Post-inhibitory rebound excitation drives extensor activity following spinal cord stimulation

Abstract

Background:

Neural circuits throughout the CNS can exhibit rebound excitation following prolonged periods of inhibition. However, the potential to control this phenomenon and harness it for clinical applications remains largely unexplored.

Objective:

We investigate rebound excitatory responses evoked by spinal cord stimulation (SCS) that can generate functional motor output, providing a testable model of circuit-level rebound excitation relevant to neuromodulation across the CNS.

Methods:

Brief (5 s) electrical stimulation trains were delivered to the lumbar spinal cord in adult cats. We recorded intracellular neuronal activity in the cord and EMG and force output from hindlimb muscles.

Results:

SCS elicited a robust and long-lasting rebound excitation selectively targeting the ipsilateral ankle extensors—a response we term Long Extension Activated Post-stimulation (LEAP). The force output during LEAP can support weight-bearing in cats, underscoring its clinical potential. Intracellular recordings revealed that extensor motoneurons received strong inhibitory inputs during stimulation, driven by reciprocal inhibition via proprioceptive afferents. Upon cessation of stimulation, a shift to a rebound excitatory synaptic input occurred, resulting in sustained firing in extensor motoneurons during LEAP. We also observed concurrent dynamic changes in interneuron firing across spinal laminae, suggesting broad circuit engagement. We systematically mapped the parameter space required to reliably evoke LEAP, including stimulation location, amplitude, and frequency, providing a framework for controlled rebound activation.

Conclusions:

LEAP represents a novel rebound response that can generate weight-supporting postural output. This mechanism not only expands the therapeutic potential of SCS in motor disorders but also serves as a model for modulating rebound excitation throughout the CNS.

1. Introduction

Post-inhibitory rebound (PIR) is a phenomenon observed throughout the CNS, where neurons exhibit enhanced excitability following hyperpolarizing events [1–4]. It often manifests as a transient depolarization or burst of action potentials after the cessation of an inhibitory input. This intrinsic property is mediated by multiple ionic mechanisms, including hyperpolarization-activated cation current (Ih) and low-threshold T-type calcium current [5–7]. In spinal circuits, PIR plays a critical role in the generation and modulation of rhythmic activities within central pattern generators by facilitating mutual re-excitation [1, 8]. This underlies oscillatory circuit activity that generates repetitive rhythmic movements such as locomotion and breathing.

Electrical stimulation of spinal cord dorsum primarily activates afferent axons close to the stimulation electrode [9]. The repeated activation of these tracts can produce excitatory and/or inhibitory downstream effects [10]. Therefore, despite the apparent excitatory responses, such as muscle activation, some neurons discreetly undergo sustained inhibition during stimulation [11]. This extended activation of inhibitory pathways can lead to prolonged post-inhibitory rebound excitation in certain circuits. The post-SCS rebound effects have not been investigated and remain largely unknown. In the present study, in vivo electrophysiological data shows that SCS can evoke potent inhibition in certain spinal circuits causing post-inhibitory rebound excitation. This rebound activity selectively targets extensor muscles eliciting powerful and long-lasting joint extension. Therefore, we termed this phenomenon “Long Extension Activated Poststimulation” (LEAP).

Extensor movements, such as ankle extension, are critical for maintaining postural stability [12], thus LEAP has great clinical potential. The restoration of stable posture represents a primary aspiration for individuals with lower limb disability [13], and confers several physiological benefits. Beyond being a prerequisite for walking, standing can improve autonomic output, including cardiovascular, bowel, bladder, and sexual functions [14–17]. Yet, there remains a scarcity of studies that specifically investigate SCS-induced postural movements. The LEAP response described here provides a potent extensor response that can facilitate sit-to-stand transition and other postural movements.

This study offers unique in vivo mechanistic insights into the effects of SCS in an adult feline model. Our in vivo intracellular recordings of spinal neurons demonstrate that LEAP is driven by the activity of a distributed network of interneurons. Some of these interneurons synapse onto and excite extensor motoneurons, with their synaptic currents further augmented by persistent inward currents in the motoneurons. In addition, we show that LEAP is mostly triggered by the activation of proprioceptive afferents during electrical stimulation.

To enhance the translational potential of LEAP, we characterized the stimulation parameters required to evoke it across multiple animals. We examined several combinations of stimulation frequencies, amplitudes, and locations along the lumbosacral cord. The data shows that LEAP is both controllable and potent, generating large isometric forces in the ankle extensors that are equivalent to those required for full weight-bearing in adult cats. Hence, this new response could be adopted clinically to facilitate postural movements in patients with motor disabilities. In addition, our study establishes a model system for controlling rebound activity through precise manipulation of stimulation parameters—a framework that may be broadly applicable across neural circuits throughout the CNS.

2. Methods

2.1. Experimental model

All experimental procedures were approved by the Northwestern University Institutional Animal Care and Use Committee. The data was obtained from 19 adult cats of either sex weighing 2.7–5.1 kg. All animals were obtained from a designated establishment for scientific research and were housed and fed within designated areas monitored daily by veterinary staff and trained personnel. Animals underwent acute terminal experiments, in which initial surgical procedures were done under deep gaseous anesthesia. A pre-collicular decerebration was then performed before data collection which allows the discontinuation of anesthesia [18], thus relieving any suppressive effects on spinal circuits. Unlike other forms of decerebration that cause rigidity, pre-collicular decerebrate cats can express coordinated locomotor activity [19,20]. The specific experimental procedures tested in each animal are reported in Table S1. Each figure legend contains the number of animals/experiments and data points used to make each figure panel.

2.2. Terminal surgical procedures

Anesthesia. Animals were initially anesthetized by inhalation of a mixture of isoflurane and nitrous oxide (4 % in 100 % O₂) in a custom chamber and then via a tracheal tube (1.5 %–3 % in 100 % O₂). The right common carotid artery and right jugular vein were cannulated to monitor blood pressure, and deliver intravenous fluids, respectively. A heating lamp and circulating water pad were used to maintain body temperature which was continuously monitored using an esophageal thermometer. Throughout the surgery, body temperature, arterial blood pressure, heart rate, respiratory rates, reflexes, and muscle tone were checked/recorded every 15 min and used to adjust the level of anesthesia.

Hindlimb preparation.

We dissected the left leg to expose the distal muscles and nerves around the ankle. A small incision was made in the front of the leg to expose the tibialis anterior (TA) muscle and separate it from surrounding tissue. The same procedure was done in the back of the leg to separate the medial gastrocnemius (MG), lateral gastrocnemius (LG), and soleus (SOL). Surgical sutures were then tied securely around the TA and SOL distal tendons at the ankle, and the tendons were disarticulated to be connected later through sutures to force transducers. The skin incision was then extended at the popliteal fossa to expose the common peroneal (CP) and tibial nerves, and a bipolar nerve cuff was placed around each for stimulation. Nerve stimulation was used to evoke antidromic action potentials in motoneurons to identify their target muscles during intracellular recordings. In some experiments, a cuff electrode was also placed around the sural nerve to test the effect of stimulating the cutaneous afferents (1 ms pulses, 40 Hz). All nerve cuff electrodes were connected to a constant voltage stimulator (Grass^® stimulator, model S8800). In addition, we inserted a pair of wires in each muscle to record EMG activity (details below). Prior to data collection, the animal’s hindlimbs were always rigidly clamped in the stereotaxic frame with the ankle at 90° relative to the tibia, the knee at 130°, and the hip at 105°.

Spinal cord preparation.

We transferred the animal to a stereotaxic frame (Kopf, Model 1530, Tujunga, CA, USA) where it remained until the end of experimental procedures. A dorsal laminectomy was carefully performed to expose the lumbosacral spinal cord (L4 to S2 spinal segments). The dura was incised and retracted laterally to expose the spinal cord tissue and allow for the identification of spinal segments and dorsal root entry zones. The cord was covered with warm mineral oil to prevent tissue drying.

Decerebration.

While the deep anesthetic plane was maintained, we performed a craniotomy and made a precollicular incision with a blade. All anterior forebrain structures were removed via an aspirator and replaced with saline-soaked cotton to help blood coagulation. At this point, animals were considered to have complete lack of sentience (Silverman et al., 2005), and gaseous anesthesia was discontinued. The preparation was left to recover for about 60 min before data collection commenced.

2.3. Experimental protocol

Force and EMG measurements.

A descriptive schematic of the experimental setup is shown in Fig. 1A. Isometric force around the left ankle were measured from the TA and SOL muscles. The distal tendons of these two muscles were severed and connected through surgical sutures to linear force transducers. Care was taken to maintain the disarticulated muscles at their physiological length. The force data was low-pass filtered at 30 kHz. EMG activity was recorded from muscles around the ankle of the left hindlimb, including the TA, SOL, MG, and LG along with the right-side SOL or LG in some experiments. We inserted pairs of insulated stainless steel fine wires (A-M Systems, WA, USA) in the belly of each muscle with 2 mm exposed wire at the tip. EMG data was acquired using custom-built differential amplifiers, band filtered between 0.1 Hz and 30 kHz, and digitized at 10 kHz (Power-1401, CED, UK). The data was acquired through Spike2^® software (CED, UK) and stored on a computer for offline analysis.

Fig. 1.

Long extension activated poststimulation (LEAP) from the cat spinal cord.

A: Schematic of the in vivo preparation setup in decerebrate cats. The response to lumbar spinal cord stimulation was recorded in ankle flexors and extensors (isometric force and EMG) and intracellularly in spinal motoneurons and interneurons. Each stimulation location was digitized using a Microscribe MLX Digitizer system and recorded alongside electrophysiology data. B: Diagram of the lower lumbar and upper sacral segments of the cat spinal cord showing different stimulation locations. The two leads of the stimulation electrode (1 mm apart) were placed on the left side of the cord at junctions between dorsal root entry zones, such as L5-L6 junction. When the roots were distinctly separated by distance, the two ends were designated as separate locations, such as the caudal end of L4 (L4C) and rostral end of L5 (L5R). C: Spinal cord stimulation electrode design and stimulation protocol. The bipolar silver ball electrode was loaded on springs that help maintain contact with the tissue while minimizing tissue damage due to any vertical movements. Stimulation trains were delivered for 5 s at 10, 20 or 40 Hz and with an amplitude of 50–600 μA (with 50–100 μA increments). D: Force and EMG recorded from soleus (SOL, black) and tibialis anterior (TA, green) of the left hindlimb in a decerebrate cat in response to a 5-s train (shaded area) of subdural electrical stimulation at the L5R dorsal root entry zone. Note that stimulation was delivered to the left side of the cord ipsilateral to the recorded muscles. SOL muscle responds with a strong and prolonged contraction shortly after stimulation train ends, a phenomenon we termed Long Extension Activated Poststimulation (LEAP, N = 19 experiments). E: Intracellular recordings from a different experiment showing typical firing behavior of an ankle extensor motoneuron (black) and ankle flexor motoneuron (green) during and after stimulation. Each of the two cells were recorded several hours apart in the same preparation. The superimposed traces (purple) are the integrated EMG activity of SOL and TA recorded before paralysis was induced for intracellular recording stability. Inset: A close-up on motoneuron membrane potential at the beginning of the stimulation train. Some triceps motoneurons fire an action potential following the first pulse of stimulation followed by hyperpolarization during most of the train time. N = 8 Experiments, triceps MNs n = 55 cells, common peroneal MNs n = 19 cells.

Intracellular recordings.

Experiments dedicated to intracellular recordings generally require high mechanical stability. Therefore, following decerebration, we induced paralysis via iv injection of a neuromuscular blocker (Gallamine triethiodide, 1 mg/kg) and connected the preparation to a ventilator (Harvard Apparatus, model 613). To further dampen movements caused by the ventilator, bilateral thoracotomy was sometimes induced. Before paralysis, however, a quick search for an optimal combination of stimulation parameters to evoke LEAP was done, and these parameters were used after paralysis.

Single motoneurons (MNs) and interneurons in the L6 and L7 segments were impaled and recorded in-vivo using sharp intracellular glass electrodes pulled using a micropipette puller (P97, Sutter instruments, CA). The electrode tip had a resistance of 5–15 MΩ when filled with electrolyte solution (1M K. acetate +1 M KCl). We advanced the electrode into the cord tissue using a micro-stepper (Burleigh, Inchworm, model 8200). The ankle extensor and flexor motoneurons were identified by their antidromic responses to peripheral stimulation (0.1 ms, 2.5x threshold, 0.5 Hz) of the tibial or common peroneal nerves, respectively. Cells were accepted for recording when they maintained a resting membrane potential below − 60 mV and their antidromic spike was ≥60 mV. Intracellular potential was recorded using an Axoclamp-2B amplifier (Molecular Devices, CA) running in bridge or discontinuous current clamp (DCC) mode. Intracellular data was low-pass filtered at 10–30 kHz, digitized at 20 kHz, and acquired into Spike2 software for offline analysis.

Spinal cord stimulation (SCS).

Electrical stimulation of the spinal cord was generated by a custom-built voltage-controlled current source (VCCS). To ensure accurate transfer of intended stimuli, both voltage command and delivered current were simultaneously recorded at 20 kHz alongside the evoked physiologic responses. Electrical stimulation was delivered using a custom-made bipolar silver/silver chloride ball electrode, with each silver ball measuring 0.9 mm in diameter and with a center-to-center spacing of 1.9 mm (DC impedance of 1.4 Ω for each side of the electrode). The electrode was fitted on a spring-loaded mechanism to prevent tissue damage whenever vertical movements occurred (Fig. 1C). The electrode was placed at each location and different combinations of stimulation frequency and amplitude were repeated. The stimulation locations were set at dorsal root entry zones in different segments (Fig. 1B) including caudal end of L4 (L4C), rostral end of L5 (L5R), caudal end of L5 (L5C), rostral end of L6 (L6R), junction between L6 and L7 (L6-L7), junction between L7 and S1 (L7-S1), and junction between S1 and S2 (S1-S2). In some animals, the distance between L4C and L5R or L5C and L6R was too small and was treated as a single location instead (i.e. L4-L5 junction or L5-L6 junction). The 3-D location of the identifiable dorsal root entry zones (L4-S2) as well as the border of laminectomy were recorded using a digitizer system (Microscribe G2X). These digitally recorded locations were used to guide the placement of the stimulation electrode on the cord. Stimulation was delivered as 1 ms monophasic square waves for 5 s (Fig. 1C). To determine optimum parameters needed to evoke poststimulation activity, we used combinations of stimulation frequencies (10, 20, and 40 Hz), and amplitudes (50–600 μA, in increments of 50 or 100).

2.4. Quantification and statistical analysis

Intracellular Data.

To identify the underlying mechanism for motoneuron firing activity during LEAP, we performed intracellular recordings of ankle extensor motoneurons. After impalement, the cells were held at different potentials by injecting DC current. The holding potential was measured as the maximum deflection in voltage following the injection of the DC current. LEAP was repeatedly evoked using the same stimulation parameters at each holding potential. Any evoked spiking was low-pass filtered to reveal the underlying excitatory potential. The peak amplitude of the evoked poststimulation potential was measured and plotted against the holding potential. This analysis is shown in Fig. 2. The delay of the EPSP and IPSP evoked by individual stimulation pulses was measured as the time between the onset of the stimulation pulse and the initial upward or downward deflection of the membrane potential, respectively.

Fig. 2.

Motoneuron firing during LEAP response is driven by a combination of synaptic excitation and intrinsic mechanisms.

A: Intracellular recording of LEAP from a triceps (TS) motoneuron at resting potential (black) and when hyperpolarized by a DC current (blue). Note that potentials in the pre-stimulation periods for both traces were overlayed to highlight the difference in LEAP amplitude poststimulation. The traces were low-pass filtered to eliminate spikes and stimulation artifacts. Membrane depolarization during LEAP was not inhibited by hyperpolarizing current injection, but instead became larger. B: Summary of the peak amplitude of LEAP potential at different holding potentials measured in 10 TS motoneurons recorded in 5 experiments (legend). Left: 7/10 MNs exhibited negative correlation between LEAP and holding potential indicating that LEAP was mediated by an excitatory synaptic input in these cells. Right: in 3/10 MNs, the correlation was positive, indicating that peak depolarization during LEAP was intrinsically-sustained in this group of cells, i.e. via voltage-gated channels.

Interneurons’ firing rates were calculated as the average firing rate in the 5-s period before, during, and after stimulation. The change in firing rate during the post-stim period was calculated as a percentage in relation to the 5-sec baseline firing rate before stimulation (Fig. 3C). A change in firing poststimulation was defined as an increase or decrease ≥10 % of the baseline firing rate. Interneuron depth from the surface was approximated based on the distance the electrode traveled within the tissue, the entry point, and the range of impalement angles used throughout the experiments (Fig. 3D). The size of spinal cord section shown in Fig. 3D is based on the anatomical study by Toossi et al. [21].

Fig. 3.

Poststimulation changes in firing of spinal interneurons.

A: Example traces of in-vivo intracellular recordings from spinal interneurons recorded using sharp glass microelectrodes. Distinct firing behaviors were observed poststimulation, with some interneurons showing increased (cell 1), decreased (cell 2), or unchanged (cell 3) firing rates compared to their pre-stimulation states. B: Distribution of interneurons based on changes in firing rate. C: Summary of poststimulation firing rate changes as a percentage of the pre-stimulation values. Individual data points are presented on each box plot for clarity. D: Schematic representation of a hemi-section of the cat lumbar L6-L7 spinal cord segments, illustrating the estimated depth distribution of recorded interneurons. Note that the precise 3D locations of these neurons within the gray matter remain undetermined. The red arrow heads indicate the three cells shown in “A”.

EMG data analysis.

The time series data was imported from Spike2 into MATLAB and zeroed at the initiation of the 5 s stimulation train. For each channel (muscle), a window ranging from − 1 to +35 s was extracted, the DC shift was removed, and data was rectified. For every trial, a moving average with a 300 ms window was computed from the series (movsum, MATLAB 2023a). Then, the root mean square of the average was taken (rms, MATLAB 2023a) from end of stimulation to end of the recording window (+5 to +35 s) to evaluate the post-stim response.

The effect of stimulation location

(This analysis is shown in Fig. 5A–C). To identify the optimum location, the data extracted from each experiment was grouped according to stimulation frequency, amplitude and location. The mean and standard deviation was taken for multiple trials of the same parameters (typically 3 trials for each combination of stimulation location-amplitude-frequency). Outliers were removed as points outside of a 0–90 percentiles range (rmoutliers, MATLAB 2023a). Experiments with less than 3 stimulation locations were excluded from this analysis. To simplify the analysis, we used 40 Hz stimulation data and searched for maximum response at each location, then the stim amplitude that evoked the largest of these values was selected to be the one used across all locations. This process was repeated for each experiment separately, and the selected amplitude is displayed in the legend of Fig. 5B. The data was then normalized to the maximum LEAP response for each experiment. The data was split on the plot into two groups, those with peak response in the lumbar segments and others which continued to show larger responses with stimulation of sacral segments.

Fig. 5.

Alternating activation of SOL and TA muscles during and after stimulation.

A: Example force traces of SOL and TA muscles from same experiment showing responses to 5 s spinal stimulation trains at different locations and frequencies (stimulation amplitude kept constant, 500 μA). B: Left: The relationship between the amplitude of poststimulation response in soleus (LEAP) and the response of TA during stimulation in 10 experiments (colors). Right: The individual data points from each of the experiments on the left. The data shows that largest LEAP responses are generated when TA is more active during stimulation. The alternating activity between the antagonistic muscle pair suggests the involvement of reciprocal inhibition circuits.

The effect of stimulation frequency

(This analysis is shown in Fig. 5D–E). To evaluate the effect of stimulation frequency (10, 20, 40 Hz) on LEAP response, we first extracted the data in each experiment, averaged trials with same parameters and excluded outliers as discussed above. Then the optimum stimulation location for each experiment was chosen from the analysis above to be used for comparing different frequencies. To select a stimulation amplitude, the algorithm searched for the amplitude closest to half the value of the maximum stimulation amplitude. This was done to make sure that the effect of stimulation frequency is not masked due to saturation. For each experiment, the data was normalized to the largest LEAP response at that selected stimulation amplitude.

The effect of stimulation amplitude

(This analysis is shown in Fig. 5F–G). The data was grouped, repetitions averaged, and outliers removed using the same technique as described above. The optimum stimulation location for each experiment was chosen (from the analysis in Fig. 5B) to evaluate the effect of stimulation amplitude. This analysis was run separately at each frequency (10, 20, and 40 Hz). The plot in Fig. 5G shows the analysis of the 40 Hz data, while analysis of 10, and 20 Hz is shown in Fig. S4. The data was normalized to the maximum LEAP response for each experiment. These plots are akin to an activation curve.

3. Results

To examine both the cellular mechanisms and systems-level output of spinal cord stimulation (SCS), we used an in vivo preparation in decerebrate adult cats (Fig. 1A). This model allowed us to study spinal circuit dynamics, including rebound excitation, without the confounding effects of anesthetic agents. Notably, the pre-collicular decerebration we induced preserves the capacity for generating coordinated locomotor activity and does not induce rigidity [19,20], making it an ideal platform for investigating SCS-induced post-inhibitory rebound responses.

Bipolar electrodes were precisely positioned on the left side of the cord dorsum targeting multiple locations at dorsal root entry zones spanning from L4 to S2 (Fig. 1B). Electrical stimulation was delivered as trains of 1 ms pulses at multiple frequencies (10, 20, and 40 Hz) and amplitudes (50–600 μA, Fig. 1C). Motor responses were monitored via EMG and force transducers, focusing on the ankle flexor tibialis anterior (TA), and the ankle extensor soleus (SOL) ipsilateral to stimulation.

Certain combinations of stimulation parameters (details in Fig. 6 below) evoked a distinctive pattern of muscle activation around the ankle, characterized by robust persistent contraction of the extensors after the stimulation train ended (Fig. 1D). Here, we term this poststimulation activity in extensor muscles Long Extension Activated Poststimulation (LEAP). This study investigates spinal circuit activity during this poststimulation response to gain insights into its mechanisms and potential clinical benefits.

Fig. 6.

Control of LEAP response via stimulation parameters.

A: Examples of SOL force measurements during and after stimulation at 4 different locations in the lumbosacral area in one experiment. The amplitude of LEAP response varies noticeably at different stimulation locations. B: Summary of the effect of stimulation location on LEAP amplitude in 10 experiments (different colors). A near maximum stimulation amplitude at 40 Hz was used. The optimum location (maximum response amplitude) varied slightly among animals. The data was divided into 2 groups with the optimum location around the lumbosacral junction (bottom panel) or strictly in the lumbar segments (top panel). C: Diagram of the lower lumbar and upper sacral segments of the cat spinal cord showing the stimulation location at which the largest LEAP response was evoked for each experiment in “B”. D: Examples of force traces measured from SOL at the optimum stimulation location for that animal (L5R) and stimulation amplitude that produces about half maximal response at 40 Hz (400 μA, see F panel). E: Summary of the effect of stimulation frequency (10, 20, and 40 Hz) on LEAP amplitude in 8 experiments (colors). Electrical stimulation was delivered at an amplitude that produced about 50 % of maximum response at 40 Hz in each experiment, and optimum stimulation location (chosen from panel B). F: Examples of SOL force measurements at the optimum stimulation location for that animal (L6R) and 40 Hz stimulation frequency. G: Summary of the effect of stimulation amplitude (50–600 μA) on LEAP response in 10 experiments (colors). Electrical stimulation was delivered at 40 Hz to the optimum stimulation location in each experiment (chosen from panel B). In B, E, and G, the amplitude of LEAP was measured as the integrated EMG response in soleus in the 30 s following the stimulation train and normalized to the maximum response evoked in each experiment. Each data point in these plots represents the average of multiple trials (typically 3).

3.1. Long extension activated poststimulation (LEAP)

During brief stimulation trains (5 s in this study) with optimized parameters (see Fig. 6), the ankle flexor tibialis anterior (TA) was consistently activated, sometimes followed by a short post-stimulation contraction (Fig. 1D). In contrast, the ankle extensor soleus (SOL) remained largely inactive during stimulation, apart from a brief twitch at train onset. Notably, after stimulation ended, the ankle extensors produced a robust, sustained contraction that persisted for several seconds (LEAP, Fig. 1D). This extensor response was ipsilateral to stimulation and primarily restricted to the ankle, and therefore distinct from a crossed extension reflex.

The same behavior recorded intracellularly from extensor and flexor motoneurons in another animal is shown in Fig. 1E. We recorded a total of 55 tibial nerve motoneurons (extensor MNs), 19 common peroneal motoneurons (flexor MNs), and 11 un-identified motoneurons. Approximately 73 % of the impaled flexor motoneurons fired repetitively and their membrane potential remained depolarized throughout the stimulation train (n = 14/19 cells, Fig. 1E). Conversely, 71 % of the impaled extensor motoneurons exhibited hyperpolarization during stimulation (Fig. 1E, n = 39/55 cells). Following each pulse of stimulation, an EPSP was observed in extensor motoneurons (latency: 1.68 ± 0.36 ms) followed by multiple IPSPs (latency of the first IPSP was 3.46 ± 0.73 ms) which resulted in an overall hyperpolarization of the membrane (Fig. 1E, inset). The EPSP elicited by the first pulse of the train preceded any hyperpolarization and thus frequently resulted in an action potential (n = 25/55 cells), accounting for the brief twitch observed in SOL at the beginning of stimulation (see example in Fig. 1D). The size of the evoked IPSPs showed slow adaptation causing the membrane potential to gradually return towards resting values, similar to what has been described before for maintained high-frequency stimulation of peripheral nerves [22]. Therefore, spiking occurred sometimes in some extensor motoneurons at the end of the train.

Shortly after stimulation ended, 89 % of the extensor motoneurons recorded (n = 49/55 cells) started to depolarize/fire for several seconds. Throughout this study, SCS was delivered directly to the spinal cord surface after opening the dura (subdural). However, LEAP responses can also be evoked by epidural stimulation (Fig. S1).

3.2. Circuit mechanisms of LEAP

The prolonged poststimulation firing in extensor motoneurons during LEAP could be caused by excitatory synaptic currents and/or intrinsic self-sustained motoneuron firing mediated by persistent inward currents (PICs) [23]. To determine the underlying mechanism, we performed intracellular recordings of spinal motoneurons and interneurons. To enhance the mechanical stability of intracellular recordings, we induced paralysis via i.v. injection of a neuromuscular blocker. Therefore, we could not concurrently record EMG or force output alongside intracellular potentials. Nonetheless, the stimulation parameters needed to evoke a stable LEAP response were determined before paralysis and used throughout the experiment.

We recorded triceps motoneurons and repeatedly evoked LEAP while systematically varying the membrane potential of impaled cells. The peak depolarization during LEAP was then measured for each holding potential. At more hyperpolarized potential, most triceps motoneurons showed increased depolarization during LEAP (Fig. 2A). In 7/10 MNs, there was a negative correlation between the holding potential and peak LEAP depolarization (Fig. 2B). This indicates that LEAP is largely mediated by an excitatory synaptic current. Nonetheless, in 3/10 cells, there was a positive correlation indicating that peak response in these cells was driven by a voltage-gated membrane conductance. This suggests that although LEAP is initiated by excitatory synaptic currents targeting extensor motoneurons, it subsequently recruits and is enhanced by their intrinsic PICs.

If LEAP is indeed driven by synaptic currents, it is expected that a subset of interneurons would exhibit altered firing patterns poststimulation. Our intracellular recordings of spinal interneurons revealed an increase in firing activity post-stim (concurrent with LEAP) in about 30 % of recorded cells (Fig. 3A–B, n = 9/30 cells). On average, the baseline firing rate of these cells was more than doubled (Fig. 2C). In addition to the increased firing post-stim, 7/9 cells increased their firing rate during the stimulation train. Conversely, 53 % of recorded interneurons showed no change in firing (Fig. 3A–B, n = 16/30 cells), while a decrease in firing poststimulation occurred in 17 % of cells (Fig. 3A–B, n = 5/30 cells). The reduction in firing rate was about 40 % on average (Fig. 3C). Because the preparations were paralyzed during intracellular recordings, it was not possible to directly relate the changes in each interneuron firing behavior to muscle activity in the hindlimb.

The vertical depth of some of the recorded interneurons is shown in Fig. 3D. Even though, we did not attempt to identify the types of interneurons recorded, the changes in their firing rate that coincide with LEAP clearly indicate that this behavior is mediated by a wide-spread spinal circuit, and not entirely intrinsic to the motoneurons.

Taken together, our intracellular data suggests that LEAP arises from diverse behaviors of populations of interneurons that collectively contribute to its full expression. Moreover, the excitatory synaptic currents received by extensor motoneurons during LEAP can recruit their intrinsic PICs, further augmenting the response.

3.3. Types of sensory afferents involved in LEAP

In our stimulation protocol, the pulse width (1 ms) could activate both cutaneous and proprioceptive afferent axons [24]. To discern the contribution of these sensory pathways to the activation of LEAP circuits, we tested the effect of shorter stimulation pulses (0.1 ms), which more selectively recruit the largest afferent fibers, including primarily proprioceptive axons [24]. Despite requiring higher stimulation amplitudes (motor threshold was about 3 times larger for 0.1 ms pulses), the shorter pulse protocol successfully elicited comparable LEAP responses (Fig. 4A–B), indicating a prominent role of proprioceptive afferents.

Fig. 4.

Contribution of proprioceptive versus cutaneous afferents to LEAP.

A: Force and EMG measurements from SOL and TA during and after stimulation at the same stimulation location and frequency but with different stimulation pulse width. Stimulation using short pulse width (0.1 ms, right), which is more selective for proprioceptive afferents, can evoke LEAP at higher stimulation amplitudes (N = 2 experiments). B: Comparison of LEAP responses evoked by spinal stimulation using 1 ms Vs 0.1 ms pulses in one experiment. The stimulation amplitude was 4X the motor threshold for both pulse durations. The response was normalized to the maximum LEAP response evoked by spinal stimulation. C: Top: A smaller LEAP response can be evoked by peripheral stimulation of the ipsilateral sural nerve (mostly cutaneous fibers) using a bipolar cuff electrode. Same pulse width (1 ms) was used for central and peripheral stimulation (N = 3 experiments). Bottom: Intracellular recordings of an ankle extensor motoneuron from a different experiment showing response to spinal cord stimulation (left) and peripheral stimulation of the sural nerve (right). Inset shows the membrane response at the beginning of stimulation with mixed EPSPs (red arrowheads) and IPSPs (blue arrowheads). D: Comparison of LEAP responses evoked by spinal stimulation Vs sural nerve stimulation in one experiment. The stimulation amplitude was 4X the motor threshold for both peripheral and central stimulation. The response was normalized to the maximum LEAP response evoked by spinal stimulation. B and D: The non-parametric Mann-Whitney test was used to compare LEAP responses. Data represented as median and interquartile ranges along with data points of multiple trials. **p < 0.01.

To explore the role of cutaneous inputs, we stimulated an ipsilateral cutaneous nerve, the sural nerve, with same duration (5 s), frequency (40 Hz), and pulse width (1 ms) using a nerve cuff. Sural stimulation generated smaller LEAP responses (Fig. 4C–D). When recorded intracellularly in extensor motoneurons, sural stimulation evoked smaller IPSPs during the train compared to SCS. In addition, the IPSPs adapted more rapidly, leading to a depolarized membrane potential for most of the stimulation train, a response that is noticeably different from that of SCS.

Importantly, LEAP manifested primarily on the ipsilateral side to SCS with very little contralateral extensor responses (Fig. S2). This indicates that while cutaneous afferents might contribute slightly to LEAP, the behavior is distinct from crossed extension responses mediated by pain pathways [25].

It was noticeable that activity in TA and SOL during and after stimulation exhibited a consistent alternating pattern. The largest LEAP responses were observed at stimulation locations that also elicited large TA responses during the stimulation train (Fig. 5A), suggesting the involvement of reciprocal inhibitory circuits between these antagonistic muscles. To investigate this possibility, we quantified the relationship between TA activity during the stimulation train and LEAP responses in SOL after stimulation ended. The data revealed a strong correlation between the two (Fig. 5B), supporting the hypothesis that reciprocal inhibition is a key mechanism underlying this rebound behavior. In addition, the data suggests that LEAP is primarily triggered by proprioceptive afferents which control these reciprocal circuits [26]. Collectively, these results indicate that while the LEAP circuit can be activated by multiple classes of sensory afferents, it is predominantly driven and controlled by proprioceptive pathways.

3.4. Stimulation parameters required to evoke LEAP

To determine the optimal stimulation parameters needed to evoke LEAP consistently in the ankle extensors, we tested combinations of different stimulation locations, frequencies, and amplitudes. A total of 10 experiments were dedicated for this purpose. The dorsal root entry zones served as reliable landmarks for precise and consistent placement of the stimulation electrode across animals (See Fig. 1B). These locations were carefully digitized in each experiment by tracing the dorsal root entry zones, the border of laminectomy, and the position of the stimulation electrode using a Microscribe MLX 3-D Digitizer system. An example of such a data set is shown in Fig. S3.

Anatomical studies have located the motor pools of the cat TA and SOL at L6 and L7 segments [27]. Hence, we stimulated multiple sites spanning from L4 to S2 segments in all 10 animals. While LEAP responses could be evoked at multiple stimulation locations (Fig. 6A), there was a noticeable variation in the response amplitude. The most pronounced LEAP responses were consistently observed either at L5 to L6 segments (N = 5 experiments) or more caudally at L7 to S1 (N = 5 experiments, Fig. 6B–C).

Additionally, we tested multiple stimulation frequencies including 10, 20, and 40 Hz while maintaining a fixed 5-s stimulation duration. Submaximal stimulation amplitudes were selected during these trials to ensure that spinal circuits are not saturated with current at higher frequencies. In 5 out of 8 experiments, the response increased linearly with increasing frequency (Fig. 6D–E), while the other 3 experiments had the maximal response at 20 Hz. Generally, the data demonstrates that LEAP can be evoked at a broad range of stimulation frequencies which might be used to control LEAP response at submaximal stimulation amplitudes.

Stimulation amplitudes exceeding the motor threshold were generally required to evoke LEAP, with the motor threshold typically ranging between 50 and 100 μA. The maximum LEAP was elicited at 2–5 times higher than the motor threshold with possible gradation of response observed between these values in SOL (Fig. 6F–G and Fig. S4) as well as other ankle extensors (Fig. S5). Overall, the amplitude of LEAP response was readily controllable via stimulation amplitude modulation. Therefore, LEAP response can be quickly established by scanning a few stimulation locations within the lower lumbar segments at an amplitude of 2–5xT and 20–40 Hz frequency, followed by creating a dose-response relation at the optimal stimulation location.

Collectively, our study has characterized LEAP, a post-SCS response mediated by PIR. In addition, we have delineated the specific stimulation protocol necessary for its induction and elucidated its underlying mechanisms. LEAP holds promising clinical potential for assisting postural movements in individuals with motor disorders.

4. Discussion

Spinal cord stimulation (SCS) has recently evolved as a rehabilitation technique for many traumatic and degenerative conditions, including spinal cord injury (SCI), stroke, spinal muscular atrophy, and Parkinson’s disease [28–31]. However, the therapeutic potential and functional repertoire of neuromodulation therapy are still constrained by the limited understanding of its cellular and circuit mechanisms. This study identified LEAP, a novel postural response to SCS, and investigated its underlying mechanisms, which provides insights into the circuit-level effects of neuromodulation. Our data indicates that LEAP is mediated by post-inhibitory rebound (PIR). Consistent with this, we have previously shown in the mouse spinal cord that repeated afferent stimulation recruits inhibitory pathways that contribute to post-activation depression of the H-reflex [10], suggesting that SCS can similarly engage inhibitory mechanisms.

4.1. Post-inhibitory rebound as the mechanistic basis of LEAP

The LEAP response exhibits the hallmarks of post-inhibitory rebound (PIR) behavior evoked by SCS-induced inhibition. Extensor force and EMG recordings showed inhibition of activity during stimulation, even silencing background activity present before stimulation. Similarly, intracellular recordings of triceps motoneurons that displayed LEAP showed marked hyperpolarization during stimulation. Although motoneurons themselves are capable of PIR [32,33], the data shows that LEAP originates primarily from PIR in spinal interneurons rather than motoneurons. This is evidenced by multiple facts: 1) The silent period that usually precedes LEAP following the cessation of stimulation, which does not happen in motoneuron intrinsic PIR behavior [32,33], 2) LEAP amplitude decreased when the cells were held at depolarized potentials, opposite to how intrinsic PIR behaves [34], 3) Widespread modulation of interneuron firing observed during LEAP, which is consistent with network-level circuit involvement. Collectively, our findings suggest that SCS activates a subset of inhibitory interneurons, which evoke inhibitory inputs onto other interneurons that subsequently exhibit PIR and, in turn, drive extensor motoneuron firing.

4.2. The source of inhibitory inputs

The TA and SOL exhibited consistently opposing patterns of activity during and after SCS. Additionally, the inhibition of extensors during stimulation was strongest at stim locations that elicited greater flexor activation during stimulation and larger LEAP poststimulation. This suggests that stimulation at these optimal locations recruits Ia afferents of the ankle flexors, leading to monosynaptic excitation of the flexors and disynaptic reciprocal inhibition of their counterpart extensors [26]. Indeed, the delay of the IPSPs recorded in triceps motoneurons during stimulation aligns with a disynaptic mechanism. Furthermore, a behavior similar to LEAP was observed during intraspinal stimulation in cats, where the LG muscle was recruited post-stim following stim-induced activation of the ipsilateral TA [35]. Similarly, a comparable phenomenon was reported by Sherrington, where the knee extensors exhibited rebound activation following stimulation of a knee flexor peripheral nerve [36]. Therefore, the poststimulation activation of extensors is likely driven by post-inhibitory rebound in interneurons receiving Ia inhibition during stimulation similar to that recorded in extensor motoneurons. These interneurons may include those involved in the locomotor central pattern generators [1,8].

4.3. Sensory afferents that trigger LEAP

Multiple lines of evidence suggest that LEAP is evoked by spinal cord stimulation (SCS) primarily through the activation of afferent fibers, rather than by direct recruitment of neurons. The bipolar configuration of our stimulation, along with the short inter-electrode distance (<2 mm), confines the electric field largely to the dorsal columns. Additionally, the observed alternation between TA and SOL activity supports the involvement of proprioceptive afferents in mediating LEAP. Furthermore, intracellular recordings revealed distinct firing modulation among interneurons located in close proximity within the gray matter, further suggesting circuit-level processing rather than nonspecific local activation.

Short stimulation pulses that are more selective to proprioceptive afferents effectively triggered LEAP. Indeed, selective activation of Ia afferents using high-frequency tendon vibrations has been shown to evoke prolonged responses in spinal interneurons [37]. Also, the inhibition of extensors during stimulation is most probably mediated by Ia inhibitory interneurons which are activated by la afferents of the antagonist flexor muscle.

Even though very high-intensity cutaneous nerve (sural) stimulation evoked some LEAP response, it was distinct from that evoked via SCS. This indicates that LEAP is not primarily mediated by cutaneous afferents. In addition, LEAP lacks fundamental features of the withdrawal and pain reflexes mediated by cutaneous afferents. For example, LEAP involves ipsilateral extensor rather than flexor responses and lacks crossed extension response in the other limb (see Fig. S2). Besides, we observed no wind up of LEAP responses with repeated stimulation, unlike C-fiber responses which do [38]. Thus, the data suggests that LEAP relies primarily on proprioceptive afferents with perhaps minor involvement of some cutaneous afferents.

4.4. Control of LEAP via stimulation parameters

The LEAP response was overall consistent and controllable through gradation of stimulation amplitude. However, stimulation effectiveness was location-dependent, as it recruits sensory axons in the vicinity of the stimulation electrode [9]. Therefore, an optimal location for LEAP would recruit ankle flexor Ia afferents during stimulation to induce post-inhibitory extensor rebound after stimulation. We have mapped the lower lumbar and upper sacral segments and found that the optimal stimulation location differs slightly among subjects but remains restricted to the lower lumbar segments (L5 to L7). These segments encompass the motor pools for the ankle flexors and extensors in the cat [27], with larger flexor responses to stimulation of the more rostral part of this range and to surface stimulation than deeper stimulation with intraspinal electrodes [39,40]. The variability of optimal location might stem from anatomical variability in the location of afferent axons. Similar variability between animals was observed when mapping the cord dorsum potential while vibrating the triceps tendon in decerebrate cats [37].

Human studies that focused on evoking lower limb extension found no location-specificity using epidural stimulation [41], but unexpectedly more specificity with transcutaneous stimulation [42]. These studies found that lower stimulation frequencies (5–15 Hz) evoke extension directly, i.e. during stimulation, while higher frequencies (20–30 Hz) evoke locomotor bursts. It is thought that low stimulation frequencies might preferentially activate extensors due to certain patterns of presynaptic inhibition and postsynaptic summation [43]. Our data shows that stimulation at 10 Hz can indeed activate ankle extensors directly, but only at specific locations, with the response to direct activation decaying quickly (see Fig. 5). Conversely, we observed potent long-lasting LEAP response across a broad range of frequencies. The possibility of evoking LEAP using higher frequencies is advantageous as they do not cause the paresthesia or discomfort common at low frequencies and can also help alleviate spasticity [44]. Therefore, LEAP might serve as a better alternative to low-frequency direct activation in a clinical setting.

4.5. Clinical significance and limitations

Postural stability in humans requires 5–20 % of the maximum voluntary torque at the ankle [12]. Moreover, ankle extensors play a major role in gait stability during walking [45]. Our study has identified LEAP as a new response to SCS that is selectively directed to ankle extensors, generating powerful contraction. LEAP is controllable via modulation of stimulation amplitude and is directed mainly to ankle extensors, offering great potential in clinical contexts. The force output from the SOL muscle during LEAP is potent, reaching up to 20 N at its peak, underscoring its utility in various motor tasks. In addition, it is coupled with synergistic effects from the gastrocnemius muscles. While we measured isometric forces without specifically testing their capacity to support weight bearing, the measured soleus forces are comparable to those required for stance in healthy cats [46,47]. On top of that, residual descending inputs in patients with incomplete motor disabilities could further augment the excitatory inputs during LEAP [48]. Therefore, LEAP responses can generate significant ankle extension torque that can help with many motor tasks. However, it is imperative to acknowledge the limitations of this study, including the need for future investigations to validate LEAP under pathological conditions, and the potential differences in muscle-specific activity between cats and humans. Nonetheless, LEAP presents a simple yet effective SCS-elicited response with multifaceted potential applications in augmenting plantar flexion and enhancing lower limb motor function.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.brs.2025.10.004.

Data availability

All animal data reported in this paper will be made available by the corresponding author upon reasonable request.