Muscle Spasms after Spinal Cord Injury Stem from Changes in Motoneuron Excitability and Synaptic Inhibition, Not Synaptic Excitation

Abstract

Muscle spasms are common in chronic spinal cord injury (SCI), posing challenges to rehabilitation and daily activities. Pharmacological management of spasms mostly targets suppression of excitatory inputs, an approach known to hinder motor recovery. To identify better targets, we investigated changes in inhibitory and excitatory synaptic inputs to motoneurons as well as motoneuron excitability in chronic SCI. We induced either a complete or incomplete SCI in adult mice of either sex and divided those with incomplete injury into low or high functional recovery groups. Their sacrocaudal spinal cords were then extracted and used to study plasticity below injury, with tissue from naive animals as a control. Electrical stimulation of the dorsal roots elicited spasm-like activity in preparations of chronic severe SCI but not in the control. To evaluate overall synaptic inhibition activated by sensory stimulation, we measured the rate-dependent depression of spinal root reflexes. We found inhibitory inputs to be impaired in chronic injury models. When synaptic inhibition was blocked pharmacologically, all preparations became clearly spastic, even the control. However, preparations with chronic injuries generated longer spasms than control. We then measured excitatory postsynaptic currents (EPSCs) in motoneurons during sensory-evoked spasms. The data showed no difference in the amplitude of EPSCs or their conductance among animal groups. Nonetheless, we found that motoneuron persistent inward currents activated by the EPSCs were increased in chronic SCI. These findings suggest that changes in motoneuron excitability and synaptic inhibition, rather than excitation, contribute to spasms and are better suited for more effective therapeutic interventions.

Significance Statement Neural plasticity following spinal cord injury is crucial for recovery of motor function. Unfortunately, this process is blemished by maladaptive changes that can cause muscle spasms. Pharmacological alleviation of spasms without compromising the recovery of motor function has proven to be challenging. Here, we investigated changes in fundamental spinal mechanisms that can cause spasms post-injury. Our data suggest that the current management strategy for spasms is misdirected toward suppressing excitatory inputs, a mechanism that we found unaltered after injury, which can lead to further motor weakness. Instead, this study shows that more promising approaches might involve restoring synaptic inhibition or modulating motoneuron excitability.

Introduction

Spinal cord injury (SCI) is a debilitating condition that causes long-term motor and sensory disabilities. In its acute phase, SCI results in decreased neuronal excitability, muscle weakness, and paralysis below injury level (Hiersemenzel et al., 2000). In contrast, chronic SCI is marked by spasticity, hyperreflexia, and spontaneous and sensory-evoked muscle spasms (Little et al., 1989; Skold et al., 1999; Hiersemenzel et al., 2000). While the underlying neuroplasticity is essential for functional recovery after injury, some maladaptive changes can nonetheless pose challenges to rehabilitation efforts and the lifestyle of many patients (Adams and Hicks, 2005).

Muscle spasms, in particular, can interfere with residual motor function, hinder rehabilitation, and increase hospitalization (Adams and Hicks, 2005; Richard-Denis et al., 2020). Spasms can manifest as multi-joint and crossed muscle activation indicating a complex and diffuse neuronal plasticity (Kuhn, 1950; Little et al., 1989). They are usually triggered by brief innocuous sensory stimuli and often last for several seconds (Bennett et al., 2004; Wu et al., 2005). The cellular and circuit mechanisms of these spasms continue to elude us, and the current rehabilitative and therapeutic strategies remain far from optimal (Adams and Hicks, 2005).

From a broad perspective, motoneurons represent the final processor of motor circuit output, and their activity is controlled by excitatory and inhibitory synaptic inputs. Therefore, the apparent increase in spinal circuit output during spasms can be attributed to increased synaptic excitation, decreased synaptic inhibition, increased motoneuron excitability, or a combination of these factors. While the role of increased motoneuron excitability in spasms post-SCI is well documented (Li and Bennett, 2003; Gorassini et al., 2004; Murray et al., 2011a; Johnson et al., 2013), the relative contribution of decreased inhibition or increased excitation is not clear. In addition, it is unknown whether different changes occur in different types of injury. Yet, clinical management of muscle spasms universally depends on pharmacological suppression of excitatory neurotransmitter release using drugs like baclofen, clonidine, and tizanidine (Adams and Hicks, 2005; Elbasiouny et al., 2010; Yoshizaki et al., 2020).

The approach of suppressing excitation is flawed for multiple reasons. Firstly, it results in global suppression of motor output with loss of remaining motor function and thwarted rehabilitation efforts (Yoshizaki et al., 2020), highlighting the need for new therapeutic strategy. Secondly, there is a lack of conclusive evidence that supports increased synaptic excitation after SCI. This is because it was usually studied as “net excitation,” that is, in the presence of synaptic inhibition (Hultborn and Malmsten, 1983; Bellardita et al., 2017). Therefore, reported increase in excitation could be secondary to disinhibition (Norton et al., 2008). This indicates a need for further investigation of the basic mechanisms of spasms.

Here, we studied mechanisms of spasms in chronic SCI using a whole-tissue ex vivo spinal cord preparation from adult mice. This preparation preserves circuit structures necessary to generate spasms and allows for pharmacological manipulations (Murray et al., 2011b; Jiang et al., 2021). We compared preparations obtained from mice with complete or incomplete chronic SCI to naive ones with no prior injury. The data showed that the ability to generate spasms was associated with injury severity. Surprisingly, when synaptic and intrinsic ionic currents were measured in motoneurons during spasms, there was no change in the excitatory drive after chronic injuries. Instead, there was correlated loss of synaptic inhibition and gain of motoneuron persistent inward currents (PICs). These results suggest that the culprits of spasms are the same across different types of chronic SCI and that they increase with injury severity. More importantly, targeting synaptic inhibition and motoneuron PICs could provide more favorable outcomes for management of muscle spasms without loss of remaining motor function.

Materials and Methods

Animals

All surgical and experimental procedures in this study were reviewed and approved by the Northwestern University Animal Care and Use Committee and were in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Adult mice of both sexes (C57BL/6J, Strain 000664, Jackson laboratory) were used in this study. Animals in the experimental groups underwent a survival procedure at 10 weeks of age to induce spinal cord injury (details below). This was followed by a terminal procedure 30–40 weeks post-injury in which the sacrocaudal spinal cord was surgically isolated, maintained ex vivo, and used for electrophysiological recordings. Age-matched control animals were used in terminal procedures only.

Experimental design and models of spinal cord injury

At 10 weeks of age, mice were anesthetized with isoflurane, and a laminectomy was performed at the T10 vertebral level (T11–T12 spinal segments). Based on whether/how a SCI was induced (Fig. 1), animals were divided into the following groups: (1) and (2) “Incomplete injury,” spinal cord was impacted with a 70 kdyn force and a 60 s dwell using an impactor system (Infinite Horizons, IH-0400, Precision Systems and Instrumentation). Animals were allowed to recover from injury with weekly assessments of locomotor recovery based on the Basso Mouse Scale (BMS; Basso et al., 2006) up to 20 weeks post-injury. The impacted mice were split into two groups: low functioning impacts (I-LF), which had a BMS score of 1–2, and high functioning impacts (I-HF), which had a BMS score of 3 or greater. (3) Complete injury (transection, Tx), spinal cords were transversely cut with spring scissors. The animals were allowed to recover and were assessed weekly for locomotor function. All the animals with complete transection had a BMS score ≤2 at 20 weeks post-injury. (4) Control, spinal cord tissue obtained from age-matched naive animals that did not go through survival surgeries or had any prior spinal injuries were used as a control (acute).

Figure 1.

Experimental design and spinal cord injury models. Left, A diagram showing the allocation of animals to different experimental groups. Complete (transection, Tx) and incomplete (impact) chronic SCI was initiated at 10 weeks old followed by functional evaluation at 20 weeks post-injury. Animals with incomplete SCI were divided into low and high recovery subgroups (I-LF and I-HF) based on their motor performance (BMS score). Right, Schematic representation of terminal experiment setup. The whole-tissue sacrocaudal spinal cord was extracted and used ex vivo to record motor output from ventral roots (using wire electrodes) and individual motoneurons (using sharp glass electrodes). Spasms were elicited via dorsal root stimulation.

It is noteworthy that all spinal cord preparations were transected at lower lumbar segments to extract sacrocaudal cord during terminal experiments (Fig. 1). This means that preparations from animals with chronic SCI were transected caudal to the chronic injury site, which makes our control preparation appropriate for comparison. This ensures that any differences found between experimental and control groups are solely due to neuroplasticity caused by the chronic injury. Moreover, any changes in spinal circuit behavior observed here is expected to be more profound in vivo.

Ex vivo spinal cord preparation

The procedures for surgical isolation of the sacrocaudal spinal cord has been previously described (Mahrous and Elbasiouny, 2017). Briefly, the animal was first deeply anesthetized by IP injection of urethane (≥0.2 g/100 g) and supplied with carbogen (95% O₂ + 5% CO₂) through a face mask. A dorsal laminectomy was performed to expose the lower half of the spinal cord. The cord was transected around the lumbosacral enlargement (L5–L6) and the whole sacrocaudal part (S1-Co2) with the attached roots was maintained ex vivo. The cord was mounted ventral side upward in a custom-built recording chamber and continuously perfused with oxygenated artificial cerebrospinal fluid (ACSF) at room temperature (∼21°C). The ACSF was composed of the following (in mM): 128 NaCl, 3 KCl, 1.5 MgSO₄, 1 NaH₂PO₄, 2.5 CaCl₂, 22 NaHCO₃, and 12 glucose. The osmolarity of the solution was ∼295 mOsm, and the pH was 7.35–7.4 when aerated with carbogen (95% O₂ + 5% CO₂). The dorsal roots (DRs) and ventral roots (VRs) were mounted on bipolar wire electrodes and covered with petroleum jelly to prevent drying. The preparation was allowed to recover for about 1 h before recordings started.

Electrophysiological recordings

Intracellular recordings

Single motoneurons were recorded in the isolated whole tissue using sharp intracellular glass microelectrodes, as described in previous studies (Mahrous et al., 2019). The electrodes were filled with a 2 M K⁺ solution (1 M acetate + 1 M Cl⁻) and had a resistance of 35–55 MΩ. A micro-stepper (Kopf Instruments) was used to advance the electrodes into the tissue. Motoneurons were identified by antidromic stimulation of the corresponding ventral root and were accepted for recording when the resting membrane potential was below −55 mV and the antidromic spike was 65 mV or larger. Recordings were performed using an Axoclamp-2B amplifier (Molecular Devices) running in either discontinuous current-clamp (DCC) or single-electrode voltage-clamp (SEVC) mode.

Ventral root recordings

The VRs were mounted on bipolar wire electrodes and connected to differential amplifiers (WPI) with 1,000× gain and 300 Hz–3 kHz bandwidth filter.

The outputs of extracellular and intracellular amplifiers were digitized using Micro3-1401 data acquisition interface (CED) at 10–20 kHz. The data were acquired into Spike2® software (version 8–10, CED) and stored on a computer for offline analysis.

Ex vivo spasm-like activity

The DRs were mounted on wire electrodes and connected to an S88 Grass stimulator (A-M Systems). Sensory-induced spasm-like activity was evoked ex vivo by electrical stimulation of the DRs. Stimulation was delivered at 2 and 10 times threshold (2×T and 10×T), either as a single pulse (0.1 ms duration, 0.067 Hz) or a five-pulse train at 25 Hz. Stimulation threshold was defined as the minimum stimulation amplitude which produced the smallest noticeable response in the corresponding VR. Sensory-induced spasm-like activity was defined as an increase in VR activity (5× the standard deviation of baseline) in response to DR stimulation occurs in at least two VRs simultaneously, lasts for more than 0.5 s, and maintains a minimum of 20 Hz of spiking frequency in the rectified VR trace (each time the rectified trace showed a peak above the 5× SD threshold, it was considered a spike; Fig. 2B).

Figure 2.

Spasm-like activity evoked ex vivo following chronic spinal cord injury. A, Example VR recordings at the third and fourth sacral segments (S3 and S4—R and L; right and left side of the cord) from a control preparation (left panel, no prior SCI) and another preparation with chronic transection (right panel). The traces show spasm-like behavior after chronic SCI evoked by DR stimulation. Note how activity spills to the other side of the cord and to multiple segments. B, Identification of sensory-evoked spasms as an increase in VR activity (>5× the SD of baseline, red dotted line) induced by DR stimulation and continuing for more than 0.5 s. The end of the spasm was determined by a drop of the spiking frequency in the rectified VR trace below 20 Hz (arrowheads) since these low-frequency spikes can appear randomly (see baseline before spasm). C, Summary of spasm scores in different models of chronic injury as well as the control. Scores were evaluated based on the intensity and frequency of DR stimulation needed to evoke spasms. Kruskal–Wallis tests with Dunn's pairwise multiple comparison were used to compare the scores. N = 38 control preparations, 19 I-HF, 18 I-LF, and 30 Tx animals. ns, nonsignificant; **p < 0.01, ****p < 0.0001.

Data analysis

Spasm score ex vivo

The ability to generate spasms was evaluated and scored before any drugs were added ex vivo. A score (0–4) was given to each preparation based on the intensity (2×T vs 10×T) and frequency (single pulse vs five-pulse train) of DR stimulation needed to evoke spasm-like activity (4, spasms with 2×T single pulse; 3, spasms with 10×T single pulse; 2, spasms with 2×T five-pulse train; 1, spasms with 10×T five-pulse train; 0, no spasms with 10×T five-pulse train). This analysis was done using an automated script written in Spike2®. This data set is presented in Figure 2.

Ventral root compound action potentials

The VR response to DR stimulation was measured as the peak-to-peak compound action potential (coAP). To evaluate rate-dependent depression (RDD) of the monosynaptic reflex, DRs were stimulated with a five-pulse train at 25 Hz with an amplitude of 2×T (threshold defined above). The ipsilateral VR monosynaptic response had a latency of 2–2.5 ms. The early contralateral VR response was also measured peak-to-peak and had a latency of 4–6 ms. This data set is presented in Figures 3–5.

Figure 3.

The disinhibited spinal cord shows less rate-dependent depression of monosynaptic root reflex on the ipsilateral side and larger polysynaptic responses on the contralateral side. A, Example traces from a control naive preparation showing the response of the ipsilateral VRs (top) and contralateral VRs (bottom) to five-pulse DR stimulation (P1–P5) at 25 Hz before (left) and after (right) application of strychnine and picrotoxin (STR–PTX, blockers of synaptic inhibitory receptors). Before drug applications, the monosynaptic coAP responses on ipsilateral side exhibited RDD, while contralateral responses were relatively steady in amplitude (no adaptation) but much smaller than ipsilateral responses (note different scale bars). STR–PTX application in the same experiment partly alleviated RDD on the ipsilateral side and resulted in larger responses on the contralateral side to all pulses. B, Summary of VR responses on the ipsilateral (left) and contralateral (right) side to five-pulse stimulation train of DRs at 25 Hz. Data were obtained from 18 control preparations (38 ipsi VRs and 60 contra VRs) before (dark color bars) and after (light color bars) STR–PTX application. Ipsilateral responses in each experiment were normalized to the amplitude of the first response (at P1) before drug application to show the change in RDD. However, contralateral responses are presented as absolute values to show change in amplitude. Repeated-measures two-way ANOVA was used to study the effect of drugs. Data represented as the mean ± SD. ns, nonsignificant; *p < 0.05, ****p < 0.0001.

Figure 4.

Spinal cord is disinhibited following chronic injuries. A, Summary of VR responses to ipsilateral DR stimulation (5 pulses at 25 Hz). Stimulation amplitude was set at 2×T in all experiments. All responses were normalized to the response at P1 in each experiment to show the change in adaptation. All chronic SCI models showed less adaptation than control preparations. B, Summary of VR responses to contralateral DR stimulation (same stimulation parameters as in A). Contralateral responses are presented as absolute values to show change in amplitude (larger in chronic SCI models than control). A mixed-effects analysis was used to test for differences between SCI models. Data represented as the mean ± SD. N = 34 control preparations (n = 86 ipsi VRs, 125 contra VRs), 20 I-HF (n = 62 ipsi VRs, 76 contra VRs), 18 I-LF (n = 44 ipsi VRs, 68 contra VRs), and 29 Tx (n = 74 ipsi VRs, 90 contra VRs). ns, nonsignificant; **p < 0.01, ***p < 0.001, ****p < 0.0001.

Figure 5.

Contribution of glycine and GABA_A receptors to rate-dependent depression of the monosynaptic reflex. The effect of synaptic inhibition was indirectly estimated from the root reflex measured before and after pharmacological blockade of inhibitory receptors. The VR responses to ipsilateral DR stimulation (5 pulses–2×T–25 Hz) were recorded. The same stimulation was repeated after synaptic inhibition was blocked using STR/PTX. In each experiment, the responses were normalized to the response at P1 before drug application. Then, the normalized responses before drug application were subtracted from their corresponding post-drug responses. The concentration of drugs and application time were the same in all experiments. A mixed-effects analysis was used to test for differences between SCI models. Data are represented as mean ± SD. N = 18 control preparations (n = 38 ipsi VRs, same data from Fig. 3), 14 I-HF (n = 44 ipsi VRs), 8 I-LF (n = 19 ipsi VRs), and 14 Tx (n = ipsi 39 VRs). ns, nonsignificant; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Spasm duration

The duration of sensory-evoked spasms was measured after synaptic inhibition was blocked using strychnine and picrotoxin (STR/PTX). Duration measurements were done at the S4 VR in all experiments. Spasms were evoked by 15 DR stimuli (0.1 ms pulse) at 2×T. The interpulse duration was set at 15 s (0.067 Hz). The first stimulus was preceded by 5–10 min period of no stimulation. This analysis was done using an automated script written in Spike2® (Fig. 2B for how spasm duration was measured). This data set is presented in Figure 6.

Figure 6.

Spinal cords with chronic injuries generate longer spasms when inhibition is blocked. Summary of duration of 15 consecutive spasms measured from the VRs. Spasms were elicited by DR stimulation at 0.067 Hz in the presence of STR–PTX. For each spasm number, chronic SCI models were compared to control preparations. Preparations with chronic injuries generated spasms of longer duration than control. Kruskal-Wallis test with Dunn's multiple comparisons was used for comparison. Data points are mean ± SD. N = 17 control preparations, 17 I-HF, 14 I-LF, and 24 Tx. ns: nonsignificant, *p < 0.05, **p < 0.01, ****p < 0.0001.

Excitatory membrane currents during spasms

To measure excitatory currents underlying spasms, inhibitory currents were first blocked using STR/PTX. Motoneurons were impaled and checked for stability, and then the amplifier was switched to the SEVC mode. Spasms were evoked by DR stimulation at 2×T and 0.067 Hz. Because we noticed variability and adaptation in duration during the first few spasms (Fig. 6), the first 10–12 spasms were evoked at the resting potential and were not included in the analysis of voltage-clamp data. Total membrane current was measured during spasms while the cell was held at different potentials (−80 to −20 mV) in 5–10 mV steps. While the membrane is held at each potential, spasms were evoked 2–3 times, and the recorded current trace was averaged and used for analysis. The current trace was divided into four phases (Fig. 8A) as suggested by Murray et al. (2011b): short polysynaptic reflex (SPR, 10–40 ms post-stimulus), longer polysynaptic reflex (LPR, 40–500 ms), and long lasting reflex (LLR, >500 ms). Here, we further divided the LLR period into LLR1 (500–1,500 ms) and LLR2 (1,500–3,000 ms) because we noticed that LLR1 contained a sizable synaptic current component (current recorded below −70 mV; Fig. 8B,C), while LLR2 was mostly self-sustained intrinsic current with little or no synaptic component.

Figure 8.

Measurement of spasm-mediated excitatory currents in motoneurons. A, Example traces showing voltage-clamp measurement in a motoneuron (bottom) along with VR response (top). Note the opposite changes during early phases of the spasm (SPR and LPR) versus later phases (LLR1 and LLR2) when the cell was held at different potentials. B, I–V relationship of different phases of the spasm recorded from a motoneuron (chronic Tx spinal cord). The linear relationships for SPR and LPR indicate synaptic origin, while LLR1 and LLR2 are mainly voltage-gated currents intrinsic to the motoneuron. For each cell, SPR and LPR were calculated from the linear fit of their I–V relation at −70 mV. C, The LLR phase sometimes included currents at hyperpolarized potentials indicating a remaining synaptic current component. To estimate and subtract the remaining synaptic currents during LLR1 or LLR2, a straight line (blue) was fit between the current values at hyperpolarized potentials (voltage-gated component = 0) and the reversal potential of LPR (arrow, synaptic component = 0). The resultant I–V relation after subtraction (gray dotted line) represents intrinsic motoneuron currents activated by the spasm (subtracted I_LLR). The peak current of this relationship was used for comparison between different injury groups. D, A scatterplot showing correlation between the values of SPR and LPR current and their respective cell input conductance (regardless of injury type, n = 82 MNs). Correlation was tested using the nonparametric Spearman test; p < 0.0001 for both SPR and LPR.

The leak-subtracted membrane current in each phase was measured as the area under the curve except for SPR where the peak current was used. A current–voltage (I–V) relation was established for each phase (Fig. 8B). To identify the voltage-gated component of LLR1 and LLR2, any remaining synaptic current during these phases was calculated and subtracted from the I–V relationship. To calculate the remaining synaptic current during LLR1 or LLR2, a straight line (Fig. 8C, blue line) was fit between the current values at −70 and the reversal potential of LPR (Fig. 8B,C, arrow), and then the line was subtracted from the I–V relationship. For comparison between animal groups, the value of SPR and LPR currents at −70 mV were calculated from the linear fit of the I–V relation, while for LLR1 and LLR2 the peak current of the I–V relationship was used. In this paper, the negative sign of membrane current measured using SEVC is used to indicate a depolarizing inward current.

Drugs and chemicals

Strychnine (STR, 1 µM), a blocker of the glycine receptors, and picrotoxin (PTX, 10 µM), a blocker of the GABA_A receptors, were used to block synaptic inhibition in vitro. These drugs as well as the chemical components of physiological solutions were purchased from Sigma-Aldrich®.

Statistical analysis

Statistical analysis was done using GraphPad Prism software (version 10.0.1, GraphPad®). The nonparametric Kruskal–Wallis test with Dunn's pairwise multiple comparisons was used to compare spasm scores, spasm durations, and motoneuron membrane currents during spasms (Figs. 2, 6, 9). Repeated-measures two-way ANOVA was used to study the change in rate-dependent depression of control preparations before and after STR-PTX (Fig. 3). A mixed-effects analysis was used to test for differences in rate-dependent depression among different SCI models (Figs. 4, 5). Correlation between synaptic currents and input conductance was tested using the nonparametric Spearman correlation test (Fig. 8C). Statistical significance was denoted as *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

Figure 9.

Motoneuron EPSC mediating spasms are not upregulated in chronic spinal cord injury. Summary of voltage-clamp data showing: A, B, Membrane current during SPR and LPR phases calculated from the linear fit at −70 mV for each motoneuron (as in Fig. 8B), C, D, Synaptic conductance for SPR (G_SPR) and LPR (G_LPR) calculated from the slope of the linear fit of their corresponding I–V relationship. E, F, Membrane current during LLR1 and LLR2 phases. The remaining synaptic currents during LLR1–LLR2 phase were estimated and subtracted to calculate the amplitude of the voltage-gated component (as in Fig. 8C). All parameters were normalized to the cell input conductance to account for differences in cell size. Kruskal–Wallis test with Dunn's multiple comparisons was used for all comparisons. N = 16 control preparations, 12 I-HF, 10 I-LF, and 21 Tx. Individual data points (motoneurons) are shown on violin plots for each group. ns, nonsignificant; *p < 0.05.

Results

Chronic SCI causes spasm-like behavior ex vivo

Muscle spasms are known to emerge in the chronic phase of SCI. Here, we studied the mechanisms of spasms in multiple chronic SCI models, partial and complete injuries. The animal groups that represent chronic SCI underwent a survival surgery to induce partial (impact) or complete (transection) injuries several months before terminal electrophysiology experiments (Fig. 1). Animals with impact injuries were then deemed to have high or low functional recovery (I-HF or I-LF) based on their BMS score (see Materials and Methods for details). We used the ex vivo whole-tissue sacrocaudal spinal cord preparation to study the mechanisms of spasms. The tissue extracted from healthy adult mice with no chronic injury was used as control. This part of the mouse spinal cord controls mainly the tail muscles. Most of our chronic SCI mice showed spasms in their tails in response to touching or pinching the tail (data not shown).

To test whether the neuronal activity underlying spasms can be detected ex vivo, we stimulated one of the DRs with 1–5 pulses and recorded the activity of the VRs at multiple segments. In control preparations (no prior injury), a brief segmental response was recorded in the VRs which typically ended within 5–10 ms (Figs. 2A, left, 3A, left; note the different time scales). On the other hand, preparations with chronic SCI generated much longer responses that were not confined to specific segments but rather spread to multiple segments and different sides of the cord (Fig. 2A, right). This pattern of bilateral multiple segments prolonged activity in response to brief sensory inputs is a hallmark of sensory-evoked spasms in vivo (Bennett et al., 2004; Wu et al., 2005; Norton et al., 2008) and was used in this study as a requirement to define spasms ex vivo. Figure 2B shows how spasms were identified in ventral root recordings.

To compare this behavior among different SCI models, we developed a scoring system based on the sensory stimulus needed to evoke spasms in vitro. Preparations that could generate spasms with less sensory stimulation were given higher scores and vice versa (see Materials and Methods). All control preparations invariably received a score of zero (no spasms), while higher scores were obtained in preparations with chronic SCI (Fig. 2C). The development of this spasm-like activity following SCI can be caused by changes in synaptic excitation–inhibition balance and/or motoneuron excitability.

Synaptic inhibition is reduced following SCI

When DRs were stimulated with multiple pulses at frequencies higher than 0.1 Hz (25 Hz throughout this study), the ipsilateral VRs respond with monosynaptic (∼2 ms latency) coAPs that decrease gradually in amplitude with subsequent pulses (Fig. 3A, top trace, left, B, black bars, left). This phenomenon is known as RDD. The 10% decrease in amplitude at first pulse (P1) after STR–PTX is probably due to blocking primary afferent depolarization which facilitates spike propagation in sensory axons (Hari et al., 2022). On the contralateral side, however, polysynaptic responses (∼4–6 ms latency) of much smaller but steady amplitude were evoked (Fig. 3A, bottom trace, left, B, dark blue bars, right).

After blocking synaptic inhibition with strychnine and picrotoxin (STR/PTX), the ipsilateral RDD was largely alleviated, and contralateral responses became larger while staying steady (Fig. 3A, right, B). Therefore, these effects can be used as an indirect measure of overall functional synaptic inhibition in spinal motor circuits. Accordingly, we compared the ipsilateral and contralateral VR reflexes among different models of SCI without any drugs to estimate their level of synaptic inhibition (Fig. 4).

Figure 4A shows the summary of VR responses to a five-pulse stimulation train to a DR at 25 Hz in different SCI models. In each experiment, ipsilateral VR responses were normalized to the response at P1 for easy comparison. All three chronic injury models had less rate-dependent depression (Fig. 4A). Complete transection (Tx) and low-function partial injury (I-LF) models showed higher responses than control starting at P2, and high-function partial injury (I-HF) started to be significantly higher at P3 (Fig. 4A). In addition, responses of the contralateral VRs were higher in all chronic injury models than their corresponding ones in the control starting at P1 (Fig. 4B). The diminished ipsilateral RDD and increased crossed responses in models of chronic SCI mimic the behavior of control spinal cords after blocking synaptic inhibition (Fig. 3). This suggests that synaptic inhibition is reduced in chronic SCI. In addition, the increase in contralateral responses in models of chronic SCI provides a new mechanism by which bilateral spasms (Fig. 2A) can occur.

To further confirm, we repeated the same stimulation after applying STR/PTX in a subset of experiments for each model. We then subtracted the normalized responses obtained before drug application from those recorded after it in the same experiment (Fig. 5). The data show that synaptic inhibition has strong effects on later pulses (P3–P5) causing 53–60% reduction in the motor pool response of the control preparations (green). This indicates that repeated sensory stimulation recruits strong inhibitory pathways that limit the buildup of excitation.

In chronic SCI, this effect was reduced to 25–29% for the complete injury model, 19–22% for low-function partial injury model, and 34–41% in high-function partial injury model (Fig. 5). This confirms that effects of synaptic inhibition on spinal motor output are decreased in chronic SCI. Furthermore, the lack of intraspinal inhibition in chronic SCI can cause the motor output to be doubled during repeated sensory stimulation.

Is loss of synaptic inhibition the only detectable difference?

Since our data indicated that there was loss of synaptic inhibition in chronic injury models, we then tested whether blocking synaptic inhibition had a differential effect on chronic SCI models versus the control. STR–PTX were added to the perfusion solution, and the tissue was left undisturbed for at least 15 min to allow the drugs to diffuse. We then used a 2×T DR stimulus every 15 s to evoke spasms. At this frequency, the monosynaptic coAP response in VRs did not show adaptation in amplitude (data not shown). Importantly, with synaptic inhibition blocked, the monosynaptic coAP was followed by much longer spasms (several seconds) in all preparations, including the control ones (Fig. 6). Using our scoring system for in vitro spasms, every preparation scored 4 (the highest score) when STR–PTX were applied (data not shown). This indicates that well-functioning synaptic inhibition is crucial for preventing spasms.

Importantly, the evoked spasm activity showed an adaptation in duration (Fig. 6), where the duration declined gradually reaching a steady state around spasm numbers 8–10. Although this adaptation was observed in all groups, the duration of the spasm remained longer in all chronic injury models as compared to the control (Fig. 6). Since synaptic inhibition was blocked in these experiments, this suggested that there are other underlying mechanisms contributing to longer spasms in chronic SCI models, namely, upregulated excitatory synaptic inputs, increased motoneuron excitability, or both.

Motoneurons can self-sustain their firing after synaptic inputs end due to the presence of PICs (Lee and Heckman, 1998a). To test if in vitro spasms can activate motoneuron PICs, intracellular recording of motoneurons was done, and spasms were evoked while the cell was held at different membrane potentials in both current and voltage clamps (Fig. 7). At more depolarized potentials, where PICs can be activated, there was a clear PIC in voltage-clamp mode after the motor pool activity ended in the VR (Fig. 7, top). This current caused the cells to continue to fire when the cell was in current-clamp mode (Fig. 7, bottom). These data confirmed that spasms can activate motoneuron PICs, which can then contribute to the duration of the spasm.

Figure 7.

Spasms activate motoneuron bistable behavior. Simultaneous VR recording and intracellular motoneuron recording in voltage-clamp mode (top) and current-clamp mode (bottom) during spasms evoked by DR stimulation in a disinhibited preparation. Top, At more depolarized potentials (red traces), synaptically evoked spasms activated PICs that were detected for several seconds after the spasm in the VR ended. Note that, in voltage clamp, downward (negative) shift in recorded current indicates an inward depolarizing current. Bottom, In current-clamp mode, when a bias current brings the membrane potential to a similar value, the PIC seems to trigger sustained firing that continues until the bias current is removed (n = 13 MNs).

Polysynaptic excitation does not change in chronic SCI

To discriminate the contribution of synaptic excitation and motoneuron PICs to spasms, we performed intracellular recordings of motoneurons in presence of STR/PTX. Spasms were evoked via DR stimulation every 15 s, and the first 10 spasms were not included in the analysis to avoid the effect of adaptation and initial variability in spasm duration (Fig. 6). Using SEVC, the spasm-evoked membrane currents were recorded in single motoneurons while the cell was held at different potentials. Figure 8A shows how the membrane current at all potentials was then divided into multiple time epochs corresponding to different phases of the spasms as suggested by previous studies (Murray et al., 2011b; see Materials and Methods for details). The first phase (SPR) starts at 40 ms post-stimulus focusing on polysynaptic excitatory inputs capable of evoking spasms (Li et al., 2004).

A current–voltage (I–V) relationship was established for different phases of the spasm in each cell (Fig. 8B). The first two phases, SPR and LPR, had a linear I–V relationship indicating that they were mediated by synaptic currents. Conversely, LLR1 and LLR2 had a U-shaped relationship indicating that the underlying conductance is voltage dependent (Fig. 8B), most probably mediated by motoneuron PICs (Murray et al., 2011b; Jiang et al., 2021). Any remaining synaptic current during LLR1 and LLR2 was estimated and subtracted (Fig. 8C; also see Materials and Methods).

The amplitude of the membrane current during SPR and LPR was correlated to the total cell conductance (p < 0.0001; Fig. 8D). Thus, we normalized currents to the cell input conductance to account for how effective these currents would be in depolarizing the cell.

The data showed no difference in synaptic membrane currents during SPR or LPR (Fig. 9A,B) nor their synaptic conductance (Fig. 9C,D) among the injury models. Conversely, the motoneuron intrinsic currents during LLR1 and LLR2 showed a trend of increase that became significant during LLR2 for the low-function partial injury (I-LF) and the complete injury (Tx) groups as compared to the control (Fig. 9E,F). This indicated that sensory-evoked excitatory synaptic currents were not different in chronic injury models, unlike motoneuron PICs which were upregulated.

Discussion

One of the prominent signs of recovered excitability following SCI is the development of muscle spasms. These spasms can be either spontaneous or triggered by brief, normally innocuous, sensory stimuli. In this study, we investigated changes in excitation–inhibition balance in spinal networks, as well as motoneuron excitability in relation to spasms. In addition, we compared these mechanisms among different models of SCI including moderate and severe incomplete injury as well as complete injury. Our data showed that the disrupted excitation–inhibition balance is driven solely by loss of inhibition, a situation that is aggravated by increased motoneuron PICs leading to generation of spasms in response to brief sensory inputs. Importantly, these maladaptive mechanisms occur in different types of SCI and seem to increase with injury severity.

Starting at the end: increased motoneuron excitability

Motoneurons are the spearhead and final pathway of the motor system. They can significantly amplify their synaptic inputs depending on their intrinsic excitability, especially PIC levels (Hounsgaard et al., 1988; Heckman et al., 2005; Harvey et al., 2006). PICs are mediated by slowly inactivating calcium and sodium channels conveniently located on the dendrites (Elbasiouny et al., 2005; Heckman et al., 2005), making them easier to activate by excitatory synaptic currents (Lee and Heckman, 1996; Bennett et al., 1998). They allow the motoneuron to be recruited earlier and to continue firing even after the excitatory input has ceased (Fig. 7) and (Lee and Heckman, 1998a). Motoneurons are thus in need of strict control mechanisms to curb their PICs.

There are two control mechanisms for PICs: (1) the level of neuromodulators that facilitate their activity (Hounsgaard et al., 1988; Lee and Heckman, 1998b) and (2) synaptic inhibition which effectively terminate them (Hultborn et al., 2003; Hyngstrom et al., 2007). In chronic SCI, both PIC control mechanisms are dysregulated. Synaptic inhibition is diminished after injury, as evident in this study. In addition, motoneurons increase their expression of constitutively active neuromodulator receptors that facilitate PICs even in the absence of neuromodulators (Murray et al., 2010). When these dysregulated PICs are activated by the long EPSPs during the initial phases of the spasm (SPR and LPR), they can generate low-frequency firing that maintains the spasm during its much longer LLR phase. Indeed, suppression of the L-type Ca²⁺ PIC in vitro has been shown to shorten the LLR phase of spasms (Jiang et al., 2021).

The difference between normal and dysregulated PIC behavior is evident in the data presented in Figure 6. It shows the duration of spasms in the chronic injury groups were prolonged several seconds in some cases, as compared to the control. Since inhibition was blocked in these experiments, the difference in PIC behavior is probably caused by the abnormal constitutively active neuromodulatory receptors in chronic SCI (Murray et al., 2010). Therefore, the manifestation of spasms relies on uncontrolled motoneuron PICs activated by prolonged depolarization, both of which could be attributed, at least in part, to the lack of synaptic inhibition.

Overall synaptic inhibition is diminished

Synaptic inhibition in the spinal cord is required to focus sensory-evoked activity (Goulding et al., 2014). For instance, when dorsal roots are electrically stimulated briefly, many excitatory neurons are activated in the dorsal and ventral horns. However, this excitatory wave is normally extinguished quickly by inhibitory interneurons (Bellardita et al., 2017), keeping the response brief. This necessary inhibition becomes impaired after SCI due to a multitude of factors which we have recently reviewed (Mahrous et al., 2023). Primarily, the main factors are loss of preferential activation by descending tracts (Lundberg and Voorhoeve, 1962; Hongo et al., 1969) and loss of the chloride transporter KCC2 with subsequent depolarization of the chloride reversal potential (Lu et al., 2008; Boulenguez et al., 2010; Mapplebeck et al., 2019).

Our data show that RDD is distinctly impaired after injury causing almost 100% increase in motor output during repeated sensory stimulation (Figs. 4, 5, compare severe models to control). This result is strikingly similar to the loss of RDD of soleus H-reflex at 10 Hz in patients with chronic SCI (Schindler-Ivens and Shields, 2000). Consequently, in chronic SCI successive sensory volleys cause greater buildup of excitation in the spinal cord and recruitment of motoneuron PICs. As a result, sensory receptive fields, which are otherwise focused, become broad after SCI (Hyngstrom et al., 2008; Johnson et al., 2013; Mapplebeck et al., 2019), allowing short localized sensory inputs to predispose multi-joint prolonged spasms. Moreover, our data showed that inhibition is needed to restrict the spread of excitation to the contralateral side. We found larger contralateral responses in chronic SCI (up to 300% increase over the control; Fig. 4B). These disinhibited crossed inputs can predispose bilateral spasms. In fact, blocking synaptic inhibition in control experiments (acute injury) caused preparations to develop multiple segments and bilateral spasms even though their excitatory inputs were presumably unaltered.

Excitatory polysynaptic currents after injury

Optogenetic activation of excitatory interneurons (VGluT2⁺ neurons) has been shown to initiate spasms in spinal cords with a chronic injury (Bellardita et al., 2017). Specific groups of excitatory interneurons, such as the bursting neurons of the deep dorsal horn, show increased firing after injury and can drive spasms (Thaweerattanasinp et al., 2020). Additionally, the involvement of V3 neurons in propagating spasms across multiple segments has been observed (Lin et al., 2019). Yet, when we blocked synaptic inhibition, we found that chronic SCI of different types did not cause a direct increase in polysynaptic excitatory inputs to motoneurons. This suggests that the inherent excitability of excitatory interneurons remains unchanged following SCI, and any apparent rise in excitation is secondary to the loss of inhibition.

Previous studies in chronic spinal cats have shown an increase in Ia monosynaptic EPSPs (Munson et al., 1986; Hochman and McCrea, 1994a,b), which can explain the increased stretch reflex after injury. However, these monosynaptic EPSPs have a short duration (decay prior to SPR phase), do not engage the long-lasting Ca²⁺ PIC, and are unlikely to contribute to spasms.

In our voltage-clamp experiments, we made a deliberate choice to exclude the first few evoked spasms to mitigate the inherent variability and adaptation in spasm duration (Fig. 6). It is plausible that this approach might have masked subtle differences in excitation occurring exclusively during the first evoked spasm. Nonetheless, it is unlikely that this would be the case if the increased excitation resulted from excitatory input sprouting or increased interneuron excitability. Such mechanisms would be expected to produce sustained increases in excitatory currents throughout subsequent spasms.

Notably, we did not investigate the impact of neuromodulators, such as serotonin and noradrenaline, on excitatory inputs post-injury. Recent studies have reported that excitatory V2a neurons have unaltered basic excitability after SCI, while their sensitivity to serotonin becomes higher (Husch et al., 2012). Whether neuromodulation would shift the excitation–inhibition balance further or not, it remains clear that direct suppression of excitatory inputs for the purpose of treating spasms is not the optimal strategy.

Clinical implications

Optimal therapeutic management of spasms can improve the quality of life of patients with SCI and accelerate motor rehabilitation. Our data show that the mechanisms that drive spasms are similar across various types of chronic SCI. Importantly, none of the injury types caused a direct increase in synaptic excitation to motoneurons, excluding it as a critical target.

Instead, our results suggest that restoration of synaptic inhibition could hold more promise in managing spasms with less adverse effects. This could be achieved through various strategies such as utilizing drugs that activate or increase the expression of KCC2 (Bilchak et al., 2021), activating the 5-HT1 receptors to disinhibit glycinergic inputs (Murray et al., 2011b), or exploring the potential of newly emerging positive modulators of glycine receptors (Gallagher et al., 2022).

Furthermore, suppression of PICs can shorten spasms or even prevent their initiation. Indeed, a recent study in mice with SCI showed that early administration of calcium channel blockers, which inhibit the Ca²⁺ PIC, prevented the development of muscle spasms (Marcantoni et al., 2020). Similar outcomes could potentially be achieved by targeting the neuromodulator receptors that facilitate PICs (Murray et al., 2011a,b).