To delineate the role of cutaneous feedback from the paws in the regulation of balance during walking, we compared the corrective responses of cats to lateral support surface translation before and after cutaneous denervation of the hindpaws. In addition, we compared characteristics of undisturbed walking before and after denervation. Electromyographic and kinematic data were collected from three cats trained to walk across a walkway, the central portion of which could be unexpectedly translated laterally in either direction. Following denervation, all of the cats changed their step width, lowered their pelvis, and spent more time with the hindlegs in double-support when walking across the walkway. When displaced by lateral support surface translations, the denervated cats made larger lateral steps and required more than a single step to regain balance. However, none of the cats fell following the denervation. The appearance and latency of the responses evoked in the hindleg muscles by the perturbations were unaffected by the denervation. However, the amplitude of these responses was affected by the loss of cutaneous inputs. Responses evoked at paw contact were significantly reduced in most muscles in the absence of cutaneous input, whereas responses evoked at end of stance revealed significant increases in gluteus medius activity with little influence on the activity of other muscles. Therefore the loss of cutaneous inputs leads to instability during gait. Although cutaneous feedback from the hindpaws is not essential for triggering corrective responses to support surface disturbances, it appears that cutaneous inputs are important for scaling the responses initiated by other cues.
Unexpected disturbances to balance are countered with rapid, automatic postural reactions (APRs) aimed at maintaining stability (Horak and Macpherson 1996). For any motor response to successfully counter a postural challenge, the disturbance must first be quickly and accurately detected. Although a variety of sensory sources are available to detect postural disturbances, including vision, vestibular sensation, and somatosensation, several studies have indicated that somatosensory cues are highly important for triggering the most rapid reactive balance responses in both human (Inglis et al. 1994) and animal models (Inglis and Macpherson 1995; Mori et al. 1970; Stapley et al. 2002). For example, in cats following the loss of large-diameter somatosensory afferents (via pyridoxine intoxication), responses to a platform perturbation are significantly delayed, often leading to falls (Stapley et al. 2002). In contrast, after surgical lesion to the vestibular system, cats showed unaltered initial activation of APRs when standing balance was disturbed (Inglis and Macpherson 1995). In that study, deficits were only found in later portions of the corrective response, most notably as an overshoot when regaining upright posture. This preservation of the initially triggered responses, suggests that vestibular inputs are not critical for evoking the response. Similarly, processing delays associated with visual feedback pose limits on visual contributions to rapid corrective responses. Reactive balance can be modulated by visual feedback; however, much of this influence is limited to visuospatial information acquired prior to perturbation (Maki and McIlroy 2007). Such feedback offers contingency plans for corrective reactions when needed, for example, where to step. However, vision is unable to provide on-line corrective adjustments once the perturbation occurs until much later in the response (Maki and McIlroy 2007). Overall it appears that visual and vestibular senses have the ability to modify somatosensory-evoked responses as they emerge; however, without this somatosensory influence, corrective reactions are significantly delayed.
Somatosensation is comprised of several different submodalities, mostly responsive to various forms of mechanical pressure or deformation. Stapley et al. (2002) provided evidence for an important role of large diameter somatosensory afferents in triggering APRs. This included signals for muscle stretch, muscle force, and large-diameter cutaneous afferents. To resolve which sensory signal could trigger rapid, direction-specific corrective reactions, Ting and Macpherson (2004) compared APRs during platform rotations to those occurring with platform translations. Their results indicated that the ratio of shear to load ground-reaction force may provide an important cue to specify the direction-specific response. Further, these authors proposed that cutaneous sensors from the paw pads were ideally situated to offer this important sensory signal.
Although it is clear that somatosensory inputs are important in activating APRs when standing balance is disturbed, it is less clear what role somatosensory information has in reactive balance during dynamic tasks, such as walking. This study will focus specifically on the contribution of cutaneous inputs from the hindpaws. In undisturbed walking, cats without cutaneous sensation in their hindpaws display a near-normal walking pattern, with only subtle deficits (Bouyer and Rossignol 2003a; Sherrington 1910). Therefore the role of cutaneous input from the paws in regulating undisturbed locomotion is minimal in the otherwise intact cat. On the contrary, when these same cats are placed in more challenging walking environments, such as crossing a horizontal ladder or negotiating slopes, greater deficits are revealed (Bouyer and Rossignol 2003a). It is also possible that information from cutaneous sensors under the paw pads may provide a particularly critical signal when posture is unexpectedly challenged while walking. These sensors are ideally situated beneath the base of support to offer a direct appraisal of sudden shifts in pressure distribution. Given the proposed role of cutaneous sensors from the paw pads in triggering directionally relevant APRs while standing (Ting and Macpherson 2004), presumably cutaneous inputs could also contribute to reactive balance while walking. It is well known that unexpected cutaneous feedback from the paws influences stepping behavior in cats. For instance, touching the dorsum of the paw during swing phase generates a stumbling-corrective reaction (Forssberg 1979) and electrical stimulation of most cutaneous nerves leads to distinct reflexes in muscles distributed throughout the legs (Abraham and Loeb 1985; Drew and Rossignol 1987; Duysens and Loeb 1980; Duysens and Stein 1978; Frigon and Rossignol 2008b; Pratt et al. 1991). Thus cutaneous feedback may play a more prominent role in detecting and correcting for unexpected disturbances during walking than in maintaining or regulating undisturbed gait.
It is currently unknown to what extent cutaneous information contributes to the corrective responses in walking cats following a disturbance to balance (Karayannidou et al. 2009; Misiaszek 2006a). Given the role of cutaneous feedback in generating APRs during standing and when walking in more challenging environments, we predict that cutaneous feedback contributes importantly to reactive balance control during walking. We hypothesized that hindlimb cutaneous denervation would lead to deficits in reactive balance responses of walking cats, notably with delayed and/or reduced electromyographic activity and larger deviations from the original walking path following perturbation. Portions of these results have been reported in abstract form (Bolton and Misiaszek 2007).
Animals, training, and testing protocol
The experimental procedures were approved by the University of Alberta Health Sciences and Animal Welfare Policy Committee in accordance with the Canadian Council on Animal Care. The animals were housed in a group facility with environmental enrichment. Daily care and maintenance was provided by the staff of the Health Sciences and Laboratory Animal Services of the University of Alberta. After surgical procedures (see following text), the animals were housed in individual cages for 48 h and monitored regularly. Three adult female cats were included in this study. All cats were trained to walk across a custom-built walkway (8 ft long, 2 ft wide) with a training time of ∼4–6 wk/cat. Training and testing protocols were similar to those previously outlined by Misiaszek (2006a) but will be briefly repeated here. Training consisted of manually placing a cat at one end of the walkway and allowing the cat to walk to the other end of the walkway to receive a food reward. To constrain cats from jumping off the walkway, clear Plexiglas walls were installed along both sides. The walkway had a rubber surface to provide traction. Training ended once cats performed consistent walking for ≥20 min/session.
After successful training, all cats were implanted with chronic indwelling electromyographic (EMG) electrodes and then allowed 2 days to recover. Subsequently, baseline testing was initiated which involved the collection of EMG and kinematic data during walking sessions. During testing, cats were periodically exposed to medial-lateral perturbations of the walking surface (∼1 of every 10–15 passes). This was accomplished using horizontal translation of the central portion of the walkway. This movable portion of the walkway was mounted on a ball-bearing slide and attached to a linear motor (model CF04B, Baldor Motors) and could slide perpendicular to the cat's direction of travel. When inactive, the platform was rigid. The platform motion was under computer control and triggered by one of the EMG signals. Translations were timed to approximate either stance onset or late stance in the right hindlimb. Stance onset was approximated using an 80-ms delay from the onset of the right medial gastrocnemius (MG) burst (Gorassini et al. 1994), and late stance using a 100-ms delay off the left MG burst (only 1 of these triggers were used on any given testing day). Platform translations consisted of a 5-cm displacement with a peak acceleration of 0.6 G. The direction of platform translation (right or left with respect to the cat) was randomized. A single-axis accelerometer (model 1210 Analog Accelerometer, Silicon Designs) attached to the platform monitored movement onset.
Once an adequate number of trials were collected, requiring several weeks of daily recording sessions, a second surgery was performed to transect the cutaneous nerves supplying both hindpaws (procedure to follow). Following a 2-day recovery period, testing resumed. Thus each cat was its own control, which minimizes the number of animals used and also allows direct comparisons before and after denervation.
EMG electrode implantation and recording
EMG electrodes were chronically implanted into several hindlimb muscles. In all cats, the right and left medial gastrocnemius (MG: ankle extensor/knee flexor) were implanted. The other implanted muscles varied between cats but all were located on the right hindlimb. These muscles included tibialis anterior (TA: ankle flexor), adductor femoris (AF: hip adductor/hip extensor), gracilis (GR: hip adductor/knee flexor), and gluteus medius (GMd: hip abductor/hip extensor). A total of six muscles per cat were implanted.
Teflon-insulated, multi-stranded stainless steel wire electrodes (Cooner Wire, No. AS632) were sown into each muscle belly in pairs to provide bipolar EMG recordings. These wires were led subcutaneously from the hindlimb to a 14-pin connector at the head that was secured to the skull using dental acrylic bonded to stainless steel screws. In addition to EMG implants, Velcro patches were sutured onto the skin over several joints. During testing sessions these Velcro patches allowed reflective markers to be attached for collecting kinematic data and removed when the animals were returned to their housing.
During recording sessions, EMG signals were amplified (1,000–8,000 times) and filtered (30–3,000 Hz, P511 amplified, Grass Instruments) prior to storage on magnetic tape (VHS, Vetter 4000A PCM recording unit). Subsequently, data were digitized at 1,000 Hz with a 12-bit A/D converter (DAQcard AI-16E-4, National Instruments). Kinematic data were collected using a four-camera video capture system (Motus 8.0, Peak Performance Technologies) with an effective sampling rate of 60 Hz. Reflective markers were positioned on the surgically attached Velcro patches over the joints of the hindlimbs, both forepaws, and a head marker. Note that due to movement of skin about the knee joint, a virtual knee position was extrapolated using ankle and hip position data, and the length of the femur and tibia obtained post mortem.
Surgical neurectomies of five nerves in both hindlimbs were performed according to the procedures outlined by Bouyer and Rossignol (2003a). Transected nerves included: the superficial peroneal (SP), tibial (Tib), caudal cutaneous sural (CCS), saphenous (Saph), and the deep peroneal cutaneous branch (DPc). Transection of these nerves eliminates cutaneous sensation from the entire paw. Transection sites were as follows: 1) SP—1 cm above the transverse crural ligament, 2) Tib—behind the calcaneal tendon, 3) CCS—surface of the belly of the lateral gastrocnemius muscle, 4) Saph—along the saphenous vein just above the knee, and 5) DPc—on the metatarsal bones, distal to the motor nerve branch going to extensor digitorum brevis muscle. After transection, the proximal stump of each nerve was folded onto itself and sutured to prevent re-growth. Completeness of the denervation was assessed via skin pinches over the hindpaws, 1–2 days after surgery. This evaluation was periodically repeated during the testing period to ensure no cutaneous sensation had returned.
Post hoc, the video records of each recording session prior to the denervation were screened for the walking behavior of the cat. Any in which the cat did not walk straight, turned its head, or varied its speed and pattern were excluded from analysis. Subsequently, those passes for which the cat walked consistently were then captured for further analysis. The markers of the cat were digitized (Motus 8.0, Peak Performance Technologies) for all four cameras and the three-dimensional location of the markers calculated. Using a custom-written routine the walking speed of the cat was calculated using the forward motion of the head marker. The average walking speed of all of the visually selected passes was then calculated and trials with a walking speed outside ± 1 SD were further excluded from analysis. Perturbation trials were also screened to exclude trials if one or more paws were in contact with a stable portion of the walkway. The trials from the postdenervation recordings were screened as described in the preceding text but were speed-matched by accepting only trials with a walking speed within ± 1 SD of the average walking speed of the predenervation trials.
The selected trials were sorted into 1) control (no perturbation), and perturbation trials with perturbation direction and onset of 2) leftward at stance onset, 3) rightward at stance onset, 4) leftward at late stance, and 5) rightward at late stance. The individual trials, in each grouping, were then averaged to produce a group average record for each of the kinematic variables calculated from the video records. The individual trials could be from several different recording sessions throughout the baseline period or following denervation. Frontal plane joint angles were determined for the hip, along with sagittal plane joint angles for the hip, knee, and ankle. These joint angles were compared pre- and postdenervation using the maximum, minimum, and range of motion during one step cycle. Also from the video data, various stepping parameters were calculated including: fore- and hindlimb step widths (distance between paw placements perpendicular to the line of travel), step length, hip height, and paw clearance during swing. In addition, stance duration, swing duration, and hindleg double support time were calculated.
EMG analysis was performed only on those trials selected in the video analysis screening. Custom-written software was used to analyze selected segments of EMG data (Labview 8.2, National Instruments). For each step analyzed a 1,500-ms segment of data was selected, beginning 400 ms prior to the right MG burst onset. The EMG traces were digitally full-wave rectified and low-pass filtered (50 Hz, 2nd-order dual-pass Butterworth filter). The EMG activity of undisturbed, control steps was used to calculate the timing, duration, and amplitude of the bursts for the various muscles studied and compared before and after denervation.
The main objective of this study was to compare the responses evoked by the lateral translation of the support surface before and following cutaneous denervation of the hindpaws. To identify the evoked responses, the group-averaged control EMG traces were subtracted from the selected perturbation trials to yield individual subtracted traces for the perturbation trials. These subtracted traces were then sorted by condition as described. Evoked responses were identified in a muscle if the subtracted trace exceeded the 95% confidence interval around the average control steps continuously for a minimum of 5 ms. Response latency was estimated as the earliest deflection of the visually identified burst. The subtracted traces were then aligned to the onset of the support surface translation and averaged to yield an average subtracted trace. From these average subtracted traces a window for analysis of the response amplitude was identified by visually identifying the time of the onset and offset of the evoked burst. Response amplitudes were then calculated from each individual subtracted trace and taken as the mean subtracted EMG over the window previously defined visually from the average subtracted trace. The burst amplitudes were then normalized to the maximum EMG observed from the control trials prior to denervation for each muscle.
All variables were compared between the intact and denervated states for each individual animal's data using Student's t-test (α = 0.05).
Effects of denervation on undisturbed, overground locomotion
Although the primary aim of this study was to investigate the role of cutaneous sensation in producing corrective adjustments following a perturbation, we also compared aspects of undisturbed walking prior to and following hindpaw denervation. Only data from speed matched trials prior to and following denervation were compared. Table 1 summarizes the number of trials included and the average walking speed of the selected trials for each of the cats.
Cutaneous denervation of the hindpaws led to adaptations in the step width and length during undisturbed walking. In Fig. 1A, the average paw placements and swing trajectories are shown for two cats during undisturbed walking viewed from overhead. The forepaws are displayed with a lateral offset for clarity. The gray lines and open paw prints represent the data before denervation, whereas the black lines and filled paw prints are derived from the data after hindpaw denervation. The traces were initiated at, and aligned to, left hindpaw contact and proceed until the second left hind lift-off, thereby creating traces of two left hind stance phases and two right hind swing phases. As can be seen from these traces, subtle differences in paw trajectory and placement were observed following denervation. Step lengths tended to be shorter across all cats and in both the fore and hindpaws. Adaptations in step width were also observed in all three cats; however, the nature of the adaptation differed between cats. Cat 1 adapted to the denervation of the hindpaws by increasing hindpaw step width (P < 0.05), while concurrently narrowing forepaw step width (although this difference was not significant, Fig. 1B). In contrast, both cats 2 (data shown in Fig. 1, A and B) and 3 displayed significantly narrower hindpaw step widths (P < 0.05) with concurrently broader forepaw step widths (P < 0.05).
The adaptation in step length is better visualized in the sagittal plane. Displayed in Fig. 2A are the average right hindpaw and hip motions for one full step cycle from one cat, aligned to swing onset. As can be seen, the step length is shorter following denervation (black trace) compared with the steps prior to denervation (gray trace). The left histogram in Fig. 2A compares the group average step lengths across all cats before and after denervation and shows the step length following denervation was significantly reduced (P < 0.05). In addition to the shorter step length, cats typically walked slightly crouched following denervation. As can be seen in the data traces of Fig. 2A, the height of the hip was lower throughout the step cycle after denervation. This was consistently observed in all cats, and Fig. 2A, right, histogram displays the group averaged maximum and minimum hip heights across the step, showing that the hip was significantly lower after denervation (P < 0.05). We also observed that, following denervation, the hindpaw makes ground contact at a less forward point relative to the hip joint (P < 0.05) at touchdown, while lifting the paw from a more posterior position relative to the hip (P < 0.05) at lift-off (Fig. 2B). These spatial adaptations in the walking behavior of the cats were accompanied by minimal changes in the temporal parameters of gait. Step cycle, stance, and swing durations were largely unaffected (P > 0.05), presumably reflecting the speed-matching of the trials analyzed. However, a significant increase in the percentage of the step cycle in which the hindlegs were in double support was observed (P < 0.05).
The more crouched posture and the more posterior position of the paw relative to the hip adopted by the cats during walking were accompanied by adaptations in the joint angles of the hindlegs. Average sagittal plane joint trajectories over one full step cycle are displayed for two cats in Fig. 3A, normalized in time to a full step cycle. Typically, the hip tended to be more extended and the ankle more dorsiflexed throughout the step cycle, as shown for both of these cats. In contrast, adaptations in knee joint trajectory varied between animals. In the examples shown, cat 1 adopted a more extended posture of the knee throughout the step cycle, whereas cat 2 adopted a more flexed posture.
Frontal plane motion at the hip also demonstrated minor adaptations following denervation. Figure 3B displays the average hip abduction angle over a full step cycle for cat 2 before (gray trace) and after (black trace) hindpaw denervation. As can be seen, this cat walked with a slightly less abducted posture of the hip throughout most of the step. Recall that this animal, cat 2, walked with a slightly narrower stance width following denervation (Fig. 1). The other cat (cat 3) that walked with a narrower stance width following denervation also walked with a less abducted hip posture. The hip abduction angle of cat 1, which walked with a wider stance following denervation, was largely unchanged following denervation and did not deviate beyond the 95% confidence band for the predenervation control steps. As displayed in Fig. 3C, the maximum and minimum hip abduction angles were significantly (P < 0.05) less abducted following denervation, when compared across cats. Moreover, as can be seen for cat 2 in Fig. 3B and for the group average data in Fig. 3D, the total range of motion of the hip in the frontal plane was also significantly (P < 0.05) greater following denervation.
In general, the timing, pattern and amplitude of EMG activity in the recorded muscles were similar before and after hindpaw denervation. However, one notable exception was observed in AF. Following denervation, a novel burst of activity was observed in late stance (see Fig. 4, *). This change in the pattern of AF activity was observed in all three cats.
Responses to medial-lateral perturbation at stance onset
Medial-lateral translation of the support surface produced corrective responses similar to those reported previously in the intact cat (Misiaszek 2006a). Figure 5 depicts average paw trajectories and placements for the rightward (A) and leftward (B) perturbation trials for one cat. Data from before the denervation are displayed to the left with gray traces and open paw prints, and data from after the denervation are to the right with black traces and filled paw prints. Intact cats typically produced a correction within one step such that the paws reacquired the original path of progression (Fig. 5, A and B, ···). This was achieved following rightward displacements of the right hindpaw (concomitant rightward displacement of the left forepaw) with a leftward placed left hindpaw. The gray histograms in Fig. 5C show the predenervation group averaged paw placements (relative to the estimated line of progression) for the step immediately following the perturbation in A. As can be seen, the displaced right hindpaw regains the path of progression within one step, whereas the left hindpaw, which was in swing at the time of the perturbation, is placed leftward of its original path of progression. The forepaws typically landed near their original path of progression with a modest rightward offset.
Similarly, leftward perturbations of the right hindpaw (Fig. 5B) were also adapted to within one step cycle. As can be seen, the right hindpaw reacquired the original pathway of progression within one step, whereas the left hindpaw, which was in swing at the time of the perturbation, swung leftward avoiding contact with the perturbed stance limb before moving rightward to be placed in a position slightly left of the original path of progression (Fig. 5D). Following the leftward perturbation of the right hindpaw, the forepaws were replaced close to their original paths of progression but were both placed slightly more laterally, thereby adopting a wider stance (Fig. 5D).
Following denervation, the cats tended to stumble sideways in the direction opposite the translation of the support surface. Thus the first step of each paw immediately after the onset of the perturbation shows a large displacement to the left for rightward support surface translations (Fig. 5A, black traces) and a large displacement to the right for leftward support translations (Fig. 5B, black traces). This effect is summarized in Fig. 5, C and D, which summarize the group averaged paw placement data. Note that all paws after denervation showed a significant displacement in paw placement, relative to before denervation, except the left hindpaw which was in swing phase at the time of the perturbation.
The frontal plane hip angular movements associated with the perturbation and subsequent corrective reactions are displayed in Fig. 6. Before denervation, rightward translation of the support surface resulted in abduction of the right hip throughout the disturbed stance phase and further abduction in early swing, before the leg was rapidly adducted during swing (Fig. 6A). After denervation, the perturbed hip motion was characterized as decreased abduction throughout most of the stance phase with an increased adduction late in stance that continued throughout the subsequent swing. Leftward translation of the support surface (Fig. 6B) produced less disturbance of the right hip motion with a modest increase in abduction in the early portion of the disturbed stance phase followed by an increased adduction relative to the undisturbed steps. This was similar for disturbances both before and after the denervation. However, during the subsequent swing phase in the intact animals, the hip abducted toward the undisturbed hip angle, whereas following denervation the hip remained adducted throughout the subsequent swing phase. The data shown in Fig. 6 are from cat 1, but these patterns of motion were consistent across animals and are consistent with data from intact animals previously reported (Misiaszek 2006a).
Medial-lateral translation of the stance paw during walking induces activation of muscles throughout the perturbed hindleg. Figure 7A displays average EMG traces from three muscles for the leftward translation trials, overlaid on the average traces of the undisturbed steps, for data from cat 2 both before and after hindpaw denervation. In this figure it is obvious that responses evoked in right MG and GMd were preserved following denervation but appear to be of smaller amplitude. In Fig. 7B, the same data are displayed but as subtracted traces whereby the average undisturbed traces (control steps) are subtracted from the perturbation trials and the perturbation trials are aligned to the onset of platform translation. As can be seen, the latencies of the evoked responses were not substantially influenced by the denervation of the hindpaws (summarized in Table 2), but the response amplitudes in right MG and GMd were reduced. Figure 7C summarizes the group average response amplitudes across cats for those recorded muscles for which an evoked response was elicited by the perturbation. Responses in right MG, GMd, and GR tended to be reduced following denervation, significantly so for right MG and GMd (P < 0.05). In contrast, the modest response evoked in AF tended to be larger, but this was not significant (P > 0.05).
Figure 8A displays the subtracted EMG traces for data from cat 2 for responses evoked by rightward displacement of the stance paw. As with the leftward displacements, responses were evoked at short latency and the denervation of the hindpaws did not influence the expression or latency of the responses but did impact the amplitude of the evoked activity. The group average response latencies are summarized in Table 2 for those recorded muscles for which a response was evoked. The group average response amplitudes across cats are summarized in Fig. 8B. Significantly smaller burst amplitudes (P < 0.05) were evoked in all muscles by the perturbations after the denervation of the hindpaws.
Responses to medial-lateral perturbations at end of stance
Perturbations of the support surface were also delivered at late stance of the right hindleg, which was also stance onset of the left hindleg. Consequently, the gross motor behavior of the animals following the perturbations was similar to that described in the preceding text for perturbations delivered at stance onset of the right hindleg but in the opposite direction. Therefore we will not repeat this description here, but limit our discussion to the behavior and responses of the right hindleg.
Leftward support surface translations (medial displacement of the right paw at the end of stance) resulted in an increase abduction angle of the right hip toward the end of swing in the intact cat (Fig. 9A, top traces). The hip angle remained abducted, relative to the undisturbed steps, throughout the subsequent stance phase. Also note that paw contact was initiated substantially earlier, relative to the control steps. Following denervation, the right hip also became more abducted following the perturbation (Fig. 9A, bottom traces). The magnitude of the hip abduction, relative to the control steps, appears to be greater following denervation. The group averaged peak deviation (maximum difference between perturbed and control steps) prior to denervation was 11.1 ± 4.3° and was significantly greater (P < 0.05) following denervation at 17.1 ± 2.5°.
During undisturbed walking, the hip abductors and adductors are generally quiet during much of the swing phase, with AF and GMd becoming active in late swing just prior to foot contact (Fig. 4). This was true prior to and following denervation. Leftward perturbations of the right hindleg at late stance resulted in a rapid evoked response in GMd and slightly later responses in GR and AF. The evoked responses (subtracted traces) to leftward perturbations are displayed for GMd and AF from one cat in Fig. 9B. The responses evoked prior to denervation are displayed as the gray traces and the responses following denervation are overlaid in black. As can be seen, comparable responses were evoked in these muscles following denervation. Moreover, the onset latencies of the evoked responses were not affected by the denervation (Table 3). However, following denervation the response evoked in GMd was substantially increased compared with the response in the intact animal (Fig. 9B). This result is emphasized in Fig. 9C, which displays the group average response amplitudes for these two muscles across cats. The response amplitude in GMd was significantly increased following denervation (P < 0.05).
Rightward support surface translations (lateral displacement of the right paw at the end of stance) resulted in little change to the ongoing right hip abduction angle during the subsequent swing phase in the intact animal (Fig. 9D, top traces). However, the right hip angle became more abducted during the subsequent stance phase compared with the undisturbed control steps. In contrast, following denervation the same perturbations resulted in a more adducted hip angle very soon after the onset of the subsequent swing phase, compared with undisturbed steps (Fig. 9D, bottom traces). The right hip remained more adducted throughout swing and for most of the subsequent stance phase.
In the intact animal, rightward perturbations of the right hindpaw at the end of stance produced a rapid response in GMd and later responses in AF (Fig. 9E, gray traces) and GR (not shown). Following denervation, similar responses were evoked by the perturbations and with comparable latencies. The group average response latencies for GMd, AF, and GR are summarized in Table 3. There was no difference in response latency in GMd and AF, whereas the response in GR was of significantly shorter latency following denervation. The amplitude of the evoked responses in AF (Fig. 9, E and F) and GR (not shown) were not different between the intact and denervated animals (P > 0.05). However, as shown in Fig. 9E for one cat and summarized for the group average data in Fig. 9F, the amplitude of the response evoked in GMd was substantially larger following the denervation of the hindpaws (P < 0.05).
Although all cats successfully maintained balance when challenged with medial-lateral perturbations, deficits were evident in reactive postural adjustments after hindpaw cutaneous denervation. Following denervation, cats tended to stumble sideways to regain their balance and consequently the correction required large displacements of all four limbs. The muscle response patterns following perturbation were largely unaffected by the denervation; however, the amplitude of these responses were markedly different following the loss of cutaneous sensation from the hindpaws. Specifically, response amplitudes in the perturbed stance leg tended to be decreased, whereas GMd of the swing leg tended to show a marked increase. We suggest that the decreased responses in the muscles of the perturbed stance leg lead to the instability noted by the increased stumbling and propose that the loss of cutaneous sensation from the hindpaws is directly related to this decreased response amplitude.
Impact of denervation on the undisturbed overground walking pattern
The main purpose of the present investigation was to elucidate contributions from cutaneous afferents in producing corrective adjustments while walking. For an accurate assessment of these corrective reactions, however, it was first important to investigate the impact of denervation on undisturbed, overground walking. Following denervation, the most prominent changes in the walking pattern included a novel burst in AF toward the end of stance phase, increased double support time, lower hip heights and altered step widths. A detailed evaluation of treadmill walking in cats after hindlimb cutaneous denervation was previously performed by Bouyer and Rossignol (2003a). Overall the present results substantiate their findings, most notably in that undisturbed walking appeared very normal on visual inspection. Moreover, several subtle features such as increased double support times, paws lifting off the ground at a point further behind the hip at swing onset, and a generally “crouched” gait (due to an increase in ankle dorsiflexion), were confirmed in our study. However, our observations did not show a change in paw ground clearance, nor was there a consistent increase in knee flexion during swing. Furthermore, Bouyer and Rossignol (2003a) noted consistent increases in the lateral ground reaction force exerted by all four limbs; this contrasts the variable adaptations noted for step widths in our study.
One obvious reason for any disparity between our results and those of Bouyer and Rossignol (2003a) is the fact that their evaluation of walking employed a treadmill, whereas the present study involved overground locomotion. Although treadmill locomotion shares similarities to overground locomotion, the difference in adapting to a constantly moving treadmill belt under the paws may impose a different demand on how walking is controlled. Another important difference between the two studies is that cats in the present study were walking in an environment where unpredictable disturbances to the support surface occasionally take place. This contextual difference may result in different walking strategies (Misiaszek 2006b). Some of the commonly observed adaptations such as crouched walking, possibly reflect a general cautionary strategy to maintain a lower center of mass after sensory loss. Moreover, the increased lateral ground reaction forces reported by Bouyer and Rossignol (2003a) differs from the variable influence on fore and hindlimb step widths in the present study. This may simply reflect the challenges inherent in responding to unpredictable medial-lateral perturbations in the present study. It is interesting to note that the cats in the current study adopted different step width strategies following the sensory loss, despite being exposed to a common walking environment.
With any study that ablates some part of the nervous system in hopes of revealing a role in the intact animal, interpretation is always difficult. Any behavior following such ablation typically reflects adaptability of intact neural regions, and these remaining sensory sources could be modified to generate appropriate corrective responses to maintain balance in the absence of cutaneous inputs from the paws. For example, Frigon and Rossignol (2007) showed that partial denervation of the ankle extensors led to adaptations in the expression of reflexes from the paw during walking in cats. In the present study, the removal of cutaneous inputs from the paws might have a comparable effect on other sensory systems controlling gait. In addition, the changes we observed in the undisturbed gait of the cats following denervation might reflect compensatory strategies to overcome or limit the impact of the lost sensory cues. That is, we observed that the animals tended to walk with a more extended hip and dorsiflexed ankle following denervation. Such a strategy might be incorporated to increase the inputs from hip flexor and ankle extensor muscle afferents; signals argued to be critical for the control of locomotion (Pearson 2004; Rossignol et al. 2006). Further, it must be recognized that compensatory adjustments often change over time, especially in the weeks immediately following sensory loss. Many early adaptations reported by Bouyer and Rossignal (2003a) during walking following denervation returned to predenervation values by the end of the testing period, which typically exceeded 1 mo. Although the present study focused on data collected 3 wk postdenervation, it is possible that some degree of gradual compensation had already emerged. Therefore revealing specific roles for hindlimb cutaneous inputs in the intact animal might be partly confounded by compensatory adjustments.
An additional concern with interpreting our results is that one of the nerves transected in our study, the tibial nerve, innervates intrinsic muscles of the foot in addition to the skin on the plantar surface of the foot. Although our preparation cannot eliminate the possible influence of afferents from the intrinsic foot muscles, there is good reason to believe cutaneous afferents are largely responsible for the results of the present study. For example, Bouyer and Rossignol (2003b) noted that innervation of the intrinsic foot muscles was not essential for normal paw placement during walking in spinal cats; however, at least some degree of hindpaw cutaneous sensation was required to preserve proper plantar stepping. Moreover, the reflex responses noted from low intensity tibial nerve stimulation were suggested to result from activating low-threshold cutaneous afferents (Frigon and Rossignol 2008a). In human studies on reactive postural control, depressed EMG responses were noted following anesthetizing the plantar foot sole, and this was attributed to loss of cutaneous sensation by the authors (Do et al. 1990). Taken together there is good evidence to suggest that the deficits noted for the present study, particularly the reduced muscle responses of the stance limb, are likely due to the loss of cutaneous sensation.
A further issue with postdenervation responses from the present study is whether or not they truly represent a deficit in postural control. Due to the reasonably wide and level walking surface, along with Plexiglas walls to remove any threat of falling off of the walkway, there was little incentive for these cats to maintain the original path of progression. We may therefore only reveal a fraction of the compensation that is possible if these cats were more motivated or constrained to regain the proper walking path. Cats in the present study are likely not encouraged to further optimize their corrective reactions because the responses they produced were adequate for maintaining balance. If the environmental challenge was greater, deficits may become more pronounced with the absence of hindpaw cutaneous sensation. Constraints on accurate paw placement for example may reveal a greater compromise to balance control. This idea would match with previous observations that cutaneous afferents become increasingly important in walking environments where accurate paw placement is necessary, such as when walking over a narrow beam or horizontal ladder (Bouyer and Rossignol 2003a).
It is possible that an auditory startle response was induced by the perturbation. Activation of the motorized slide was accompanied by a noticeable sound for the duration of the translation. However, this is unlikely to affect the results as the animals experienced repeated exposures of the perturbations daily for several weeks. In addition, most trials included for analysis occurred midway through any individual recording session, after a number of perturbation exposures had been experienced that session. Startle responses typically habituate rapidly with repeated exposure (Davis 1984; Koch 1999). Also auditory startle responses during walking in humans typically involve activation of TA with more modest responses in soleus, an ankle extensor (Nieuwenhuijzen et al. 2000; Schepens and Delwaide 1995). In contrast, the evoked responses in our study were prominent in the extensors of the leg but with no responses evoked in the ankle flexor TA. Moreover, similar recording protocols were used before and after denervation. Therefore any influence of a startle response within the analysis would be expected to be consistent in our pre- and postdenervation comparison.
Role of hindlimb cutaneous sensation in the corrective responses to medial-lateral perturbation
Following a perturbation, the intact cat makes a seamless transition back to the original path of progression within one step cycle as indicated by paw placements and swing trajectories (Fig. 5). These findings confirm our previous observations (Misiaszek 2006a). This smooth return to the original walking path was not observed following denervation, where instead paw placements were markedly displaced opposite to the direction of perturbation (Fig. 5). Importantly, platform translations were imposed when the entire body of the cat was positioned over the platform, thus postural disturbances were body-wide. It is noteworthy, that although cutaneous denervation was only performed on the hindpaws, deviation from the original walking path was not limited to the hindlimbs but instead affected the trajectories of all four limbs. Therefore the loss of cutaneous sensation from the hindpaws has a clear body-wide impact. Moreover, intact sensory feedback from the forelimbs was insufficient to compensate for this loss, and the forelimbs themselves were incapable of correcting for the disturbance.
Responses in muscles throughout the perturbed stance limb were reduced in amplitude following denervation. Taken together with the exaggerated paw displacements this suggests that the stance limb responses were insufficient to stabilize the body following denervation and presumably the more laterally displaced paws act to catch a falling center of mass. The impact of denervation on the behavior of the cats following perturbation was also apparent in the joint angles. For example, following denervation, the hip was not abducted subsequent to the perturbation. Presumably, this is related to the reduction in hip abductor activity evoked by the perturbation following denervation. This suggests that in the intact cat the response evoked in the hip abductors (including GMd) of the stance limb serves to support the pelvis and prevent the contralateral hip from falling. How this is achieved is not apparent from the limited number of muscles recorded in this study. However, it has been previously demonstrated that several muscles in the cat hindlimb produce significant torques that are not limited to the sagittal plane (Lawrence et al. 1993). In addition, passive tissue such as fascia connecting muscles throughout the hindlimb has been shown to have a clear role in limb stability across multiple planes (Stahl et al. 2007). Given the complicated interaction between the various muscles and passive tissue, a detailed study of mechanical actions is likely required to resolve how specific muscle patterns influence the observed movements. What is clear; however, is that a loss of cutaneous sensation from the hindpaws results in a less effective corrective response and that this is related to the decreased muscle activity in the stance limb.
Concomitant changes were noted in the swing limb where the postdenervation hip movement about the frontal plane was significantly exaggerated when perturbations were imposed. In particular, denervated cats showed greater adduction with medial perturbations (Fig. 9D) despite an increased amplitude of the evoked response in GMd. These increases in hip movements of the swing leg might be related to the decreased stability in the contralateral stance limb as described in the preceding text. That is, if the stance leg is less capable of counteracting the perturbing forces and the gravitational acceleration, then presumably the center of mass will experience a greater displacement. This in turn would result in greater acceleration of the swing limb, resulting in a passive lateral displacement of the paw and the exaggerated lateral paw placements. The increased response amplitude in GMd in this situation is likely related to the increased stretch or rate of stretch of the abductors and presumably assists in limiting the impact of these passive effects.
Intact cats demonstrate responses that are directionally specific to counter either a medial or lateral perturbation. This was exemplified in the abductors and adductors of the stance limb where GMd but not AF produced responses during lateral perturbations, whereas medial perturbations triggered responses in both muscles. Denervating the hindpaws did not change the pattern of these corrective responses, and response latencies were largely unaffected. In contrast, response amplitudes were altered following denervation. Thus it appears that cutaneous sensations have a role in scaling the evoked responses but are not required to trigger the responses. Clearly other sources of afferent input are able to provide a trigger for the responses that do emerge. In fact there is evidence to suggest that muscle spindle feedback can activate the direction-specific responses that occur with horizontal support surface translation in standing cats (Honeycutt et al. 2007, 2008; Nichols et al. 1999). Indeed Stapley et al. (2002) showed that loss of large-diameter somatosensory afferents resulted in significantly delayed corrective responses to disturbances of standing balance. In the present study, such feedback would still be largely available throughout the hindlimbs, thus offering a sensory cue for corrective reactions.
Excitatory connections from cutaneous afferents to ankle extensor motoneurons have been identified (LaBella et al. 1989; Wilson 1963), but these appear to have little influence on the normal locomotor output (Bouyer and Rossignol 2003a; Duysens and Stein 1978; Engberg 1964; Forssberg et al. 1977; Prochazka et al. 1978; Wand et al. 1980). Our findings would suggest that the excitatory connections from cutaneous afferents to extensor motoneurons might be more relevant in contributing to the corrective responses than the ongoing locomotor activity. As shown in Fig. 4, the amplitude of the ongoing locomotor activity was little affected by the loss of cutaneous inputs, consistent with previous findings. However, the reduced amplitude of the evoked responses might result from a loss of such facilitation to extensor motoneurons with the absence of cutaneous feedback related to ground contact. For example, Schieppati and Crenna (1984) demonstrated that physiological stimulation of cutaneous afferents (via hair bending, skin indentation or stretch, but not pinprick) in unanesthetized spinal cats produced a late onset, but long-lasting facilitation of ankle extensor motoneurons as detected by testing the amplitude of the monosynaptic reflex. In other words, ground contact may provide cutaneous sensory cues that serve to facilitate, or scale, responses evoked from other sensory triggers.
Given the persistence of rapid response latencies following denervation, a somatosensory source is likely. Thus the remaining capacity to register rapid changes in muscle load and stretch could trigger the observed hindlimb responses as indicated earlier (Honeycutt et al. 2007, 2008; Nichols et al. 1999). This may seem contradictory to Ting and Macpherson (2004), who argued that cutaneous sensors provided the only unambiguous signal to cue a directionally relevant response. However, in their study, platform perturbations consisted of both rotation and translation, thus imposing a special context for postural corrections. It is not that muscle stretch and load receptors are inherently unable to cue a directionally specific response. Rather, in the context of rotations and translations, these sensors may alone be inadequate for providing an unambiguous cue to maintain balance. Even the directional cues provided by cutaneous receptors are subject to modulation by vestibular-based references to earth-vertical. In cats with surgically lesioned vestibular systems, somatosensory inputs produce APRs that are referenced to the support surface but not gravity during surface tilts, resulting in highly inappropriate responses (Macpherson et al. 2007). In this case, the vestibular systems' detection of earth-vertical would dictate the directional relevance of these somatosensory cues.
It is important to recognize that all cats in our study showed adaptations in the undisturbed walking trials following sensory loss, which likely reflects a more cautious gait. Adaptations such as lowered hip heights and widening either the fore or hindpaw widths may reflect compensation for medial-lateral instability especially in an environment where this stability is periodically challenged. Accordingly, any postdenervation reactions need to be considered in the context of this altered walking strategy. Increasing stance width for example can significantly bolster resistance to lateral perturbations by offering greater biomechanical advantage. Here enhanced protection against lateral falls would be conferred by simply altering body configuration instead of requiring a change in the active response (Scrivens et al. 2006). The same would apply to walking with a lower center of mass by crouching. Through repeated exposure, these cats likely learn that the perturbations will displace them to one side or the other. Therefore the resultant walking strategies may reflect proactive adjustments to remain stable in the midst of medial-lateral disturbance. Overall, the corrective responses always occur in the context of an adjusted walking strategy primed for medial-lateral disturbance. This altered strategy therefore represents a compensation that will likely influence responses cued by the remaining sensory sources.
Summary and conclusions
Following the loss of cutaneous sensation from the hindlimbs, cats revealed altered muscle responses with consequent behavioral deficits. Although general response patterns and latencies were preserved, the loss of cutaneous sensation was associated with a reduction in muscle response amplitude in the stance limb. Cutaneous input appears to scale responses triggered by other somatosensory sources, although this may indicate a generalized loss of somatosensory feedback as opposed to a cutaneous-specific contribution. The diminished stance limb responses resulted in deviated limb trajectories from the original path of progression. Our results indicate that the remaining sensory cues can trigger sufficient responses to maintain upright balance; however, these responses are less effective when cutaneous sensation is absent. It is possible that environments with greater constraints on corrective limb placement would reveal even more pronounced and distinct deficits.
This work was funded by a grant from the Canadian Institutes for Health Research to J. E. Misiaszek.
The authors thank R. Gramlich for technical assistance with this work and Dr. K. G. Pearson for comments on the manuscript.
Present address of D.A.E. Bolton: Department of Kinesiology, Faculty of Applied Health Sciences, University of Waterloo, 200 University Ave, Waterloo, Ontario, Canada N2L 3G1.
- Copyright © 2009 the American Physiological Society
- Abraham and Loeb 1985.↵
- Bolton and Misiaszek 2007.↵
- Bouyer and Rossignol 2003a.↵
- Bouyer and Rossignol 2003b.↵
- Davis 1984.↵
- Do et al. 1990.↵
- Drew and Rossignol 1987.↵
- Duysens and Loeb 1980.↵
- Duysens and Stein 1978.↵
- Engberg 1964.↵
- Forssberg 1979.↵
- Forssberg et al. 1977.↵
- Frigon and Rossignol 2007.↵
- Frigon and Rossignol 2008a.↵
- Frigon and Rossignol 2008b.↵
- Gorassini et al. 1994.↵
- Honeycutt et al. 2007.↵
- Honeycutt et al. 2008.↵
- Horak and Macpherson 1996.↵
- Inglis et al. 1994.↵
- Inglis and Macpherson 1995.↵
- Karayannidou et al. 2009.↵
- Koch 1999.↵
- LaBella et al. 1989.↵
- Lawrence et al. 1993.↵
- Macpherson et al. 2007.↵
- Maki and McIlroy 2007.↵
- Misiaszek 2006a.↵
- Misiaszek 2006b.↵
- Mori et al. 1970.↵
- Nichols et al. 1999.↵
- Nieuwenhuijzen et al. 2000.↵
- Pearson 2004.↵
- Pratt et al. 1991.↵
- Prochazka et al. 1978.↵
- Rossignol et al. 2006.↵
- Schepens and Delwaide 1995.↵
- Schieppati and Crenna 1984.↵
- Scrivens et al. 2006.↵
- Sherrington 1910.↵
- Stahl et al. 2007.↵
- Stapley et al. 2002.↵
- Ting and Macpherson 2004.↵
- Wand et al. 1980.↵
- Wilson 1963.↵