiollier et al., 2018; Dyhrfort et al., 2019). Future assessments of neuron-derived EVs in plasma at these acute timepoints could further illuminate peripheral-central biomarker interactions that may be important for recovery (Ledreux et al., 2020). Since the current study does not discriminate microglia from invading peripheral macrophages (Zhang Z. et al., 2022; Zhang et al., 2019), further characterization of distinct subpopulations of immune cells at the transcriptomic and proteomic level would also be important (Anttila et al., 2017; Butler et al., 2024). Collectively, these types of parallel multi-modal datasets would be critical to understand the interactions between peripheral and central inflammation. Overall, we have found strong evidence to support the role of MSC-EVs in modulating acute peripheral inflammation, which may lead to a long-term reduction in neuroinflammation, facilitating functional recovery from cortical injury.