Neural recovery after cortical injury: Effects of MSC derived extracellular vesicles on motor circuit remodeling in rhesus monkeys

Abstract

Reorganization of motor circuits in the cortex and corticospinal tract are thought to underlie functional recovery after cortical injury, but the mechanisms of neural plasticity that could be therapeutic targets remain unclear. Recent work from our group have shown that systemic treatment with mesenchymal stem cell derived (MSCd) extracellular vesicles (EVs) administered after cortical damage to the primary motor cortex (M1) of rhesus monkeys resulted in a robust recovery of fine motor function and reduced chronic inflammation. Here, we used immunohistochemistry for cfos, an activity-dependent intermediate early gene, to label task-related neurons in the surviving primary motor and premotor cortices, and markers of axonal and synaptic plasticity in the spinal cord. Compared to vehicle, EV treatment was associated with a greater density of cfos⁺ pyramidal neurons in the deep layers of M1, greater density of cfos⁺ inhibitory interneurons in premotor areas, and lower density of synapses on MAP2+ lower motor neurons in the cervical spinal cord. These data suggest that the anti-inflammatory effects of EVs may reduce injury-related upper motor neuron damage and hyperexcitability, as well as aberrant compensatory re-organization in the cervical spinal cord to improve motor function.

Introduction

Every year, over 795,000 people in the United States experience a stroke, of which 140,000 will die while many of the survivors suffer long-term disability (AHA stroke.org; Benjamin et al., 2017). However, the only FDA approved stroke therapies target removal of blood clots to re-perfuse brain tissue during the acute post-stroke phase (Koroshetz, 1996; Leon-Jimenez et al., 2014; Saver et al., 2013). Other than physical rehabilitation, there are no effective treatments to facilitate recovery (Grefkes and Ward, 2014) and a full return of function rarely occurs in humans. Therefore, new therapeutic agents are critically needed to reduce overall damage following a stroke and enhance re-organization.

After cortical injury, there is an increase in oxidative stress and pro-inflammatory processes leading to apoptosis within the lesion (Lipton, 1999) and hyperexcitability of surviving neurons in the surrounding areas (Carmichael, 2012, Kohman and Rhodes, 2013, Lai et al., 2014, Jayaraj et al., 2019). We and others have reported that stem cell-based therapies reduce inflammation and neuronal damage while enhancing plasticity after cortical injury (Chopp et al., 2009, Lakhan et al., 2009, Mahmood et al., 2005, Stonesifer et al., 2017, Liu et al., 2010, Lambertsen et al., 2018, Orczykowski et al., 2018, Go et al., 2020). In particular, mesenchymal stem cells (MSC) release endosome derived microvesicles, known as extracellular vesicles (EVs). MSC derived EVs (MSCdEVs) contain microRNAs (miRNA) and proteins that can penetrate the blood brain barrier (Matsumoto et al., 2017; Banks et al., 2020) and appear to promote functional recovery, anti-inflammatory responses, and plasticity in rodent models (Katakowski et al., 2013, Xin et al., 2012, Aleynik et al., 2014, Marei et al., 2018, Di Trapani et al., 2016; Phinney et al., 2017; Zhang et al., 2014, Zhang et al., 2016). We further tested the efficacy of these MSCs derived EVs in our non-human primate (NHP) model of cortical damage, which resulted in significant functional recovery within the first few weeks after the injury (Moore et al., 2019). Further, this EV-mediated enhanced recovery is associated with an increase in ramified surveying microglia (Go et al., 2020), and myelin maintenance (Go et al., 2021), as well as preservation of excitatory:inhibitory (E:I) balance within pyramidal neurons (Medalla & Chang et al., 2020). Yet, it is unknown how this functional recovery relates to neuronal laminar cortico-cortical and cortico-spinal motor circuitry.

Studies have shown that compensatory activation of premotor cortices (PMC) and their spinal cord projections help support motor recovery after injury in M1 (Barbas and Pandaya 1987; Johansen-Berg et al., 2002; Kantak et al. 2012; Morecraft et al., 2019). Specifically, recruitment of activity from the premotor areas can occur after injury (Johansen-Berg et al., 2002, Carmichael, 2003; Dancause et al., 2005; Carmichael, 2006; Darling et al., 2011; Nishimura and Isa, 2012) to compensate for lost signals. However, pre-motor hyperactivation needs to be properly regulated otherwise hyper-excitability and glutamate toxicity can further damages neurons (Carmichael, 2012, Kohman and Rhodes, 2013, Lai et al., 2014, Jayaraj et al., 2019). Additionally, corticospinal projections of motor cortices sprout new connections in the spinal cord to compensate for damaged M1 projections and re-innervate lower motors neurons (LMN) (Wiessner et al., 2003; Weidner et al., 2001; Lee et al., 2004; Zai et al., 2009). Yet, new compensatory synaptic connections also need to be tightly controlled as aberrant neuronal plasticity after injury associates with functional impairments (Niv et al., 2012, Kim et al., 2015, Beauparlant et al., 2013, Wahl et al., 2014, Gennaro et al., 2017) and prevention of this abnormal plasticity after stroke can improve recovery (Cuartero et al., 2019).

cfos is an activity dependent early intermediate gene upregulated after depolarization (Morgan et al., 1987, Saffen et al., 1988, Link et al., 1995, Lyford et al., 1995) and can label recently active neurons following a sustained behavioral task (Guzowski et al., 1999, Vann et al., 2000, Hall et al., 2001, Ramirez-Amaya et al., 2005). Our recent work in the ventral PMC (Medalla & Chang et al., 2020) reports increases in cfos⁺ activation of calbindin-positive (CB⁺) inhibitory neurons with MSCd EVs. However, whether distinct layers and motor cortical areas exhibit cell-type specific cfos⁺ activation to support compensatory activity for recovery of function after injury is unknown. The relevance of cortical layers is that they give rise to distinct cortical and cortico-spinal connections, which play different roles in motor control (Morecraft et al., 2019), and likely have selective vulnerability to injury. Building on this, we have now assessed the functional remapping of laminar cortical and spinal motor circuits after injury and treatment with MSCd-EVs. Specifically, we quantified layer specific neuronal numbers and activity via cfos labeling in the perilesional M1 cortex (periM1) and dorsal premotor cortices (dPMC), along with markers for synaptic plasticity in the cervical spinal cord. Here we provide evidence that EV treatment resulted in greater pyramidal neuron cfos+ activity in the periM1 and reduced presumably aberrant synapses in the ventral horn (VH) of the cervical spinal cord (CSC) compared to vehicle treatment. Additionally, EV treatment promoted inhibitory neuron activity in the dPMC, likely downregulating injury-related compensatory hyperexcitability.

Methods

Subjects

Ten female rhesus monkeys (Macaca mulatta) ranging from 16 to 26 years of age (approximately equivalent to 48- and 78-year-old humans; Tigges et al., 1988) were obtained from the National Primate Research Facilities or Private Vendors and had known birthdates with complete health records (Table 1). Monkeys received medical examinations and magnetic resonance imaging to ensure there was no occult health problems or neurological damage. Monkeys were housed in the Animal Science Center of Boston University School of Medicine which is AAALAC accredited. All procedures were approved by the Boston University Institutional Animal Care and Use Committee.

Table 1.

Subject List. Details on subjects used for this study. There were 5 females in each treatment group with no significant differences in age or weight between groups.

Animal	Group	Age	Weight	Sex
AM320	EVs	21.83	11.4	F
AM332	EVs	24.08	12.1	F
AM338	EVs	16.42	6.1	F
SM061	EVs	21.75	6.2	F
SM062	EVs	20.92	8.2	F
Mean		21.00 (+/- 1.26)	8.80 (+/- 1.27)
AM323	Vehicle	23.67	11	F
AM331	Vehicle	26.08	13.1	F
AM335	Vehicle	20.33	9.5	F
AM337	Vehicle	24.33	7	F
AM339	Vehicle	21.42	10.3	F
Mean		23.17 (+/- 1.02)	10.18 (+/- 0.99)
Ttest		P = 0.109	P = 0.208

Cortical Injury

All surgical procedures were carried out under aseptic conditions following the description in Moore et al., 2019. Each monkey was anesthetized with intravenous sodium pentobarbital (15–25 mg/kg) to effect, and antibiotics and analgesics were given prior to and following surgery. To create reproducible cortical injury and motor deficits, the precentral gyrus was systematically explored using electrical stimulation while a trained observer noted muscle movements in specific areas of the digits, hand, forearm or arm, both visually and by palpation. Electrophysiological stimulation was used to create a cortical surface map of the hand area that was used to guide placement of the lesion. Cortical injury was induced as described in Moore et al., 2019 via a small glass suction pipette that was inserted under the pia and used to bluntly transect the small penetrating arterioles. Suction and irrigation with sterile saline were sufficient to stanch bleeding and maintain a clear field. This pial dissection of penetrating vessels removes the blood supply to the cortex of the hand representation, inducing degeneration that extends down to the underlying white matter. Immediately following surgery and for 3–7 days post-op, antibiotics and analgesics were administered. As previously reported (Moore et al., 1 Jan. 2019, Go et al., 2020) an independent samples Student’s T-test comparing the volume of the lesion between the vehicle control and treated monkeys revealed no significant differences between groups (T = –0.732, P = 0.488).

Motor Testing

As described in detail previously (Moore et al., 2010, Moore et al., 2016), monkeys were tested on a task of fine motor function of the hand, the Hand Dexterity Task (HDT), using a testing apparatus that controls, quantifies, and video records responses from each hand. Monkeys were trained pre-operatively to asymptotic performance with equal trials to both hands and lesions were made in the cortex controlling the preferred hand. Postoperative re-testing began two weeks after surgery and continued for 12 weeks. Seventy percent of the trials required the use of the impaired hand, which is similar to the constraint-induced therapy that is frequently used in human rehabilitation (Corbetta et al., 2015; Kwakkel et al., 2016; Souza et al. 2015). The length of time required for monkeys to achieve asymptotic or pre-operative function was recorded. The criterion on this task for successful return to pre-operative performance was five consecutive days at or below the pre-operative time to retrieve the food reward.

Final Testing and cfos activation: On the day of euthanasia at 14 weeks post-injury, monkeys were tested on the HDT task for 1 h. For this final testing session all trials were administered to the impaired hand to maximize activation of neurons participating in function of the impaired hand. Two hours after the completion of the final testing session, monkeys were euthanized, an interval sufficient to allow expression of the cfos protein in activated neurons. (Zhong et al., 2014, Barros et al., 2015).

Grasp assessment

We used video recordings of the pre- and post-operative performance on the HDT and our Non-Human Primate Grasp Assessment Scale (GRAS) to quantify impairments and recovery of fine motor function of the hand. Using the GRAS, we evaluate recovery of function of individual digits and precise finger-thumb pinch to distinguish between compensatory grasp function and a return to pre-injury grasp patterns (Pessina et al., 2019).

Extracellular vesicle preparation and administration

EVs were extracted from MSCs harvested from the bone marrow of one young adult monkey as detailed in Moore et al., 2019. The EVs given to all animals in this study came from the same bone marrow sample and MSC extraction batch and therefore should be identical in content. Treatments were prepared at a concentration of 4 × 10¹¹ particles/kg in 10 mL of PBS via an Izon qNano particle counter. Monkeys received the treatment or a PBS vehicle intravenously at 24 h and 14 days post injury. Dosing was determined based on our rodent studies (Xin, Li, Cui, et al., 2013; Y. Zhang et al., 2017). Dosing intervals were based on the rodent studies 24 h administration. However, we chose to also administer a second dose of EVs at 14 days since larger animals have a slower neural recovery (Agoston, 2017) and based on previous experience with functional improvements in our rhesus monkey model.

Tissue collection and preparation

Brain and cervical spinal cord (CSC) were collected at 14 weeks post-injury following 12 weeks of behavioral testing. For perfusion-fixation, monkeys were sedated with Ketamine (10 mg/kg IM) deeply anesthetized with sodium pentobarbital (25 mg/kg IV to effect) and were euthanized by exsanguination during transcardial perfusion-fixation of the CNS. The brain was blocked, in situ, in the coronal plane and removed from the skull. The CSC was removed from the vertebral column and cut with a 45-degree angle to denote rostral from caudal and notched on the ipsilesional side for later processing. Tissue was post-fixed overnight in 4% paraformaldehyde and then transferred to cryoprotectant solution. Cryoprotected blocks were flash frozen at − 75 °C and stored at − 80 °C. Frozen sections were cut on a sliding microtome while frozen in dry ice into interrupted series of coronal sections (10 series of 30 µm thick sections) with a 300 µm spacing between sections. Sections were collected in phosphate buffer with 15% glycerol and stored at − 80 °C. Spinal cord tissue processed in the same way but cut into interrupted series of transverse sections (12 series of 40 µm thick sections) with a 480 µm spacing between sections. Sections were collected in phosphate buffer with glycerol and 30% sucrose stored at − 20 degrees.

Immunohistochemistry

Antibodies labeling synapses, dendrites, axons, neuron activation, inhibitory interneurons, and pyramidal neurons were used to assess CNS tissue. Microtubule associated protein 2 (MAP2) was used to label lower motor neurons and their associated dendrites. Synaptophysin (SYN) is a presynaptic integral membrane glycoprotein used to label synapses. Neurofilament (clone:SMI32) and calcium binding proteins calretinin (CR), parvalbumin (PV) and calbindin (CR) were used to label subpopulations of pyramidal neurons and inhibitory interneurons, while cfos labeled neuronal nuclei of cells that were active during recent behavioral tasks. All assays were batched processed and simultaneously thawed to room temperature.

Fluorescence

For immunofluorescence, sections were incubated in a blocking solution of 10% normal goat serum (NGS), 1 M glycine, and 0.4% triton for two hours and then incubated in the appropriate primary antibodies (1:500 MAP2 abcam; 1:500 SYN abcam; 1:400 cfos Synaptic Systems; 1:1000 PV; 1:1000 CB Swant; 1:2000 CR Swant) overnight at room temperature. The next day tissue was rinsed and incubated with the corresponding secondary antibodies (1:600; Alexa Fluor Invitrogen or Jackson Immunoresearch) in the dark for two hours. Sections were mounted on glass slides and cover-slipped with Prolong anti-fade medium (Life Technologies) and left at room temperature in the dark for 36 h before being imaged. Control experiments where the primary antibody was omitted showed no labeling.

Brightfield

Double labeled chromogen sections were sequentially run 1 label at a time. First sections were exposed to 3% hydrogen peroxide for 30 min, followed by a 1 h 10% NGS block and overnight primary antibody incubation (1:1000 SMI32 or 1:1000 cfos). The next day sections were incubated with the corresponding biotinylated secondary antibodies (Vector) for 1 h followed by a 1-hour amplification step (ABC Elite, Vector) and exposure to enzyme mediated color reactions. For the second label sections were rinsed and incubated in Superblock (Thermofisher) for 1 h. Antibody, amplification, and enzyme steps were repeated for the second label.

Data analysis

Brightfield cell counts

Cell density was assessed in the perilesional M1 cortex (periM1) as well as the dorsal premotor cortex (dPMC) on the ipsilesional hemisphere using the optical fractionator probe of Stereoinvestigator (MBF Bioscience) (Fig. 1B,C). Pyramidal neurons (SMI32⁺), active pyramidal neurons (SMI32⁺cfos⁺ cells), or other active neurons (SMI32^-cfos⁺) (Fig. 2B) were counted using 3–5 sections per monkey and the sampling scheme was optimized based on Gundersen’s Coefficient of Error (<0.1). Unfortunately, cortical injury made the upper layers very difficult to define in the periM1 (Fig. 1E), hence we only counted deep (4−6) layers. However, in the dPMC counts were segregated by upper (1−3) layers and deep (4−6) layers.

Fig. 1.

Histological Methods: A) Schematic of experimental design and sampling areas. After lesioning the hand area of the motor cortex, we assessed whether extracellular vesicle (EV) treatment enhanced neuronal integrity and repair in the ipsilesional cortex and contralesional cervical spinal cord. B, C) Brightfield images of Nissl-stained coronal sections showing cortical ROIs, which include L5–6 of the perilesional cortex (periM1, B) and L1–6 of the dorsal premotor cortex (dPMC, C) separated into upper (L1–3) and deep layers (L4–6). D) Nissl-stained coronal sections showing cervical spinal cord ROIs including the Ventral Horn (VH), as well as the Dorsal Horn (DH) as a control region. E, F) Brightfield images showing dual labeling of SMI32 (blue) and cfos (brown, not readily distinguished at this objective). Note the disrupted layer architecture within the periM1 near the lesion (E), compared to dPMC (F) further way from the lesion. Scale bars 500 µm.

Fig. 2.

MSCd-EV Effects on Task-relevant cfos⁺ Activation of Pyramidal Neurons in the Motor Cortices: A) Representative brightfield image of coronal section in dorsal premotor cortex (dPMC) showing dual labeling of SMI32 (blue) and cfos (brown). We analyzed the number of SMI32⁺ pyramidal neurons (black arrow; A) and SMI32⁺cfos⁺ presumably task-activated pyramidal neurons (white arrow) as well as the number of SMI32-cfos⁺ task activated non-pyramidal neurons (red arrows). B) Box and whisker plots showing the density of SMI32⁺ neurons, C) the subset of double labeled SMI32⁺cfos⁺ pyramidal neurons and D) double labeled SMI32-cfos⁺ neurons in periM1. E) Box and whisker plots showing the density of double labeled SMI32⁺cfos⁺ pyramidal neurons and F) double labeled SMI32-cfos⁺ neurons in dPMC. Gray/White circles represent outliers on the graphs. Scale bar 25 µm.

Fluorescent cell counts

To estimate the density of cfos⁺ neurons co-labeled with inhibitory markers (CB, PV or CR), 4-channel confocal image stacks were acquired (Zeiss 710; 20x, 0.8 N.A; 0.3 ×0.3 ×1 µm voxel). From each monkey, 8 fields, spaced 500 µm apart, in the upper (1−3) and deep (4−6) layers of the dPMC, were randomly selected and imaged throughout the entire z extent of two 30 µm section. In each z-stack, the cells single- and double-labeled with the markers of interest were counted using FIJI cell counter [(http://imagej.nih.gov/ij/, 1997–2016); RRID:SCR_003070; (Schindelin et al., 2012)]. Stereological counting rules were applied with inclusion and exclusion borders, as described by Fiala and Harris, (2001).

CSC analysis

FIJI software (National Institutes of Health) was used to analyze fluorescent images. To assess dendritic changes, the MAP2 channel was separated and z-projected onto a 2D plane. Background was subtracted and percent area assessed via FIJI measurement tools. We used MAP2 and SYN colocalization to estimate synaptic density along the dendritic shafts and somas of lower motor neurons as previously done (Skup et al., 2012; Maxwell et al., 2018; Lee et al., 2021). Briefly, co-labeled z-stacks of the VH and DH were split into two channels and run through the FIJI Coloc2 tool. Mander’s colocalization coefficient (CLC) was used for analysis.

Statistics

R Studio was used to run all statistical analyses. Cell densities and proportion of active neuronal subtypes were run through mixed two-way ANOVAs comparing between factors “group” (untreated, treated) and within subject factors “layers” (upper, deep). periM1 densities were run through Student T-tests since only deep layers were counted. Percent area or CLCs from CSC images were run through a two-way mixed ANOVA with between factors “group” (treated, untreated) and within subject factor “hemisphere side” (ipsilesional, contralesional). Post Hoc Student T-tests with Bonferroni corrections were run when appropriate. Outliers were determined based on 1.5x the interquartile range. Removal of outliers did not significantly change any of the statistics so we did not remove them from the final analyses, but they are noted on the graphs. Additionally, all outcomes were analyzed through a multiple linear correlation model against grasp recovery endpoints (days to return to pre-operative grasp and mean grasp scores during the final week) and HDT latency endpoints (days to return to pre-operative latency and retrieval latency during the final week) to assess associations with measures of functional recovery.

Results

MSCd EVs increase the density of cfos activated pyramidal neurons in the periM1

We examined the density of pyramidal neurons marked by SMI32, a non-phosphorylated neurofilament protein enriched in L3 and L5 pyramidal neurons (Campbell and Morrison, 1989), in the periM1 directly below and surrounding the lesion, where the pia is not intact. In all cases, the pia and upper layers L2–3 were not intact, thus pyramidal neurons were counted in L5–6 of periM1. Student T-test showed a trend of higher SMI32⁺ neuron density in the deep periM1 layers in EV treated monkeys (T(9,1)= 2.186, P = 0.070; Fig. 2B). Using dual immuno-labeling (Fig. 2 A), we then assessed the density of SMI32⁺ neurons expressing cfos, an intermediate early gene that is upregulated in task-related neurons, about 2 h after performing the motor task (Orczykowski et al., 2018). In the periM1, a Student T-test showed a significantly larger number of cfos activated SMI32 pyramidal neurons (SMI32⁺cfos⁺) in EV treated monkeys (T(9,1)= 2.50, P = 0.020; Fig. 2 C). Student T-test of SMI32^-cfos⁺ neurons showed no significant difference between groups (T(9,1)= 1.361, P = 0.212; Fig. 2D). These data indicate a higher number and task-related activity of SMI32 pyramidal neurons in the periM1 of EV treated monkeys after cortical injury.

MSCd EVs decrease the density of cfos activated pyramidal neurons in the dPMC

Neurons in premotor cortices have been shown to exhibit compensatory hyperactivation after injury in M1 (Carmichael, 2012, Kohman and Rhodes, 2013, Lai et al., 2014, Jayaraj et al., 2019). To investigate whether damage in M1 resulted in compensatory changes in pyramidal neurons and task-related activation in premotor areas, we assessed the density of SMI32⁺ and cfos⁺ neurons in layers 2–3 and 5–6 of dorsal PMC (dPMC). Neurons in upper L2–3 of dPMC are one source of cortico-cortical projections to M1, while the deep L5–6 neurons contain subsets that project to the spinal cord and other subcortical structures (Dum and Strick, 1991). A two-way ANOVA revealed no difference in the number of SMI32⁺ pyramidal neurons between groups (F(9,1)= 1.401, P = 0.256), no effect of layer (F(9,1)= 3.093, P = 0.10) and no interactions (F(9,1)= 0.519, P = 0.483). SMI32⁺cfos⁺, presumably task-activated pyramidal cells, show no significant differences between groups (F(9,1)= 3.31, P = 0.092; Fig. 2E), no effect of layer (F(9,1)= 2.171, P = 0.164) and no interactions (F(9,1)= 2.25, P = 0.158). SMI32^-cfos⁺ neurons showed no significant difference between groups (F(9,1)= 1.03, P = 0.328; Fig. 2 F), or layers (F=0.216, P = 0.649) and no significant interaction (F(9,1)= 0.182, P = 0.676).

To further investigate the laminar distribution of these presumably task-activated neurons, we mapped the percent of cfos⁺ active SMI32⁺ and SMI32^- neurons across upper and deep layers of the cortex. In both groups SMI32^-cfos⁺ neurons show a higher concentration in the upper layers (Vehicle:59.5%, EVs:59.1%), while SMI32⁺cfos⁺ neurons show a high concentration in deep layers (Vehicle:71.1%, EVs:78.12%). A subset of these deep layer pyramidal neurons likely contributes to long-range cortico-subcortical motor projections, while neurons in the upper layer are either cortico-cortical projections or local inhibitory interneurons (Ralston et al., 1985; Dum and Strick, 1991, Hof et al., 1995; Morecraft et al., 2013, Morecraft et al., 2019). Hence, we further subtyped the active cfos⁺ neurons.

MSCd EVs increase the density of cfos activated inhibitory interneurons in the dPMC

A cortical insult to the motor cortex, can result in hyperexcitability of the surviving motor and premotor neurons (Carmichael, 2012, Kohman and Rhodes, 2013, Lai et al., 2014, Jayaraj et al., 2019). Inhibitory interneurons are recruited to dampen hyperexcitability and prevent excitotoxicity to promote functional recovery (Carmichael et al., 2012; Medalla & Chang et al. 2020). Thus, we investigated the density of cfos⁺ activated inhibitory neuron sub-types labeled for the calcium binding proteins, parvalbumin (PV), calbindin (CB), and calretinin (CR) in upper versus deep layers of dPMC (Fig. 3D, E, F). These inhibitory neuron subtypes constitute three non-overlapping populations that have distinct innervation patterns and modes of inhibition in the primate cortex (DeFelipe, 1997 review). Two-way ANOVAs showed that the density of CR, CB, and PV total cells did not significantly differ between groups (CR: F(9,1)= 0.644, P = 0.438; CB: F(9,1)= 1.254, P = 0.285; PV: F(9,1)= 2.965, P = 0.111; Fig. 3A-C). However, post-hoc Student T-tests with Bonferroni corrections show CR⁺ (P = 0.00004; Fig. 3 A) and CB⁺ (P = 0.019; Fig. 3B) neurons were denser in the upper (1−3) layers compared to deeper (4−6) layers, while PV⁺ neurons had more equal distribution across all layers (P = 0.310; Fig. 3 C), which is consistent with previous literature (Condé et al., 1994).

Fig. 3.

MSCd-EV Effects on Task-relevant cfos⁺ Activation of Inhibitory Interneurons in the dPMC: A-C) Box and whisker plots showing the density of single labeled inhibitory interneurons expressing the calcium binding proteins calretinin (CR), calbindin (CB), or parvalbumin (PV). D-F) Box and whisker plots showing the density of double labeled, presumably task-related, neurons with cfos and calcium binding proteins (CR, CB, PV). G-I) Representative confocal images of interneuron markers and cfos. Gray/White circles represent outliers on the graphs. Scale bars 25 µm.

The proportion of cfos⁺ cells for each subtype was further evaluated. A two-way ANOVA of CR⁺cfos⁺ neurons (Fig. 3D) reveal no effect of group (F(9,1)= 0.079, P = 0.784), or layer (F(9,1)= 0.007, P = 0.937), and no interactions (F(9,1)= 0.389, P = 0.545). Similarly, two-way ANOVA of PV⁺cfos⁺ neurons (Fig. 3 F) shows no significant difference between groups (F(9,1)= 1.523, P = 0.241), no effect of layers (F(9,1)= 0.278, P = 0.608), and no significant interactions (F(9,1)= 0.001, P = 0.994). A two-way ANOVA of CB⁺cfos⁺ neurons (Fig. 3E) shows no difference between groups (F(9,1)= 2.731, P = 0.124), but a significant effect of layer (F(9,1)= 21.651, P = 0.0005) and significant interactions between group and layer (F(9,1)= 7.238, P = 0.019). Post-Hoc Student T-tests with Bonferroni corrections show treated monkeys have significantly more CB⁺cfos⁺ neurons in the upper layers (P = 0.003), but no treatment related differences in deep layers (P = 0.864). Consistent with our previous findings of the vPMC (Medalla & Chang et al. 2020), these data indicate an increased activity of the inhibitory CB⁺ interneurons in the upper layers of the dPMC in treated monkeys.

MSCd EVs reduce LMN synapses in ventral horn LMNs

Since evidence suggests EV treatment prevented upper motor neuron loss and dysfunction in the periM1, we further investigated the reorganization of cortico-spinal projections at the level of the cervical spinal cord (CSC). We then assessed the density of putative synaptic connections in the ventral horn (VH) of the spinal cord during recovery (Fig. 4E) by quantifying the CLC of the presynaptic protein SYN, localized on synaptic active zones, and MAP2, a protein expressed by the LMN somas and dendrites. In the VH, where LMNs reside, a two-way ANOVA of SYN⁺MAP2⁺ CLC (Fig. 4 A) showed a significant difference between treatment groups (F(9,1)= 11.164, P = 0.0049), no significant effect of hemisphere (F(9,1)= 0.264, P = 0.615) and no interactions (F(9,1)= 0.181, P = 0.676). Post-hoc Student T-tests with Bonferroni corrections reveal that EV treated monkeys have significantly fewer SYN⁺MAP2⁺ putative synapses in the contralesional VH (lesioned pathway; P = 0.019) but no differences in the ipsilesional VH (non-lesioned pathway; P = 0.183).

Fig. 4.

MSCd-EV Effects on Axonal Sprouting, Synapse Density and Dendrite Density in Spinal Cord: (A,B) CLC of synaptic maker, SYN, with dendritic marker, MAP2, as an estimate of putative synapses in the contralesional (lesioned pathway) and ipsilesional (non-lesioned pathway) ventral horn (VH; A) as well as the dorsal horn (DH; B) in order to assess changes outside the motor circuit. (C,D) Percent area label of MAP2 expression in the contralesional (lesioned pathway) and ipsilesional (non-lesioned pathway) VH (C) and DH (D). E (E) Representative confocal image and inset in higher magnification, showing SYN⁺ axon terminals in the red channel, MAP2⁺ somas and dendrites in the green channel, and the SYN⁺MAP2⁺ colocalization in the merged channel, with colocalization in yellow (white arrows) in VH of the cervical spinal cord. Gray/White circles represent outliers on the graphs. Scale bars 25 µm.

To see if these UMN injury effects are specific to the VH, we also examined the dorsal horn (DH), where the sensory neurons of the CSC receive inputs from peripheral somatosensory receptors. In the DH (Fig. 4B) a two-way ANOVA showed no significant difference between hemispheres (F(9,1)= 1.27, P = 0.292) or groups (F(9,1)= 0.855, P = 0.382) and no interactions (F(9,1)= 1.85, P = 0.211). These data show that EV treated monkeys have significantly fewer SYN⁺MAP2⁺ putative synaptic interactions in the VH compared to vehicle controls.

MSCd EVs increase MAP2 staining in the ventral horn

Finally, dendritic complexity and MAP2 expression are reduced when LMNs are denervated (Jones and Thomas, 1962, Kolb and Winshaw, 1998). Hence, we evaluated density of dendritic MAP2 expression in the VH and DH of the CSC. A two-way ANOVA of MAP2 percent area labeled in the VH (Fig. 4 C) showed a significant difference between groups (F(9,1)= 21.069, P = 0.0004), no significant effect of hemispheres (F(9,1)= 0.395, P = 0.538), and no interactions (F(9,1)= 0.398, P = 0.538). Post-Hoc Student T-tests with Bonferroni corrections revealed that compared to vehicle, EV treated monkeys had significantly more MAP2 staining in the contralesional (lesioned pathway; P = 0.044) and ipsilesional VH (non-lesioned pathway; P = 0.007). Additionally, outside of the motor circuit a two-way ANOVA of MAP2% area in the DH (Fig. 4D) showed a significant difference between groups (F(9,1)= 15.295, P = 0.0016), no effect of hemispheres (F(9,1)= 0.144, P = 0.710) and no interactions (F(9,1)= 0.112, P = 0.743). Post-Hoc Student T-tests with Bonferroni corrections show increased MAP2 expression in the contralesional DH (lesioned pathway; P = 0.049) of EV versus vehicle treated monkeys, and no differences in the ipsilesional DH (non-lesioned pathway; P = 0.13156). In summary, we found total MAP2 staining in the EV treated monkeys was significantly greater in both the ipsilesional and contralesional VHs as well as in the contralesional DH, compared to those in vehicle treated monkeys.

Cortical and cervical spinal neuronal distribution and structure correlate with measures of recovery

Multiple linear correlation analysis showed an association between both cortical neuronal distribution and spinal cord structural outcomes with measures of faster motor recovery. The number of periM1 SMI32⁺cfos⁺ task-related neurons were negatively correlated with days to return to pre-op retrieval latencies (R=0.579, P = 0.015; Fig. 5 A) and trending with days to pre-op grasp patterns (R=0.392, P = 0.058; Fig. 5 A). In the dPMC upper layers, the presumably task-related CB+cfos⁺ inhibitory interneuron exhibited a significant negative correlation with days to pre-op retrieval latencies (R=0.679, P = 0.017; Fig. 5B). In the VH, the contralesional SYN⁺MAP2⁺ CLC showed a positive trend with days to pre-op retrieval latencies (R=0.626, P = 0.083; Fig. 5 C). Lastly, contralesional MAP2 staining in the VH was negatively correlated with days to pre-op retrieval latencies (R=0.769, P = 0.026; Fig. 5D). In summary these data show that increasing cfos⁺ activity of pyramidal neurons in the periM1 as well as the activity of inhibitory interneurons in the upper dPMC correlated with faster recovery of motor function. Additionally, fewer SYN⁺MAP2⁺ synapses in the VH and increased MAP2 staining also correlated with improved functional recovery.

Fig. 5.

Relationships with Motor Performance: Multiple linear correlations show significant correlations between histological measures and motor recovery measured as the number of days it took animals to return to pre-op functions. A) Task activated pyramidal neurons (SMI32+cfos+) in the periM1 and B) task activated calbindin interneurons (CB+cfos+) in the dPMC are associated with recovery of pre-op functions. Additionally, in the contralesional VH C) fewer LMN synapses and D) great MAP2+ expression associated with faster recovery of retrieval latencies.

Discussion

In this study, we investigated the efficacy of MSCd EVs to reduce the cascade of secondary damage that follows cortical injury and how it contributes to neuronal recovery of the cortical and spinal motor circuits. Our previous study reported that monkeys treated with EVs evidence a significant recovery of fine motor function of the hand during the first few weeks after cortical injury in M1 and attain a full return to pre-operative retrieval latencies and grasp patterns by 14 weeks post-injury (Moore et al., 2019). In the present study, histological examination 14 weeks after injury revealed that EV treatment was associated with 1) greater densities of SMI32⁺ pyramidal neurons in deep layers 5–6 of periM1, including the subset double-labeled with cfos⁺ and 2) in the dPMC, greater densities of presumably task-related cfos⁺ inhibitory interneurons in the upper layers, compared with vehicle treatment. In the CSC, EV group exhibited 1) lower LMN synaptic density in the contralesional VH with 2) a bilateral increase in dendritic MAP2 expression. These treatment-related differences in upper and lower motor neurons were associated with improved functional outcomes.

Cortical remodeling

In this study, we showed that increased task-related cfos activation of SMI32⁺ pyramidal neurons in periM1 were significantly correlated with improved functional recovery of treated monkeys. Specifically, deep layer 5–6 pyramidal neurons in PMC and M1 contribute to cortico-spinal projections (Ralston et al., 1985; Dum and Strick, 1991, Hof et al., 1995; Morecraft et al., 2013, Morecraft et al., 2019) and are enriched with SMI32, a non-phosphorylated neurofilament heavy chain protein (Tsang et al., 2000). In the current study, EV treated monkeys had a greater density of intact and task-related SMI32⁺ pyramidal neurons in periM1 deep layers. This suggests that after injury, EV-treated monkeys may have had less secondary damage to M1 promoting survival and function of UMNs, particularly those that send long distance projections to the spinal cord.

Our previous studies have shown that increased cfos activation in the vPMC following injury in M1 is correlated with functional recovery (Orczykowski et al., 2018; Medalla & Chang, 2021). In the current study, we specifically showed a strong correlation between recovery and increased cfos activity of inhibitory interneurons in upper cortical layers of dPMC. Hyperexcitability is a common consequence of cortical injury (Carmichael, 2012, Kohman and Rhodes, 2013, Lai et al., 2014, Jayaraj et al., 2019) due to increased extracellular glutamate concentration and stimulation of the neuronal death cascade (Lai et al., 2014). Inflammatory astrocytes and microglia have been shown to further promote excitotoxicity (Vinet et al., 2012, Pekny et al., 2016). However, MSCd EVs can modulate inflammatory responses in rodent models (Di Trapani et al., 2016; Phinney et al., 2017; Zhang et al., 2014), including reduction of astrogliosis (Xian et al., 2019) and microglia reactivity (Ruppert et al., 2018; Howe and Barres, 2012). We previously reported in monkeys that MSCd EVs encourage ramified microglia after cortical injury (Go et al., 2020) and using single-cell in vitro electrophysiology of vPMC pyramidal neurons, we showed that EV treatment promoted restoration of E:I balance (Medalla & Chang et al., 2020). We expand on this work by showing inhibitory task-related cfos+ activity in the dPMC was also associated with laminar specific pyramidal neuron activity in periM1. Greater cfos activation of CB⁺ inhibitory neurons with EV treatment suggests a mechanism by which increased inhibitory inputs onto pyramidal neurons may help maintain the excitatory and inhibitory balance and is consistent with our previous electrophysiology (Medalla & Chang et al., 2019). CB⁺ neurons, which target distal dendrites of pyramidal neurons to modulate dendritic excitability, are highly resistant to hyperexcitability after injury (Mattson et al., 1991; Phillips et al., 1999; D’Orlando et al., 2002) and may play an important role in recovery after injury. Although further investigation of molecular effects on inhibitory mechanisms is needed, the combination of greater number of intact pyramidal neurons in the deep layers of periM1 and potential balancing of excitatory and inhibitory activity in PMC with EV treatment may allow greater control of LMNs and motor function.

Spinal cord plasticity

After injury to the motor cortex, upper motor neurons die leading to reduced signals to LMNs (Lipton, 1999). During recovery, plasticity in the cortex can occur to reorganize motor circuits (Carmichael et al., 2003; Nudo, 2006; Dancause et al., 2005; Kantak et al., 2012). Yet, even for UMNs that survive the lesion, their long-distance cortico-spinal projections into the lateral corticospinal tract (LCST) could be susceptible to damage and therefore limit the recovery of fine motor movements. However, plasticity in the LCST axon terminations and LMNs within the spinal cord may compensate for loss of cortical input. Indeed, studies have shown, in both rodents and primates, intact CST axons send collaterals and increase synaptic connections with denervated neurons after injury (Wiessner et al., 2003; Weidner et al., 2001; Lee et al., 2004; Zai et al., 2009; Wahl et al., 2014; Fouad et al., 2004; Freund et al., 2006; Rosenzweig et al., 2009, Rosenzweig et al., 2010). However, it has also been noted that aberrant plasticity after nerve injury can be harmful to functional recovery (Kim et al., 2015, Beauparlant et al., 2013). Specifically, in rodents, Kim et al. (2015) showed a negative correlation between cortical synaptic density and functional improvement after stroke, and Beauparlant et al. (2013) reveals increased synaptic plasticity after axonal injuries correlated with motor neuron dysfunction. As follows, our study shows EV treated monkeys have significantly fewer VH synapses compared to vehicle-treated control monkeys. Additionally, increased VH synapses correlated with worse behavioral outcomes suggesting too many aberrant connections could interfere with motor recovery. EV treatment may have reduced undirected aberrant plasticity either by preserving LMN innervations due to reduced cortical damage, as evidenced by a greater number of intact and task-activated M1 pyramidal neurons, or acted independently in the spinal cord to direct LMN plasticity in a more specified manner. Moreover, MAP2 expression is increased in the CSC of EV treated monkeys, which is consistent with our previous data showing increased dendritic branching complexity in cortical pyramidal neurons (Medalla & Chang et al., 2020). MAP2 is required for dendritic growth and stabilization (Harada et al., 2002) and studies have shown that when neurons lose their afferent connections, dendrites begin to atrophy (Jones and Thomas, 1962; Kolb and Winshaw, 1998). Therefore, maintenance of upper motor neuron connections may have prevented dendritic atrophy of LMNs in EV-treated monkeys. Further investigation is needed to determine the mechanism and specific effects of EVs on spinal cord circuits.

Bilateral effects of EV treatment on CSC MAP2 expression may suggest contributions of UMN pathways from both contralesional (lesioned) and ipsilesional (non-lesioned) motor areas, of which both pathways have been shown to play a role in spinal cord reorganization after cortical injury (Buetefisch, 2015, Zai et al., 2009, Gonzalez et al., 2004, Hoff and Hoff, 1934, Tigges et al., 1979, Nakagawa, 1980 Jun 9; Ralston & Ralston, 1985; Alawieh et al., 2017). Interestingly, we also saw significant differences of MAP2 staining between EV and vehicle treated monkeys in the contralesional somatosensory neurons of the DH as well. These data suggest that EVs may have a more global effect on spinal cord dendrites or that injury to the motor cortex may have induced changes in the complementary sensory regions. Indeed, studies indicate that motor learning affects both motor and sensory areas of the brain (Ostry et al., 2010, 2016) and recovery after stroke can recruit activity from sensory and motor areas (Zemke et al., 2003). Further investigation is needed to understand these recovery- and treatment-related changes in dendritic structure in distinct sensory and motor cortical and spinal cord regions.

Conclusion

The consequences of cortical injury can lead to long lasting disability and loss of fine motor function. Although there are treatments to remove vessel blockages to re-perfuse tissue (Koroshetz, 1996), there are currently no FDA approved treatments to improve neural recovery after injury. EVs derived from mesenchymal stem cells are filled with immune modulating and growth signaling miRNA, mRNA, and proteins (Chopp et al., 2009; Di Trapanni et al., 2016; Casado et al., 2017), which may be beneficial for neural repair. While future molecular and proteomic studies are underway to assess the specific content of these MSC-EVs derived from monkey bone marrow, our series of studies show that EV treatment after cortical injury improves functional fine motor recovery (Moore et al., 2019), decreases perilesional inflammatory microglia and myelin damage (Go et al., 2020, Go et al., 2021), reduces hyperexcitability, and maintains E:I balance in vPMC neurons (Medalla & Chang et al., 2020). In the current study we showed that EV treatment is further associated with reorganization of cortical and spinal motor circuits. Specifically in the cortex, EVs appear to promote pyramidal neuron survival and task-related cfos⁺ activation in L5/6 of periM1 and inhibitory interneuron activity in the dPMC. In the spinal cord, EVs reduced potential aberrant synaptic plasticity in the VH as evidenced by fewer SYN⁺MAP2⁺ synapses. Hence, the immune and growth signals from MSCd EVs may contribute to the observed faster recovery of motor function after cortical injury by reducing neuronal damage and denervation of LMNs as well as dampening dysfunctional plasticity. Overall, our data demonstrate that MSCd EVs are a potential therapeutic for increasing functional recovery after cortical injury, likely through multifaceted mechanisms of inflammation and neural plasticity.

Funding

This work was supported by the National Institutes of Health [NIH/NINDS R56 NS112207, NIH/NIA R21 NS111174, NIH/NINDS R21-NS102991].