Ischemic Stroke-Induced Polyaxonal Innervation at the Neuromuscular Junction is Attenuated By Robot-Assisted Mechanical Therapy

Abstract

Ischemic stroke is a leading cause of disability world-wide. Mounting evidence supports neuromuscular pathology following stroke, yet mechanisms of dysfunction and therapeutic action remain undefined. The objectives of our study were to investigate neuromuscular pathophysiology following ischemic stroke and to evaluate the therapeutic effect of Robot-Assisted Mechanical massage Therapy (RAMT) on neuromuscular junction (NMJ) morphology. Using an ischemic stroke model in male rats, we demonstrated longitudinal losses of muscle contractility and electrophysiological estimates of motor unit number in paretic hindlimb muscles within 21 days of stroke. Histological characterization demonstrated striking pre- and postsynaptic alterations at the NMJ. Stroke prompted enlargement of motor axon terminals, acetylcholine receptor (AChR) area, and motor endplate size. Paretic muscle AChRs were also more homogenously distributed across motor endplates, exhibiting fewer clusters and less fragmentation. Most interestingly, NMJs in paretic muscle exhibited increased frequency of polyaxonal innervation. This finding of increased polyaxonal innervation in stroke-affected skeletal muscle suggests that reduction of motor unit number following stroke may be a spurious artifact due to overlapping of motor units rather than losses. Furthermore, we tested the effects of RAMT – which we recently showed to improve motor function and protect against subacute myokine disturbance – and found significant attenuation of stroke-induced NMJ alterations. RAMT not only normalized the post-stroke presentation of polyaxonal innervation but also mitigated postsynaptic expansion. These findings confirm complex neuromuscular pathophysiology after stroke, provide mechanistic direction for ongoing research, and inform development of future therapeutic strategies.

Introduction

Ischemic stroke is a leading cause of chronic disability world-wide (1,2). Motor impairment after stroke is attributed to cortical or subcortical lesions (3-6) with secondary damage from cerebral edema and neuroinflammation (7, 8). Additional factors such as maladaptive compensatory reorganization, uncoupling of central and peripheral nervous systems, and loss of descending input to skeletal muscle further compound the severity and evolution of functional disability (9-11). There is some evidence for complex peripheral pathophysiology in the neuromuscular system following ischemic stroke. Termed “stroke-related sarcopenia,” this disease-specific state of rapid muscle decline prompts maladaptive responses at the neuromuscular interface and structural changes in skeletal muscle cells (12, 13).

Recent data suggest skeletal muscle dysfunction directly contributes to the burden of stroke (12-15). Task-related coupling between the motor cortex and skeletal muscle is reduced in the acute and subacute phases (16). Chronic stroke patients exhibit up to 24% reduction in paretic muscle volume compared to the nonparetic limb (17), with resulting weakness and atrophy that amplify impairment and complicate recovery (18-20). Paretic muscle fibers shift to more fatigable isoforms, the degree of which inversely correlates with patient walking speed (14). Additionally, stroke-affected muscle shows altered myokine expression related to catabolic, inflammatory, and reparative pathways (15, 21, 22). Muscle function and size are dependent on factors intrinsic to muscle (e.g., sarcolemmal excitability, contractility) as well as upstream neurological factors (e.g., synaptic transmission, motor unit number) (23-26). The peripheral nervous system has received limited attention in preclinical and clinical stroke research, but reports of variable EMG-force relations (30, 31) and motor unit remodeling (32-34) indicate peripheral nerve involvement.

Following acute phase interventions, stroke disability is largely managed through rehabilitation (35, 36), with the goal of promoting beneficial neuroplastic reorganization that strengthens both neural pathways and paretic muscle (37-39). However, best practices across the spectrum of post-stroke therapies are unclear, and evidence-based recommendations are few (35, 36). To that end, we previously evaluated the effects of paretic hindlimb massage on rat functional recovery after stroke (40). This prior work established preclinical evidence that mechanical massage, delivered via Robot-Assisted Mechanical Therapy (RAMT), improved post-stroke sensorimotor behavior, preserved gait, and attenuated subacute expression of inflammatory and metabolic myokines (40). Massage in various contexts has shown therapeutic benefit to motor and neuromuscular systems by modulating inflammation, cellular stress, perfusion, myokine secretion, and signal transduction (41-45). Still, the peripheral mechanisms that contribute to stroke-specific pathophysiology and hold potential for clinically-relevant therapeutic modification remain undefined (35, 36, 46). Taken together, the current work sought to characterize maladaptive changes to the neuromuscular system following ischemic stroke and explore the effect of RAMT treatment on stroke-induced neuromuscular alterations. We first examined neuromuscular pathophysiology in a rat model of acute ischemic stroke using in vivo longitudinal assessments of muscle contractility and motor unit connectivity, and histochemical analysis of myofiber oxidative capacity (Experiment 1). Second, we characterized stroke-induced changes to pre- and postsynaptic elements of the neuromuscular junction (NMJ) and evaluated the associated therapeutic potential of RAMT on NMJ morphology (Experiment 2).

Materials and Methods

Experimental Design

All experiments were approved by the Institutional Animal Care and Use Committee of The Ohio State University (Columbus, OH, USA). Male Wistar rats (N=48) (Harlan Laboratories, Indianapolis, IN, USA) were used for all studies. Animals were housed in a controlled vivarium with 12-hour light/dark photoperiods and access to food and water ad libitum.

Experiment 1 was designed to assess changes in muscle contractility, motor unit electrophysiology, and myofiber type following ischemic stroke (Table 1; Figure 1A). Rats were subjected to the middle cerebral artery (MCA) occlusion (MCAO) method of ischemic stroke induction (Stroke group) or sham surgery (Sham group). In vivo longitudinal testing of contractility (plantarflexors) and motor unit function (gastrocnemius) was performed at pre-stroke baseline (BL) and repeated at 7, 14, and 21 days post-stroke. Each testing session measured (a) twitch and tetanic plantarflexion contraction torque and (b) electrophysiological measures of compound muscle action potential (CMAP) amplitude and average single motor unit potential (SMUP) amplitude, from which motor unit number estimation (MUNE) was calculated. Animals were weighed at the beginning of each testing session. All contractility and electrophysiology measurements were performed on the left (stroke-affected) hindlimb blinded to group. An additional cohort of rats with (D14 Stroke group) and without stroke (D14 Sham group) were used for myofiber analysis (soleus).

Table 1.

Objectives and Design

Objectives by Experiment	Control Groups	Stroke Groups	Analysis Parameters
Experiment 1: In vivo: Assess longitudinal impact of stroke on muscle contractility, electrophysiology, and motor unit connectivity	Sham n=3	Stroke n=7*	Maximum plantarflexion twitch and tetanic torques following tibial nerve stimulation (in vivo) MUNE, derived from electrophysiological output of gastrocnemius after sciatic nerve stimulation (in vivo)
Histology: Examine stroke-induced changes in myofiber phenotype	D14 Sham n=3	D14 Stroke n=8	Fiber type distribution, determined by histochemical visualization of soleus myofiber oxidative capacity
Experiment 2: Characterize NMJ morphology and innervation following stroke	Naïve n=5 (158 NMJs)	Stroke−RAMT n=6 (233 NMJs)	Presynaptic (nerve terminal), postsynaptic (AChR), and derived variables of gastrocnemius NMJ morphology
Test therapeutic modulation of NMJ alterations via RAMT		Stroke+RAMT n=6 (279 NMJs)	Number of axonal inputs at each junction; identification of either mono- or polyaxonal innervation

Experiment 1: Sham = sham MCAO; Stroke = MCAO. [*n=4 for contractility].

Experiment 2: Naïve = healthy; Stroke^−RAMT = MCAO, no treatment; Stroke^+RAMT = MCAO and RAMT.

Figure 1. Experimental Investigation of Neuromuscular Effects Following Stroke and RAMT.

(A) Experiment 1 scheme for in vivo longitudinal study of muscle contractility and motor unit connectivity with myofiber analysis. (B) Experiment 2 scheme for histological characterization of NMJ morphology following stroke and RAMT treatment. Representative coronal MR images from (C) Experiment 1 (Stroke, top; Sham, bottom) and (D) Experiment 2 (Stroke^−RAMT. top; Stroke^+RAMT , bottom). (E) Diagram showing RAMT administration.

Experiment 2 examined NMJ morphology and innervation after stroke and evaluated NMJ parameters as potential therapeutic targets of RAMT (Table 1; Figure 1B). Three experimental groups were assessed: Stroke^−RAMT rats (MCAO only, no treatment), Stroke^+RAMT rats (MCAO followed by daily RAMT), and Naïve rats (healthy age-matched controls, no stroke, no RAMT). Gastrocnemius muscles from the left (stroke-affected) hindlimb of Stroke^−RAMT and Stroke^+RAMT rats were harvested at post-stroke D14 and frozen for histological analysis of NMJs. Healthy gastrocnemius muscles were harvested from the left hindlimb of age-matched Naïve rats for comparison.

In vivo Procedures

Animal Preparation and Anesthesia.

Rats were anesthetized for MCAO or sham surgery, MRI, electrophysiology and contractility testing, and RAMT treatment by inhaled isoflurane (5% induction, 1.5-2% maintenance) delivered in medical air. Artificial tear ointment was applied to the eyes. Body temperature was maintained at 37°C using a heated table or warming pad, and respiration rate was monitored for the duration of each procedure. Rats recovered from all anesthetic events in their cage on heated surface maintained at 37°C under observation. Following stroke induction, animals received 3mL warm (37°C) sterile NaCl via subcutaneous injection and were closely evaluated for body temperature, body weight, incision healing, and overall health.

Ischemic Stroke Induction.

Transient (60min) cerebral ischemia was induced in the right hemisphere of 8- to 10-week-old rats using the intraluminal suture method of MCAO (40, 47). Following permanent distal ligation of the right external carotid artery, a 4-0 monofilament nylon suture with silicon tip was inserted and directed to the common carotid bifurcation, then advanced into the right internal carotid artery to the origin of the MCA. Laser Doppler Flowmetry (DRT4; Moor Instruments, Wilmington, DE, USA) monitored cortical blood flow and confirmed occlusion (>40-70% decrease in MCA-supplied territory). Sham MCAO surgery was performed as presented above, but without suture placement and MCA occlusion. Animals were monitored for health through the duration of the study. One rat from the D14 Stroke group (n=9) and two rats from the Stroke^+RAMT group (n=10) were excluded per IACUC protocol early removal criteria, specifically respiratory distress, inability to eat or drink, or >20% body weight loss (Table 1).

Brain MRI.

Brain imaging (9.4-T MRI, Bruker BioSpec 94/30, Billerica, MA, USA) was performed at 48hr post-MCAO (± 24hr) to confirm MCA stroke (Figure 1C,D). After localizer scans, T2-weighted spin echo with Rapid Acquisition with Relaxation Enhancement (RARE) sequence providing 8 echo train length was applied (FOV = 30×30 mm; acquisition matrix = 256×256; TR = 3500ms; TE = 46.924ms; flip angle = 162 degrees; images in acquisition = 30; resolution = 8.533 pixels/mm; number of averages = 2). Four rats from the Stroke^−RAMT group (n=10), and two additional rats from the Stroke^+RAMT group, were excluded due to unsuccessful MCA-territory infarction (Table 1).

Motor Unit and Muscle Electrophysiology.

CMAP and average SMUP amplitudes were recorded from the left (stroke-affected) gastrocnemius after stimulation of the sciatic nerve using a clinical electrodiagnostic device (Cadwell, Kennewick, WA, USA), as previously described by our group (48, 49). In brief, the hindlimb and posterolateral thigh were shaved with clippers. Rats were placed prone with hindlimbs abducted at the hips, extended at the knees, and secured with tape. One active and one reference surface disk recording electrode were placed over the gastrocnemius mid-belly and the calcaneus, respectively (TECA 6030-TP, Natus Neurology, Middleton, WI, USA). To reduce skin/electrode electrical impedance, both disks were coated with electrode gel (Spectra 360, Park Labs, Fairfield, NJ, USA) before placement on the skin. A disposable tab adhesive electrode (Natus Neurology, Middleton, WI, USA) on the tail served as a ground. Two disposable 28-gauge insulated monopolar needle electrodes (TECA, Oxford Instruments Medical, NY, USA) were inserted at the proximal hindlimb subcutaneously to stimulate the sciatic nerve. The low and high frequency filters were set to 10Hz and 10kHz, respectively. The sciatic nerve was stimulated at increasing intensities until the maximum CMAP response was observed (square wave pulse, 0.1-0.2ms duration, from 1mA to 20mA intensity). A supramaximal stimulation (120% stimulus required for maximum response) was then delivered, and baseline-to-peak and peak-to-peak CMAP amplitudes were recorded. For SMUP measurements, also optimized by our group (48, 49), gradually-increasing submaximal sciatic nerve stimulations (adjusted in 0.027mA steps) were delivered to obtain ten all-or-none increments, all of which were averaged to determine the average SMUP amplitude. To calculate MUNE, the CMAP peak-to-peak amplitude was divided by the average SMUP peak-to-peak amplitude.

Muscle Contractility.

After electrophysiological recording, plantarflexion torque was assessed using an in vivo contractility apparatus, with a force footplate on rotating axis connected to a force detection motor (Model 1300A, Aurora Scientific Inc, Canada), akin to methods described previously by our group in mice (50-53) and rats (48). The left hindpaw was taped to the force plate such that the hindpaw dorsum was angled 90° to the anterior tibia. The leg was extended with the knee joint securely clamped, non-invasively, at the femoral condyles, avoiding compression of the fibular nerve at the fibular head. Two disposable 28-gauge insulated monopolar needle electrodes (Natus Neurology, Inc, Middleton, WI, USA) were inserted subcutaneously over the tibial nerve, just posteromedial to the knee.

To identify the maximum stimulus intensity, a series of increasing square stimuli (0.2ms pulse duration) were delivered at 0.5Hz frequency until the twitch response no longer increased. Maximal twitch torque was then recorded following a single supramaximal stimulus. Next, a 200ms train of supramaximal square stimuli was delivered at 125Hz frequency to acquire the maximum tetanic contraction torque. Three rats from the Stroke group (n=7) were excluded from contractility analyses after confirming insufficient stimulation during baseline acquisition based on tetanic waveform visualization; thus, for contractility data, the Stroke group was n=4 (Table 1).

RAMT Treatment.

As previously published (40), the RAMT device was designed to facilitate objective study of mechanical physiotherapy on preclinical stroke outcomes. Following MCAO to induce stroke, Stroke^+RAMT rats received 30min of RAMT treatment, 5 days per week, beginning post-stroke D1 and continuing to D14, for a total of 10 treatments. For each treatment session, rats were maintained under inhaled isoflurane (1.5-2%). The paretic hindlimb was then positioned under an 8mm-diameter contact head connected to a load cell on a z-axis stage, controlled for mechanical force, frequency, and pattern of motion. Administered to the left (stroke-affected) hindlimb over the medial aspect of the gastrocnemius muscle, RAMT was delivered at 0.5N force across a 10mm linear path (proximal-to-distal pattern, 1Hz) (Figure 1E). To control for potential effects of anesthesia during RAMT sessions, the Stroke^−RAMT group received a matching dose, schedule, and duration of anesthesia.

Histology and Immunohistochemistry

Nicotinamide adenine dinucleotide (NADH) tetrazolium reductase (-TR) histochemistry.

Right and left soleus muscles were fixed and cryoprotected in a series of overnight incubations (formalin, 15% sucrose, 20% sucrose) and frozen in optimal cutting temperature embedding medium (OCT) floated on liquid nitrogen, then stored at −80°C until sectioning. For histochemical analysis of myofiber type distribution, transverse 10μm sections were incubated 30min at 37°C in a solution of 1.5mM nitrotetrazolium blue chloride (NBT; Sigma), 1.5mM β-NADH-reduced disodium salt hydrate (Sigma), and 0.2M Tris HCl at pH 7.4. After a series of 2min acetone washes (30%, 60%, 90%, 60%, 30%), slides were mounted with Vectamount AQ (Vector Labs, CA, USA).

Immunofluorescence.

The left gastrocnemius muscles were frozen in OCT floated on liquid nitrogen and stored at −80°C until sectioning. Samples were acclimated 30min in the cryostat and sectioned at −20°C in 50μm mid-belly longitudinal sections onto Superfrost Plus glass slides (Fisher Scientific). Slides were dried 1hr at room temperature (RT) and stored at −20°C until histological processing. At RT, slides were thawed 30min, fixed in 4% PFA 30min, and washed (3×10min) in 1X PBS. Slides were incubated in blocking solution (10% goat serum, 4% BSA, 3% Triton-X-100, in 1X PBS) 2hr at RT, followed by overnight incubation at 4°C in primary antibody solution containing chicken α-NF-200 (Abcam, Ab72996, [1:5000]) and rabbit α-Synapsin-1 (Cell Signaling, 5297S, [1:200]). After washes (3×10min) in 1X PBS, slides were incubated 2hr at RT in secondary antibody solution containing goat anti-Chicken Alexa Fluor 594 (Life Technologies, A11042, [1:1000]), goat anti-Rabbit Alexa Fluor 546 (Life Technologies, A11010, [1:1000]), and α-bungarotoxin (-BTX) Alexa Fluor 488 (Life Technologies, B13422, [1:1000]). Slides were then washed (3×10min) in 1X PBS and mounted with Fluoromount-G (Invitrogen).

Myofiber Phenotype Identification.

Soleus myofibers were imaged on an Axio Scan Z.1 (brightfield, 20X objective; Zeiss, Oberkochen, Germany) and viewed in ZEN 2.3 imaging software (blue edition; Zeiss, Oberkochen, Germany). ROIs measuring 800μm x 800μm were selected, each containing 100-150 myofibers. The phenotype of each myofiber was then determined by observer in ImageJ (54) (v. 1.51h, NIH, Bethesda, MD, USA) based on resulting myofiber staining intensity (55, 56). Mitochondrial NADH-TR enzyme activity transfers a hydrogen from reduced NADH to NBT dye; the resulting tetrazolium reduction yields a purple/blue formazan precipitate present heavily in highly oxidative fibers but sparsely in glycolytic fibers.

NMJ Imaging and Quantification.

Gastrocnemius NMJs were imaged on an Andor Revolution WD spinning disk laser confocal microscope system (Inverted Nikon TiE microscope, 40X water immersion objective, 488nm and 561nm lasers) with MetaMorph 7X Premier software (v. 7.7, Molecular Devices, San Jose, CA, USA). Individual channel images were merged and processed as Z stacks in ImageJ (54) (v. 1.51h, NIH, Bethesda, MD, USA), after which maximum intensity projections were quantified using a custom-designed workflow, including modifications from NMJ-morph (57). It was discovered upon imaging that sections from one rat in the Naïve group (n=6) were photobleached. Therefore, for NMJ analysis, the Naïve group was n=5. The total number of NMJs imaged per group are shown in Table 1.

Neuronal components at the NMJ were identified by presence of neurofilament (axon) and synapsin (axon terminal). The quantified presynaptic variables included area and perimeter of the nerve terminal (distal branchpoint of innervating axon and associated axon terminal branches). Postsynaptic AChRs were stained with α-BTX. Postsynaptic variables of interest were area and perimeter of the AChRs (all stained clusters) and the motor endplate (ellipsoid receptor footprint; all stained clusters and unstained spaces). Additional observations included the number of AChR clusters and the number of axons innervating each NMJ. Finally, a number of variables were derived using ImageJ functions or calculated from existing pre- or postsynaptic variables. NMJ quantification variables are defined in Table A.1.

Statistical Analysis

Sample sizes were estimated based on our previous experience employing the MCAO model of ischemic stroke (40, 47, 58). The sample size per group, per data set is reported in the figure legends. Statistical tests were performed in GraphPad Prism (v. 8.4.3, GraphPad Software, San Diego, CA, USA). Differences in group means from our in vivo longitudinal study (2 groups, 4 timepoints) were tested via Repeated Measures (RM) 2-way ANOVA, followed by Sidak’s multiple comparisons post hoc test to evaluate Stroke vs. Sham groups at each timepoint (BL, D7, D14, D21). For fiber type histology data, a 3-way ANOVA was performed to test interactions between the following variables: group (Stroke vs. Sham), myofiber type (type I, type IIA, type IID/X), and hindlimb (stroke-affected/left vs. contralateral/right). ANOVA data are presented as mean ± SEM, unless otherwise noted. Histological NMJ data (3 groups) were analyzed using the nonparametric Kruskal-Wallis (KW) test and are presented graphically as median ± 95% confidence interval (CI), unless otherwise noted. Differences between the mean rank of each group and that of a statistical control (Stroke^−RAMT group) were identified using Dunn’s multiple comparisons post hoc test. Group means ± SEM and KW test data (Kruskal-Wallis statistic (H), p value) are provided in Table 2; Dunn’s post hoc results of comparisons between groups are provided in the figure legends (Stroke^−RAMT vs. Naïve; Stroke^−RAMT vs. Stroke^+RAMT). Statistical significance was set at p 0.05.

Table 2.

Quantified Variables of NMJ Morphology Across Experimental Groups

Variable	Naïve	Stroke−RAMT	Stroke+RAMT	Kruskal-Wallis test p value, H statistic
Presynaptic:
Nerve terminal area (μm2)	85.6 ± 3.1****	121.4 ± 4.9	117.9 ± 3.9	p<0.0001, H=28.60
Nerve terminal perimeter (μm)	108.1 ± 3.3****	160.5 ± 10.5	124.7 ± 3.2*	p<0.0001, H=28.26
Postsynaptic:
Motor endplate area (μm2)	269.8 ± 9.4**	303.9 ± 7.9	269.0 ± 6.5**	p=0.0008. H=14.35
Motor endplate perimeter (μm)	76.5 ± 1.5	79.5 ± 1.2	76.7 ± 1.1	p=0.0789. H=5.08
AChR area (μm2)	167.8 ± 5.6****	210.3 ± 5.6	190.9 ± 4.7**	p<0.0001, H=30.89
AChR perimeter (μm)	127.5 ± 3.4	160.0 ± 12.7	122.0 ± 2.6***	p=0.0019, H=12.53
No. AChR clusters	2.11 ± 0.099****	1.43 ± 0.047	1.55 ± 0.051	p<0.0001, H=44.95
Derived:
Compactness (%)	63.07 ± 0.662****	69.76 ± 0.766	71.41 ± 0.466	p<0.0001, H=101.4
Fragmentation index (0 to 1)	0.37 ± 0.025****	0.18 ± 0.017	0.21 ± 0.017	p<0.0001, H=44.95
Colocalization (% AChR area)	35.36 ± 1.469****	44.68 ± 1.430	45.97 ± 1.194	p<0.0001, H=30.22
Vacancy (% AChR area)	63.94 ± 1.532**	62.14 ± 3.021	54.28 ± 1.201	p<0.0001, H=24.64

Data represent variables of NMJ morphology quantified from the left (stroke-affected) gastrocnemius of Naïve (n=5), Stroke^−RAMT (n=6), and Stroke^+RAMT (n=6) groups. KW test: p value and H statistic shown (degrees of freedom = [total number of values – number of groups] = [670 – 3] = 667); Dunn’s multiple comparisons (adjusted p values):

^*p<0.05

^**p<0.01

^***p<0.001

^****p<0.0001 vs. Stroke^−RAMT group. Data are shown as mean ± SEM.

Results

Stroke Reduces Nerve-Evoked Muscle Contractility

To assess the effects of stroke on paretic muscle contractility, we evaluated twitch and tetanic plantarflexion torque production longitudinally in vivo following supramaximal tibial nerve stimulation. Plantarflexion twitch torque, evoked by a single supramaximal stimulus, showed no difference between Stroke and Sham groups (Figure 2 legend; data not shown). Plantarflexion tetanic torque following high-frequency (125Hz) supramaximal stimuli, however, was markedly reduced after stroke. By D21, Stroke rats displayed >20% reduction in tetanic contractility compared to Sham (Figure 2A,B).Tetanic torque normalized to body weight was also assessed. Animal weights were documented longitudinally in Stroke rats and age-matched Sham controls (Figure 2C). At BL, body weights did not differ across groups. Weekly weight gain was demonstrated in Sham rats, increasing by 26% from BL to D21. Rats after MCAO exhibited significant loss of body weight by post-stroke D7, at which time they began to regain weight. By D14, weight increased enough to surpass BL. Because Stroke body weight was considerably less than Sham at each post-stroke timepoint, normalized tetanic torque data were less affected and showed no difference between groups until D21 (Figure 2D).

Figure 2. Loss of In Vivo Plantarflexor Contractility Following Stroke.

Plantarflexion torque was measured following both single (twitch) and train of stimuli (tetanic; 125 Hz) of the tibial nerve. Twitch torque was not different between Stroke and Sham groups at any timepoint (RM 2-way ANOVA: F _(3,15) = 1.65, p=0.221; Sidak’s: Stroke vs. Sham at D7 p=0.9995, D14 p=0.805, D21 p=0.501); Time (p=0.011) but not Group (p 0.950) was a source of variation. Data not shown. (A) Tetanic torque was reduced after stroke (RM 2-way ANOVA: F _(3,15) = 4.44, p=0.020) at post-stroke D7 (Sidak’s: Stroke vs. Sham p=0.069). D14 p=0.014). and D21 (p=0.001): Time (p<.0.0001) and Group (p=0.010) were additional sources of variation. Data shown as individual subject trajectories. (B) Data presented as percent of Sham group average at each timepoint. (C) Body weights of Stroke and Sham rats were compared across time (RM 2-way ANOVA: F _(3,24) = 11.41, p<0.0001). Sham rats gained weight continuously (Sham BL vs. D21 p<0.0001). while Stroke rats lost weight after MCAO (Stroke BL vs. D7 p<0.0001) but regained above BL by D14 (Stroke BL vs. D14 p=0.015). Stroke rats weighed less than Sham at each post-stroke timepoint (Stroke vs. Sham at BL p>0.9999, D7 p=0.0002, D14 p=0.002, D21 p=0.004). (D) Stroke tetanic data normalized to body weight were not different from Sham until D21 (RM 2-way ANOVA: F _(3,15) = 2.42, p=0.107; Sidak’s: Stroke vs. Sham at D7 p=0.9898, D14 p=0.357, D21 p=0.044). (B-D) Data shown as mean ± SEM. RM 2-way ANOVA: ‡p<0.05 Time x Group; Sidak’s multiple comparisons (adjusted p values): *p<0.05, **p<0.01. ***p<0.001 Stroke vs. Sham; †p<0.0001 within Stroke. Sham (gray, n=3), Stroke (red, n=4).

Motor Unit Functionality is Compromised in Stroke-Affected Skeletal Muscle

To elucidate the degree and progression of stroke-induced alterations to muscle excitation (CMAP), motor unit sizes (SMUP), and motor unit numbers (MUNE), we performed in vivo motor unit electrophysiology longitudinally in the paretic hindlimb of Stroke rats and compared to age-matched Shams. Stroke had no effect on muscle excitation as measured by the CMAP response (Figure 3A). There was no significant effect on average SMUP amplitude after stroke (Figure 3B). Calculation of MUNE revealed significant reduction in the number of functional motor units at post-stroke D21 compared to Sham (Figure 3D,E).

Figure 3. Stroke-Induced Alteration of In Vivo Motor Unit Function.

Electrophysiological motor unit assessment of the paretic gastrocnemius was measured following supramaximal (CMAP) and submaximal series (SMUP) stimulation of the sciatic nerve. (A) CMAP amplitude (RM 2-way ANOVA: F _(3,24) = 0.67, p=0.579; Sidak’s: Stroke vs. Sham at D7 p=0.9997, D14 p=0.995, D21 p=0.686) and (B) average SMUP size (RM 2-way ANOVA: F _(3,24) = 0.63, p=0.601: Sidak’s: Stroke vs. Sham at D7 p=0.884. D14 p=0.980, D21 p=0.843) were unchanged by stroke. (C) MUNE, however, declined by post-stroke D21 compared to Sham (RM 2-way ANOVA: F _(3,24) = 2.70, p=0.068; Sidak’s: Stroke vs. Sham at D7 p>0.9999, D14 p=0.536, D21 p=0.026). (A-C) Data shown as individual animal trajectories. (D) MUNE data presented as percent of Sham group average at each timepoint (means ± SEM) show an almost 20% reduction by post-stroke D21. RM 2-way ANOVA; Sidak’s multiple comparisons test: *p<0.05, Stroke vs. Sham. Sham (gray, n=3), Stroke (red, n=7).

Stroke Prompts Bilateral Myofiber Transition to More Fatigable Phenotype

To analyze mitochondrial distribution and, thus, myofiber phenotype as related to oxidative capacity, we stained stroke-affected and non-stroke-affected soleus muscles using NADH-TR reactivity and compared to corresponding left and right soleus muscles from healthy Sham rats (Figure 4). Sham soleus presented with a mix of myofiber types, with a heavy bias toward dark-staining, highly-oxidative myofibers (type I, IIA; Figure 4A). Stroke soleus, however, exhibited a robust shift in myofiber phenotype to more fatigable, glycolytic isoforms. By post-stroke D14, the darkest-staining type IIA fibers were undetectable (Figure 4B). Analysis revealed significant main effects of group (Stroke vs. Sham) and myofiber type (type I vs. type IIA vs. type IIX/D), as well as a significant interaction between the two factors. Stroke inverted the proportions of type I-to-type IIX/D myofibers, reducing type I myofiber counts by 73% while increasing fatigable type IIX/D presence by 72% compared to Sham.

Figure 4. Myofiber Shift in Oxidative Capacity Following Ischemic Stroke.

NADH-TR reactivity in left and right soleus muscles of (A) Sham and (B) Stroke rats at D14. Scale bar = 200 μm. (C) Distribution of myofiber type across left (solid shape) and right (open shape) soleus muscles from Sham (gray squares; n=3) and Stroke rats (red circles; n=8). There was no significant three-way interaction involving group, hindlimb, and myofiber type (3-way ANOVA: F _(2,18) = 1.77, p=0.1981). There was a significant main effect for group (F _(1,9) = 8.37, p=0.0178), a significant main effect for myofiber type (F _{(1.807,16.26)} = 57.05, p<0.0001), and a significant group x myofiber type interaction (F _2,18 = 53.67, p<0.0001). The main effect of hindlimb was not significant (F _{(1.000,9.000)} = 0.86, p=0.3781) and did not interact with group (F _(1,9) = 0.37, p=0.5573) or myofiber type (F _{(1.902,17.11)} = 0.59, p=0.5556). Data points indicate individual animals by hindlimb; lines denote mean ± SEM. (D) NADH-TR reactivity is indicative of mitochondrial distribution with an intensity gradient in rats of dark >>> light corresponding with myofiber types IIA > I > IIX/D > IIB (55, 56). Rat soleus does not contain type IIB (55).

Of note, the stroke-induced shift in myofiber metabolic phenotype presented in both the stroke-affected (left) and non-stroke-affected (right) soleus muscles. Analysis confirmed no significant main effect of hindlimb (left vs. right) or any associated two-way interactions. The three-way interaction across group, myofiber type, and hindlimb was also not significant.

Stroke Alters Pre- and Postsynaptic Elements of NMJ Morphology

A detailed examination of paretic gastrocnemius NMJs at post-stroke D14 was performed to determine (a) whether stroke caused morphological restructuring of NMJ elements and (b) whether RAMT treatment modified any changes observed after stroke (Table 2; Figures 5-7).

Figure 5. Presynaptic Alteration of NMJs Following Stroke and RAMT Treatment.

(A) Representative single-channel confocal micrographs of stained presynaptic elements (NF/synapsin). Scale bar = 10μm. (B,C) Quantification data for NMJ nerve terminals in stroke-affected gastrocnemius: Naïve (dark gray, n=5), Stroke^−RAMT (red, n=6), Stroke^+RAMT (light gray, n=6). (B) Nerve terminal area was increased after stroke (Naïve vs. Stroke^−RAMT p<0.0001) and remained unchanged by RAMT treatment (Stroke^+RAMT vs. Stroke^−RAMT p>0.9999). (C) Nerve terminal perimeter increased after stroke (Naïve vs. Stroke^−RAMT p<0.0001) but was reduced by RAMT (Stroke^+RAMT vs. Stroke^−RAMT p=0.01.3). Raw data in Table 2. KW test; Dunn’s multiple comparisons (adjusted p values): *p<0.05, **p<0.01, ****p<0.0001 vs. Stroke^−RAMT. Data are shown as median ± 95% CI.

Figure 7. Changes to Endplate Configuration Following Stroke and RAMT Treatment.

(A) Representative confocal micrographs showing merged images of nerve terminals (NF/synapsin, red) and AChRs (α-BTX, green). Scale bar = 10μm. (B-E) Quantification data for derived variables of NMJ morphology: Naïve (dark gray, n=5), Stroke^−RAMT (red, n=6), Stroke^+RAMT (light gray, n=6). Stroke-affected NMJs were more compact (B; Naïve vs. Stroke^−RAMT p<0.0001) and exhibited less fragmentation (C; p<0.0001) compared with healthy NMJs. Stain colocalization was greater after stroke (D; p<0.0001): correspondingly, AChR area vacant of nerve terminal stain was less (E; p=0.003). RAMT had no effect on derived variables (B-E; Stroke^+RAMT vs. Stroke^−RAMT compactness p=0.55, fragmentation p=0.37, colocalization p=0.53, vacancy p=0.12). Raw data in Table 2. KW test; Dunn’s multiple comparisons (adjusted p values): *p<0.05, **p<0.01, ****p<0.0001 vs. Stroke^−RAMT. (B,D,E) Data are shown as median ± 95% CI. (C) Bars span min-to-max values; black lines denote the median.

Compared to age-matched healthy NMJs (Naïve group), NMJs after stroke (Stroke^−RAMT group) exhibited multiple pre- and postsynaptic alterations. Presynaptic analysis (Figure 5) revealed increased nerve terminal area and perimeter in paretic muscle. Similarly, postsynaptic changes (Figure 6) after stroke included enlargements of both AChR area and motor endplate area. Stroke-affected NMJs also had fewer AChR clusters.

Figure 6. Postsynaptic Alteration of NMJs Following Stroke and RAMT Treatment.

(A) Representative single-channel confocal micrographs of stained postsynaptic AChRs (α-BTX). Scale bar = 10μm. (B-F) Quantification data for NMJ AChRs and motor endplates in stroke-affected gastrocnemius: Naïve (dark gray, n=5), Stroke^−RAMT (red, n=6), Stroke^+RAMT (light gray, n=6). (B) Naïve vs. Stroke^−RAMT p<0.0001) and motor endplate area (D; p=0.004) were enlarged at post-stroke D14 compared to naïve healthy NMJs. Stroke-affected NMJs also had fewer AChR clusters (F; p<0.0001). RAMT prevented the post-stroke expansion of AChR area (B; Stroke^+RAMT vs. Stroke^−RAMT p=0.007) and motor endplate area (D; p=0.001). Changes to number of AChR clusters (F; p=0.37) after stroke remained unchanged by RAMT treatment. (C) Stroke did not alter AChR perimeter (Naïve vs. Stroke^−RAMT, p=0.17); Stroke^+RAMT AChR perimeter was smaller than Stroke^−RAMT (p=0.0008). (E) Motor endplate perimeter presented no change after stroke (p=0.12 vs. Naïve) or RAMT (p=0.09 vs. Stroke^−RAMT). Raw data in Table 2. KW test; Dunn’s multiple comparisons (adjusted p values): *p<0.05, **p<0.01, ****p<0.0001 vs. Stroke^−RAMT. (B-E) Data are shown as median ± 95% CI. (F) Bars span min-to-max values; black lines denote the median.

Calculated variables derived from pre- and postsynaptic data provided further insight into endplate configuration (Figure 7). Compactness (57) (sometimes referred to as dispersion (59)) indicates what percentage of the motor endplate is occupied by stained AChRs and was significantly higher following stroke. Fragmentation index (57) expresses endplate clustering as a value ranging from 0 (solid) to 1 (highly fragmented). AChRs in paretic muscle were far less fragmented than healthy receptors. The ImageJ Colocalization function calculates the amount of staining overlap in staining between pre- and postsynaptic structures. The percentage of AChR area occupied by nerve terminal stain (Colocalization) was notably larger in stroke-affected NMJs compared to healthy junctions, while the percentage unoccupied by nerve terminal stain (Vacancy) was reduced. Taken together, data support stroke-induced expansion of pre- and postsynaptic structures, resulting in larger NMJs with diminished receptor clustering.

RAMT Treatment Modulates Stroke-Induced Changes at the NMJ

To explore the impact of mechano-stimulation on NMJ morphology after stroke, we investigated stroke-affected gastrocnemius from RAMT-treated rats (Stroke^+RAMT) and compared with untreated rats (Stroke^−RAMT) (Table 2; Figures 5-7). RAMT prevented the post-stroke expansion of nerve terminal perimeter, AChR area, and motor endplate area that were observed in Stroke^−RAMT NMJs. Changes to nerve terminal area, number of AChR clusters, colocalization, vacancy, fragmentation, and compactness following ischemic stroke remained unchanged by RAMT treatment. AChR perimeter did not change after stroke, though values were smaller with RAMT. Motor endplate perimeter was the only variable to show no change after stroke or RAMT.

Induction of Novel Polyaxonal Innervation After Stroke is Prevented by RAMT

Our investigation of NMJ morphology revealed differences affecting the number of axonal inputs at each NMJ. Stroke-affected NMJs in paretic muscle exhibited an abnormally high frequency of polyaxonal innervation, with two or more axons supplying 12.4% of NMJs compared to just 5.7% in healthy NMJs. Most interestingly, RAMT treatment after stroke reduced the frequency of polyaxonal innervation to 3.9% (Figure 8).

Figure 8. Maladaptive Post-Stroke Polyaxonal Innervation and Attenuation by RAMT.

(A) Frequency distribution showing percent of NMJs exhibiting 1 (blue), 2 (purple), or 3+ (orange) axonal inputs from each group: Naïve (n=5), Stroke^−RAMT (n=6), Stroke^+RAMT (n=6). KW test: H=14.45, p=0.0008. By post-stroke D14, the frequency of polyaxonal innervation in paretic gastrocnemius was increased significantly (Dunn’s: Naïve vs. Stroke^−RAMT p=0.0228) compared to healthy NMJs but was reduced by RAMT treatment (Stroke^+RAMT vs. Stroke^−RAMT p=0.0004). (B) Images of Stroke^−RAMT NMJs exhibiting one (bottom), two (middle), and three (top) axonal inputs. Merged confocal images show nerve terminals (NF/synapsin, red) and AChRs (α-BTX, green), with split-channel micrographs shown to the right of each (NF/synapsin, top; α-BTX, bottom). Scale bars = 10μm.

Discussion

Prior work has identified ischemic stroke-induced pathophysiology in the neuromuscular system (13, 17, 29, 31-33). Here, we showed rapid declines in paretic muscle contractility and estimated motor unit numbers following stroke. Examination of NMJ morphology identified striking stroke-induced alterations, many of which were attenuated with implementation of RAMT.

Novel observation of polyaxonal innervation as possible explanation for MUNE reduction.

Using clinically-relevant methodologies optimized by our group, we showed reduction in MUNE after stroke, suggesting motor unit losses similar to reports from clinical studies (32-34). We found a 20% reduction in MUNE in rat paretic gastrocnemius over a period of 21 days. The magnitude of our findings was similar to those found by Arasaki and colleagues (60) in stroke patients acutely (within 4-30 hours) but was less than their findings in patients in the sub-acute to chronic phases of stroke (40% reduction). Proposed mechanisms for motor unit loss following stroke have included impaired axonal transport, altered neuromuscular transmission, and myofiber atrophy (30, 32, 34, 61-63). Some claimed decreased MUNE is evidence of transsynaptic degeneration (61, 64-67), though anterograde loss of LMNs after stroke has been equally reported (68) and refuted (69). Others have instead proposed functional inactivation of otherwise-healthy LMNs (32, 33). Processes of rapid motor unit loss, as observed in our study, might be expected to show parallel reductions of both CMAP and MUNE, pending motor unit enlargement via NMJ remodeling, but it is possible that losses or inactivation of motor units are compensated by collateral reinnervation or parallel increases in motor unit sizes (70). Additional experiments are required to determine whether transsynaptic degeneration is occurring within our subacute timeline, or if other processes are responsible for the observed reduction in MUNE.

Importantly, MUNE calculation is dependent on myofibers being singly-innervated. Our prior work showed spurious MUNE reduction during polyneuronal innervation in neonatal mice, which resolved following the expected timeline of synaptic pruning (70). Thus, our observation of post-stroke reduction of MUNE, in the context of stable CMAP amplitudes, prompted us to examine innervation at the NMJ. We demonstrated multiple alterations at the NMJ, including individual motor endplates innervated by multiple axons.

Using the techniques employed in the current study, we cannot determine whether the multiple axonal inputs are attributable to different motor neurons and their respective motor units. Therefore, we have used the terminology of polyaxonal innervation. If these redundant inputs are attributable to polyneuronal and not simply polyaxonal inputs, this could suggest that the post-stroke MUNE reductions observed here and by other groups (32-34) may be driven by motor unit overlap rather than true loss of motor neurons. Another distinct possibility to consider is that structural features at the NMJ (axonal counts) may not correlate with function integrity (number of functional axonal inputs). The relationships between MUNE reduction and polyaxonal innervation of the NMJ, and the mechanisms driving these processes, deserve additional attention and investigation.

Prior work has demonstrated that disuse atrophy is associated with rapid onset of alterations at the NMJ, including sprouting (71). Additionally, we have recently observed that C57BL/6J mice demonstrate ~20% losses of MUNE in the gastrocnemius following 14 days of hindlimb immobilization (unpublished observation). Together, these observations suggest that reduced muscle activity may be driving changes at the NMJ, which may explain reductions of MUNE. This might also support why the aforementioned studies by Arasaki et al. showed that patients without weakness presented no evidence of MUNE change, yet in patients with weakness, MUNE reduction correlated with the severity of upper limb motor impairment (60).

Morphological alterations of stroke-affected NMJs.

In addition to polyaxonal innervation, other prominent pre- and postsynaptic alterations were observed following stroke. NMJs in paretic muscle had larger nerve terminals and motor endplates. AChR stain was greater, though more diffuse, with fewer receptor clusters and minimal fragmentation. Area increased without changing overall perimeter, due to expanded AChR coverage at the endplate, and a smoother, less convoluted endplate border. Altogether, stroke-affected NMJs broadened their postsynaptic footprint while losing the pretzel-like configuration characteristic of healthy endplates (72, 73). Aside from our work, one other group recently investigated postsynaptic NMJ size following endothelin-1-induced stroke using nonspecific esterase technique and reported similarly enlarged endplate size in forelimb muscles (74).

During development, polyneuronal-innervated NMJs show diffuse AChR-rich areas with plaque-like configurations, while single-innervated endplates have focal AChR labeling with a pretzel-like appearance (75). NMJs remodeling after denervation often mimic developing NMJs (76). Akin to postnatal NMJ maturation (75), labeled AChR area was enlarged post-stroke. Unlabeled areas within the endplate did not concomitantly increase, resulting in more homogenous staining. Perhaps AChR turnover is impacted, with new AChRs being inserted into the postysynaptic membrane faster than others are removed (75); this could explain the apparent increase in AChR density (Compactness) after stroke. NMJ maturation also involves AChR aggregation into clusters, which produces greater stability (77) and relies on transcriptional regulation (78). The reduced AChR clustering and other NMJ alterations noted in our study are consistent with a recent transcriptome analysis of stroke-affected muscle that showed altered expression of genes related to NMJ remodeling and ECM maintenance (79). Drivers of NMJ configurational changes and effects on NMJ function remain to be determined, and it is possible that NMJ remodeling after stroke involves compensatory reversion to developmental pathways.

Linking loss of muscle contractility and motor unit function after ischemic stroke.

The rapid decline of MUNE in paretic gastrocnemius was preceded 1 week by reduction of tetanic plantarflexion torque. When absolute tetanic data were normalized to body weight, the differences in contractility between groups were less overt, though still present at D21. This observation suggests loss of contractility may in part be related to loss of lean muscle tissue. Indeed, body weight was considerably less in rats with stroke compared with sham at each post-stroke timepoint, but whether reduction in body weight was associated with paretic muscle atrophy versus systemic weight loss secondary to health decline was not assessed. Experiments designed to assess the impact of stroke on nerve- versus muscle-elicited contraction force production and muscle contractility per unit tissue (i.e., specific force) could provide additional insight. Aside from atrophy, contractility can be compromised due to malfunctioning motor units, impaired NMJ transmission, or calcium dysregulation and excitation-contraction decoupling (52). Calcium kinetics differ between oxidative and glycolytic myofibers (80, 81). Our tests were performed on ankle plantarflexors, which include gastrocnemius and soleus. The gastrocnemius contains mixed myofiber types with many glycolytic myofibers, whereas the soleus is comprised of more oxidative fibers. The stroke-induced shift of myofiber phenotype is well-characterized as a bias toward more fatigable isoforms (14, 82). To best confirm this shift, we chose soleus for its high oxidative capacity at baseline. Indeed, myofiber phenotype in stroke-affected soleus changed drastically to reveal a global, more fatigable, type IIX/D predominance compared to soleus from healthy rats. Interestingly, fatigable myofibers are more susceptible to denervation and inadequate collateral reinnervation (83). Future plans will consider the stroke-induced shift in oxidative capacity in muscle with various myofiber type compositions and potential roles of myofiber-specific fatigue and calcium kinetics in the observed physiological phenotype.

Of particular interest was our observation of a bilateral myofiber type shift in the soleus muscle. Little is known about the mechanisms or extent of impact on the body ipsilateral to the brain lesion, but studies support systemic impacts of ischemic stroke across various organ systems (46). Our observation, like others (21, 84, 85), supports systemic impacts of stroke on muscle structure and function, beyond motor pathway disruption. Future work is required to understand how a unilateral brain lesion results in a myofiber type shift bilaterally. Though MUNE (86) and NMJ morphology (57) in healthy subjects have shown no sex-related differences, the influence of sex on post-stroke neuromuscular effects must also be examined.

Therapeutic benefit of RAMT on stroke-affected NMJs in the gastrocnemius.

Previously, we reported striking therapeutic effects of RAMT on motor recovery after stroke (40). We were interested in whether benefits of RAMT were mechanistically-associated with protection at the NMJ. RAMT treatment of the paretic gastrocnemius mitigated pathological alterations to NMJ pre- and postsynaptic elements after stroke, including polyaxonal innervation. These findings, coupled with the fact that RAMT has previously been shown to improve sensorimotor behavior and gait after stroke(40), provides reasonably strong evidence that polyaxonal innervation is maladaptive following stroke. In humans, the general consensus remains that polyinnervation does not occur in healthy muscle after birth (87). The increase in nerve terminal perimeter was suppressed by RAMT, though the enlarged nerve terminal area persisted, suggesting RAMT may reduce terminal branching complexity, thereby reducing perimeter while maintaining overall area through thicker terminals. Though configurational changes to AChR clustering and fragmentation were not countered, RAMT did attenuate stroke-induced expansion of AChR labeling and motor endplate size.

Therapeutically, massage is known to improve tissue perfusion, signal transduction, cellular stress, and inflammation (41-44). In addition to our findings related to stroke, a number of non-stroke studies present evidence for motor/neuromuscular system remodeling following mechanical stimulation (45). Cyclic compressive loading improved muscle remodeling in rats after unloading (88). Vibration applied to healthy subject muscle during voluntary contraction prompted lasting change in primary motor cortex excitability (89). Manual therapy in a murine model of facial nerve injury provided positive results on recovery and, intriguingly, decreased incidence of injury-induced polyinnervation at NMJs (90). With CNS injury, massage therapy enhanced balance and concussion recovery after traumatic brain injury (91), reduced resting and postural tremors in advanced Parkinson’s patients (92), and improved muscle strength and range of motion following spinal cord injury (93). Drivers of neuromuscular changes following massage are not completely understood but may be related to artificially-imposed muscle activity, activation of mechanotransduction pathways, anabolic signaling, altered inflammatory or metabolic myokine expression, or extracellular matrix modification (40, 88, 90, 94, 95). In this current work, we did not investigate the electrophysiological effects of RAMT, but future studies should interrogate the impact of RAMT treatment on measures of MUNE and gastrocnemius contractility to understand whether resolution of polyaxonal innervation is paralleled by other physiological benefits. Additionally, future experiments should specifically investigate the effects of RAMT at different stages of stroke recovery and in conditions of flaccid versus spastic paresis.

Though skeletal muscle is a key effector organ in post-stroke disability, less than 3% of ischemic stroke literature over the past decade has addressed the muscular response to stroke (46). Preclinical stroke rehabilitation studies have previously assessed benefits of whole-body vibration (96), tactile stimulation (97), or mechanical therapy (40) on recovery, but only one considered skeletal muscle effects (40). Data regarding clinical effects of massage are limited, with prior studies primarily focusing on spasticity, anxiety, pain, or holistic well-being (98-103), not motor outcomes. Accordingly, massage is not considered in current stroke rehabilitation guidelines (36). A 2013 trial sought to study effects of touch massage in subacute stroke (104) but was terminated due to lengthy inclusion and recruitment difficulties (clinicaltrials.gov: NCT01883947). Whole body vibration has been explored as a means of increasing muscle activation in chronic stroke patients during exercise (105). Similarly, Toscano, et al. (106) described post-stroke functional improvement following focal muscle vibration, proposing central neuroplasticity and bidirectional feedback as potential mechanisms, though they were not explored. Massage offers a non-invasive intervention widely practiced across therapeutic settings and could therefore readily support existing stroke rehabilitation programs. Further work is needed to understand therapeutic mechanisms by which mechanical massage therapy might convey benefit. Previously, RAMT was shown to blunt the increase in myostatin and reduction of BDNF observed in paretic muscle after stroke (40). Future studies should be designed to specifically test whether BDNF- or myostatin-related pathways contribute directly to polyaxonal innervation induction and the impact of mechanical massage intervention.

Conclusion

This work provides novel insight into the neuromuscular response following ischemic stroke. We identified that stroke-induced disruption to the motor cortex and associated pathways initiates a maladaptive response at the neuromuscular unit, alters motor unit organization, and reduces muscle contractility. Complex motor endplate alterations and induction of polyaxonal innervation are novel observations in stroke-affected muscle that may explain the in vivo electrophysiological phenotype of reduced motor unit numbers. Furthermore, mechanical physiotherapy provided robust protection against the impacts of stroke on NMJs and notably prevented the stroke-induced rise in polyaxonal innervation. These findings offer insight into potential peripheral mechanisms of functional disability and neuromuscular decline after stroke and, importantly, present a potential therapeutic strategy that could be readily translated to clinical studies.

Significance Statement:

Ischemic stroke is a leading contributor to chronic disability, and there is growing evidence that neuromuscular pathology may contribute to the impact of stroke on physical function. Following ischemic stroke in a rat model, there are progressive declines of motor unit number estimates and muscle contractility. These changes are paralleled by striking pre- and postsynaptic maladaptive changes at the neuromuscular junction, including polyaxonal innervation. When administered to paretic hindlimb muscle, Robot-Assisted Mechanical massage Therapy – previously shown to improve motor function and protect against subacute myokine disturbance – prevents stroke-induced neuromuscular junction alterations. These novel observations provide insight into the neuromuscular response to cerebral ischemia, identify peripheral mechanisms of functional disability, and present a therapeutic rehabilitation strategy with clinical relevance.

Highlights.

• Ischemic stroke incited peripheral neuromuscular pathophysiology in a rat model

• Stroke reduced paretic muscle contractility and motor unit estimates

• Stroke induced novel polyaxonal innervation at paretic neuromuscular junctions

• Striking changes to pre-/postsynaptic junction morphology were evident post-stroke

• Mechanical therapy attenuated stroke-induced neuromuscular junction alterations

Abbreviations:

AChR

acetylcholine receptor

BL

baseline

BTX

bungarotoxin

CMAP

compound muscle action potential

MCAO

middle cerebral artery occlusion

MUNE

motor unit number estimation

NADH-TR

nicotinamide adenine dinucleotide-tetrazolium reductase

NBT

nitrotetrazolium blue chloride

NMJ

neuromuscular junction

RAMT

robot-assisted mechanical therapy

RARE

Rapid Acquisition with Relaxation Enhancement

SMUP

single motor unit potential

Appendix A

Table A.1.

Definitions for NMJ Quantification Variables

Variable	Definition	Method	Sample Images from Workflow
Axonal Innervation	Number of axons innervating a single NMJ	Manual count
Nerve Terminal Perimeter (μm)	Composite length of tracings around nerve terminals at the endplate	ImageJ
Nerve Terminal Area (μm2)	Cumulative area of all stained nerve terminals at the endplate	ImageJ
AChR Perimeter (μm)	Composite length of tracings around all AChR clusters	ImageJ
AChR Area (μm2)	Cumulative area of all stained AChR clusters	ImageJ
Endplate Perimeter (μm)	Length of outermost border encompassing entire ellipsoid footprint of endplate, within which lies the stained AChR clusters and unstained spaces in between	ImageJ
Endplate Area (μm2)	Total area of ellipsoid endplate footprint, including both stained AChR clusters and unstained intercluster spaces	ImageJ
Compactness (%)	Quantity of AChRs within a given area; % of endplate area occupied by stained AChRs (also referred to as ‘dispersion’(59, 107))	= ( AChR area / endplate area ) * 100
AChR Clusters (count)	Number of morphologically distinct, stained groupings of AChRs	Manual count; ImageJ
Fragmentation Index	‘Fragmentation’ index calculated whereby a solid plaque-like endplate has an index of zero (0) and a highly fragmented endplate has an index towards numerical value of one (1)	= 1 – ( 1 / # AChR clusters)
% Vacancy	% AChR area not exhibiting nerve terminal colocalization; where AChR stain is not overlapped by nerve terminal stain	ImageJ
% Colocalization	% AChR area exhibiting nerve terminal colocalization; where AChR stain and nerve terminal stain overlap	ImageJ, Colocalization plugin % Colocalization = (Colocalized area / AChR area) *100