Electrical stimulation of the sural cutaneous afferent nerve controls the amplitude and onset of the swing phase of locomotion in the spinal cat

Abstract

Sensory feedback plays a crucial role in the control of locomotion and in the recovery of function after spinal cord injury. Investigations in reduced preparations have shown that the locomotor cycle can be modified through the activation of afferent feedback at various phases of the gait cycle. We investigated the effect of phase-dependent electrical stimulation of a cutaneous afferent nerve on the locomotor pattern of trained spinal cord-injured cats. Animals were first implanted with chronic nerve cuffs on the sural and sciatic nerves and electromyographic electrodes in different hindlimb muscles. Cats were then transected at T12 and trained daily to locomote on a treadmill. We found that electrical stimulation of the sural nerve can enhance the ongoing flexion phase, producing higher (+129%) and longer (+17.4%) swing phases of gait even at very low threshold of stimulation. Sural nerve stimulation can also terminate an ongoing extension and initiate a flexion phase. A higher prevalence of early switching to the flexion phase was observed at higher stimulation levels and if stimulation was applied in the late stance phase. All flexor muscles were activated by the stimulation. These results suggest that electrical stimulation of the sural nerve may be used to increase the magnitude of the swing phase and control the timing of its onset after spinal cord injury and locomotor training.

studies in fictive locomotion preparations and in decerebrate spinal cats with a locomotor pattern induced pharmacologically have shown that afferent input has profound effects on the locomotor cycle and that the responses to afferent stimulation are modulated by the phase of the gait cycle (reviewed in Pearson 2004; McCrea 2001; Rossignol et al. 2006). Muscle or cutaneous nerves can be used to enhance extension or terminate an ongoing flexion phase [e.g., electrical stimulation of group I extensor afferents enhances extensor activity when delivered during the extension phase and initiates an early extension when delivered during the flexion phase (Conway et al. 1987; Guertin et al. 1995)], whereas others can be used to enhance flexion or terminate an ongoing extension phase, most notably the sural cutaneous nerve (Schomburg et al. 1998).

In an acute spinal fictive preparation (with locomotion induced pharmacologically), stimulation of the sural nerve with trains of low-intensity pulses (below 2 times threshold of the volley recorded in the dorsal root) enhanced and prolonged the flexion phase or terminated an ongoing extension, whereas stimulation at higher strength (2–5 times threshold) during late flexion or early extension enhanced extensor activity on occasions (Schomburg et al. 1998). In intact and premammilary decerebrate cats, Duysens and Stein (1978) found that stimulation of the sural nerve with low-intensity stimulus trains enhanced extensor activity, especially for stimulation delivered during stance, but had little effect on the durations of the extension or flexion phases. With single-stimulus pulses at the same intensity (<2 times threshold), Pratt et al. (1991) found no effects on extensors for sural nerve stimulation.

In addition, spinal reflex pathways are plastic and undergo changes after injury and training regimen (Cote and Gossard 2004; Cote et al. 2003; Frigon et al. 2009; Frigon and Rossignol 2008 and 2009; Skinner et al. 1996). Responses to sural nerve stimulation may thus be different in trained spinalized animals, which could impact the potential use of afferent stimulation as a rehabilitation tool. Furthermore, studies in spinal cord-injured cats have looked at changes in reflexes with single-stimulus pulses, but no studies have explored the functional changes in kinematics and muscle activity using the stimulation of afferent nerves in chronic spinal cats trained to locomote on a treadmill. In those animals, afferent stimulation could potentially be used to correct the remaining deficits in their gait, which include reductions in step length, height, and duration (Belanger et al. 1996).

We therefore investigated the phase-dependent (stance vs. swing) effects of stimulation of the sural nerve on the kinematics and electromyographic (EMG) characteristics of hindlimb movements in chronic spinal cats trained to locomote on a treadmill.

METHODS

Animals and Experimental Protocol

Five adult domestic short-haired female cats (weight: 2.6–3.2 kg) were used in these experiments. Before any surgery, cats were trained (5 days/wk, 15–20 min/day) using food rewards to walk on a motorized treadmill at speeds ranging from 0.2 to 0.6 m/s. Cats were then chronically implanted with nerve cuffs and bipolar EMG electrodes in their hindlimb(s). Once we determined that the implants did not affect gait, the animals' spinal cords were transected at the low thoracic (T12) level. They were housed individually posttransection and carefully monitored to prevent medical complications. One week postspinalization, treadmill training was initiated for 15–30 min/day, 5 days/wk, with the forelimbs standing on a platform raised ≈1 cm above the treadmill. Walking speed was fixed at 0.3 or 0.4 m/s depending on the cats. In the first few weeks, manual stimulation of the perineum, base of the tail, and/or abdominal regions was necessary to induce or maintain locomotion in the hindlimbs. In addition, the hindquarters needed to be lifted by the tail and supported by the experimenter. The amount of weight support needed decreased with time, and locomotion could be induced with minimal tail stimulation after 2–4 wk of training. Profiles of muscle activation and kinematics of gait were measured in four of the five animals. In cat 5, bilateral muscle activation patterns and video were recorded but detailed kinematics were not measured. Modifications in EMGs and kinematics produced by stimulation applied to the cutaneous sural nerve were measured to determine the phase and amplitude dependence of afferent stimulation effects on the locomotor pattern of spinal cats (see Fig. 1).

Fig. 1.

Experimental setup. Bipolar eletromyographic (EMG) electrodes (7–9 muscles) and nerve cuffs (around the sural nerve for delivering stimulation and around the sciatic nerve for monitoring nerve activity) were implanted chronically in the right hindlimb. In two cats, EMG electrodes were also implanted in the left hindlimb. All wires were tunneled subcutaneously to exit between the shoulders and attach to a connector mounted on a small jacket (or to exit at a head-mounted connector). After transection, cats were trained to walk on a motorized treadmill with the forelimbs standing on a platform raised ∼1 cm above the treadmill. Reflective markers were taped on the skin over selected hindlimb joints during kinematics and EMG recording sessions.

Cats were studied for several weeks but only after their locomotor performance had plateaued to a stable spinal locomotor pattern. Thus, the timing of the recordings postinjury was not considered as a dependent variable that could affect the results. Stable spinal locomotion was reached after a minimum of 27 days of training (see details for each cat in the results). For each recording session, cats were solicited to walk on the treadmill for 15–20 min while ∼15 records (45 s long) of the kinematics, EMGs, and video were taken. In total, 152 posttransection recordings including stimulation deliveries were used for analysis. The number of recordings varied depending on the animal's health and the quality of the EMGs (cat 1: 37 recordings over 7 wk, cat 2: 23 recordings over 2 wk; cat 3: 16 recordings over 2 wk; cat 4: 55 recordings over 7 wk, and cat 5: 21 recordings over 3 wk). At the termination of the experiments, cats were euthanized.

The protocol was approved by local Institutional Animal Care and Use Committees and animal care followed National Institutes of Health guidelines. All surgical procedures were conducted under general anesthesia and aseptic conditions in a surgical suite. Cats were anesthetized using ketamine HCl (Ketaset, 15 mg/kg im), atropine sulfate (0.05 mg/kg im), and inhaled isoflurane (5%). They were then intubated and maintained at a surgical level of anesthesia with isoflurane (1–3%) while the heart rate, blood pressure, temperature, and respiration were monitored. After surgery, analgesia was administered for 72 h using a transdermal fentanyl patch (Duragesic, 25 μg/h). Antibiotics (ampicillin, 250 mg/day) were administrated subcutaneously for 10 days after surgeries.

Electrode Implantation

Bipolar EMG electrodes (Cooner Wire AS633) were implanted unilaterally in eight right hindlimb muscles for cats 1 and 4 [biceps femoris anterior (BFA), rectus femoris (RF), sartorius medial (SartM), biceps femoris posterior (BFP), vastus lateralis (VL), tibialis anterior (TA), soleus (Sol), and medial gastrocnemius (MG)]. For cats 2, 3, and 5, EMG electrodes were implanted bilaterally in seven muscles [sartorius anterior (SartA), BFA, BFP, VL, TA, Sol, MG, and lateral gastrocnemius (LG)]. All wires were tunneled subcutaneously to either exit between the two shoulders and attach to a connector mounted on a small jacket worn by the cat or exit at a head-mounted multipin connector (Gregor et al. 2006). In addition, nerve cuff electrodes (Haugland 1996) for monitoring nerve activity (tripolar cuff on the sciatic nerve) or delivering stimulation (bipolar cuff on the sural nerve) were implanted unilaterally on the right leg. The sciatic nerve was mobilized for ∼4 cm just proximal to its branching point into tibial and common peroneal components. Nerve diameter was determined by measuring its circumference with a piece of suture, and a cuff was selected that provided a cuff-to-nerve diameter ratio of 1.5. The cuff was installed with its distal end lying ∼1 cm from the branching point of the sciatic. The cuff was sutured shut around the nerve using silk sutures. A similar fitting procedure was used for the sural nerve, with the cuff installed around the nerve approximately halfway between the knee and ankle joints as the nerve is passing over the LG muscle.

Spinalization

A midline incision was made over the dorsal spinous processes between T9 and L1. The overlying muscles were separated using blunt dissection, and a narrow laminectomy was performed over two vertebral bodies (T11–T12). The dura was opened with a midline incision, xylocaine (1%) was applied to the spinal cord, and the cord was transected using microscissors under a surgical microscope at the T11–T12 level. The dura, muscles, and skin were closed in layers. After transection, animals were housed in cages lined with an orthopedic foam mattress. Bladders were expressed manually, and reflex defecation was initiated twice daily.

Data Collection

After spinalization, single 100-μs-duration biphasic pulses were delivered to the sural nerve at different current intensities to evoke action potentials in the sciatic nerve and establish the recruitment threshold of the sural nerve cuff. The recruitment thresholds were stable and did not change through the recording period. Kinematics and muscle EMGs were collected synchronously before and after spinalization using a high-speed video motion analysis and data-acquisition system (Vicon Motion System). Three cameras were used to capture the kinematics of the hindlimb and forelimb on the right side of the body. All EMG signals were amplified and bandpass filtered (10–1,000 Hz) via a differential alternating current amplifier (model 1700, A-M Systems, Carlsborg WA). Muscle activity was sampled at 2,400 Hz and kinematics data at 300 Hz. Markers for the kinematics of locomotion were placed unilaterally (right side) on the skin overlying the ischium, femoral head, estimated knee joint center, lateral malleolus, metatarsophalangeal joint, and tip of the digits of the hindlimb as well as on the humeral head, estimated elbow joint center, lateral epicondyle, metacarpophalangeal joint, and tip of the digits of the forelimb (see details in Boyce et al. 2007).

Afferent stimulation (biphasic pulses, 20- to 200-ms train duration, 100-Hz frequency, 100-μs pulse duration) was applied postspinalization via an isolated pulse stimulator (model 2100, A-M Systems) during different phases of gait (6 equidistant phases, i.e., the step cycle was divided into 6 segments of the same length): early stance, midstance, late stance, early swing, midswing and late swing (see Fig. 3A). Timing of the stimulation to the step cycle was accomplished by varying the delay between Sol muscle activity onset and stimulation delivery. Sural stimulation amplitude level was set between 1.0 × T (1 times threshold) up to 7.0 × T (grouped into 4 stimulus amplitude levels: = 1.0 × T, (1.0–2.0] × T, (2.0–3.0] × T, and >3.0 × T). Grouping of the stimulus amplitude levels was chosen based on previous studies (Schomburg et al. 1998; Duysens and Stein 1978; Pratt et al. 1991) reporting changes in the nature of the response to sural nerve stimulation for stimuli around 2.0 × T and the fact that all responses above 3.0 × T were flexion withdrawals in our experiments. In some animals, the effects of stimulus train durations were evaluated by measuring the responses to two different train durations at the same stimulation amplitude level. Overall, four different stimulus train durations were used: 20 or 50 ms (considered as short train duration) and 100 or 200 ms (considered as long train duration). As we did for stimulus amplitude, the durations of the stimulus trains were chosen based on previously published literature with sural nerve stimulation (Schomburg et al. 1998; Duysens and Stein 1978). Cat 1 was tested at 20 and 200 ms, cats 2 and 3 were tested at 20 ms only, cat 4 was tested at 50 and 100 ms, and cat 5 was tested at 50 ms only. Stimulation was only applied during steady bouts of walking.

Fig. 3.

Illustration of the kinematics of locomotion recorded in one cat. A: before transection. Stance length (a) was defined as the horizontal distance traversed by the metacarpophalangeal marker between touchdown of the paw and lift off, whereas swing length (b) was defined as the horizontal distance covered between lift off and touchdown. Swing height (c) was measured by the maximal vertical displacement of the metacarpophalangeal marker during the swing phase of the step. The step cycle was divided into six equidistant phases (early, mid, and late stance and early, mid, and late swing). B: after transection (9 wk of training). C: after transection and training with sural nerve stimulation. Stimulation (biphasic, 200- or 20-ms train duration, 100-Hz stimulation frequency, 100-μs pulse duration) was delivered either during the swing or stance phase at different level of stimulation (1.0 and 7.0 × T, where T is the pulse amplitude threshold of sural nerve stimulation producing a response in the sciatic nerve). Arrows indicate when the stimulation was applied, and thick solid lines at the ankle joint indicate the period during which stimulation was delivered. Note the enhancement of the swing phase when stimulation was delivered during the swing (left). Sural stimulation can also terminate the stance phase and initiate the swing phase if delivered during stance at high intensity (7.0 × T, right). Treadmill speed: 0.3 m/s. Stance is indicated by the black line; swing is indicated by the gray line.

Data Analysis

For the kinematics of locomotion, step length and swing height of the right hindlimb with/without stimulation were calculated. Step length was defined as the stance length for stimulation delivered during the stance phase and as the swing length for stimulation delivered during the swing phase. Stance length was measured as the distance between touch down and lift off of the right foot, whereas swing length was measured as the distance between lift and contact of the right foot. Swing height was determined using the maximal vertical displacement of the metatarsophalangeal joint during the swing phase (Boyce et al. 2007). All parameters were measured before and after transection over repeated steps. In posttransection animals, we measured those parameters for the step preceding stimulus delivery, the stimulated step (step 0), and the two following steps (steps 1 and 2). When stimulation was applied during the stance phase, the percentage of early swing onset, defined as the instantaneous and premature swing initiation due to the stimulation, was also determined. Early termination of the stance phase was marked by an instantaneous and simultaneous premature termination of the raw EMG activity in all extensors. During nonstimulated steps, termination of activity in the extensors was never concurrent (i.e., termination was different between each of the extensors). The effects of phase of stimulus delivery and intensity of stimulus were evaluated by varying those parameters over multiple trials with each animal (see results).

Raw EMGs of each muscle were highpass filtered (30 Hz), rectified, and then lowpass filtered (Butterworth, zero lag, fourth order, cutoff frequency: 10 Hz) to obtain EMG linear envelopes. Rectified lowpass-filtered (70 Hz) EMG data were used to measure the times of onset and offset of activity in each muscle. EMG burst onsets and terminations were detected by identifying the mean resting level of each muscle's EMG and setting a threshold of mean + 2SD of the mean resting level to detect onset and termination. Burst onsets and terminations were then visually confirmed using a custom graphical interface routine. Burst duration, average EMG, integral EMG, and the peak EMG value were calculated. Each parameter corresponding to the stimulated steps were compared and normalized with the average of all nonstimulated steps (by excluding at least 3 steps after stimulation) recorded in the same session. The step cycle duration was also calculated and averaged for all nonstimulated steps using the Sol muscle EMG as the step cycle onset marker (number of nonstimulated steps used for the averages was 19 ± 8). The poststimulation phase shift in the locomotor cycle was then measured as the time between the actual onset of Sol muscle activity and the predicted onset (obtained by adding the average step cycle duration of all nonstimulated steps to the Sol onset of the stimulated step) for the two steps after the stimulation.

Statistical Analysis

For the kinematics of locomotion parameters (step length and swing height), the effects of intensity and duration of stimulation as well as the effects of phase of stimuli delivery were analyzed. A repeated-measure linear mixed model with stimulus intensity [expressed as one of four stimulus amplitude level groups: 1.0 × T, (1.0–2.0] × T, (2.0–3.0] × T, and >3.0 × T], stimulus train duration (in ms), and phase of locomotion (1–6) as fixed effects and cats as random effect with steps within a cat as the repeated measure was used to establish the effects of the three parameters of stimulation on step length and swing height (no interaction terms, only main effects, PASW Statistics, SPSS, Chicago IL). The data of steps 0 (stimulated step), 1, and 2 were first normalized with the data of the step before stimulation. Post hoc pairwise comparisons (with Bonferroni's adjustment for multiple comparisons) of the estimated marginal means of the models were used to assess the significance of the differences between stimulation parameters levels with a significant effect on model output. The 95% confidence intervals (CIs) of the estimated marginal means were also used to estimate if the step length or swing height with stimulation were significantly increased or decreased compared with the step before stimulus delivery (i.e., we verified if CIs included 100%, which was the normalized level of the preceding nonstimulated step). For the early stance termination and swing initiation cases, the intensity and phase of stimulation effects were tested with Pearson's χ²-tests.

For EMG parameters (average, integral, peak, and burst duration), the effect of sural nerve stimulation was evaluated for each phase of stimuli delivery using paired t-tests (when the number of trials was ≥10) or Wilcoxon signed rank-sum tests (when the number of trials was <10) between stimulated steps and nonstimulated steps. Only trials with short train duration stimulation were included in the EMG analyses. For each recording session, data (average, integral, peak, and burst duration for the flexor muscles and Sol cycle duration) from many nonstimulated steps (19 steps on average) were averaged and used for comparison and to normalize the data of each stimulated step in the same session (i.e., the stimulated step EMG characteristics were normalized to the average EMG characteristics of nonstimulated steps collected at least 3 steps after a stimulated step).

The poststimulation phase shift in the Sol EMG onset was analyzed using repeated-measure linear mixed models to establish the effects of the two parameters of stimulation varied (phase of stimulation delivery and intensity) on the delay between predicted and actual Sol burst onsets (fixed effects were phase of stimulation delivery from 1 to 6 and stimulus intensity as multiple of threshold, with steps within cats as the repeated measure; main effects only). The dependent variables in each of the eight different models used were 1) right Sol onset delay [i.e., time between the actual onset of the right Sol muscle activity and the predicted onset (obtained by adding the average step cycle of nonstimulated steps to the Sol onset of the stimulated step)] for the 1st and 2nd step after stimulation; 2) left Sol onset delay for the same two steps; 3) difference between right Sol onset delays for the 1st and 2nd step; 4) difference between left Sol onset delays for the 1st and 2nd step; 5) difference between Sol onset delays of the 1st step for the left and right legs; and 6) difference between Sol onset delays of the 2nd step for the left and right legs. The 95% CIs of the estimated marginal means for each of the six phases of gait were used to evaluate if the delays or differences were significantly different from zero. Significance was set at P < 0.05 for all statistical tests. All values reported, unless otherwise stated, are means ± SE.

RESULTS

In the present study, the effect of sural nerve stimulation on locomotor parameters and muscle activity was quantified during locomotion after spinalization for the 5 animals studied over 19 sessions, with an average of 37 trials of stimulation per session. Details on the number of trials under each condition (phase of gait, intensity, etc.) are given below. Recordings of kinematics and EMG did not commence until the animals could maintain stable hindlimb locomotion uninterrupted for prolonged periods (>10 min; meaning a training period of 30, 122, 27, 65, and 41 days postspinalization for cats 1, 2, 3, 4, and 5, respectively). Cat 4 was highly spastic and showed marked flexor and extensor spasms in response to cutaneous stimulation. Thus, a higher divider (10.7 cm tall instead of 6.9 cm tall for the other animals) had to be placed between its legs to prevent the crossing of the legs that occur in spinal cord-injured cats due to increased adductor activity (Belanger et al. 1996).

Classes of Afferents Activated by Sural Stimulation

The estimated conduction velocities of cutaneous afferents activated at different multiples of threshold in cat 5 are shown in Fig. 2. Conduction velocities were measured by dividing the distance between the sural and sciatic nerve cuffs (obtained at autopsy) by the latencies of the potentials evoked in the sciatic nerve from sural nerve activation. The shortest latencies measured in two other cats were even faster at 0.7 and 0.8 ms, suggesting that conduction velocities may have been higher in those two animals (no length measurements available). Threshold excitatory responses were mediated by fibers in the Aβ-range [40–90 m/s (Burgess et al. 1968; Hunt and Mc 1960; Whitehorn et al. 1974)], indicating that the largest fibers in the sural nerve were not affected by the cuff. Our data also suggests that, as in Boyd and Kalu (1979), stimulation below ∼2.4 × T activates primarily Aβ-fibers, whereas Aδ-fibers (5–30 m/s) start being recruited above that level.

Fig. 2.

Nerve fibers activated at different sural stimulus intensities in cat 5. Each trace shows the averaged evoked potentials recorded in the sciatic nerve for 64 stimuli delivered to the sural nerve at the multiple of threshold (T) indicated to the right of the trace. Stimulus onset is indicated by the descending edge on the left of each trace. The numbers above each trace are the estimated conduction velocities for the associated waveform (marked by a solid circle). Distance between the cuffs was measured at autopsy. All traces are displayed on the same vertical and horizontal calibrations.

Sural Nerve Stimulation Effects on Kinematics

General effects of stimulation on kinematics of locomotion.

One example of the kinematics observed before transection, after transection without stimulation, and posttransection with sural nerve stimulation is shown in Fig. 3, A–C, which demonstrates that longer and higher steps were produced when the sural nerve was stimulated during the swing phase of the gait. Stimulation at low intensity delivered early in the stance phase had no effects on kinematics, whereas high-intensity stimulation could terminate the stance phase and initiate swing when delivered at any phase of stance (Fig. 3C).

Average swing length and height are shown in Fig. 4 for cats 1–4. These parameters were measured for pretransection steps, posttransection steps without stimulation, and posttransection steps with sural nerve stimulation applied during the swing phase. Posttransection steps with stimulation during the swing phase were chosen for the analysis because the effects on swing length and height were larger for stimulation delivered during the swing phase (see Fig. 3 and also Fig. 5). For each cat, sural nerve stimulation delivered during the swing phase of gait promoted longer (17.4 ± 5.8% increase in average) and higher (129 ± 12.4% increase) steps compared with steps without stimulation in the same animal posttransection [see Fig. 4; Bonferroni-corrected pairwise comparisons of estimated marginal means of linear mixed model with swing length or height as the dependent variable and condition (pretransection, posttransection, and posttransection with sural stimulation using short or long stimuli trains) as the factor with step as the repeated measure; condition had a significant effect on swing height or length in each cat]. One exception to the increase in swing length with sural nerve stimulation was cat 4, where the posttransection swing length was not significantly different from the posttransection swing length with long stimulus trains. However, cat 4 was the only cat that presented longer steps after transection compared with before (Bonferroni-corrected pairwise comparisons, P < 0.05). For the three other cats, swing length and step height after transection were smaller than before transection (Bonferroni-corrected pairwise comparisons, P < 0.05) but increased with stimulation. Stimulation of the sural nerve increased swing heights to above or equal to pretransection values in all animals (Bonferroni-corrected pairwise comparisons, P < 0.05 or no significant difference).

Fig. 4.

Swing length and height results for each cat. Swing length (left) and height (right) were measured before transection, after transection without stimulation, and posttransection with short or long train duration sural nerve stimulation applied at any time during the swing phase of the gait. Short train durations were 20 ms (cats 1–3) and 50 ms (cat 4), whereas long train durations were 200 ms (cat 1) and 100 ms (cat 4). Averages were normalized to the pretransection data for each cat (means ± SE, in %). Both parameters decreased after transection compared with the uninjured situation (except for the very spastic cat 4). Sural nerve stimulation delivered during the swing phase of the gait after transection increased both the swing length and height (Bonferroni-corrected multiple pairwise comparisons of estimated marginal means of a linear mixed model, P < 0.05). For cat 1, the stimulus amplitude ranges used were (1.0 to >3.0] × T for the short stimulus trains and (1.0–3.0] × T for the long stimulus trains. For cats 2 and 3, the amplitude range used was (2.0 to >3.0] × T, and, finally, for cat 4 the amplitude range was (1.0, 3.0] × T for both short and long stimulus trains.

Fig. 5.

Changes in step length and height as a function of the phase of stimulation. Data were normalized to the step preceding the stimulation (means ± SE, in %). A and B: graphs representing averages for the stimulated steps for each cat [stance/swing length (A) and swing height (B)]. C and D: graphs representing averages (of all cats together) for the stimulated step (step 0) and the two following steps (steps 1 and 2) [stance/swing length (C) and swing height (D)]. The 100% line represents normal length or height in spinalized animals (without stimulation). The two plus symbols (⁺⁺) indicated that the 95% confidence interval (CI) of the estimated marginal means of the linear model relating step length or swing height as a function of phase of stimulus delivery, stimulus amplitude, and train duration does not include 100% for a specific phase of stimulation; the asterisk (*) indicates a within-step significant difference in step length- or swing height-estimated marginal means between the phase of stimulation (Bonferroni-corrected multiple pairwise comparisons, P < 0.05). The number of trials (n) is shown for each phase of stimulus delivery.

Effects of the phase of locomotion in which stimulation was delivered.

Stimulation was delivered at different phases of gait. Stimulus intensity and train duration were varied within and between sessions (stimulus intensity from 1 × T to >3 × T and train duration from 20 to 200 ms). For the stimulated step, only phase of stimulus delivery had a significant effect on step length and swing height (linear mixed-model P < 0.05; Fig. 5). If the stimulation was delivered during the stance phase, the stance length of the stimulated step was shorter (Fig. 5, A and C), especially if stimulation was delivered during mid or late stance. Effects on swing height were minimal as the 95% CIs of the estimated marginal means for stimulation in the early, mid, and late stance included 100% (i.e., the normalized swing height of the preceding nonstimulated step).

If delivered during the swing phase, both swing height and length were increased compared with the previous nonstimulated step (lower bounds of the 95% CI of the estimated marginal means above 100% except for swing length with midswing stimulus delivery). The phase of swing at which stimulation delivery produced the longest steps varied between animals (Fig. 5A), but the highest steps were always observed when stimulation was delivered in early swing or midswing (Fig. 5B). The average effects of sural nerve stimulation on three consecutive steps over the four animals were evaluated using linear mixed models for the two steps after the stimulated steps and are shown on Fig. 5, C and D. For step 1 (the step after stimulation), none of the parameters of stimulation (phase of delivery, stimulus train duration, or intensity) had a significant effect on step length, but all had an effect on swing height. For step 2, only stimulus train duration and stimulus intensity had a significant effect on swing height, and no parameter had a significant effect on step length. While the stimulated step (step 0) was longer for stimulation delivered during swing, the following step (step 1) was shorter than a normal step (upper bound of 95% CI < 100%). This occurred because the ipsilateral flexion phase was interrupted when the contralateral leg initiated a flexion phase (see EMG data below). For stimulation during stance, the effects on the length of the following step were not significant (95% CI included 100% except for stimulation in early stance, which was slightly shorter). Step 2 was always back to the average posttransection swing length (95% CIs included 100% for all marginal means at the different phase of stimulus delivery of the linear mixed model). Although a number of stimulation parameters had significant effects on the heights of step 1 and 2 (as per the linear mixed models), the 95% CI of the estimated marginal means of step 1's height included 100% except for stimulation delivered in the early swing phase of the previous step. In addition, the estimated marginal means of step 2's height were not significantly different from one another at any of the phases of gait (Fig. 5D), although the 95% CI did not include 100% for stimulus delivered in midstance to late swing.

Effects of stimulation intensity on early swing initiation.

The effects of the stimulation intensity on early swing initiation were also investigated. For stimulation trains delivered during stance, the percentage of premature swing initiation due to the stimulation was evaluated (Fig. 6). This percentage increased with the intensity of stimulation (χ²-test, P < 0.001). We also observed that the stimulation was more effective (which means that switching to the swing phase occurred at a lower intensity of stimulation or more frequently for the same intensity of stimulation) if delivered during late stance instead of early stance (χ²-test, P < 0.001).

Fig. 6.

Stance phase termination and swing phase initiation. The percentage of premature swing phase initiation due to stimulation given either in early, mid, or late stance phase is shown for four different levels of stimulation. The number of trials (n) for each situation is shown at the bottom of each bar. Note that the incidence of stance termination and swing initiation increases with the intensity of stimulation (χ²-test, P < 0.001) and if stimulation is given later in the stance phase (χ²-test, P < 0.001).

Sural Nerve Stimulation Effect on EMG

Examples of the EMG recordings observed after transection on the stimulated and contralateral legs are shown in Fig. 7. A number of channels were defective in each cat, but we obtained 10 and 13 channels of EMG (right TA, right SartA, left VL, and left TA muscles were missing on cat 2 and the left BFP muscle was missing on cat 3) for the cats implanted bilaterally. For the cats implanted unilaterally, the TA muscle was missing on cat 1 and the BFA and VL muscles were missing on cat 4. Sural and sciatic nerve cuffs were functional for the duration of the study in all cats.

Fig. 7.

Examples of EMG recordings with stimulation of the sural nerve given either during the swing (A) or stance (B) phase of locomotion on a treadmill. Sural stimulation is indicated on the bottom trace and by a vertical dashed line. Dotted lines on the top show the real onset of the soleus (Sol) burst, whereas arrows indicate the expected onset of the Sol burst if the rhythm had not been perturbated. A: cat 3. Left, recordings of the right leg before spinal transection. Middle, recordings of the right (stimulated) leg after transection. Sural stimulation was delivered in late swing (2.5 × T). Note the enhancement in the activity of the flexor muscles inducing a delay in the Sol burst onset. Right, recording of the contralateral leg for the same stimulation. The extensor muscle activity was prolonged, inducing a delay in the next Sol burst onset as well. B: cat 2. Sural nerve stimulation (3 × T) was delivered in early stance. The ongoing extensor muscle activity was terminated and replaced by flexor muscle activity on the stimulated leg, whereas the ongoing flexor burst on the contralateral leg was interrupted by a short extensor burst. The rhythm interruption induced a phase shift on both legs; the original rhythm continued, but with a maintained time shift. EMGs were rectified and lowpass filtered at 30 Hz. Stimulus train duration: 20 ms; frequency: 100 Hz; treadmill speed: 0.3 m/s. Extensor muscles were as follows: Sol, vastus lateralis (VL), biceps femoris anterior (BFA), medial gastocnemius (MG), and lateral gastrocnemius (LG). Flexor muscles were as follows: sartorius anterior (Sart), biceps femoris posterior (BFP), and tibialis anterior (TA).

Effects for stimulation delivered during the swing phase.

When stimulation of the sural nerve was delivered during the swing phase of gait, flexor muscles activity (BFP, Sart, TA) was enhanced and the next extension phase lasted longer (since the step taken was longer) introducing a delay in the rhythm, as evidenced by a shift in the Sol activity onset (Fig. 7A). EMG averages, peaks, and integrals as well as the burst durations of all flexor muscles were increased with the stimulation (paired t-test, P < 0.001) compared with nonstimulated steps for the stimulated leg (Fig. 8). The perturbations observed on the contralateral leg were minimal with a slight delay in the Sol activity onset. This delay was induced by a prolongation of the ongoing stance phase in the contralateral leg, which occurred as a result of the stimulated leg still being in the swing phase (Figs. 7A and 9A).

Fig. 8.

Flexor muscle activity characteristics of the stimulated step as a function of the phase of stimulation (means ± SE, in %). EMG average, peak, integral, and burst duration of the BFP, TA, and Sart (anterior or medial) were calculated. Raw data of the stimulated steps were normalized and compared with regular nonstimulated steps of the ipsilateral leg for each trial. The number of trials (n) is shown under the top left graph. *P < 0.05; **P < 0.01; ***P < 0.001 (paired t-tests when n ≥10 or Wilcoxon signed rank sum tests when n < 10).

Fig. 9.

A: illustration of Sol and TA activity with a right sural nerve stimulus (20-ms train duration, 2.5 × T) delivered during late swing (cat 3). Note the shift in the left Sol burst onset (dotted lines) compared with the right Sol activity after the stimulation (vertical dashed line) creating an overlap in Sol activity between both legs. The amount of overlap in the activity of the extensors is limited by the flexors' activity since right and left flexors cannot be activated simultaneously. The arrow shows that the right TA activity was shortened and terminated when the left TA started to be activated for the first step after stimulation. B: illustration of Sol activity with sural nerve stimulation (20-ms train duration, 3 × T) delivered during early stance (cat 2). The onset of the left Sol bursts (dotted lines) coincides with the end of the right Sol bursts before and after the stimulation of the right sural nerve, meaning that both legs underwent the same phase shift after the stimulation.

Effects for stimulation delivered during the stance phase.

EMG analysis was performed only on trials showing an early switch from the stance to swing with stimulation (as we did not observe significant effect on kinematics for the stimulus train delivered during the early stance and midstance phase at intensity that did not produce a switch of stance to swing). When stimulation was delivered during the stance phase of the gait and caused an early termination of the stance phase, the ongoing extensor muscles activity (Sol, BFA, VL, RF, MG, or LG) was terminated and replaced by flexor muscles activity, inducing a permanent shift in the rhythm (Fig. 7B). Steps subsequent to the stimulation were of similar duration, but the cycle was reset. EMG averages and peaks of the TA and BFP muscles were increased compared with normal steps when stimulations were applied during mid and late stance (P < 0.01; Fig. 8). The perturbations observed on the contralateral leg were as important as the ones observed on the stimulated leg (Fig. 7B). While the stimulated step was in early stance or midstance phase, the contralateral leg was in the swing phase. Since the stimulation promoted an early swing in the ipsilateral leg, both legs were now in the swing phase, making the cats fall (total fall prevented by the tail support). This near fall then caused contralateral foot contact with the treadmill belt and the onset of stance in the contralateral leg (Fig. 7B). A new step cycle was then reinitiated: the stimulated leg proceeded to a stance phase while the contralateral leg proceeded to a swing phase in our example (Figs. 7B and 10B).

Fig. 10.

Time shift of the Sol burst onset after sural nerve stimulation. A: time difference between the real and expected Sol burst onset was calculated by subtracting the actual/real Sol burst onset time minus the expected onset time for the 1st and 2nd steps after stimulation for both legs (stimulated and contralateral). B: data of three cats (implanted bilaterally) were averaged for each phase of stimulation (means ± SE). The Sol burst onset was advanced (negative values) when stimulation was delivered during stance, whereas it was delayed (positive values) when stimulation was delivered during swing. Comparisons between both legs were performed on the 1st and 2nd steps and showed a shorter delay for late swing stimulations on the contralateral leg. A linear mixed model with stimulated or contralateral leg Sol onset delay (step 1 and step 2) as the dependent variables, phase of stimulus delivery and stimulus amplitude (grouped into 4 levels, as described in methods) as fixed factors and steps as repeated measure showed that phase of stimulus delivery was a significant factor on Sol onset delay for both steps of the stimulated leg after the stimulation and on the onset delay of step 2 of the contralateral step. Stimulus amplitude was not a significant factor for any of the Sol onset delays. Bonferroni-corrected multiple pairwise comparisons of estimated marginal means of the Sol onset delays showed significant differences in the delay between early stance and midstance versus midswing and late swing for the stimulated leg, but no significant differences were found between any of the Sol onset delays of the contralateral legs. The asterisk (*) indicates that the 95% CI of the estimated marginal means of the model (at each phase of stimulus delivery) do not include zero. The linear mixed models with the difference in Sol onset delays between the 1st and 2nd steps of each leg or between the onset delays of the stimulated and contralateral legs (one for each step) as dependent variables and the same factors as above demonstrated a significant effect of phase of stimulus delivery and no significant effect for stimulation amplitude. The dagger (†) indicates that the 95% CI of the estimated marginal means for these models (at each of stimulus delivery) do not include zero. The Sol cycle duration for these three cats was 747 ± 65 ms (mean ± SD calculated over 24 nonstimulated steps in average on 38 recordings).

Perturbations in ipsilateral and contralateral limb step cycles.

The poststimulation phase shift in the locomotor cycle was measured in three cats (cats 2, 3, and 5 implanted bilaterally) as the time between the actual onset of the Sol muscle activity and the predicted onset (obtained by adding the average step cycle to the Sol onset of the stimulated step) for the 1st and 2nd step after the stimulation (Fig. 10A). When stimulation was delivered during stance (only cases with a stimulus causing early termination of the stance phase were included in the analysis), Sol onset was advanced (negative values) for both limbs, and when stimulation was delivered during swing, the onset was delayed (positives values) for both limbs. The delays were significantly higher when the stimulation was given during late swing (Bonferroni corrected pairwise comparisons of the estimated marginal means, P < 0.05), whereas stimulus intensity had no significant effect on the delays (linear mixed models of Sol onset delays for steps 1 and 2 of each leg, with phase of stimulation and stimulus intensity as fixed effects). The phase of stimulus delivery had no significant effect on the Sol onset delay of the step after stimulation (step 1) for the contralateral leg, although it was significant for the step 2 delay. The onset delays did not vary much between the 1st and 2nd step after stimulation for each leg, and the 95% CI of the estimated marginal means of the difference between the 1st and 2nd step Sol onset delays included zero for most phases of stimulation except for stimulation in late swing for the ipsilateral leg and early stance in the contralateral leg. The phase shifting present at the onset of the 2nd step even remained over many steps, as shown in Fig. 7, A and B.

The time delays between the predicted and actual Sol onsets were compared between both legs on the 1st and 2nd steps after stimulation (Fig. 10B; linear mixed model). The Sol onset was more delayed on the stimulated leg versus the contralateral leg for both steps, but this difference was significant only during stimulation in late swing (95% CI of the estimated marginal means of the delays between the predicted and actual Sol onsets for the ipsilateral and contralateral hindlimbs included zero except for stimulation in late swing). This led to a change in the delay between the onsets of the right and left Sol muscle that was maintained over several steps (Fig. 9A), indicating that the relative timing between the two legs could be disturbed over several steps with sural stimulation.

DISCUSSION

The present study demonstrates that, in spinal cats trained to locomote on a treadmill, electrical stimulation of the cutaneous sural nerve can enhance the ongoing flexion phase or terminate an ongoing extension phase and initiate flexion. The most profound effects were observed when stimulation was applied during stance to terminate the extension phase or applied during the early swing or midswing to enhance the length and height of the swing. For stimuli applied during the stance phase, stimulus intensity had a positive effect on the occurrence rate of the stance to swing switching.

Sural Nerve Stimulation Effects on Spinal Locomotion and Similarities and Differences With Other Preparations

Our results matched most of those reported by (Schomburg et al. 1998) for stimulation of the sural nerve in anemically decapitated, high-spinal-curarized cats, where stimulation of the sural nerve at low intensity (below 2 times threshold) enhanced the flexion phase, mainly through activation of a flexion reflex pattern, with a dependence of the phase of stimulus delivery. In that report (Schomburg et al. 1998), when low-intensity stimulation was delivered during the extension phase, the extensor activity was interrupted and a flexion phase was initiated. This was not observed in our preparation. For stimulation delivered during the flexion phase, a prolongation of that phase with a variable change in the duration of the following extension phase was induced (Schomburg et al. 1998). With high stimulus strengths (2–5 × T), they and others (Perreault et al. 1995) observed extensor activity that we did not obtain. Our results in spinal cats are also contrary to those seen in spontaneously walking premammilary decerebrated cats, where Duysens and colleagues (Duysens 1977b; Duysens and Stein 1978) obtained mostly extension promotion at low intensities (using short trains) and only observed flexion promotion at high intensities. Other authors who have studied the responses to single pulses applied to the sural nerve have reported no clear effect on the short latency excitatory postsynaptic potential of motoneurons in fictive locomoting cats (Schmidt et al. 1989). Using single stimulus pulses in intact cats, authors have found no effects (Pratt et al. 1991) on EMGs of extensor muscles for stimuli below two times threshold and inhibition at higher levels or short latency inhibition and longer latency excitation for low intensity pulses (Abraham et al. 1985). In contrast to some of the reports mentioned, we did not see any increase in extensor muscle activity at any level of stimulation or phase of gait delivery, and we obtained consistent enhancement of the swing phase with stimulation of the sural nerve. In our experiments, higher stimulus intensities facilitated flexors even for stimulation delivered during the stance phase. This illustrates some of the differences that may occur between fictive locomotion and actual spinal locomotion on a treadmill where all sensory inputs from the hindlimb interact with the cutaneous afferent stimulus. Our conduction velocity measurements indicate that nerves were not damaged by the cuffs, and differences in the results are unlikely to be attributable to differences in the population of afferent fibers recruited with chronic cuffs.

After stimulation amplitude, the second parameter of importance in facilitating swing initiation during the stance phase was the step cycle phase of stimulus delivery. Stimulation of the sural nerve was found to be more effective at initiating swing during late stance (extended hip) than early stance (flexed hip). Our results showed a dependence of the stimulus response on the phase of the step cycle, as demonstrated by others, but not a complete reversal in the reflex's response, as reported by others. In premammillary cats, stimulation of the sural nerve at the onset of the flexion phase also prolonged the flexor burst, but stimulation of the sural nerve at low intensity prolonged extension or switched the locomotor phase from flexion to extension when delivered at other phases of the step cycle (Duysens 1977b; Duysens and Pearson 1976; Duysens and Stein 1978). This reversal in the response to sural nerve stimulation is likely dependent on afferent signals from the hip flexors that are known to contribute to the onset of the swing phase in cats (Pearson 1993 and 2008; Whelan 1996). Although we did not observe a complete reversal in the reflex response, our results and those above indicate that the best time to stimulate the sural nerve to enhance the flexion phase seems to be early swing or mid-swing.

At higher stimulus intensity, we found that stimulation of the sural can terminate an ongoing stance phase and initiate the swing phase and that the early termination is easier to obtain in late stance and at higher stimulation levels (see Fig. 6). The termination affects the locomotor rhythm of both legs (as shown in Fig. 7); extensor onset is delayed for both legs if stimulation is delivered during late swing (Fig. 7A) or the rhythm is advanced (extensor onset is advanced) if stimulation is delivered during stance (Fig. 7B). In premammilary cats with the ipsilateral leg held isometrically and for high-intensity stimulus (75 × T), Duysens (1977a) found that for stimuli delivered during swing, the ipsilateral extensor onset was delayed and the contralateral extensor activity was advanced, whereas for stimuli delivery during late stance, the rhythm was largely undisturbed in either limbs (see Fig. 1 in Duysens 1977a). More surprising was the fact that in that preparation, stimulation at 75 × T did not terminate extensor activity when stimulation was delivered during the extensor phase. In our preparation, both legs were free to move and modifications in the rhythm of one limb generally affected the rhythm of the other, although not in strict proportion, as discussed below.

Modifications in the Timing of the Locomotor Cycle

Due to its effect on the timing of the ipsilateral leg stance to swing transition, sural stimulation showed that the onset of Sol activity in each leg could be controlled independently within some limits. Stimulation had larger effects on the ipsilateral hindlimb, whereas the contralateral rhythm and step cycle were typically very steady with only a slight delay due to the extended flexion on the ipsilateral leg when stimulation was delivered during the ipsilateral swing (see Fig. 7A). Nevertheless, stimulation of the sural nerve induced a shift in the rhythm, which was different between both legs and not corrected over the following steps after stimulation. We observed an overlap (up to 40%) of the ipsilateral and contralateral extensor activity during locomotion after stimulation. In this extreme case, the contralateral flexor activity started right after the ipsilateral flexor activity ended. Typically, the swing length of the first ipsilateral step after the stimulation was shorter than normal (Fig. 5C) because the contralateral leg initiates its flexion phase, thereby inhibiting the swing phase of the ipsilateral leg and terminating it early (Fig. 9A, arrow). Our data suggest a reciprocal inhibition between the right and left flexor-related neurons in the spinal cord, but no (or less) reciprocal inhibition between the right and left extensor-related neurons during locomotion, as previously suggested (Kjaerulff and Kiehn 1997). Split-belt treadmill locomotion experiments during which the belts operate at different speeds also showed that the extensor phases may readily overlap to varying degrees but that flexor phases do not overlap (Forssberg et al. 1980; Yang et al. 2005). This variability in the overlap between the left and right extensor phase might allow for corrections in the balance that would not be possible if the spinal circuits were too rigid. This flexibility might also explain why “deletions” (burst disappearances) occur more frequently for extensor bursts (Duysens 1977b; Faist et al. 2006). However, when high-intensity stimulation was applied in early stance, the ipsilateral flexion reflex made the cats fall, and a new step cycle was then initiated with a standard synchronization between the right and left flexors and extensors. In this case, i.e., a reinitialization of locomotion, the sural reflex seems to be acting on the ipsilateral and contralateral half-centers and “re-synchronizing” their operation.

Sural Nerve Stimulation as a Rehabilitation Tool After Spinal Cord Injury

Our results support earlier results in reduced preparations that demonstrate how a cutaneous nerve like the sural nerve can have powerful actions on the locomotor cycle. In this study, we demonstrated that cutaneous afferent nerve (sural nerve) stimulation during the swing phase significantly increases both step length and step height in chronic spinal cats trained to locomote on a treadmill. Sural nerve stimulation also makes it possible to initiate a swing phase when applied during the stance phase. Stimulation of the sural could thus be used to augment the swing phase of the gait cycle on a cycle to cycle basis. Functional electrical stimulation of afferent nerves, most notably the common peroneal nerve, has been used in clinical settings to initiate stepping (Barbeau et al. 2002; Field-Fote 2001; Graupe 2002). This further demonstrates that stimulation of only one cutaneous nerve is enough to promote flexion of all the leg joints (hip, knee, and ankle). Studies in intact individuals have shown that stimulating the sural nerve during swing increases flexor activity in humans (Duysens et al. 1992; Zehr et al. 1998). A similar enhancement of the flexion reflex evoked by sural nerve stimulation is seen in body weight-supported spinal cord-injured patients trained to walk on a treadmill [although this study (Knikou et al. 2009) concentrated strictly on EMG responses].

In conclusion, our findings suggest that electrical stimulation of the sural nerve may be used to improve the recovery of walking after spinal cord injury (in animals and humans). It should not be used as an alternative but as a supplement to body weight-supported treadmill training and/or neuroregenerative factors. In future experiments, it would be interesting to stimulate cats for many weeks during the training sessions and see if this prolonged stimulation regimen can “train” the spinal cord to generate longer and higher steps even after termination of the stimulation regimen, as suggested by Barbeau et al. (2002).

GRANTS

This work was supported by National Institute of Neurological Disorders and Stroke Grants NS-048844, NS-055976, and NS-007440 (to K. Ollivier-Lanvin) and by The Craig H. Neilsen Foundation.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).