Brain Advance Access originally published online on February 26, 2004
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Brain, Vol. 127, No. 5, 1019-1034, 2004
© 2004 Guarantors of Brain
doi: 10.1093/brain/awh115
Distributed plasticity of locomotor pattern generators in spinal cord injured patients
1 IRCCS Fondazione Santa Lucia, via Ardeatina 306, 00179 Rome, 2 Institute of Neurology, Catholic University, 00197 Rome, 3 Biomedical Engineering Laboratory, Istituto Superiore di Sanità, 00168 Rome and 4 Department of Neuroscience and Centre of Space Bio-medicine, University of Rome Tor Vergata, Via Montpellier 1, 00133 Rome, Italy
Correspondence to: Professor Francesco Lacquaniti, University of Rome Tor Vergata and IRCCS Fondazione Santa Lucia, via Ardeatina 306, 00179 Rome, Italy E-mail: lacquaniti{at}caspur.it Deceased
 |
Summary |
|---|
 Top Summary IntroductionMethodsResultsDiscussionConclusionsReferences |
|---|
Recent progress with spinal cord injured (SCI) patients indicates that with training they can recover some locomotor ability. Here we addressed the question of whether locomotor responses developed with training depend on re-activation of the normal motor patterns or whether they depend on learning new motor patterns. To this end we recorded detailed kinematic and EMG data in SCI patients trained to step on a treadmill with body-weight support (BWST), and in healthy subjects. We found that all patients could be trained to step with BWST in the laboratory conditions, but they used new coordinative strategies. Patients with more severe lesions used their arms and body to assist the leg movements via the biomechanical coupling of limb and body segments. In all patients, the phase-relationship of the angular motion of the different lower limb segments was very different from the control, as was the pattern of activity of most recorded muscles. Surprisingly, however, the new motor strategies were quite effective in generating foot motion that closely matched the normal in the laboratory conditions. With training, foot motion recovered the shape, the step-by-step reproducibility, and the two-thirds power relationship between curvature and velocity that characterize normal gait. We mapped the recorded patterns of muscle activity onto the approximate rostrocaudal location of motor neuron pools in the human spinal cord. The reconstructed spatiotemporal maps of motor neuron activity in SCI patients were quite different from those of healthy subjects. At the end of training, the locomotor network reorganized at both supralesional and sublesional levels, from the cervical to the sacral cord segments. We conclude that locomotor responses in SCI patients may not be subserved by changes localized to limited regions of the spinal cord, but may depend on a plastic redistribution of activity across most of the rostrocaudal extent of the spinal cord. Distributed plasticity underlies recovery of foot kinematics by generating new patterns of muscle activity that are motor equivalents of the normal ones.
Key Words: human paraplegia; muscle synergies; motor equivalence; human locomotion; central pattern generators
Abbreviations: ASIA = American Spinal Injury Association; BF = long head of biceps femoris; BIC = biceps brachii; BWS = body weight support; BWST = body-weight-support on treadmill; CPG = central pattern generator; ES = erector spinae; GCL = gastrocnemius lateralis; GM = gluteus maximus; GT = greater trochanter; IL = ilium; LD = latissimus dorsi; LE = lateral femur epicondyle; LM = lateral malleolus; MAS = Modified Ashworth Scale; MN = motor neuron; OE = external oblique; OI = internal oblique; RAM = middle rectus abdominis; RAS = superior rectus abdominis; RF = rectus femoris; SCI = spinal cord injury; TA = tibialis anterior; TRAP = trapezius; TRIC = triceps brachii; VL = vastus lateralis; VM = fifth metatarso-phalangeal joint; VMA = normalized tolerance area of VM; WISCI = Walking Index for Spinal Cord Injury
Received October 14, 2003. Revised December 18, 2003. Accepted December 19, 2003.
|
Introduction |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionConclusionsReferences |
|---|
In traditional schemes the human spinal cord is assigned a subservient function for the production of complex movements, being viewed as an inflexible conduit for information transmitted to and from supraspinal systems. However, recent data on spinal cord injury (SCI) patients challenge this view by showing that the spinal cord has the potential to generate rhythmic motor activity in a flexible, task-dependent manner (Bussel et al., 1988
; Barbeau et al., 1993; Brown et al., 1994; Calancie et al., 1994; Nathan, 1994; Dobkin et al., 1995; Wernig et al., 1995; Harkema et al., 1997; Dimitrijevic et al., 1998; Shapkova and Schomburg, 2001; Dietz et al., 2002; Ivanenko et al., 2003). Patterned sensory inputs play a key role in facilitating and modulating the spinal rhythmic output, and daily locomotor training with body-weight-support on a treadmill (BWST) often results in significant improvements in locomotor function in motor-incomplete SCI patients (Dietz et al., 1995, 1998; Dobkin et al., 1995; Wernig et al., 1995; Barbeau et al., 1999a, b; Barbeau and Fung, 2001). Locomotor improvements often extend outside the laboratory, some form of walking becoming possible in these patients.
In motor-complete SCI patients, unsupported walking seldom, if ever, recovers. However, the peripheral sensory inputs associated with BWST can influence the motor patterns in these patients under laboratory conditions (Dobkin et al., 1995; Wernig et al., 1995; Harkema et al., 1997; Maegele et al., 2002). It has been shown that the isolated lumbosacral spinal cord can interpret load (Harkema et al., 1997; Dietz et al., 2002) and speed (Patel et al., 1998) information in a state-dependent manner. Under optimal conditions of limb loading, treadmill speed and appropriate kinematics, patients with clinically complete SCI could generate three to 10 consecutive steps without assistance on at least one leg (Harkema et al., 2000; Maegele et al., 2002).
Understanding the mechanisms of locomotor responses after a spinal lesion is fundamental to the development of improved rehabilitation strategies (Barbeau et al., 1999a, b; Harkema et al., 2000; Wernig et al., 2000: Edgerton et al., 2001; Dietz et al., 2002; Edgerton and Roy 2002; Dietz, 2003; Ivanenko et al., 2003). There is a growing consensus that recovery largely depends on plasticity phenomena induced by the lesion (Calancie et al., 1994; Wernig et al., 1995; Harkema et al., 1997; Barbeau et al., 1999a,b; Dietz et al., 1999; Dobkin 2000; Raineteau and Schwab, 2001; Calancie et al., 2002). What is the functional outcome of the plastic reorganization of the lesioned spinal cord? An especially important but unresolved question is whether locomotor responses depend on the re-activation of the normal motor patterns or do they depend on learning new motor patterns (Dietz et al., 1999; Pearson 2000; de Leon et al., 2001). In the experimental model of cats trained with BWST after a complete low-thoracic spinal cord transection, the patterns of muscle activity are very similar to those in the normal animal (Belanger et al., 1996). This similarity indicates that recovery mainly depends on the re-activation of the neuronal circuits involved in generating the motor patterns in normal animals. In spinalized cats, however, re-activation is contingent on the afferent feedback (Pearson, 2001) and the specific training task (de Leon et al., 1998).
The picture in human SCI-subjects is much less clear. Improved performance in BWST-trained patients is associated with an increase in the level and extent of modulation of activity in leg extensor muscles (Dietz et al., 1995), larger than can be voluntarily recruited from resting positions (Wernig et al., 1995; Maegele et al., 2002). However, several motor neuron (MN) pools located below the lesion may remain unable to generate normal patterns and levels of activity sufficient to support body weight and to propel the limbs and body forward (Dietz et al., 1999; de Leon et al., 2001). This probably depends on the loss of facilitatory inputs from supraspinal centres. Cortico-spinal and other supraspinal descending systems are more dominant for the control of locomotion in higher primates than in the other mammals such as the cat (Duysens and Van de Crommert, 1998). Therefore, one might hypothesize that, in contrast with spinalized cats, SCI-subjects cannot entirely re-activate the normal motor patterns but must develop new compensatory strategies to replace lost function.
If so, a new question arises: what aspects of movement, if any, are regained after spinal injury? Although the motor output from the spinal cord consists of the waveforms of muscle activity, major locomotor goals are defined in terms of foot kinematics (trajectory and speed) and kinetics (contact forces). The control of foot position requires more global coordination than the control of the position of a single joint, as the former depends on the spatiotemporal coordination of multiple muscles acting on several body and limb segments (Winter, 1991; Ivanenko et al., 2002a). There are some indications that the human spinal cord can interpret global limb parameters such as foot loading and translation (Harkema et al., 1997; Dietz et al., 2002). However, the kinematic determinants that can be expressed by the human spinal cord are still poorly understood (Barbeau et al., 1999a; Harkema et al., 2000; Ivanenko et al., 2003).
Here we addressed these questions by applying quantitative analysis to kinematic and EMG data recorded in detail both in SCI patients trained with BWST and in healthy subjects. Our aim was not to assess the efficacy of BWST as a therapy, but to explore the mechanisms involved in locomotor improvements associated with this training.
|
Methods |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionConclusionsReferences |
|---|
Subjects
Eleven SCI patients (Table 1) and 11 healthy age-matched subjects were studied. In a previous study we reported factor analysis from these subjects (Ivanenko et al., 2003
). In most patients the neurological level of the injuries was located in the thoracic cord and no patient had signs of conus medullaris syndrome. At hospital admission, they were submitted to neurological evaluation, routine radiological and neurophysiological tests. No signs of denervation were found in the leg muscles by EMG. They were classified according to the American Spinal Injury Association (ASIA) impairment scale (Maynard et al., 1997). Five patients were classified as ASIA-A (complete paraplegia, no sensory or motor function below the neurological level including S4–S5 segments), two as ASIA-B (sensory but not motor function is preserved below the neurological level), and four as ASIA-C (motor function is preserved below the neurological level, and more than half of key muscles below this level have a muscle grade less than 3 out of 5, i.e. they cannot be actively contracted against gravity). It should be noticed that the assessment of completeness of a spinal lesion is based on clinical, radiological and neurophysiological tests indicating the absence of motor and sensory function below the injury site, but it does not necessarily imply that there are no axons that cross the injury site.
|
At discharge, two ASIA-C patients were re-classified as ASIA-D (at least half of key muscles below the neurological level have a muscle grade equal to or higher than 3 out of 5, i.e. they can be actively contracted against gravity or some resistance), whereas the classification of the other patients did not change. Written informed consent was obtained from each subject according to the Declaration of Helsinki; the experiments and training procedures were approved by the Ethical Committee of IRCCS Fondazione Santa Lucia.
Experimental set-up
Subjects stepped on a treadmill (EN-MILL 3446.527; Bonte Zwolle BV, Netherlands) at different controlled speeds. They placed the abducted arms on horizontal rollbars located at the side of the treadmill, at breast height. Body weight support (BWS) was obtained by suspending the subjects in a parachute harness (Reha, BONMED, Germany) connected to a pneumatic device that applied a controlled upward force at the waist (Ivanenko et al., 2002a). The overall constant error in the applied force and dynamic force fluctuations monitored by a load cell were <5% of the residual body weight (Gazzani et al., 2000). Three-dimensional motion of selected body points was recorded at 100 Hz by means either of the Optotrak system (Northern Digital, Waterloo, Ontario) (±3 SD accuracy better than 0.2 mm for x, y, z coordinates) or of 9-TV cameras Vicon-612 system (1-mm accuracy) during about 20–100 s depending on treadmill speed. Five infrared markers were attached on the right side of the subject to the skin overlying the following landmarks: the mid-point between the anterior and the posterior superior iliac spine (ilium, IL), greater trochanter (GT), lateral femur epicondyle (LE), lateral malleolus (LM) and fifth metatarso-phalangeal joint (VM). In all controls and nine patients EMG activity was recorded by means of surface electrodes from leg muscles [tibialis anterior (TA), gastrocnemius lateralis (GCL), long head of biceps femoris (BF), rectus femoris (RF), vastus lateralis (VL), gluteus maximus (GM)], axial muscles [middle rectus abdominis (RAM) and superior rectus abdominis (RAS), external oblique (OE), internal oblique (OI), latissimus dorsi (LD), erector spinae (ES)] and shoulder girdle muscles [trapezius (TRAP), triceps brachii (TRIC), biceps brachi (BIC)]. In six controls and three patients, EMG was also recorded from peroneus longus, semitendinosus, adductor longus, sartorius, tensor fasciae latae and deltoid. In two controls and two patients, EMG was also recorded from soleus, gastrocnemius medialis, gluteus medius and ilio-psoas. In two patients, EMG was only recorded from leg muscles (TA, GCL, BF, RF, VL, GM). EMG signals were preconditioned at the recording site (active electrodes from BTS, Milan, Italy or DelSys, Boston, MA, USA), digitized, transmitted to the remote amplifier (20-Hz high-pass and 200-Hz low-pass filters), and sampled at 500 or 1000 Hz (synchronized with kinematic sampling).
Protocol
Patients performed daily sessions of BWST training for 1–3 months, starting from 1–6 months after the lesion. Training began shortly after admission and continued till discharge. Patients were assisted to step as necessary by two physiotherapists. Under their guidance, patients underwent progressive training with increasing treadmill speed, decreasing BWS and decreasing manual assistance from the therapists. Each physiotherapist initially held one patient’s leg at the ankle to assist with swing and foot placement, but patients were encouraged to step independently as soon as possible. BWS was set at 75% of body weight at the beginning and was subsequently decreased by 5% steps according to the patient’s improvement. Three ASIA-C/D patients reached 0% BWS and 2–3 km/h at the end of training, whereas one ASIA-C (SCI-C4) and all ASIA-A/B patients never went below 60–75% and faster than 1–2 km/h. Treadmill speed was set at 0.7 km/h in the first session, and was increased to 1, 1.5, 2 and 3 km/h whenever possible. This was done because it has been suggested that SCI patients may execute the swing phase of stepping more independently at faster than at slower speeds (Harkema et al., 2000). However, the initial training condition (75% BWS, 0.7 km/h) was included in each session till the end of training. Kinematic and EMG data were collected during stepping attempts 1 day before and then every 15 days after the start of training. The recording session after 15 days could not be performed in patient SCI-C1. Patients were tested at all previously trained values of BWS/speed. The default condition (75% BWS, 0.7 km/h) was always included. During the recording sessions, patients stepped by themselves, being helped by the physiotherapists only when they stumbled. The specific strategies used to step differed widely among patients, and will be described under Results. Patients performed several such trials, each comprised eight to 15 consecutive step cycles, and paused between trials when fatigued. During the pauses, the unloading system was released and the patient sat on a chair. Training and recording sessions were interrupted at subjects’ request or when heart rate or blood pressure (constantly monitored) reached attention levels (this occurred very rarely). All control subjects were tested in one experimental session at 0.7, 1, 1.5, 2 and 3 km/h, and at 0, 35, 50 and 75% BWS. Controls did not undergo daily training with BWST but, to verify repeatability of results, seven of them were tested in additional experiments separated by 1 or more months.
Clinical evaluation
We measured mobility by means of Rivermead Mobility Index (Collen et al., 1991, 0–15 score), and ambulation by means of the Walking Index for Spinal Cord Injury (WISCI) (Ditunno et al., 2000, 2001; 0–20 scale) and the Garrett scale (Garrett et al., 1987; 0–6 scale). Leg spasticity was evaluated by the Modified Ashworth Scale (MAS) modified by Bohannon and Smith (1987). In MAS, f denotes flaccidity, 0 denotes ‘no increase in muscle tone’, whereas 1–5 denote increasing levels of spasticity. MAS scores 0, 1, 2, 3, 4, 5 are sometimes also indicated as 0, 1, 1+, 2, 3, 4, respectively, and MAS has one score (1+) intermediate between scores 1 and 2 of the original Ashworth scale. Here we summarize the general trend, while individual data are reported in Table 1.
In motor-complete SCI patients, the mean Rivermead score was 0.3 ± 0.7 (n = 7), and both WISCI and Garrett scores were 0 before training (meaning that patients could not ambulate outside the BWST apparatus). After training, stepping remained non-functional outside the laboratory conditions in five patients; their mean Rivermead score was 3.2 ± 0.4, Garrett 0.4 ± 0.5 and WISCI 0. Two patients could ambulate with support, their scores being: Rivermead 4, Garrett 1 and WISCI 9 (meaning that they could ambulate for 10 m with walker, braces and no physical assistance) (Ditunno et al., 2000, 2001).
In motor-incomplete SCI patients, the mean Rivermead score was 1 ± 1.1 (n = 4), WISCI 0.5 ± 1 and Garrett 0 before training. After training, community walking became possible in three of these patients, their mean scores being: Rivermead 10 ± 4.6, WISCI 19, and Garrett 5.3 ± 0.6. In one patient (SCI-C4), stepping remained non-functional outside the laboratory (WISCI 0).
Patients had a variable degree of spasticity, and no patient was flaccid.
Data analysis
The body was modelled as an interconnected chain of rigid segments: IL-GT for the pelvis, GT-LE for the thigh, LE-LM for the shank and LM-VM for the foot. The elevation angle of each segment in the sagittal plane corresponds to the angle between the projected segment and the vertical. These angles are positive in the forward direction (i.e. when the distal marker is located anterior to the proximal marker). The limb axis was defined as GT-LM. Gait cycle was defined as the time between two successive maxima of the elevation angle of the limb axis. The time of maximum and minimum elevation of the limb axis corresponds to heel-contact and toe-off (stance to swing transition), respectively, in healthy subjects (Bianchi et al., 1998). These time markers were used to identify stance and swing phases. In previous experiments in which a force platform (Kistler 9281B) was used to monitor the contact forces during ground walking, we found that this kinematic criterion predicts the onset and end of stance phase with an error smaller than 2% of the gait cycle duration (Borghese et al., 1996). This observation was confirmed in the present experiments in four control and two SCI-subjects by monitoring in-shoe forces (PEDAR-mobile system, Novel, Germany). The insole contains 99 capacitive sensors interposed between the subject’s foot and the shoe to measure the external vertical contact forces. Before each trial, the mean level of each sensor was measured while the foot was unloaded for a few seconds and this value was used as a zero level. Pressure threshold was 2 N/cm2. We found that the resultant vertical force derived from the pressure sensors went above threshold (corresponding to foot contact) and below threshold (foot take-off) in coincidence with the maximum and minimum elevation of the limb axis, respectively (with a precision of about 2% of the gait cycle). For some gait cycles, notably some of those performed during the first recording session in SCI patients, the swing phase could not be separated reliably from the stance phase. These cycles were excluded from further quantitative analysis. Analyses were carried out on the pooled data of all gait cycles of a given trial. To this end, the data were time-interpolated over individual gait cycles on a time base with 200 points. Averages were constructed over all gait cycles of a trial. Ensemble control averages were constructed by pooling the data recorded under comparable BWS/speed conditions from all healthy subjects. Comparisons between patients and controls were performed for the default condition of 75% BWS, 0.7 km/h (recorded in every session of all patients and controls) and the illustrations refer to this condition except when explicitly indicated. When available, trials at higher speeds and lower BWS were also analysed.
We analysed the trajectory of the distal-most marker (VM) of the foot relative to: (i) a moving intrinsic frame attached to the IL; and (ii) a fixed extrinsic frame attached to the treadmill. In the former case, foot trajectory appears as though the IL was fixed in space. In the latter case, foot trajectory appears as recorded, except that marker’s position was corrected by subtracting the mean horizontal IL position cycle by cycle in order to account for possible drifts of the subject along the treadmill during the recording epoch. All results, except those of Fig. 2, will be illustrated using this latter method.
|
End-point path
The mean area described by foot trajectory was derived by computing the surface of the polygon defined by the x, y coordinates of VM marker (in the fixed extrinsic frame) for each cycle and by taking the average value over all cycles of each trial. In order to compare foot trajectories between patients and controls, we represented each VM trajectory as a time series of vectors and then compared the two resulting vector fields through a correlation measure. The directional correlation coefficient is given by the ratio of the co-variance of the times series and the product of their standard deviations.
End-point variability
Foot-trajectory variability was quantified in terms of instantaneous spatial density and normalized tolerance area of VM (VMA), computed separately over the swing and the stance phase (Ivanenko et al., 2002a). VM trajectories were first re-sampled in the space domain by means of linear interpolation of the x, y time series (1.5-mm steps). Spatial density was calculated for each trial as the number of points falling in 0.5 x 0.5 cm2 cells of the spatial grid divided by the number of gait cycles. The density of each cell was depicted graphically by means of a colour scale (empty cells were excluded in the plot). Normalized tolerance area was derived as follows. The mean length of foot trajectory over the swing phase (or over the stance phase) was calculated over each trial as the corresponding path integral. Every 10% of the horizontal excursion we computed the 2D 95% tolerance ellipsis of the points within the interval. The typical number of points in each interval ranged from 300 to 1000 (depending on the stepping speed and the number of gait cycles). The areas of all tolerance ellipses were summed and normalized by the mean length of foot trajectory. This total area provides an estimate of the mean area covered by the points per 1 cm of path (VMA) and is measured in cm2/cm.
Velocity–curvature power law
To study the velocity–curvature relationship of VM trajectory, all samples corresponding to the swing phase of a trial were pooled together (Ivanenko et al., 2002b). Then we performed a linear regression analysis in log-log scales of equation 
(t) = K·C(t)ß, where (t) and C(t) are the instantaneous values of the angular velocity and path curvature of VM, respectively, K is a velocity gain factor that depends on overall movement duration, and ß is the power exponent. In logarithmic scales, a power function becomes a straight line whose slope corresponds to the exponent.
Inter-segmental coordination
The inter-segmental coordination was evaluated as described previously (Borghese et al., 1996; Bianchi et al., 1998; Grasso et al., 1998). Briefly, the changes of the elevation angles of the thigh, shank and foot co-vary linearly throughout the gait cycle. When these angles are plotted one versus the others in a 3D graph, they describe a gait loop that can be fitted by a plane computed by means of orthogonal linear regression. The 3D orientation of the covariance plane is directly related to the phase-relationship of inter-segmental coordination, and is measured by the plane normal, i.e. the vector orthogonal to the plane (Bianchi et al., 1998). As reference data, we computed the mean normal and its 95% confidence cone from all healthy subjects (Mardia, 1972).
EMG analysis
Raw data were numerically rectified, low-pass filtered with a zero-lag Butterworth filter with cut-off at 15 Hz, time-interpolated over a time base with 200 points for individual gait cycles and averaged. Factor analysis of a subset of these data has been previously reported (Ivanenko et al., 2003).
Spatiotemporal patterns of MN activity in the spinal cord
The recorded patterns of EMG activity were mapped onto the rostrocaudal location of ipsilateral MN pools in the human spinal cord (for a related application to cat locomotion data see Yakovenko et al., 2002). This reconstruction is based on the approximate rostrocaudal location of MN pools innervating different muscles in the human spinal cord based on published charts of segmental localization. In general, each muscle is innervated by several spinal segments. Kendall et al. (1993) compiled reference segmental charts for all body muscles by integrating the anatomical and clinical data of several different sources. A capital X in Kendall’s chart denotes a localization agreed upon by five or more sources, a lower-case x denotes agreement of three to four sources, and an x in brackets (x) denotes agreement of only two sources. In our maps, X and x were weighted 1 and 0.5, respectively, whereas we discarded (x). On the whole, 0.5 weights were 19% of the total. We assumed that our population of subjects has the same spinal topography as this reference population. To reconstruct the output pattern of any given spinal segment, all rectified EMG waveforms corresponding to that segment were averaged and normalized to the maximum during the gait cycle after subtraction of the minimum. For this analysis three sets of ensemble averages were constructed from the pooled data of all controls, ASIA-C/D patients and ASIA-A/B patients.
Statistics
Statistical comparisons between patients and controls were performed at matched values of treadmill speed and BWS, using t-statistics. Analysis of variance designs were used when appropriate to test for the effect of different conditions on locomotor parameters. Reported results are considered significant for P < 0.05. Statistics on correlation coefficients were performed on the normally distributed, Z-transformed values. Spherical statistics of directional data were used to compare the normal to the covariance plane (see above) between patients and controls (Mardia, 1972).
|
Results |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionConclusionsReferences |
|---|
Prior to training, some SCI patients were completely unable to step. However, all patients could be trained to step with BWST in laboratory conditions. The specific strategies used to step differed widely among patients. Most incomplete paraplegics recovered independent control of leg muscles sufficient to propel the limbs in swing and to support body weight in stance. Complete paraplegics instead used their arms and body to assist the leg movements via the biomechanical coupling of limb and body segments. In the following we detail kinematic and EMG changes associated with training in both sets of patients.
Foot kinematics
In those patients who could step in the first recording session prior to training, the spatial path of the foot was highly irregular, jerky and variable from step to step (Fig. 1A). With training, foot path tended to regain gradually the shape typical of normal stepping, with reduced step-by-step variability. The foot traced a loop in the sagittal plane whose extent and regularity was assessed by computing the mean area of VM, the distal-most marker: the larger the surface, the longer is the step length and the higher is the foot clearance from ground, for any given treadmill speed. VM surface increased with training in eight out of 11 patients. On average in this group, VM surface in the last session was significantly higher than in the first session (by 3.37 ± 1.28 times, P < 0.01 paired t-test). In three patients, VM surface did not differ significantly between these two sessions (mean ratio = 0.96 ± 0.04).
|
The time course of foot trajectory was compared between patients and controls by computing the correlation coefficient between the time series of VM position vectors in each patient and the ensemble average in controls. At the end of training, the mean correlation coefficient was 0.98 ± 0.03 across all patients. Foot trajectory also was analysed separately in the vertical direction (foot lift, VMy) and in the horizontal direction (foot translation, VMx). In ASIA-C/D patients, the mean correlation coefficient was 0.94 ± 0.04 and 0.99 ± 0.01 for VMy and VMx, respectively. In ASIA-A/B patients, the mean correlation was 0.74 ± 0.07 and 0.95 ± 0.04 for VMy and VMx, respectively. (These values have been previously reported in Ivanenko et al., 2003
.) The lower correlation in the vertical direction is due to foot-drop in paraplegics. Foot-trajectory variability was quantified in terms of the instantaneous spatial density and the normalized tolerance area of VM (VMA), computed separately over the swing and stance phase (Fig. 1B and C). All patients exhibited a significant reduction of the variability during both phases (P < 0.005, paired t-test between first and last session), although the degree of improvement differed markedly among patients (from 5 to 100% of the first session, on average 76 ± 42%). Changes of performance in patients with training can be contrasted with the stereotypical and stable performance of control subjects. Thus, VMA exhibited limited variability both across subjects (green area in Fig. 1C denotes mean ± SD over all controls), and within subjects. Seven controls were tested several times over a period of 1 to several months between sessions. On average, VMA was 2.3 ± 0.6 cm2/cm over all subjects. VMA was 2.4 ± 0.3 cm2/cm across eight different sessions performed by one subject over a time span of 4.8 years.
At the end of training, foot trajectory also obeyed the two-thirds power relationship between instantaneous curvature and angular velocity that characterizes normal gait (Ivanenko et al., 2002b). Figure 1D shows the time course of VM angular velocity () and curvature (C) in one patient and one control. These variables are widely modulated but they co-vary throughout the gait cycle: increases (decreases) with increasing (decreasing) C. The –C relationship obtained in the patient is plotted in Fig. 1E. In all patients, the correlation was high (r = 0.95 ± 0.03) and the exponent close to the nominal two-thirds value (ß = 0.68 ± 0.05). Neither ß nor r differed significantly from the control values at any tested value of BWS (Fig. 1F).
Patients often restored foot kinematics by implementing new coordinative strategies that involved the trunk in addition to the lower limbs. In Fig. 2 foot trajectories are plotted relative to the extrinsic space (as in Fig. 1) or relative to the intrinsic frame attached to the IL. The former trajectories appear as recorded, except for the correction of subject’s drifts along the treadmill. The latter trajectories, instead, describe the foot motion that would occur without any contribution by trunk and pelvis motion. Therefore extrinsic and intrinsic trajectories almost coincide when subjects step with limited excursion of the trunk and pelvis; this was the case in normal subjects and in less severe SCI patients. By contrast, motor-complete paraplegics stepped with considerable excursion of the pelvis that shifted in synchrony with leg motion (notice the wide excursion of IL in Fig. 2). Thus, the vertical IL-displacement in ASIA-A/B patients was 8.7 ± 2.3 times higher than in controls. As a result, only the extrinsic trajectory of the foot resembled the normal one, whereas the intrinsic trajectory differed substantially. In some cases the trajectory was actually reversed between swing and stance, foot position being higher during stance than during swing (see patient SCI-B2 in Fig. 2). These results indicate that control of foot trajectory in severe SCI patients was not accomplished by means of the normal pattern of coordination of the lower limb segments. This point is taken up in the following section.
Inter-segmental kinematic coordination
In healthy subjects, the main segments of the lower limbs oscillate back and forth with a stereotypical waveform and a progressive phase-shift from the thigh to the shank to the foot (Fig. 3D). This pattern of inter-segmental coordination was altered in SCI patients. Prior to training, angular oscillations were often of small amplitude, especially at the thigh, with considerable step-by-step variability (Fig. 3A–C, left panels). With training, the amplitude increased and the variability decreased, but some important features of the waveform were never restored during our observation period. The minimum value of each segment angle occurred later in the gait cycle than in controls. Moreover, the phase-relationship between limb segments remained abnormal. This can be best appreciated by considering the 3D gait loops obtained by plotting the elevation angles one versus the others (cubes of Fig. 3). The loops evolve close to a plane in both controls and patients indicating a strong linear covariance between the temporal changes of the segment angles. The planar regression accounts for 99.1 ± 0.3% of the variance in controls and 98.4 ± 1.1% in SCI patients. The 3D orientation of the covariance plane (given by the plane normal) measures the phase-relationship of inter-segmental coordination (Bianchi et al., 1998). The plane orientation varies very little among normal subjects (compare the controls of Fig. 3D): the angle of the 95% confidence cone (denoting the angular dispersion) for the mean reference normal is only ±8° over all controls. By contrast, the plane orientation in patients systematically differed from this reference. Thus, at the end of training, it deviated by 54 ± 22° (range 20 ÷ 86°) in ASIA-A/B patients, and by 31 ± 26° (range 15 ÷ 71°) in ASIA-C/D patients.
|
Muscle activity patterns
The extent of modulation of activity of limb and body muscles during the gait cycle increased with training (Fig. 4). The ratio of maximum to minimum of rectified EMG in the last session was significantly higher than in the first session (by 3.24 ± 1.13, P < 0.01). In ASIA-C/D patients, the mean amplitude of activity of leg muscles (TA, GCL, BF, RF, VL, GM) over the gait cycle in the last session did not differ significantly from controls, whereas the mean activity of axial muscles (RAM, RAS, OE, OI, LD, ES, TRAP) was significantly greater than in controls (by 3.19 ± 1.59, P < 0.01). In ASIA-A/B, instead, the mean activity of leg muscles was significantly smaller than in controls (mean ratio = 0.29 ± 0.24, P < 0.001), and the mean activity of axial muscles was significantly greater (4.86 ± 1.87, P < 0.005).
|
In ASIA-C/D patients, activity in ankle extensors (GCL) regained a quasi-normal waveform (Fig. 4C). However, the pattern of activity of most other recorded muscles often remained altered in SCI patients during our observation period. Thus, whereas controls activated reciprocally knee flexors (BF) and extensors (RF, VL) (Fig. 5A), patients co-activated knee flexors and extensors throughout stance (Fig. 5B), or activated knee flexors briskly only in early stance and late swing with little modulation of knee extensors (Fig. 4A and 5C). SCI patients largely relied on proximal and axial muscles to lift the foot and to project the limb forward (Fig. 5). The correlation coefficient between the time series of activation of each muscle in a patient and the corresponding ensemble average in controls varied widely among patients but was generally low (Fig. 6; r = 0.13 ± 0.36, range –0.63 ÷ 0.89).
|
|
In association with the abnormal patterns of leg muscle activity in ASIA-A/B patients, also the time course of changes of the limb joint angles often remained poorly related to that of normal subjects (Fig. 5). The hip normally extends during stance and flexes during swing in healthy subjects (Fig. 5A; Winter, 1991
; Borghese et al., 1996). In motor-complete paraplegics, instead, the hip extended little during stance; it initially extended during swing due to inertial coupling with trunk translation and rotation and then flexed (Fig. 5C). Also, knee flexion in mid-stance was faster and more prolonged than in controls. Finally, the ankle extended in stance later than in controls and flexed less during swing. On average, the correlation coefficient between the time series of joint angles in ASIA-A/B patients and the corresponding ensemble average in controls was –0.47 ± 0.24, 0.82 ± 0.15 and –0.01 ± 0.42 for hip, knee and ankle, respectively (Fig. 6). In ASIA-C/D patients, the correlation coefficients were higher (mean r = 0.90 ± 0.30, 0.96 ± 0.03, and 0.56 ± 0.32 for hip, knee and ankle, respectively). The data of Fig. 6 summarize the trend previously described: at the end of training, the time series of foot position in space is roughly comparable to that of the controls (high correlation coefficients), but the corresponding time series of changes of limb segment angles, joint angles, and EMG activities deviate further and further from those of the controls (low correlation coefficients).
Spatiotemporal patterns of MN activity in the spinal cord
The approximate map of activity of MN pools during locomotion was reconstructed by mapping the recorded EMG waveforms on the published charts of segmental localization. The derived spatiotemporal patterns of MN activity along the rostrocaudal axis of the spinal cord are plotted in Fig. 7. The map is limited to levels between C3 and S2 in relation to the set of recorded muscles. It reflects the relative amplitude of activity in each given spinal segment as a function of the gait cycle, because it is based on averaged, normalized EMG waveforms (see Methods). This map does not provide any information about the absolute amount of activity in the spinal cord.
|
In the lumbosacral spinal cord of healthy subjects, a brief burst of activity occurs just prior to and during heel strike in L1–L4 segments. This burst is associated with EMG activity in hip flexors, knee extensors and ankle dorsiflexors. It is responsible for extending the leg and foot prior to heel strike, and for weight acceptance at the beginning of stance. The focus of activity then shifts to L5–S2 segments resulting in a prolonged burst of activity with a peak in mid-stance. This focus is mainly associated with activity in hip extensors and ankle plantar-flexors, providing support moment and forward thrust. A lower amplitude focus appears at the time of transition between stance and swing, and is responsible for pulling the swinging limb forward. This focus starts with relatively low intensity in more caudal segments of the lumbosacral spinal cord, and then jumps to cranial segments (where it has higher intensity). At cervical and thoracic levels of the spinal cord, a burst of activity occurs in late stance and stance-to-swing transition in T1–T4 segments, and another burst occurs in late swing in C3–C8 segments and in T5–T12 segments. These bursts are related to the trunk stabilization activity of different trunk muscles. Notice, however, that the cervical-thoracic foci would appear of much smaller amplitude than those in the lumbosacral segments on an absolute scale of activity.
The corresponding maps of MN activity are very different in SCI patients. In ASIA-C/D patients, the focus of activity over L2–L4 segments starts later and is much more prolonged than in controls: it begins at foot strike and extends through mid-stance. This focus partially overlaps in time with the following burst in L5–S2 segments, whereas the latter lasts less than in controls. Extensive regions of the spinal cord (C3–L1 and L4–L5) become very active at the transition between stance and swing, contributing to pulling the swinging limb forward.
In ASIA-A/B patients the pattern is still different. There are four very brief, almost impulsive bursts of activity. A first burst centred in L5–S2 segments is responsible for weight acceptance after foot strike and involves the activation of hip extensors and ankle plantar-flexors. A second burst occurs at mid-stance to provide support moment and forward propulsion; it involves C3–C6 and, with smaller relative amplitude, C8–L3 segments. A third burst occurs in extensive regions of the spinal cord (C3–L5) at the transition between stance and swing, as in ASIA-C/D patients. Finally, a relatively smaller burst occurs at late swing, prior to foot strike, and involves L5–S2 segments.
|
Discussion |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionConclusionsReferences |
|---|
There are three main points in this study: (i) locomotor responses in SCI patients mainly depend on learning new motor strategies to replace lost function; (ii) these new strategies are motor equivalents of the normal ones in so far as they produce roughly normal kinematics of the foot; (iii) the reconstructed maps of activity of MN pools show major spatiotemporal changes involving a plastic redistribution of activity across most of the rostrocaudal extent of the spinal cord.
Control hierarchy
The patients could be trained to step with BWST, but they used new coordinative strategies. Patients with more severe lesions stepped with considerable excursion of the pelvis position in synchrony with leg motion. In all patients, the phase-relationship (in some also the waveform) of the angular motion of the different lower limb segments was very different from the control, as was the pattern of activity of most recorded muscles. Surprisingly, however, the new motor strategies were quite effective in generating foot motion that closely matched the normal in the laboratory conditions. With training, foot motion of SCI patients tended to regain the shape, the step-by-step reproducibility, and the two-thirds power relationship between curvature and velocity that characterize normal gait (Ivanenko et al., 2002a, b). A correlation analysis between the gait waveforms of the patients and those of the controls yielded the following ranking (from high to low correlation): foot position, limb segment angles, joint angles and EMG patterns (Fig. 6). This ranking is congruent with current ideas on the hierarchy of control in locomotion (Lacquaniti et al., 1999, 2002; Poppele and Bosco, 2003). A control hierarchy is defined operationally by the extent to which different gait parameters vary under different walking conditions: parameters that vary the least are placed at the highest control level, whereas those varying the most are placed at the lowest level. Healthy subjects accurately regulate foot kinematics across wide changes of body load and stepping speed (Winter, 1991; Ivanenko et al., 2002a, b). At the following level of control, the limb segment angles are not specified independently of each other, but they co-vary in time on a loop confined close to a plane (Borghese et al., 1996; Bianchi et al., 1998). In contrast with foot kinematics, the timing of segment angle kinematics does change with body load and stepping speed (Ivanenko et al., 2002a). At a still lower level of control, the temporal patterns of muscle activity vary the most across loads and speeds, adapting to the varying biomechanical requirements (Winter 1991; Ivanenko et al., 2002a).
Motor equivalence
The control hierarchy obeys the principle of motor equivalence stating that an invariant task goal can be achieved with variable means (Lashley, 1933; Hebb, 1949; Lacquaniti, 1989). Thus, our handwriting is recognizable regardless of whether the pen is held between the fingers, the toes or the teeth. Lashley (1933) introduced the concept of motor equivalence in the context of lesion studies by showing that monkeys with motor cortex lesions were still capable of opening a box despite the paresis. A lesioned nervous system might take advantage of the natural redundancy in the neuromuscular system to accomplish a given motor goal at the limb end-point (hand or foot) by means of new compensatory muscle synergies (Winter, 1991; Kazennikov et al., 1998; Cirstea and Levin, 2000). To our knowledge, the present study is the first to show quantitatively that trained SCI patients use motor equivalence. Patients learn to produce new temporally tuned patterns of muscle activity, resulting in the desired kinematics of the foot via the biomechanical coupling of the angular motion of different limb and body segments (Bianchi et al., 1998; Lacquaniti et al., 1999). Indeed, patients used extensively their arm and/or axial muscles to assist the swing phase (Fig. 5). It has been noticed that upper extremity paralysis has a restrictive effect on independent ambulation (Maegele et al., 2002). Coupled angular motions also generate sensory stimulation that can entrain both supra- and sublesional segments of the cord and result in appropriately patterned activity of muscles (Pearson, 2001). The specific procedures used to train SCI patients might be instrumental in leading to restoration of foot kinematics. Thus, physiotherapists (or robotic orthoses; see de Leon et al., 2002; Dietz et al., 2002) act as external teachers by minimizing the output error of foot trajectory relative to a predefined template. Patients might then try to reproduce the learnt template during unassisted stepping. This procedure is equivalent to supervised learning in recurrent networks (Doya, 2003).
Locomotor pattern generation
We computed the spatiotemporal maps of spinal MN activation (Fig. 7) by combining two data sets: (i) averaged, rectified EMG waveforms were derived from the simultaneous recordings of EMG activity of several limb and trunk muscles during many step cycles; (ii) the approximate rostrocaudal location of MN pools innervating the corresponding muscles was derived from published charts of segmental localization. First we discuss methodological issues.
Locomotor pattern generators output command signals directed to MN pools. Each action potential in a MN propagates along the efferent axon and gives rise to a motor unit action potential in the innervated muscle. All motor unit action potentials generated by all active MNs sum to produce the recordable EMG signal. The rectified EMG then provides an indirect measure of the net firing of MNs of that muscle in the spinal cord at any given moment during locomotion. The exact quantitative relationship between the net motor unit action potential rate and the amplitude of EMG waveforms cannot be established uniquely. However, this was not a serious drawback for the current application. Because we were comparing subjects with very different levels of muscle activity, our interest was in the temporal pattern of relative activation at a given segmental level, rather than in the absolute intensity of the signal. Therefore averaged EMG waveforms were normalized to the maximum during the gait cycle. Finally, the maps were constructed based on a large but incomplete sample of muscles. In particular, no foot muscle was recorded. Further studies will be needed to fill in the gaps.
The rostrocaudal location of MN pools was derived from the charts of spinal segmental localization complied by Kendall et al. (1993), under the assumption that our population of subjects has the same spinal topography. Kendall et al. (1993) compiled reference segmental charts for all body muscles by integrating the anatomical and clinical data of several different sources. Functional MR imaging of the human cervical spinal cord has confirmed so far the anatomical localization of published segments (Stroman et al., 2001). Despite likely anatomical variability, the data of these charts appear sufficiently robust for the spatial resolution currently available in our reconstruction technique.
As this is the first study to report spatiotemporal maps of spinal MN activation in man, we can only compare the results between our different groups of subjects, and with the map obtained by Yakovenko et al. (2002) for the cat lumbosacral spinal cord. They constructed the map from anatomical data on MN localization obtained by Vanderhorst and Holstege (1997) and from a compilation of published records of EMG activity during locomotion of intact cats. Interestingly, the lumbosacral map we obtained in healthy subjects roughly agrees with that of Yakovenko et al. (2002), taking into account the inter-species anatomical differences. In both sets of maps, the focus of activity oscillates rostrocaudally during the gait cycle. Rostral and caudal parts of the lumbosacral enlargement are active during swing and stance, respectively, and activity jumps from one region to the other at the transition times. The foci of activity could correspond to waves of activation propagating back and forth along the spinal cord or to abrupt switching between distinct burst generators (Kiehn et al., 1998; Orlovsky et al., 1999; Yakovenko et al., 2002). We have also been able to detect patterned activity at cervical and thoracic levels, presumably related to trunk stabilization in different phases of the gait cycle. This finding cannot be compared with Yakovenko et al. (2002), as they did not investigate trunk muscles.
The corresponding spatiotemporal maps of MN activity are very different in SCI patients. The main foci of activity in lumbosacral cord seen in healthy subjects are also present in motor-incomplete paraplegics. However, activity switches between the foci in the former, whereas the foci are co-active for extended periods of stance in the latter. The map in motor-complete paraplegics departs even more radically from the control map. It shows regions of complete silence or low-amplitude activity (black or blue in Fig. 7) much more extensive than in the other groups of subjects. Silence is interrupted by impulsive bursts of activity appearing sparsely during the gait cycle, with a location and timing quite different from the normal. In both motor-complete and motor-incomplete paraplegics, extensive regions of the spinal cord including cervical, thoracic and lumbar segments are briskly active at the stance-to-swing transition.
The term central pattern generator (CPG) designates a spinal network that can generate patterns of rhythmic activity for locomotion even in the absence of external feedback or supraspinal control (Grillner and Wallen, 1985). Normally, however, the spinal network is modulated by peripheral and supraspinal inputs (Orlovsky et al., 1999). The exact organization of CPGs is still largely unknown in mammals (Kiehn et al., 1998; Barbeau et al., 1999b). They are thought to comprise a network of inter-neurons linked to the output stage of 
-motor neurons, but opinions diverge as to whether the vertebrate CPGs are localized or distributed (Kiehn et al., 1998; Orlovsky et al., 1999; Yakovenko et al., 2002). The present data do not provide any information about the activity of inter-neurons, but they show the organization of the output stage. The network of -motor neurons actively oscillating during the step cycle appears widely spaced over extensive regions of the spinal cord. This implies that individual burst generators are coupled by long propriospinal neurons projecting to different pools of -motor neurons across several spinal segments (Nathan et al., 1996).
Distributed plasticity
We showed that, after spinal lesion, the locomotor network can reorganize to an extent not previously reported. The reorganization involved all investigated segments, both supralesional and sublesional ones, extending from the cervical to the sacral cord. Lesion- and training-induced plasticity might be responsible for changes in the connections of the network. These changes are probably adaptive and learnt (being specific to the trained task; de Leon et al., 1998) and involve a major redistribution of activity to different limb and body muscles (Pearson, 2001), creating new muscle synergies (Barbeau et al., 1999b). The specific re-organization of the network might also depend on the level of the lesion (Dietz et al., 1999), but here the limited number of patients did not allow a correlation between the lesion level and the motor patterns.
Spinal lesions probably trigger multiple forms of plasticity. Synaptic strength could be modified in pre-existing circuits (synaptic plasticity), and new circuits might develop through sprouting and anatomical reorganization, including growth of axonal branches and dendrites (anatomical plasticity; Raineteau and Schwab, 2001). Evidence for plastic reorganization caudal to the level of injury has been recently provided by Calancie et al. (2002). They demonstrated novel upper limb reflexes evoked by lower limb stimulation, emerging more than 6 months after a high cervical spinal cord lesion. In addition to plasticity of intrinsic spinal networks, plasticity of unlesioned descending pathways can also contribute (Giszter et al., 1998). Beneficial plasticity often involves undamaged neural areas that may take over the function of damaged ones (Raineteau and Schwab, 2001). This is the case for plasticity induced by sensory stimuli at the cortical level after cerebral injury (Fraser et al., 2002). In the case of spinal lesion, the CNS should be capable of substantial reorganization because cortical, sub-cortical and much of the intrinsic spinal cord circuitry remain largely intact and still partially interconnected by unlesioned fibres. Cortical reorganization in SCI patients may result in enhanced excitability of motor pathways targeting muscles rostral to the level of a spinal lesion, reflecting reorganization of motor pathways either within cortical motor areas or at the level of the spinal cord (Topka et al., 1991; Dobkin, 2000; Curt et al., 2002). In particular, it can be hypothesized that stepping after a severe spinal lesion depends on cortical (and voluntary) control much more heavily than it does in healthy subjects (where locomotion is more automatic). The cortical motor areas, that encode distal leg movements and become disconnected from their target pools of motor neurons in the spinal cord after spinal injury, might re-direct their command signals to the adjacent cortical motor areas that control more proximal body segments. Thus, the plastic reorganization of pattern generation in the spinal cord we demonstrated might mirror a similar reorganization of the control centres in the motor cortex.
|
Conclusions |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionConclusionsReferences |
|---|
We argued that locomotor improvement in SCI patients may not be subserved by changes localized to limited regions of the spinal cord, but may depend on a plastic redistribution of activity across most of the rostrocaudal extent of the spinal cord. Distributed plasticity underlies recovery of foot kinematics by generating new patterns of muscle activity that are motor equivalent of the normal ones. The locomotor programmes encrypted in the reorganized networks allowed functional recovery of unsupported gait in most incomplete paraplegics, whereas they remained non-functional in most complete paraplegics outside laboratory conditions, as they could not walk without body support. Lack of functional recovery in these patients is due to, among other factors, the low level of activity in leg muscles and lack of adequate postural control. However, the demonstration of extensive distributed plasticity may prove relevant also for future rehabilitation of the latter patients.
|
Acknowledgements |
|---|
We dedicate this paper to the memory of Dr Renato Grasso who devoted his best energies to the success of the project. We thank the therapists D. Angelini, B. Morganti and M. Piccioni for training the patients, Dr L. Ercolani and D. Prissinotti for help with experiments, Dr J. F. Ditunno and Dr J. Fung for advice on the project. The financial support of Italian Health Ministry, Italian University Ministry (MIUR), Italian Space Agency (ASI), and C.N.R. (Progetto Strategico Neuroscienze) is gratefully acknowledg