Somatosensory control of balance during locomotion in decerebrated cat

Abstract

Postmammillary decerebrated cats can generate stepping on a moving treadmill belt when the brain stem or spinal cord is stimulated tonically and the hindquarters are supported both vertically and laterally. While adequate propulsion seems to be generated by the hindlimbs under these conditions, the ability to sustain equilibrium during locomotion has not been examined extensively. We found that tonic epidural spinal cord stimulation (5 Hz at L5) of decerebrated cats initiated and sustained unrestrained weight-bearing hindlimb stepping for extended periods. Detailed analyses of the relationships among hindlimb muscle EMG activity and trunk and limb kinematics and kinetics indicated that the motor circuitries in decerebrated cats actively maintain equilibrium during walking, similar to that observed in intact animals. Because of the suppression of vestibular, visual, and head-neck-trunk sensory input, balance-related adjustments relied entirely on the integration of somatosensory information arising from the moving hindquarters. In addition to dynamic balance control during unperturbed locomotion, sustained stepping could be reestablished rapidly after a collapse or stumble when the hindquarters switched from a restrained to an unrestrained condition. Deflecting the body by pulling the tail laterally induced adaptive modulations in the EMG activity, step cycle features, and left-right ground reaction forces that were sufficient to maintain lateral stability. Thus the brain stem-spinal cord circuitry of decerebrated cats in response to tonic spinal cord stimulation can control dynamic balance during locomotion using only somatosensory input.

postmammillary decerebrated animals are not able to stand or step by themselves because of isolation from the neuronal structures above the lesion (Bard and Macht 1958; Magnus 1924; Musienko et al. 2008). Locomotion with partial weight bearing, however, can be initiated and well controlled by tonic brain stem (Kazennikov et al. 1988; Shik et al. 1966) or spinal cord (Gerasimenko et al. 2005) electrical stimulation. Adequate propulsion generated by the hindlimbs under these experimental conditions has allowed a comprehensive study of the brain stem, spinal, and sensory mechanisms for locomotor control (Gerasimenko et al. 2008; Shik and Orlovsky 1976; Whelan 1996). Since the trunk of the animals was supported both vertically and laterally in these studies, the ability to sustain equilibrium during locomotion has not been investigated.

Successful stepping requires tight coordination between the trunk and the limbs to move the body forward (propulsion) while maintaining dynamic equilibrium (balance). Since the body mass during locomotion is displaced from step to step in normal subjects (Hof 2008; Misiaszek 2006a), maintaining balance when stepping demands dynamic control of the center of mass within a critical equilibrium range that is tightly integrated to locomotor activity. Results from studies using experimental models of standing ability (Deliagina et al. 2006; Honeycutt et al. 2009) suggest that the control of body posture and balance depends on a multisensory system that is organized hierarchically (Bernstein 1967; Roberts 1978). This closed-loop control system is driven by sensory feedback signals and compensates for deviations from the desired body orientation by producing corrective motor responses (Deliagina et al. 2006; Horak and Macpherson 1996; Massion 1994). During standing, the postural system restores equilibrium normally by moving the center of mass closer to the center of support enabled by ground reaction forces (GRFs) produced by the limbs against the supporting surface (Macpherson 1988a, 1988b; Moor et al. 1988). The situation is different during walking when the equilibrium has to be maintained dynamically during active propulsive movements (Karayannidou et al. 2009a). Although balance is a crucial element of successful locomotor behavior, the neural control of balance during walking has been neglected. Recent studies on intact cats have shown that locomotor features, such as step width and medial-lateral toe position and the EMG activity of extensor, abductor, and adductor muscles, depend on the postural status at each moment during each step with and without perturbations (Karayannidou et al. 2009a; Misiaszek 2006a). The neuronal mechanisms associated with this posture-locomotion integration are unidentified. The forebrain, brain stem, and spinal cord all appear to be involved in this process, but the specific roles at different central nervous system (CNS) levels have not been determined (Deliagina et al. 2006).

Tonic electrical stimulation of the mesencephalic locomotor region in postmammillary decerebrated animals initially generates a standing posture followed by weight-bearing stepping on a moving treadmill belt (Mori 1987; Mori et al. 1983; Shik et al. 1966). These movements are accompanied by an increased tone in the antigravity muscles and by an improved ability to perform postural corrections during perturbed standing before the initiation of locomotor activity (Musienko et al. 2008). This suggests that the neural control of posture and locomotion is highly integrated below the level of the decerebration and can be provided by the activation of brain stem-spinal cord circuitries. The degree to which motor and somatosensory mechanisms can control posture and locomotion, however, remains unclear.

Tonic epidural stimulation (ES) of the spinal cord can facilitate locomotor activity that is effectively controlled by somatosensory peripheral feedback from the moving limbs in decerebrated (Gerasimenko et al. 2005; Iwahara et al. 1991) and spinal (Courtine et al. 2009; Musienko et al. 2007) animals. It has been suggested that electrical stimulation of the dorsal surface of the spinal cord influences the spinal locomotor circuits though direct activation of the afferent fibers in the dorsal roots (Gerasimenko et al. 2006). It is not known, however, whether activation of the somatosensory input by ES can facilitate the ability to sustain posture and balance during stepping. The postural system is normally dependent on a number of interconnecting control loops that utilize visual, vestibular, and somatosensory inputs (Beloozerova et al. 2003; Deliagina et al. 2006; Macpherson et al. 1997). Under certain conditions, however, the system dissociates into the subsystems independently controlling the head and trunk (Barberini and Macpherson 1998; Deliagina et al. 2006). The head orientation is stabilized mainly on the basis of vestibular and visual information (Berthoz and Pozzo 1988; Boyle 2001), and somatosensory inputs are most important for trunk stabilization during standing (Beloozerova et al. 2003; Deliagina et al. 2000). The specific contributions of somatosensory inputs providing proprioceptive and exteroceptive information from the limbs in the dynamic regulation of equilibrium during stepping are not well understood.

In the present study we investigated the ability of the brain stem-spinal cord circuitry of postmammillary cats to acquire and sustain dynamic balance during stepping after isolation from the forebrain, vestibular, and other supraspinal descending systems. We found that decerebrated cats can efficiently control equilibrium during locomotion in response to tonic ES of the spinal cord, using only somatosensory input from the hindquarters. Preliminary results have been published in abstract form (Musienko et al. 2009a).

MATERIALS AND METHODS

The experiments were performed on 11 adult cats (2.5–3.0 kg body wt). All procedures were conducted according to European Community Council Directive (24 November 1986, 86/609/EEC) in accordance with a protocol approved by the Animal Care Committee of the Pavlov Institute of Physiology and followed the guidelines of the National Institutes of Health Guide for the Care and Use of Laboratory Animals.

Surgical procedures.

The cats were anesthetized deeply with a mixture of xylazine (1 mg/kg im) and ketamine (40 mg/kg im) and supplemented (30–50% the initial dose) as needed to keep the cat at a surgical level for the entire surgery. Systemic arterial pressure was monitored continuously and maintained above 80 mmHg via a carotid artery cannula. Rectal temperature was controlled and kept constant at 37 ± 0.5°C with heat irradiation. The level of anesthesia was monitored based on blood pressure, applying pressure to the paw to detect limb withdrawal, and checking the size and reactivity of the pupils. Tracheal intubation, ligation of the carotid arteries, and a precollicular postmammillary decerebration were performed as previously described and usually took 30–40 min (Gerasimenko et al. 2005, 2009; Musienko et al. 2007). The brain tissue rostral to the transection was removed. At this point, the animals are considered to have complete lack of sentience (Silverman et al. 2005). Anesthesia was discontinued after the surgical procedures, and the experiments were initiated 2–3 h thereafter.

A middorsal skin incision was made, the paravertebral muscles were reflected, and a partial laminectomy was performed between the L3 and L7 vertebral levels. The exposed dorsal surface of the lumbosacral spinal cord was covered with warm paraffin oil. An electrode for monopolar ES was made by removing a small length (∼2-mm notch) of Teflon coating to expose the stainless steel wire (AS632; Cooner Wire, Chatsworth, CA). The electrode was secured at the midline of the L5 spinal segment by suturing the wire to the dura mater above and below the electrode. An indifferent ground electrode was made by removing ∼1 cm of Teflon at the distal end of a similar wire that then was inserted into the paravertebral muscles. Bipolar intramuscular EMG electrodes were implanted into the midbelly of the following muscles bilaterally: vastus lateralis (VL, knee extensor), medial gastrocnemius (MG, ankle extensor), tibialis anterior (TA, ankle flexor), adductor femoris (Add, hip adductor and extensor), and gluteus medius (Glut, hip abductor and extensor) as described previously (Gerasimenko et al. 2009).

Experimental groups.

Two groups of cats were studied: 1) decerebrated cats (n = 7) that were used to investigate dynamic postural balance control during bipedal hindlimb locomotion after removal of all forebrain structures above the mesencephalon and 2) control cats (n = 4) that were tested during bipedal hindlimb locomotion on a treadmill to obtain a representative baseline of bipedal stepping characteristics of noninjured cats.

Experimental setup.

To monitor the dynamic signals associated with balance while stepping bipedally, we designed an experimental setup that allowed independent recording of kinematics, kinetics, and EMG data from the left and right sides of the body during locomotion on a treadmill. The treadmill was equipped with two separate belts each instrumented with a force sensor placed underneath the belt (Musienko et al. 2011a) (Fig. 1A) for quantifying contact GRFs (Di Fabio et al. 1982; Haour et al. 1976; Musienko et al. 2010). The lower trunk and hindlimbs, but not forelimbs, were positioned on the treadmill without any artificial support prior to ES and movement of the treadmill belt. The head and upper trunk (T5–T7) of the decerebrated cats were secured with a stabilizing apparatus.

Fig. 1.

Weight-bearing hindlimb stepping in decerebrated cats facilitated by epidural electrical stimulation (ES). A: cat secured in a stereotaxic frame. Stimulating electrode is sutured epidurally (at spinal level L5). Position sensors are attached to the pelvis, and displacement and force sensors are placed beneath each belt to record ground reaction forces (GRFs) from the right (R) and left (L) hindlimbs. B: R and L vastus lateralis (VL) and L tibialis anterior (TA) rectified EMGs and GRFs are shown during initiation of locomotor activity by ES (B1), during continuous stepping (B2), and after ES is turned off (B3). C: stick diagrams (60 ms between sticks) of joint movements after the initiation of ES during the transition from sitting to standing and for the initial step cycle [swing (red) and stance are indicated by brackets]; crest, iliac crest; mtp, metatarsophalangeal. D: sequence of R-L GRFs for consecutive steps showing a close relationship between left and right motor responses that gradually increase followed by a progressive decrease during a series of continuous steps. E: average correlation for the L and R total GRFs within the entire duration of the stepping trial for 10 experiments in decerebrated cats (n = 7 cats, P < 0.01). F: average correlation ratios for L vs. R GRFs plotted on a step-to-step basis during bipedal hindlimb stepping in decerebrated (Decer, n = 4 cats, 10–15 steps from each limb per cat) and intact (n = 4 cats, 10–15 steps from each limb per cat) animals. G: average amplitude of GRFs in decerebrated (mean ± SE, n = 4 cats, 10–15 steps from each limb per cat) and intact (n = 4 cats, 10–15 steps from each limb per cat) animals.

Intact cats were trained to step bipedally on the treadmill belt at 25 cm/s. Positive reinforcement was used to habituate the cats to the experimental conditions and to motivate locomotor behavior. When walking on the treadmill, the cats continuously licked food from a feeder positioned in front of the animal. All four intact cats maintained equilibrium during hindlimb stepping on the treadmill when the forelimbs were positioned on a stationary support. Mechanical sensors were used to select and measure lateral displacements of the pelvis relative to the treadmill as well as the extent of the perturbations during the postural tests in all cats (Karayannidou et al. 2009a; Musienko et al. 2008). Hindlimb movements and trunk displacements were monitored with three digital video cameras, i.e., placed to the left, right, and behind the cat. Reflective markers were placed on the iliac crest, femoral head, lateral condyle of the femur, lateral malleolus, and fifth metatarsal joint. Motion capture software was used to obtain the coordinates of the markers and to reconstruct the kinematics of the stepping movements. The signals from the EMG electrodes and the position and GRF sensors were differentially amplified (A-M Systems, model 1700, bandwidth of 30 Hz to 5 kHz), digitized at 2 kHz with a National Instrument A/D board, rectified, and integrated by computer programs developed with the LabVIEW package.

ES of spinal cord.

Monopolar electrical stimulation was delivered between the wire electrode secured to the dura mater at the L5 spinal segment and the indifferent ground wire electrode inserted into the paravertebral muscles. Electrical stimulation was applied with a stimulator (A-M Systems, model 2100 isolated pulse stimulator). Continuous stimulation at a frequency of 5 Hz, pulse duration of 0.5 ms, and intensity of 100–300 μA was used to induce and facilitate stepping movements in the hindlimbs on a motorized treadmill moving at a rate of 25 cm/s as described previously (Gerasimenko et al. 2009).

Data analyses.

For consistency, the first stable step with the left hindlimb was chosen as the starting point for each analysis. The GRF signals from the left and right hindlimbs were compared by cross-correlation. In addition, the GRF signals from both sides were transformed as the area under the curve (integral, total GRF) per step and compared. The lateral displacement output from the position sensor was measured to determine the maximal amplitude of the left and right pelvis lateral displacements. An imposed sinusoidal wave was used to calibrate the maximum amplitude in one direction by measuring half the distance between the peak and the following trough, while the maximum distance in the opposite direction was measured as half the distance between the trough and the following peak. The peak-to-peak (P-P) values for the lateral pelvis displacements were used for comparison and correlated with the width of the step, i.e., the distance between the stance paws in the frontal plane (Karayannidou et al. 2009a; Fig. 2A). In each case, we have used a representative example of a recording session from one cat to qualitatively illustrate the findings in the figures. The quantitative data are derived from all animals tested in each group.

Fig. 2.

Kinematics, kinetics, and EMG patterns of the hindquarters dynamic destabilization during unrestrained hindlimb stepping in decerebrated cats. A: movement viewed from the L side and from behind. The hindquarters deviated vertically and laterally during stepping initiated by ES, with each successive step compensating for the previous step. Peak-to-peak (P-P) lateral displacements of the pelvis and the step width (distance between the stance paws in the frontal plane) are shown. B: averaged correlation ratios for L vs. R displacements of the trunk during bipedal hindlimb stepping in decerebrated (n = 4 cats, 10–15 steps per cat) and intact (n = 4 cats, 10–15 steps per cat) animals. C: there is no correlation when all R and L displacements are randomized with respect to their order of occurrence. D: cumulative R and L pelvis displacements plotted in order of occurrence (red line) or randomized (Monte Carlo 500 times, gray line). Note that the actual cumulative sequence quickly falls outside the probability that would be expected if the R-L limb sequences were random. E: R and L medial gastrocnemius (MG), tibialis anterior (TA), adductor femoris (Add), and gluteus medius (Glut) EMG activity, GRFs, and lateral trunk displacements (Lat Displ) of decerebrated cats during weight-bearing unrestrained locomotion.

Statistical analyses.

All quantitative data are presented as means ± SE. Student's t-tests were used to determine statistically significant differences between means. Pearson correlation coefficients were used to determine the relationships between pairs of variables. Bootstrap confidence intervals were computed for the confidence bands on the cumulative pelvis displacement graph. Linear regression was calculated with total least squares, assuming errors in both x and y. Fourier transform was used to determine the periodicity of the signals. To estimate the delay between body displacements and GRF responses, we calculated the time shift between maximum lateral displacement of the pelvis and the first positive value of the GRF after this lateral motion. An autocorrelation function was used to quantify the degree of similarity from step to step for a given signal (Kim et al. 2007). Self-similarity was defined as the peak amplitude of the autocorrelation function. A nonparametric multivariate analysis of variance (MANOVA) was used to determine statistically significant differences when comparing the conditions (intact, decerebrate) with the correlation coefficients measured for GRF, displacements, and GRF vs. displacements. This was followed by Mann-Whitney U-tests to compare measures between conditions in a pairwise manner. Statistically significant differences were determined at P < 0.05.

RESULTS

Experiments were performed on decerebrated cats with the hindlimbs and lower trunk in a position to freely move on a treadmill (Fig. 1A). A precollicular postmammillary decerebration essentially precluded visual input, and the fixation of the head and spine at the thoracic level eliminated vestibular and proprioceptive head-neck-trunk reflexes as a source of ongoing dynamic control. Therefore, somatosensory input arising from the hindlimbs and lower trunk was the only source of modulation for the control of stepping and balance in our experimental model.

Weight-bearing stepping facilitated by ES.

Previous studies have reported that decerebrated cats are able to generate stepping-like movements when facilitated by ES. In these studies the animal's hindquarters were secured in a frame (Gerasimenko et al. 2005) or supported in a hammock (Iwahara et al. 1991). In contrast, in the present study no support was provided for the hindquarters. We first tested whether ES enhances not only locomotor activity but also postural tone and the ability to support body weight and maintain posture when stepping.

After postmammillary decerebration the cats showed reduced muscle tone and did not step spontaneously. The hindlimbs were unable to support the hindquarters, and the cats remained passively seated with no weight bearing (Fig. 1C). Continuous ES (5 Hz, 100–300 μA, 0.5-ms pulse duration) applied at spinal segment L5 significantly increased hindlimb muscle tone (P < 0.01) in all tested animals (n = 7 cats) such that GRFs sufficient to attain a standing posture and to step in place were produced (Fig. 1, B1 and C; Supplemental Video S1).¹ When treadmill belt motion (25 cm/s) was initiated in the presence of ES, the cats displayed weight-bearing hindlimb stepping (Fig. 1, B2 and C; Supplemental Video S1) for as long as 200 s with left-right alternation of flexor and extensor EMG activity and GRFs. Well-coordinated weight-bearing locomotor activity was associated with a wide range of GRFs across and even within an experiment (Fig. 1D). The successive GRFs generated by the left and right hindlimbs, however, were highly correlated (r = 0.98, n = 7 cats) within the entire duration of the stepping trial (P < 0.01; Fig. 1E) and on a step-to-step basis (Fig. 1D). The left-right correlations of GRFs in decerebrate (r = 0.94 ± 0.02, n = 4) and intact (r = 0.90 ± 0.01, n = 4) cats were similar (Fig. 1F). The average GRFs generated during stepping also were similar in decerebrated and intact animals (Fig. 1G). Weight-bearing stepping was sustained for up to 10–15 s after the cessation of ES, gradually becoming less robust (Fig. 1B3; Supplemental Video S1).

Dynamic balance control during locomotion.

Weight-bearing unrestrained locomotion requires lateral stability. Although the decerebrated cats were able to step with full weight bearing of the unrestrained hindquarters during ES, it was not clear whether this was due to some passive mechanical consequence of the experimental paradigm or to the decerebrated cats utilizing neural control mechanisms driven by somatosensory input to actively maintain balance as observed in intact animals (Karayannidou et al. 2009a). To test this, we compared the kinematics of the pelvis and the hindlimb stepping patterns of intact and decerebrated cats. The hindquarters were displaced in the sagittal plane and laterally and vertically in the frontal plane in both groups (Fig. 2A). There was a strong correlation between the left-right maximal lateral displacements of the pelvis in decerebrated (P < 0.02; n = 4 cats, r = 0.69 ± 0.08) and intact (P < 0.02; n = 4 cats, r = 0.67 ± 0.06) cats (Fig. 2B). To ascertain whether the relationships between the left and right body deviations during consecutive steps reflected active control mechanisms, we compared the cumulative left-right step-to-step lateral displacements within a series of consecutive steps (Fig. 2D) to a cumulative sequence in which we randomized the order of the steps (Fig. 2C). With bootstrapping as a means of testing for statistical significance, the cumulative left-right displacements for consecutive steps fell outside the range of displacements that would be expected if the order of the cumulative displacements were randomized (P < 0.001; Fig. 2, C and D).

The EMG patterns during unrestrained weight-bearing hindlimb stepping were similar to those observed during unrestrained locomotion of intact cats (Karayannidou et al. 2009a; Rasmussen et al. 1978), i.e., alternating activity in flexors (TA) and extensors (MG) and abductor (Glut) and adductor (Add) activity predominantly during the extensor phase of the step cycle (Fig. 2E). As reported previously in mesencephalic cats (Gambaryan et al. 1971), the Glut showed either a flexor pattern or separate bursts during stance and swing. This activity pattern may be due to biomechanical reasons such as fixation of the rostral aspect of the body in the frame, mechanical dissociation from the forequarters, and/or the necessity for lateral rotation and abduction of the femur during swing to control the lateral displacements of the trunk. We found that, similar to that observed in intact animals, the toe position in decerebrated cats continuously moved during the step sequence in both anterior-posterior and medial-lateral directions (Fig. 2A), resulting in variable step widths (9.8 ± 2.6 cm, n = 4 cats; Fig. 3A). Step width also varied among cats but was correlated with P-P lateral oscillations of the pelvis (P < 0.05; n = 4 cats, r = 0.50 ± 0.06; Fig. 3, B1 and C), similar to that observed in intact cats (P < 0.05; n = 4 cats, r = 0.49 ± 0.07; Fig. 3, B2 and C). By precise positioning of the toe and adjusting the step width, the hindquarters center of mass predominantly remained between the left and right toe positions and did not displace beyond the base of support, resulting in maintenance of lateral stability (Fig. 2A).

Fig. 3.

Balance-related responses in decerebrated and intact cats. A: averaged values of P-P lateral displacements (Lat Displ) of the pelvis and step width in decerebrated (n = 4 cats, 9–15 steps per cat) and intact (n = 4 cats, 9–15 steps per cat) cats. B: correlation ratios for P-P lateral displacements vs. step width plotted on a step-to-step basis in decerebrated (B1) and intact (B2) cats. Graphs were created based on data from 1 cat (n = 48 steps for decerebrated, n = 35 steps for intact). C: comparison of the average correlation ratios for P-P lateral displacements vs. step width in decerebrated (n = 4 cats, 10–15 steps per cat) and intact (n = 4 cats, 10–15 steps per cat) cats. D: R-L correlations between GRFs and pelvis displacements in decerebrated (D1) and intact (D2) cats. Graphs were created based on data from 1 cat from each group (n = 10–15 steps). E: comparison of the average correlation ratios for lateral displacements vs. ipsilateral GRFs during bipedal hindlimb stepping in decerebrated (n = 4 cats, 10–15 steps from each limb per cat) and intact (n = 4 cats, 10–15 steps from each limb per cat) cats. F: delay (d) between the peak lateral displacement (black dashed lines) and the following GRF response. The vertical red dashed line in the gray shaded area marks the initiation of the increase in the GRF. G–I: average delay between lateral displacements (Displ) and the ipsilateral GRF (G), self-similarity coefficients in normalized units for lateral displacements and GRFs (H), and standard deviation of lateral displacements and step width (I) in decerebrated (n = 4 cats, 10–15 steps per cat) and intact (n = 4 cats, 10–15 steps per cat) animals. Significant difference between conditions: *P < 0.05, **P < 0.01, ***P < 0.001.

We then determined the temporal relationships between continuous lateral pelvis displacements and left or right GRFs during weight-bearing stepping of decerebrated cats facilitated by ES (Fig. 3, D and E). The recorded contacting GRFs, mainly vertical, are cumulative values reflecting the activity of all muscles during the stance phase. In this regard we used the GRFs to characterize the corrective kinetics antigravity motor responses of the entire left or right limb following the corresponding lateral destabilization of balance during gait. As observed in intact cats (P < 0.05; n = 4 cats, r = 0.93 ± 0.02; Fig. 3, D2 and E), there was a tight relationship between the continuous lateral pelvis displacements and the associated GRFs for both hindlimbs in decerebrated cats (P < 0.05; n = 4 cats, r = 0.74 ± 0.1; Fig. 3, D1 and E). Although less than in the intact condition (Fig. 3E), the magnitude of the right and left GRFs in decerebrated cats closely matched the ongoing variations of and increases in the lateral displacements of the pelvis. Therefore, there was a highly coordinated and continuous modulation of the kinematics and kinetics of the hindlimbs and trunk during stepping in decerebrated cats similar to intact animals (Fig. 3, A–E). We also detected a delay between the lateral displacements of the pelvis and the compensatory increase in GRF responses that preserve balance (d in Fig. 3F). This delay was significantly longer in decerebrated compared with intact cats (Fig. 3G) as well as higher amplitudes of cyclic lateral oscillations (Fig. 3A). Moreover, self-similarity analysis and the standard deviation of the recorded values over all tested animals showed a substantially larger variability in the lateral displacements from step to step (P < 0.05; Fig. 3, H and I) and step width (Fig. 3I) in decerebrated compared with intact cats.

Adaptive postural responses to perturbations during gait.

As a next step, we evaluated whether the brain stem-spinal cord motor systems of decerebrated cats are able to control balance not only during unperturbed stepping but also during a disturbing influence artificially applied during gait. We introduced a prolonged lateral force to the trunk during continuous stepping facilitated by ES (n = 4 cats; Fig. 4). Such a perturbation deviated the hindquarters to the side (7–10 cm, to the right in Fig. 4A), repeatable from trial to trial, and constrained the trunk movements to the opposite side (to the left on Fig. 4A) for as long as 15–20 s. We found that decerebrated cats could still continuously step in this unstable condition and efficiently maintain their lateral stability because of specific and asymmetric adaptations of the motor patterns in the left and right hindlimbs. Sustained modulation of GRFs (Fig. 4, A–D), EMG activity, and step cycle duration (Fig. 4, A, B, and E–J) in both legs was observed and lasted for the entire duration of the perturbation. The GRFs in the limb contralateral to the perturbation were reduced significantly in the four tested cats (left GRF in Fig. 4D; P < 0.01). The GRFs on the ipsilateral side increased [GRF (R) in Fig. 4, A and B] or did not change (no significant differences over all tested cats; Fig. 4C). Both extensor and flexor activity as well as changes in stance and swing phase durations of both limbs during the perturbation were observed (Fig. 4, A and B). The stance phase was slightly, but significantly, longer on the side ipsilateral to the perturbation (Fig. 4, B and E; P < 0.01). On the side contralateral to the perturbation, the swing phase was lengthened while the stance phase was shortened (Fig. 4, B and F; P < 0.001). In all tested cats we observed bilateral modulation of adductor and abductor activity that could have contributed to the control of medial-lateral displacement of the hindquarters to maintain balance. More specifically, abductor activity ipsilateral to the perturbation increased [Glut (R) in Fig. 4, A, B, and G; P < 0.01], whereas the contralateral abductor muscle activity decreased [Glut (L) in Fig. 4, A, B, and I; P < 0.001]. In addition, adductor activity ipsilateral to the perturbation was decreased [Add (R) in Fig. 4, A, B, and H; P < 0.001]. In the four cats tested, the bilateral adjustments in the kinematics, EMG, and kinetics were sufficient to maintain balance of the hindquarters despite the pronounced disrupting lateral force. When the perturbation was removed, after 2 to 3 steps the cats exhibited locomotor patterns with EMG activity and GRFs that did not differ significantly from those recorded during the stepping sequence preceding the perturbation (Fig. 4, A–J).

Fig. 4.

Adaptive postural responses to prolonged lateral perturbation during hindlimb stepping in decerebrated cats. A: EMG activity, GRFs, and lateral body displacements before, during, and after the pelvis is being perturbed laterally to the right (shaded areas). B: average (1 SE, shaded area) rectified EMG signals from the L and R Glut, Add, MG, and TA, L and R GRFs, and lateral pelvis displacements during the stance (St) and swing (Sw) phases of gait cycles before, during, and after perturbation for the raw data in A; 9–10 gait cycles were averaged for each period. C–F: average GRF magnitudes generated by L and R hindlimbs and the duration of the swing-stance phases of L and R hindlimb step cycles before, during, and after lateral perturbation for all tested animals (n = 4 cats, 10–15 steps from each limb per cat). G–J: bilateral modulation of EMG amplitudes [R Glut (G) and Add (H), L Glut (I) and Add (J)] after perturbation averaged from all experiments (n = 4, 10–15 steps from each limb per cat). Significant difference between conditions: **P < 0.01, ***P < 0.001.

Finally, we tested the ability of decerebrated cats to reestablish an equilibrium state sufficient to sustain well-coordinated stepping during ES after a sudden collapse (n = 4 cats; Fig. 5). The cat pelvis was initially restrained in a frame by an additional clamp. After recording of the stepping pattern in the restrained condition (10–20 steps) the pelvis clamp was suddenly released and the hindquarters collapsed (3–5 cm down; Fig. 5, A and C).

Fig. 5.

Adaptive postural responses after collapse during hindlimb stepping in decerebrated cats. A: EMG activity, GRFs, and lateral pelvis displacements during stepping with restrained pelvis before collapse (Restr stepping before) from the clamps (shaded area) and unrestrained stepping soon after (Unrestr stepping soon after). B: average (1 SE, shaded area) rectified EMG signals from L and R Glut, Add, MG, and TA, L and R GRFs, and lateral pelvis displacements during the stance (St) and swing (Sw) phases of gait cycles during restrained stepping before collapse (Restr), unrestrained stepping soon after (After collapse), and stabilized unrestrained stepping (Unrestr) for the raw data in A from 1 cat; 9–10 gait cycles were averaged for each period. C: stick diagrams (50 ms between sticks, red lines represent the swing phase) of the joint movements for a sequence of stepping under the restrained condition and unrestrained stepping soon after the release of the pelvis clamp resulting in a collapse of the hindquarters and a stumble (blue). D: average duration of EMG bursts in the ipsilateral flexor (TA) and extensor (MG) muscles during the stance and swing phases of gait cycle before the collapse (Restrain), unrestrained stepping soon after (After collapse), and stable unrestrained stepping (Unrestrain); 7–10 gait cycles from 1 cat were averaged for each period. E and F: average GRF magnitudes generated by both the L and R hindlimbs and the duration of swing-stance phases of the step cycle in restrained and stable unrestrained stepping for 4 tested animals (n = 4 cats, 15–20 steps per cat). G: differences in standard deviation of the step width value during restrained compared with unrestrained locomotion (n = 15 steps per condition in 4 cats). Significant difference between conditions: **P < 0.01, ***P < 0.001.

The restrained stepping was characterized by clear left-right alternation, reciprocity in flexor and extensor EMG activity (Fig. 5, A–C) similar to that observed in intact cats (Rasmussen et al. 1978), accurate plantar foot placement with good interlimb coordination, and high magnitudes of contact forces during the stance phase (Fig. 5E). The abductors and adductors were normally active at the same time as the extensors (Restr in Fig. 5, A and B). The abrupt absence of pelvis support initially resulted in a collapse of the hindquarters and a stumble (Fig. 5, A and C), followed by changes in the EMG, kinetics, and kinematics features that allowed stabilization of the stepping pattern and dynamic equilibrium in the unrestrained condition. After a few dragging movements and hobbling of the cats on one or two limbs, plantar placement weight-bearing stepping was recovered (Fig. 5C). The motor pattern, however, was still variable, and it was several steps after the collapse that the stepping became stable. During this stabilizing period there was a decrease in the GRFs and EMG amplitudes in extensors, abductors, and adductors and coactivation of flexors and extensors (After collapse in Fig. 5, A, B, and D). In addition, the Glut was active predominantly during the swing phase (Fig. 5, A and B). During these adjustments the toe touched down in a more lateral position, thus keeping the body mass within the base of support. There was more coactivation of the flexor and extensor muscles (After collapse in Fig. 5D) that increased the stiffness of the limbs. These temporary adjustments began soon after release of the pelvis clamp and continued until the locomotor pattern was optimally modulated for stable stepping in the unrestrained condition (Figs. 2E and 5B). Most of the measures of the quality of stable stepping were similar in the unrestrained (Fig. 2E, Unrestr in Fig. 5B) and restrained conditions, although the peak GRFs were lower, the stance phase longer, and the swing phase shorter in the unrestrained than in the restrained condition (Fig. 5, B, E, and F). In addition, foot placement and step width were more variable in the unrestrained than the restrained condition (Fig. 5G) and highly dependent on lateral body displacements (Fig. 3, B and C). The active balance control under the unrestrained condition and necessity to react properly to dynamic trunk deviations were reflected in the fine features of the locomotor pattern, i.e., reciprocity between different muscles in a step cycle, kinetics and kinematics that were optimal for maintaining dynamic equilibrium during stepping.

DISCUSSION

The present results demonstrate that, in the presence of tonic spinal cord stimulation, postmammillary decerebrated cats can step with weight bearing and balance and adapt their stability to external perturbations using only somatosensory input.

Somatosensory control of dynamic balance during locomotion.

Our data indicate that the level of control of balance and equilibrium of the hindquarters that is necessary to sustain locomotion can be derived solely from hindlimb somatosensory input and to some degree from the lower trunk. The level of decerebration and the fixation of the head and spine at the thoracic level precluded visual input, thus functionally eliminating vestibular and proprioceptive head-neck-trunk reflexes as a source of dynamic control. Using left-right pelvis displacements as a measure of dynamic lateral stabilization (Karayannidou et al. 2009a), we found that decerebrated cats can generate the postural corrections needed to sustain balance. In addition, when a prolonged lateral force was imposed, strategies to maintain balance included a redistribution of the abductor-adductor activity and GRFs between the left and right hindlimbs and adjustments in the duration of the extensor and flexor phases of the step cycle. Furthermore, recovery from a sudden collapse resulted in coactivation between flexor and extensor muscles and changes in the pattern and timing of abductor activity. Comparison of restrained versus unrestrained locomotion further demonstrated a crucial role of the somatosensory information arising from the hindquarters in specific modification of the fine features of the locomotor pattern necessary to maintain dynamic equilibrium during ongoing destabilization of the body mass in the unrestrained condition (Fig. 5).

To maintain balance during locomotion there must be tightly linked mechanisms or interactions of forces generated by multiple combinations of muscles, each generating unique force vectors. Presumably this is accomplished by online regulation of the levels of recruitment of muscles generally considered in the categories of flexors-extensors and abductors-adductors. Any change in the recruitment of the combination of motor pools and the level of recruitment in these motor pools will be reflected in the net force vectors manifested by the whole hindquarters. In the present data, we cannot precisely interpret the magnitude of these vectors with respect to specific muscles, but it is clear that all of these muscles are modulated in the unperturbed and perturbed locomotor conditions. Interestingly, however, in unperturbed locomotion there is rather tight coactivation of the Glut, Add, and MG within each leg, although anatomically these muscles would be considered to have different effects on the level of forward and backward versus lateral movements of the limb.

While vestibular and other supraspinal systems normally are important sources of control in maintaining balance during stepping and standing (Beloozerova et al. 2005; Matsuyama and Drew 2000; Zelenin et al. 2010), our results show that somatosensory input is sufficient to accomplish these complex motor tasks with an astonishing degree of precision. Although the decerebrated cats successfully sustained equilibrium during locomotion for extended periods using only cutaneous (Bolton and Misiaszek 2009) and proprioceptive somatosensory information from the hindquarters, the postural adjustments were slower and comparatively less consistent than in intact cats. Therefore, these results do not exclude the forebrain motor centers (Karayannidou et al. 2009b), as well as vestibular and other inputs, contributing to the fine-tuning of lower-level neuronal networks in maintaining dynamic balance in intact cats.

Brain stem-spinal cord circuits can control balance during gait.

Similar to balance control in intact animals (Misiaszek 2006a, 2006b; Karayannidou et al. 2009a), postural adjustments in decerebrated cats can be attained in a variety of ways including a redistribution of activity in adductor-abductor and flexor-extensor muscles resulting in precise positioning of the paw and modulation of the GRFs. Our data confirm that this selection can be efficiently carried out within the somatosensory circuitries in the spinal cord and the brain stem. These results are consistent with recent modeling studies demonstrating that sensorimotor control of balance during multijoint behavior in mammals can be achieved with flexible neural feedback strategies involving the brain stem and spinal cord (Lockhart and Ting 2007; Misiaszek 2006b).

The present data do not define the specific location of the circuits necessary to generate these continually adapting responses to changes in physical conditions. For example, although somatosensory inputs are the source of the ongoing ensemble of adaptive signals from multiple receptors and locations from the hindquarters, whether all of these inputs can be processed online totally by the spinal circuitry versus the spinal cord and brain stem circuitry remains to be determined. Indeed, various sensorimotor loops, including both brain stem and cerebellar motor systems, contribute to the regulation of interactions between posture and locomotion (Armstrong 1986; Arshavsky et al. 1983; Whelan 1996). Based on a number of experiments using chronic and acute spinal cord injured animals (Barbeau and Rossignol 1987; Lovely et al. 1990; Musienko et al. 2010, 2011a), to a certain extent the postural control during standing and locomotion might be executed by the lumbosacral spinal circuits alone. The relative roles of spinal and supraspinal mechanisms in the control of balance during stepping need to be investigated further to begin to understand the functional circuits that can execute this dynamic postural control.

Interaction of tonic ES of spinal cord and somatosensory control of dynamic balance.

Our findings confirm that when the spinal cord is tonically stimulated with epidurally placed electrodes the acute decerebrated cat can process ongoing somatosensory input to dynamically update the motor commands required to sustain weight-bearing locomotion. The locomotor activity of postmammillary cats also can be evoked by stimulation of the mesencephalic locomotor region (Shik and Orlovsky 1976), which projects to reticulospinal neurons in the pons and medulla, with the latter projecting throughout the spinal cord to activate the spinal neuronal networks (Jordan et al. 2008). In the case of ES, two possible mechanisms can be considered: direct activation of spinal postural-locomotor networks and indirect activation of these networks via spinal cord-brain stem loops (Musienko et al. 2007). Previous experiments show that after a partial or complete spinal cord transection ES can facilitate locomotion (Musienko et al. 2009b) or postural reflexes (Musienko et al. 2010) by direct action on intraspinal motor circuits. In the intact or partially injured spinal cord, however, the brain stem systems may play a role in processing the somatosensory input to correct and stabilize the motor patterns.

How can tonic spinal cord stimulation enhance the spinal motor circuitries? Although the mechanisms underlying the facilitation of motor activity with ES are not yet fully understood (Gerasimenko et al. 2008), neurophysiological recordings (Gaunt et al. 2006; Gerasimenko et al. 2006) and computer simulations (Rattay et al. 2000) suggest that electrical stimulation engages spinal circuits in part by recruiting dorsal root fibers carrying somatosensory signals from the limbs at their entry into the spinal cord as well as along the longitudinal portions of the fiber trajectories (Gerasimenko et al. 2008; Musienko et al. 2011b). Sensory input, notably via the flexor afferent reflex system, has widespread access to spinal circuits and has been shown to initiate stepping in mammals (Jankowska et al. 1967a, 1967b; Kostyuk 1983; Sherrington 1910). Locomotion and posture enabled by ES may rely on similar mechanisms. Although experimental evidence is incomplete, ES appears to play a crucial role in augmenting the excitability of the spinal circuitries that underlie the control of motor behaviors (Edgerton et al. 2008). From this perspective, the present observations have significant implications for the potential of humans with a severe CNS injury to regain a significant level of functional standing and walking, particularly if clinically safe technologies can be developed to raise the general excitability of the spinal cord (Fong et al. 2009; Harkema et al. 2011; Musienko et al. 2009c).

Conclusion.

In summary, the present data demonstrate that when tonic ES of the spinal cord is applied in a postmammillary decerebrated cat deprived of vestibular and other supraspinal sensory input, weight-bearing stepping with active balance control can be performed. These results imply that the sources for regulation of equilibrium during walking can be attributed to the ensembles of sensory input from the hindquarters to the spinal cord-brain stem neuronal circuits. The strong facilitating effect of spinal cord activation by ES on maintaining equilibrium during locomotion further demonstrates an important role of the spinal circuits in postural control during stepping.

The present data, however, do not determine whether the proprioceptive and exteroceptive information from the hindlimbs is processed by the spinal cord independent of the brain stem since the brain stem could provide a necessary source of tonic excitation and important modulatory control. The relative contribution of spinal and supraspinal mechanisms, as well as the specific role of different afferent systems in the regulation of balance during stepping, needs further investigation. The novel model of minimally restrained locomotion in the decerebrated cat used in the present study, however, can serve as a useful tool in understanding the multiple components that contribute to the control of posture and locomotion.

GRANTS

This work was supported by grants from the Russian Foundation for Basic Research Grants (08-04-00688, 10-04-01172, 11-04-01669, 11-04-12074-OFI-M-2011), the International Paraplegic Foundation (P106), and the National Center of Competence in Research “Neural Plasticity and Repair” of the Swiss National Science Foundation and NIH Grants NIBIB EB-007615 and NS R01062009.

DISCLOSURES

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

AUTHOR CONTRIBUTIONS

Author contributions: P.E.M., V.R.E., and Y.G. conception and design of research; P.E.M., A.S., and Y.G. performed experiments; P.E.M., G.C., J.E.T., V.K., A.S., A.G., and Y.G. analyzed data; P.E.M., A.G., R.R.R., V.R.E., and Y.G. interpreted results of experiments; P.E.M., G.C., and Y.G. prepared figures; P.E.M. and Y.G. drafted manuscript; P.E.M., G.C., R.R.R., V.R.E., and Y.G. edited and revised manuscript; P.E.M., R.R.R., V.R.E., and Y.G. approved final version of manuscript.