Extracellular vesicles from mesenchymal stem cells reduce microglial-mediated neuroinflammation after cortical injury in aged Rhesus monkeys

Abstract

Cortical injury, such as injuries after stroke or age-related ischemic events, triggers a cascade of degeneration accompanied by inflammatory responses that mediate neurological deficits. Therapeutics that modulate such neuroinflammatory responses in the aging brain have the potential to reduce neurological dysfunction and promote recovery. Extracellular vesicles (EVs) from mesenchymal stem cells (MSCs) are lipid-bound, nanoscale vesicles that can modulate inflammation and enhance recovery in rodent stroke models. We recently assessed the efficacy of intravenous infusions of MSC-EVs (24-h and 14-days post-injury) as a treatment in aged rhesus monkeys (Macaca mulatta) with cortical injury that induced impairment of fine motor function of the hand. Aged monkeys treated with EVs after injury recovered motor function more rapidly and more fully than aged monkeys given a vehicle control. Here, we describe EV-mediated inflammatory changes using histological assays to quantify differences in markers of neuroinflammation in brain tissue between EV and vehicle-treated aged monkeys. The activation status of microglia, the innate macrophages of the brain, is critical to cell fate after injury. Our findings demonstrate that EV treatment after injury is associated with greater densities of ramified, homeostatic microglia, along with reduced pro-inflammatory microglial markers. These findings are consistent with a phenotypic switch of inflammatory hypertrophic microglia towards anti-inflammatory, homeostatic functions, which was correlated with enhanced functional recovery. Overall, our data suggest that EVs reduce neuroinflammation and shift microglia towards restorative functions. These findings demonstrate the therapeutic potential of MSC-derived EVs for reducing neuroinflammation after cortical injury in the aged brain.

Significance

Extracellular vesicles from mesenchymal stem cells (MSC) are nanoscale vesicles that can modulate central and peripheral inflammation, and have been shown to enhance functional recovery in rodent models of stroke. Our recent study (Moore et al. 2019) demonstrated the remarkable efficacy of MSC-derived extracellular vesicles in augmenting the degree and time course of recovery of fine motor hand function after cortical injury in aged monkeys. The current study presents evidence that this extracellular vesicle-mediated enhancement of recovery after injury is associated with a phenotypic switch of inflammatory hypertrophic microglia towards anti-inflammatory, homeostatic functions. These findings are the first to demonstrate the therapeutic potential of MSC-derived extracellular vesicles to reduce neuroinflammation after cortical injury in aged monkeys.

Introduction

Cortical injury triggers an ischemic cascade and induces tissue damage due to oxygen and nutrient deprivation (Chavez et al. 2009). Within minutes, oxidative stress and excitotoxicity rapidly initiate cell death, and subsequent gliosis releases cytokines and chemokines within the first 12 h of injury (Chavez et al. 2009). In the days following injury, gliosis intensifies and can exacerbate cell damage in the ischemic core, which can be contained or spread to surrounding tissue depending on the predominance of distinct markers of inflammation or repair (Kaushal and Schlichter 2008). One critical component for the survival of tissue after injury is the activation of microglia, the innate macrophages of the brain. When the brain is uninjured, microglia are “ramified” and provide homeostatic, surveilling functions (Dubbelaar et al. 2018). However, upon detecting cell injury, microglia become activated and can exhibit pro-inflammatory, damaging functions or anti-inflammatory, restorative functions (Patel et al. 2013). Currently, the nature of microglial responses after cortical injury is not well understood, especially in aged primates.

While pro-inflammatory microglia are important for phagocytosis of cellular debris immediately after injury, the persistence of these activated microglia can perpetuate damage in the brain via pro-inflammatory cytokines, chemokines, reactive oxygen species, and proteases (Kaushal and Schlichter 2008). Further, chronic pro-inflammatory activation can inhibit restoration of normal cellular functions (Villeda et al. 2011). However, microglia are dynamic and can also inhibit damage. When microglia switch to an anti-inflammatory state, they promote tissue repair by secreting growth factors that enhance neurogenesis, synaptogenesis, and angiogenesis (Patel et al. 2013). Therapeutics that can dampen the neuroinflammatory response and promote a microglial switch from pro- to anti-inflammatory functions may facilitate recovery of function by limiting secondary damage and facilitating neurorestorative events. Given that cortical injury continues to be one of the leading causes of long-term disability worldwide, with the majority of the injuries occurring in the elderly, it is critical to develop new therapeutics that may have immunomodulatory and neurorestorative properties after injury in the aged brain.

One therapeutic approach that may enhance recovery of function after stroke or cortical injury is treatment using mesenchymal stem cells (MSCs) (Chopp and Li 2002; Bang et al. 2005; Chopp et al. 2009; Liu et al. 2010; Jeong et al. 2014; Williams et al. 2019). Previous studies have demonstrated that stem cells can enhance recovery, likely through immunomodulation and neurorestorative events (Chen et al. 2001; Jeong et al. 2014; Orczykowski et al. 2019; Moore et al. 2019). In terms of mechanism, MSCs release extracellular vesicles (EVs), which are membrane-bound nanovesicles that contain miRNA, RNA, and proteins important for cell signaling (Xin et al. 2012; Phinney and Pittenger 2017). Due to their size, EVs are therapeutically advantageous, since they do not produce vaso-occlusions or form teratomas, which can be associated with whole cell therapies. Hence, they are an exciting new class of therapeutics for neurotrauma (Baker 2009).

Previous studies have demonstrated the efficacy of MSC-derived EVs in enhancing neurorestorative events including neurogenesis, angiogenesis, and axonal sprouting, both in vivo and in vitro in rodent and porcine models of brain injury and stroke (Xin et al. 2012, 2013; Chopp and Zhang 2015; Casado et al. 2017; Williams et al. 2019). Recently, we showed that systemic administration of MSC-EVs enhanced motor recovery in aged rhesus monkeys with cortical injury that involved damage to the hand representation of the primary motor cortex (Moore et al. 2019). Compared to aged monkeys that received phosphate-buffered saline (PBS) vehicle, aged monkeys treated with EVs showed remarkable recovery, exhibiting full restoration of fine motor functionality within the first 3–5 weeks of recovery (Moore et al. 2019).

While these results demonstrate the therapeutic potential of MSC-EVs in the aged brain, the cellular changes underlying this effect are still unknown. Here, we assess neuroinflammatory markers and describe cellular changes that may enhance recovery in the brains of aged monkeys with and without EVs. To our knowledge, this is the first study to test MSC-EVs as a therapeutic in aged monkeys with brain damage and to assess biospecimens after EV treatment for cortical injury. Our findings reveal that EVs promote a functional shift of microglia from pro-inflammatory to anti-inflammatory or homeostatic phenotypes after cortical injury. These data support the translational value of MSC-EVs in promoting recovery in aged humans after cortical injury.

Materials and methods

Subjects

Brain tissue used in this study was from the nine aged female rhesus monkeys used in Moore et al. 2019. They ranged in age from 16 to 26 years old (analogous to humans approximately 48 to 78 years old) (Tigges et al. 1988). Monkeys were acquired from a private vendor (WorldWide Primates, Inc.), trained on a fine motor task, then randomly assigned to a vehicle control or EV treatment group before receiving a targeted cortical injury in the hand representation of the motor cortex. Based on previous dosing studies using MSC-EVs (Xin et al. 2013), EV treatment was administered intravenously at 4 × 10¹¹ particles/kg first at 24 hours after injury and again 14 days after injury, as described in (Moore et al. 2019). Vehicle control (PBS) was administered IV at a similar volume at 24 hours after injury and 14 days after injury. Monkeys then resumed motor testing for 12 weeks following injury and were assessed for rate and degree of recovery. All staff and researchers were blinded to the treatment groups for all procedures and experiments (Table 1).

Table .

Summary of subjects and behavioral measures of recovery of motor hand function

Subject	Origin	Age	Sex	Treatment	Lesion	Return to pre-op retrieval	Days to return to grasp	Fluor. Iba-1, and LN3 staining	Iba-1 brightfield staining	Cytokine analysis
AM331P	WWP	26.08	F	Vehicle	L	104	105	+	+	+
AM337P	WWP	24.33	F	Vehicle	R	105	46	+	+	+
AM339P	WWP	21.42	F	Vehicle	R	133	63	+	+	+
AM335P	WWP	20.33	F	Vehicle	L	134	47	+	+	+
AM323P	WWP	23.67	F	Vehicle	R	109	44	+	+	+
AM332P	WWP	24.08	F	EVs	L	32	25	+	+	+
AM338P	WWP	16.42	F	EVs	L	30	23	+	+	+
SM061E	WWP	21.75	F	EVs	L	73	37	+	+	+
SM062E	WWP	20.92	F	EVs	L	57	41	+	+	+

Data were adapted from Moore et al. (2019)

Motor testing and lesion of the M1 hand representation

To test fine motor function, monkeys were pre-trained on a modified version of the Klüver board, our Hand Dexterity Task (HDT), for 5 weeks, as detailed in Moore et al. 2019. At the completion of pre-training, the dominant hand was determined using free-choice trials.

To create an injury limited to the hand representation of M1, a craniotomy was performed on the hemisphere contralateral to the dominant hand. Then, the precentral gyrus containing the M1 hand representation was electrophysiologically mapped using a small silver ball surface electrode to determine the precise area of M1 that controlled the function of the hand and digits. Using this map, a small incision was made in the pia above the hand representation, and a small glass suction pipette was inserted under the pia to bluntly separate penetrating arterioles from the underlying cortex. This approach disrupts the blood supply to the underlying cortex without physically damaging the underlying gray matter, which subsequently degenerates down to the white matter. Following injury, monkeys were given 2 weeks to recover and then resumed fine motor testing for 12 weeks, to assess the degree of impairment as well as the rate and degree of recovery of fine motor function.

Mesenchymal stem cell-derived extracellular vesicle preparation and administration

While Moore et al. 2019 referred to the treatment as “exosomes,” we chose to use the terminology “extracellular vesicles” in the present study for three reasons. First, while we measured the particles to be approximately 111 nm, exosomes have size overlap with other types of EVs. Second, we analyzed the treatment and found that the classic exosome markers were represented in our treatment but we also did not exclude other EVs. Therefore, we felt it was more appropriate to use the more encompassing term of EVs rather than exosomes. Finally, researchers in the field have not yet developed a reliable method for separating large amounts of exosomes from other EV types, so as a result, it is considered more appropriate to refer to the treatment as EVs rather than exosomes. EVs were isolated from bone marrow-derived MSCs of a single young monkey, as described (Moore et al. 2019). Briefly, the monkey was sedated with ketamine (10 mg/kg IM) then anesthetized with sodium pentobarbital (15–25 mg/kg IV), and bone marrow was extracted from the iliac crest, then immediately shipped to Henry Ford Health Systems.

Upon arrival, marrow was spun at 4000×g for 15 min to separate cells. The buffy coat was carefully discarded, and the remaining cells were washed in a culture medium. Cells were then plated in a T75 flask using media containing 20% FBS and alpha-mem, grown to confluence, and passaged as necessary. To grow enough cells for EVs, 10 × 10⁶ cells were seeded into a Quantum Incubator (Terumo BCT, Lakewood, CO) and grown in alpha-MEM with 10% EV-depleted FBS. To collect EVs, media was collected every other day for 4 days, then every day for 2 days. Media was then centrifuged in multiple stages as described (Zhang et al. 2015, 2017). Briefly, media was centrifuged at 250×g for 5 min, then 3000×g for 30 min, then filtered through a 0.22-μm filter, centrifuged a final time at 100,000×g for 2 h to pellet EVs, then stored temporarily at 4 °C. EVs were quantified and measured using an Izon qNano, and had a mode diameter of 111 nm (data not shown). Finally, EVs were adjusted and resuspended in PBS at 4 × 10¹¹ particles/kg for each monkey, then shipped back to BU for intravenous administration in monkeys 24 h after surgery and again at 14 days after surgery.

Brain perfusion and tissue section preparation and storage

At the completion of the post-operative motor testing (Moore et al. 2019), monkeys were sedated with ketamine (10 mg/kg IM) then anesthetized with sodium pentobarbital (25 mg/kg IV to effect) followed by euthanasia by exsanguination during transcardial perfusion of the brain. This two-stage perfusion began with 2–4 l of cold Krebs-Heinsleit buffer (4 °C, pH 7.4) for collection of fresh tissue biopsies followed by 4 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 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 isopentane and stored at − 80 °C until cut on a microtome in the coronal plane into interrupted series (8 series of 30-μm sections, and one 60-μm section series). Sections were then stored at − 80 °C in a cryoprotectant of 15% glycerol in buffer until thawed for immunohistochemistry (Rosene et al. 1986; Estrada et al. 2017).

Lesion volume

The lesion volume of each animal was quantified as described in Moore et al. 2019. Briefly, a calibrated photograph of the lesion was used to measure the lesion surface area using the Scale and Measurement tool in ImageJ. Then, five representative thionin sections, equally spaced through the lesion, were digitized using a Nikon Microscope, and the depth of the lesion was demarcated using gliosis as lesion boundaries. For both lesion surface area and depth measurements, three measurements were taken per image, then averaged to obtain an averaged surface area and depth. Finally, the averaged surface area and depths were multiplied for each animal to calculate the total lesion volume. As reported in Moore et al. 2019, no differences were found for lesion volume between treatment groups.

Regions of interest

Sections containing the lesion were first identified in a series of thionin-stained coronal sections (spaced 2400 microns apart) from each monkey. The lesion was identified based on the presence of tissue damage in the hand area of primary motor cortex, indicated by glial scarring, disrupted neuronal profiles, and discontinuity of the pial surface and cortical lamination, as described previously (Orczykowski et al. 2019). From sections containing the lesion, regions of interest (ROI) were delineated, which included the perilesional gray matter (PG) and sublesional white matter (SW). The PG matter was identified as the surviving gray matter directly adjacent to the edge of the lesion, extending medially approximately 100 microns tangential to the pial surface and radially, perpendicular to the pial surface, to include the entire depth of the cortex underlying the lesion (Fig. 1b). The SW was delineated as the white matter directly beneath the PG, extending from the gray-white matter boundary and ventrally to white matter surrounding the dorsal bank of the cingulate. Parallel lines to the SW lateral and medial boundaries were traced to the cortex. Boundaries were then connected using lines that traced the cortex and the gray-white boundary to create the PG ROIs (Fig. 1b) (Orczykowski et al. 2019). Coronal sections from adjacent series were matched to thionin-stained sections containing the lesion and selected for immunofluorescent labeling of markers (Fig. 1b).

Fig. 1.

Experimental timeline and representative images of lesion and microglial morphologies. a A timeline of the experimental workflow for these experiments. b Thionin section showing the perilesional gray (PG) and sublesional white (SW) regions of interest. c, d Representative micrographs of microglia in the ramified (c) and hypertrophic/amoeboid (d) morphological forms. Scale bar = 25 microns

Immunofluorescent double-labeling

To assess the total population and morphology of microglia, immunohistochemistry (IHC) for Iba1 (pan-microglial marker) was combined with IHC for P2RY12 and both were visualized with the same fluorescent probe to enhance staining of ramified microglia processes (Haynes et al. 2006). The combination of these two markers has been shown to label the total population of microglia across various morphological subtypes (Haynes et al. 2006). Labeling these two markers with the same probe allows visualization in one channel and increases signal in fine microglial processes to allow for semi-automated 3D reconstruction of microglia ramification, using the Rayburst algorithm developed for neuronal reconstruction (Rodriguez et al. 2006). To identify the immune-activated subpopulation of microglia, we used IHC for the MHCII marker, LN3 in a double immunofluorescence experiment with Iba1/P2RY12 staining. This was done on one or two 60-μm tissue sections through the lesion, as described in Medalla and Luebke (2015).

For IHC, sections from all cases were thawed at one time and batch-processed together by washing in 0.01 M PBS followed by incubation in 50 mM glycine in 0.01 M PBS for 1 h at room temperature to block unreacted aldehydes. To remove cross-links that may have occurred during fixation, antigen retrieval was performed by placing tissue in 10 mM Sodium Citrate Buffer (pH = 8.6) in a 60 °C water bath for 20 min, then cooled for 15 minutes at room temperature. Sections were pre-blocked in 5% Bovine Serum Albumin (BSA, Sigma-Aldrich), 5% Normal Donkey Serum (NDS, Abcam), and 0.2% Triton-X (TX) in 0.01 M PBS, then co-incubated in primary antibodies to mouse LN3 (1:100, MP Biomedicals, cat# 69303, RRID# AB_2314641) and a combination of rabbit Iba1 (1:250, Wako, cat# 019-19741, RRID# AB_839504) with rabbit P2RY12 (1:250, Abcam, cat# ab82725, RRID# AB_1859855) to enhance staining of processes, in diluted 0.01 M PBS, 0.2% acetylated BSA (BSAc), 5% NDS, and 0.2% TX. To enhance antibody penetration, sections were incubated in a low-wattage microwave (150 W, 40 °C, 2 × 10 min; Ted Pella Biowave), then incubated at 4 °C for 24 h with gentle agitation. After rinsing, sections were co-incubated in AlexaFluor 488 donkey anti-mouse IgG (1:200) and AlexaFluor 647 biotinylated donkey anti-rabbit (1:200; in 0.01 M PBS, 0.2% BSAc, 5% NDS, and 0.2% Tx), microwaved 2 × 10 min at 150 W, then placed overnight at 4 °C with gentle agitation. To amplify Iba1/P2RY12 signal, sections were incubated in Strepatvadin 568 (1:200) overnight at 4 °C. To reduce autofluorescence, sections were incubated in a cupric sulfate solution (10 ml of deionized water, 50 mM ammonium acetate, and 10 mM cupric sulfate) (Schnell et al. 1999) for 30 min at room temperature. After a final rinse in PBS, sections were mounted on slides, coverslipped with Prolong anti-fade mounting medium, and cured at room temperature in the dark for 2 days.

Confocal imaging

Slides were imaged using a Leica TCS SPE laser scanning confocal microscope, using 488 and 561 diode lasers. For each section, four images through the entire laminar depth of the PG cortex and four images in the SW were acquired systematically with increasing radial distance from the center of the pial edge of the lesion area. Each image field was an area of 75,625 μm² and spaced one field apart. Thus, for all subjects, sites were sampled across the entire cortical depth of the PG and SW, and obtained at equivalent distances from the lesion. For microglial reconstruction, each site was imaged through the z-stack using a 63x 1.4 N.A. oil objective lens at a resolution of 0.113 × 0.113 × 0.035 μm. For immunofluorescent quantification of colocalized Iba1/P2RY12+ and LN3+ staining, each site was imaged through the z-stack using a 40x 1.3 N.A. oil objective lens at a resolution of 0.268 × 0.268 × 1.0 μm. Confocal Z-stack images were deconvolved and converted to 8-bit images using AutoQuant (Media Cybernetics) to improve the signal-to-noise ratio, as described (Medalla and Luebke 2015).

Quantification of LN3+ and Iba1/P2RY12+ microglia

To assess Iba1+/LN3+ microglia, LN3 and Iba1/P2RY12-positive cells were marked in Neurolucida and counted using stereologic counting rules, as described (Fiala and Harris 2001). Microglial somata within the volume and touching inclusion borders were counted; cell bodies touching exclusion borders were not counted. All microglia (identified by Iba1+ staining) were separated into four categories of co-localization: ramified LN3-negative cells (ramified LN3−), ramified LN3-positive cells (ramified LN3+), hypertrophic/amoeboid LN3-negative cells (hypertrophic/amoeboid LN3−), and hypertrophic/amoeboid LN3-positive cells (hypertrophic/amoeboid LN3+; Fig. 2a–d). For each subject, microglia were counted in PG (n = 4 images) and SW (n = 4 images; Fig. 2e, f).

Fig. 2.

Imaging of microglia with distinct morphologies and MHCII expression. a–d High-resolution confocal maximum z-projection images showing Iba1/P2RY12+ (green) and LN3+ (red) label in microglia with distinct ramified vs. hypertrophic morphologies. The combined staining of Iba1/P2RY12 enhances labeling of processes in ramified microglia. e, f Representative images of vehicle and EV treatment groups taken using a 40x oil objective to assess densities of Iba1+ microglial morphologies, with and without LN3 staining. Scale bar = 25 microns

Reconstruction and 3D analyses of microglia

Each confocal z-stack was loaded in the Neurolucida morphometric software (MBF Biosciences) for microglial reconstructions. Complete microglia with the soma contained within the z-extent of the image stack, with at least three-fourths of the arbors intact without truncation were included in the analysis. Microglial somata and processes were traced and reconstructed in 3D. 3D analysis of total process length, number of processes, nodes and ends, 3D convex hull perimeter, and 3D convex hull volume coverage were quantified as described (Chiu et al. 2019). Microglial branching index was calculated by dividing the number of ends by the quantity of processses as an additional measure of microglial branching complexity. Sholl analyses using concentric circles (radii) spaced 2 μm from the soma were performed for each flattened reconstructed cell to assess branching length and complexity (number of intersections) as a function of distance from the soma.

Data analysis and statistics

Data were analyzed using R Studio (RStudio, Inc., Version 1.1.456) and Prism (GraphPad) holding α ≤ 0.05 for all analyses. Unpaired, two-sample Student’s t tests were used to analyze treatment-specific differences in the density of microglia in fluorescent stereological quantification, as well as treatment differences found in Sholl analyses and branching complexity measures. Finally, Pearson’s correlations were used to assess relationships between recovery data from Moore et al. (2019) and microglial morphological measures.

Results

Extracellular vesicle treatment and microglial densities

Microglia are immune cells in the brain that change in morphology and protein expression in response to injury. The density of total Iba1 positive (Iba1+) microglia, classified by morphological subtypes, was quantified in PG and SW matter in brain tissue of EV-treated and vehicle control monkeys. Overall, we found no differences between groups in the total density of Iba1+ microglia in PG (t(7) = 0.4275, p = 0.68, two-sample t test) or SW regions (t(7) = 1.274, p = 0.2432, two-sample t test). When Iba1+ microglia were stratified into ramified and hypertrophic/amoeboid microglial morphologies to assess activation status (Karperien et al. 2013), there were no group differences in SW in overall ramified (t(7) = 0.5857, p = 0.5765, two-sample t test) or hypertrophic microglial densities (t(7) = 1.518, p = 0.1728, two-sample t test). In contrast, there was a trend towards an increased density of ramified microglia in the PG of the EV group (t(7) = 2.38, p = 0.087, unpaired two-sample Student’s t test) but there were no differences in the hypertrophic/amoeboid microglial densities (t(7) = 1.62, p = 0.15, two-sample t test; Fig. 3).

Fig. 3.

Microglial densities in perilesional gray and sublesional white matter. a, b Total densities of microglia stained with Iba1+ in PG and SW quantified by unbiased stereology. c, d Densities of microglia by morphological subtypes: ramified vs. hypertrophic/amoeboid microglia in the perilesional gray and sublesional white matter. There was a trend towards increased densities of ramified microglia in the perilesional gray matter (p = 0.08), but no statistically significant differences in the remaining analyses. Vehicle control group: n = 5; EV group: n = 4

Greater density and proportion of ramified MHCII+ microglia in EV-treated animals

In conjunction with their dynamic morphological state, microglial phenotypes are associated with distinct expression of markers and receptors associated with different stages of the immune response. The MHCII marker, detected by the LN3 antibody, is expressed when microglia are antigen presenting and hence, immune activated (Shobin et al. 2017). Thus, we assessed whether the density of Iba1+ microglial cells expressing LN3 was different between treatment groups. Microglia were double-labeled with Iba1 and LN3, imaged with a 40x oil objective, then classified into four subgroups: ramified Iba1+/LN3−, ramified Iba1+/LN3+, hypertrophic/amoeboid Iba1+/LN3−, and hypertrophic/amoeboid Iba1+/LN3+ (Fig. 2a–d). Compared to vehicle control monkeys, EV-treated animals had a greater density of ramified Iba1+/LN3+ microglia (t(7) = 2.380, p = 0.049, two-sample t test) and a trend towards fewer hypertrophic Iba1+/LN3+ microglia (t(7) = 1.939, p = 0.0937, two-sample t test; Fig. 4a) in PG. In SW, there was a trend towards lower densities of Iba1+/LN3+ hypertrophic microglia in the EV group (t(7) = 2.205, p = 0.0633, two-sample t test) but no differences in ramified Iba1+/LN3+ subtype were found (t(7) = 0.8753, p = 0.4105, two-sample t test; Fig. 4b). Finally, we found no significant differences in density across Iba1+/LN3− subtypes for both PG (ramified: t(7) = 0.775, p = 0.4633; hypertrophic: t(7) = 0.9865, p = 0.3567, two-sample t test) and SW (ramified: t(7) = 0.93, p = 0.9284; hypertrophic: t(7) = 0.5814, p = 0.6248, two-sample t test; Fig. 4b).

Fig. 4.

Density and proportion of ramified versus hypertrophic microglia with distinct degrees of MHCII (LN3) expression. a, b Density of ramified and hypertrophic/amoeboid Iba1+ microglia with and without LN3 staining (Iba1+/LN3−, Iba1+/LN3+) in PG and SW. EV animals had greater densities of Iba1+/LN3+ ramified microglia in PG, and a trend towards reduced densities of Iba1+/LN3+ hypertrophic microglia. In the SW, there was a trend towards reduced densities of Iba1+/LN3+ hypertrophic microglia in the EV group, but no other significant differences were found in PG or SW. c–f Proportion of Iba1+ microglial sub-populations stratified by morphology and LN3 staining in PG and SW. There was a significantly greater proportion of ramified, and lower proportion of hypertrophic, microglia in the EV group in PG (p = 0.028). g We assessed the ramified:hypertrophic ratio of total Iba1+ (both LN3−/LN3+) microglia. There was a significantly greater ratio of ramified:hypertrophic microglia in the EV group in PG. h Within the Iba1+/LN3+ group, we found a greater LN3+ ramified:hypertrophic ratio in the EV group than the vehicle control group (p = 0.048). Vehicle control group: n = 5; EV group: n = 4

Interestingly, we found a significantly greater proportion of ramified microglia and a concomitantly lower proportion of hypertrophic microglia in the PG of EV monkeys compared to vehicle control (Fig. 4c–f). Thus, treatment had a significant effect on the overall ratio of total Iba1+ (both LN3− and LN3+ combined) ramified:hypertrophic microglia in PG (t(7) = 2.766, p = 0.028, two-sample t test) but not in SW (t(7) = 0.0947, p = 0.9272, two-sample t test) (Fig. 4g). This difference was due specifically to the proportion of Iba1+/LN3+ ramified microglia, which was significantly greater in EV-treated monkeys compared to vehicle control (Iba1+/LN3+ ramified:hypertrophic ratio: t(7) = 2.393, p = 0.048, two-sample t test; Fig. 4c, d, h). These data show that EV treatment is associated with a phenotypic shift of microglial morphology from hypertrophic to ramified microglia, which was specific to the LN3+ subpopulation.

EV treatment is associated with greater microglial ramification complexity

While cell density and the proportion of microglial morphological subtypes can give some insight into the overall status of microglial activation, homeostatic and activation states are also reflected by the degree and complexity of ramification of individual microglia (Karperien et al. 2013, p. 1; Patel et al. 2013; Orihuela et al. 2016, p. 2; Shobin et al. 2017; Dubbelaar et al. 2018). This was assessed using high-resolution confocal imaging, 3D reconstruction, and Sholl analyses to quantify the 3D structure and branching topology of microglia between groups (Fig. 5a). Microglia in the PG of EV-treated monkeys exhibited significantly greater branching complexity (greater number of branching nodes (t(7) = 2.468, p = 0.043, two-sample t test) and process intersections (t(7) = 3.082, p = 0.018, two-sample t test); Fig. 5b, c), compared to microglia in vehicle control monkeys. In addition, we assessed 3D convex hull area and perimeter, which are determined by outlining microglia at their most protruded points in the 3D space. Thus, the more ramified a microglial cell is, the longer the processes are, and hence the larger the convex hull area and perimeter. Indeed, microglia in the EV-treated monkeys had a larger 3D convex hull area (t(7) = 3.487, p = 0.01, two-sample t test) and convex hull perimeter (t(7) = 2.505, p = 0.041, two-sample t test; Fig. 3d, e) than vehicle control monkeys. However, none of these morphological indices were statistically significant in the SW.

Fig. 5.

Sholl analysis of microglia demonstrate greater branching complexity in the EV-treated group. a Representative micrographs of microglia imaged using a 63x oil objective stained with Iba1 and LN3. Microglia in the EV group appeared more ramified and morphologically more complex. b–e 3D reconstructions were performed using Neurolucida, where microglia were manually traced then assessed for branching complexity. The EV group had greater mean total number of intersections and number of branch nodes, consistent with greater branching complexity and increased ramification, as well as greater mean Convex 3D Hull Area and perimeter indicative of a greater coverage area. These measures were significant in PG, but not in SW. Vehicle control group: n = 5; EV group: n = 4. f, g Sholl analyses were performed on reconstructed microglia to assess process length as a function of distance from the soma. The EV group had significantly longer processes in both the proximal and distal locations to the soma in PG but not SW (16 μm, 17 μm, 18 μm, 22–34 μm p < 0.05). Vehicle control group: n = 20 from 5 cases; EV group: n = 16 from 4 cases. Scale bar = 25 microns

Finally, Sholl analysis, which quantifies the arbor length and complexity (number of processes) as a function of distance from the soma, showed greater length and complexity at proximal and distal segments in the PG, but not in the SW in the microglia of the EV-treated monkeys when compared to vehicle control monkeys (Fig. 5f, g). These findings demonstrate that EV-treated monkeys exhibited a greater extent of microglial ramification and complexity, which are morphological features consistent with greater microglial surveillance capacity and homeostatic functions (Karperien et al. 2013; Patel et al. 2013; Orihuela et al. 2016; Shobin et al. 2017; Dubbelaar et al. 2018; Orczykowski et al. 2019).

Correlations of microglial and inflammatory markers and recovery of function

The previous study of these monkeys showed that EV treatment significantly enhanced recovery of function (Moore et al. 2019). Recovery measures included mean latency to retrieve the food reward on the Hand Dexterity Task (HDT) during the first week of post-operative testing, mean days to return to pre-operative response latency, the degree of recovery of grasp after week 1 of post-operative testing, and the mean days to return to pre-operative grasp patterns (Moore et al. 2019). To determine if measures of microglial activation were associated with functional recovery, we used Pearson’s correlation tests to determine whether the “mean days to return to pre-operative latency” were related to microglial measures. Interestingly, morphological measures of microglia were significantly correlated with recovery. In PG, we found a negative relationship of ramified:hypertrophic microglia ratio and days to return to pre-operative latency (r(7) = − 0.7365, p = 0.0236; Fig. 6a), indicating that greater densities of ramified, in contrast to hypertrophic microglia, were associated with a more rapid recovery. Further, correlations of functional recovery with morphological indices of microglial branching complexity suggested that greater ramification of microglia is associated with a more rapid recovery (Fig. 6b–d). Specifically, increased branching complexity, as determined by the number of ends divided by the quantity of processes, had a negative correlation with days to return to pre-operative latency (r(7) = − 0.6643, p = 0.0510, Pearson’s correlation), and Convex 3D Hull Area significantly negatively correlated with days to return to pre-operative latency (r(7) = − 0.8176, p = 0.0071, Pearson’s correlation; Fig. 6d–f). Taken together, these data suggest that increased densities of ramified microglia and increased ramification of microglia in the gray matter enhanced return to pre-operative motor functions.

Fig. 6.

Microglial measures in perilesional gray matter is correlated with days to return to pre-operative latency. a Relationship of ramified:hypertrophic microglia ratio vs. “days to return to pre-operative latency” shows that a greater ratio of ramified:hypertrophic microglia is correlated with reduced time to recovery. b–d Greater branching complexity of microglia have a relationship with a shorter recovery period. Vehicle control group: n = 5; EV group: n = 4

Discussion

The present study used histological analyses to assess changes of neuroinflammatory markers in brain tissue from aged monkeys after cortical injury followed by treatment with EVs or vehicle. The overall findings in this study are as follows: (1) EVs increase the density of ramified Iba1+/LN3+ microglia and shift MHCII+ microglia towards a higher ratio of ramified:hypertrophic phenotypes. (2) EVs increase the degree of microglial ramification in PG but not SW. (3) Increased ramified:hypertrophic ratios and enhanced ramification of microglia significantly correlated with a more rapid recovery across all aged monkeys. These results suggest that while there is chronic neuroinflammation in the vehicle control group which likely produces secondary damage and impedes recovery, MSC-derived EV treatment likely shifts microglia towards non-damaging, homeostatic functions in the EV group, thus reducing the secondary damage cascade that follows cortical injury in the aged brain. Finally, shifting microglia towards homeostatic and anti-inflammatory functions appears to be associated with enhanced functional recovery after cortical injury in the aged brain.

Dynamic microglial morphologies and protein expression

Microglia can assume a spectrum of functions, ranging from surveilling functions, which include normal phagocytosis and synaptic pruning in a healthy brain, to cytotoxic and damaging functions in a diseased or aged brain (Kaushal and Schlichter 2008; Patel et al. 2013; Dubbelaar et al. 2018). Microglial morphologies are determined by their activation states and biological functions (Karperien et al. 2013). When microglia are not stimulated by cell injury or cell damage, microglia have a “ramified” phenotype, which is considered the “surveilling” state as they “surveil” the surrounding neuropil (Karperien et al. 2013). Upon the detection of cell injury, microglia rapidly activate and shift towards a hypertrophic/amoeboid phenotype (Karperien et al. 2013). In the activated state, microglia can be pro-inflammatory or anti-inflammatory, leading to the exacerbation or prevention of secondary cell damage, respectively (Kaushal and Schlichter 2008; Patel et al. 2013; Zhou et al. 2014; Dubbelaar et al. 2018). While the pro-inflammatory phenotype is initially beneficial for the clearance of cell debris, chronic activation can be detrimental and result in the release of excess pro-inflammatory cytokines/chemokines (such as TNF-α and IL-6), reactive oxygen species (ROS), proteases, etc. which can contribute to chronic neuroinflammation and secondary damage (Wake et al. 2011; Tang and Le 2016; Shobin et al. 2017). Conversely, when microglia shift towards the anti-inflammatory phenotype, they release growth factors and anti-inflammatory cytokines and chemokines to resolve inflammation and initiate neurorestoration (Wake et al. 2011; Karperien et al. 2013; Patel et al. 2013; Tang and Le 2016). While microglia in a young brain can easily switch microglial phenotypes, microglia in an aged brain are often dysregulated, especially after an injury (Von Bernhardi et al. 2015). Without any treatments, aged microglia are less efficient than young microglia at seamlessly switching from maladaptive functions to restorative and homeostatic functions, and often remain in a pro-inflammatory activated state for prolonged periods of time after the initial injury (Von Bernhardi et al. 2015; Shobin et al. 2017). This chronic activation of microglia can contribute to further cytotoxicity and impede recovery over time (Von Bernhardi et al. 2015; Shobin et al. 2017). Thus, a phenotypic switch mediated by MSC-EVs from pro-inflammatory to anti-inflammatory and homeostatic functions in the recovery period may serve to limit secondary damage and enhance recovery.

In this study, we demonstrated that in the PG compared to vehicle controls, EV-treated aged monkeys had greater densities of ramified microglia, and those microglia had greater degrees of ramification. Since ramified microglia are associated with increased surveilling and homeostatic functions, these findings suggest that EVs shifted microglia from activated pro-inflammatory states towards homeostatic and surveilling functions in the PG.

When we correlated microglial measures with measures of recovery, we observed that an increased ratio of ramified:hypertrophic microglia in the PG correlated with enhanced recovery. Specifically, ramified:hypertrophic ratios had a linear relationship with recovery time, where greater ramified:hypertrophic ratios were related to shorter recovery periods. In addition, increased branching complexity and ramification of microglia were also correlated with a more rapid recovery. Together, these results suggest that ramified microglia and increased degrees of ramification in the perilesional gray matter may enhance recovery after cortical injury in the aged brain.

Overall, our results suggest that the EVs reduce neuroinflammation and promote a microglial switch from pro-inflammatory to anti-inflammatory or homeostatic functions in the aged monkey brain. Ultimately, reducing neuroinflammation may reduce further damage and promote neurorestoration after cortical injury in the aging brain.

Additional actions of extracellular vesicles in neurorestoration and implications of EV therapeutics for cortical injury, age, and other neurotraumatic injuries

In our study, monkeys first received EVs at 24 h after the initial injury and again at 14 days after injury. Given this time frame, EVs were in a position to act on neuroinflammation in the immediate post-damage cascade of neurodegeneration. Alternatively, it is possible that EVs could have directly reduced ongoing neurodegeneration so that the change in microglia is secondary. However, since lesion volumes did not differ between groups in our most recent study (Moore et al. 2019), it seems more likely that EVs are acting on neuroinflammatory processes.

There are likely other effects late in the neuroinflammatory cascade that could have also been modulated by MSC-derived EVs, such as astrocytosis and regeneration. Previous studies have demonstrated a reduction of the activation of GFAP+ astrocytes and CD68+ microglia/macrophages after MSC-EV delivery (Zhang et al. 2017). Since activated microglia/macrophages and astrocytes can release pro-inflammatory cytokines, reduced activation of these cells could ultimately reduce cytokines/chemokines and subsequently reduce neuroinflammation and secondary damage (Zhang et al. 2017).

The reduction of neuroinflammation is critical for the initiation of neurorestoration, as it is well known that excess neuroinflammation can inhibit processes known to promote neurorestoration (Kohman and Rhodes 2013). In addition, acute neuroinflammation that transitions into chronic neuroinflammation can promote further damage in the lesion and recruit peripheral inflammatory cells to exacerbate cellular damage (Kempuraj et al. 2017; Xiong et al. 2018). Indeed, recruitment of T cells via MHCII on microglia can exacerbate inflammation and potentially contribute to even more damage in the brain (McColl et al. 2004). Thus, suppressing neuroinflammation and preventing chronic inflammation after injury both reduces further damage to the brain and allows for regenerative processes to initiate, such as synaptogenesis, neurogenesis, angiogenesis, and reorganization (Chen et al. 2001; Liu et al. 2010; Cui et al. 2012; Xin et al. 2012, 2013; Anderson et al. 2016; Yang et al. 2017). Previous studies have demonstrated that MSCs can enhance synaptic stability, axonal regeneration, neurite remodeling, neurogenesis, and angiogenesis (Anderson et al. 2016; Phinney and Pittenger 2017; Zhang et al. 2017).

Our earlier study (Moore et al. 2019) demonstrated the therapeutic potential of MSC-EVs for motor recovery in aged monkeys. The current study builds on the initial study to show that these EVs may limit the cascade of cortical injury by reducing neuroinflammation. Thus, MSC-derived EVs may have similar therapeutic benefits in other neurotraumatic diseases (Jarmalavičiūtė and Pivoriūnas 2016), such as stroke (Zhang and Chopp 2016), traumatic brain injury (TBI) (Xiong et al. 2017; Yang et al. 2017), and spinal cord injury (SCI) (Wright et al. 2011) as well as aging, since it is well known that neuroinflammation is exacerbated with age (Shobin et al. 2017). These therapeutic benefits are likely mediated by the wide variety of miRNA, proteins, and mRNA cargo present within the MSC-EVs to reduce neuroinflammatory events and encourage neurorestoration (Chen et al. 2001; Xin et al. 2013; Chopp and Zhang 2015; Zhang and Chopp 2016). Our findings here support the role of MSC-EVs in reducing neuroinflammation after cortical injury in the aging brain and shed light on the cellular changes that enhance recovery after a neurotraumatic brain injury.

Conclusions

In the present study, we demonstrate, for the first time in aged monkeys, one potential mechanism of MSC-derived EVs as a treatment for cortical injury, where EVs appear to reduce neuroinflammation. These results suggest that microglia, which shape the brain’s microenvironment after cortical injury, are one of the primary biological targets of EVs. However, the question of temporal sequence of microglial dynamics after injury cannot be answered from the current data but will be an important focus of future experiments that examine earlier time points to confirm whether neuroprotective effects were present in the parenchyma to limit the damage. Furthermore, while this study focused on the area surrounding the lesion (PG and SW regions), rodent and human literature has previously shown that inflammation is concentrated in the lesion area initially, but may migrate towards other regions in the brain in more chronic stages of inflammation (Justicia et al. 2008; Walberer et al. 2014). Thus, a careful assessment of longitudinal changes and the spread of inflammation within additional regions of interest likely to be critical to recovery will be important in the future to further elucidate dynamic inflammatory changes after cortical injury. Additionally, other glia may be affected by EVs (Zhang and Chopp 2016), so future work will also assess the effects of EVs on oligodendrocytes and astrocytes. Further, since EVs likely interact with cells in the peripheral immune system, it may also be important to elucidate modulations of inflammatory cells in the periphery (such as T cells). Finally, since EVs contain a treasure trove of miRNA, mRNA, and proteins, it will be important to elucidate the molecular profiles of these EVs to determine which components may enhance recovery after injury. Taken together, further investigations are warranted to provide a more comprehensive understanding of potential mechanisms of MSC-derived EVs that abet recovery after neurotraumatic injury in the aged brain.

Abbreviations

EVs

Extracellular vesicles

MSCs

Mesenchymal stem cells

PG

Perilesional gray matter

SW

Sublesional white matter

Funding

This work was supported by the NIH grants R21-NS102991, R21NS111174, and U01-NS076474 and the National Center for Advancing Translational Sciences, National Institutes of Health, through BU-CTSI Grant Number 1UL1TR001430.

Data availability

The data that support the findings in this study are available from the corresponding author upon request.

Compliance with ethical standards

Competing interests

The authors declare that they have no competing interests.