Age-Related Modifications of Muscle Synergies and Spinal Cord Activity During Locomotion
Abstract
Recent findings have shown that neural circuits located in the spinal cord drive muscular activations during locomotion while intermediating between descending signals and peripheral sensory information. This relationship could be modified by the natural aging process. To address this issue, the activity of 12 ipsilateral leg muscles was analyzed in young and elderly people (7 subjects per group) while walking at six different cadences (40–140 steps/min). These signals were used to extract synergies underlying muscle activation and to map the motoneuronal activity of the pools belonging to the lumbosacral enlargement (L2–S2). The comparison between the two groups showed that neither temporal patterning of motor primitives nor muscles loading synergies seemed to be significantly affected by aging. Conversely, as the cadence increased, spinal maps differ significantly between the groups, showing higher and scattered activity during the whole gait cycle in elders and well-defined bursts in young subjects. The results suggested that motor primitives lead the synchronization of muscle activation mainly depending on the biomechanical demand of the locomotion; hence they are not significantly affected by aging. Nevertheless, at the spinal cord level, biomechanical requirements, peripheral afference, and descending inputs are differently integrated between the two groups, probably reflecting age-related changes of both nervous system and motor control strategies during locomotion.
INTRODUCTION
As widely described in literature, aging involves modifications of gait patterns and muscle activations in healthy subjects (Judge et al. 1996; Kerrigan et al. 1998; Schmitz et al. 2008; Winter 1991). From a biomechanical viewpoint, aging is supposed to mainly affect the ability of calf muscles to produce rapid force (Thelen et al. 1996). This modification could decrease their propulsive action and consequently reduce their walking speed (Judge et al. 1996; Kerrigan et al. 1998; Winter et al. 1990). Concerning muscle activity, some authors observed that elderly people adopt intense coactivation of leg muscles, inducing greater stiffening of leg joints. They also tend to generate highly repeatable EMG signals, which is supposed to reflect that aging involves lack of neural plasticity (Schmitz et al. 2008; Winter 1991). Moreover, aging of the neuromuscular control skills does not allow the fine tuning of stride-to-stride fluctuations, thus increasing fall risks in elderly people (Hausdorff 2005). On the whole, age-related muscle weakness and poor neural control negatively affect human locomotion, reducing the performance of the elderly while walking.
Some authors have recently hypothesized that neuromuscular adaptation could take place based on the idea that the functional reformation of gait patterns occur in conjunction with aging, reduced physical activity, or pathological aging (DeVita and Hortobagyi 2000; McGibbon 2003). For instance, because of the declined performance of calf muscles in elders, kinetics would be functionally redistributed among leg joints because both proximal and distal leg extensors contribute to the same functional tasks (Burnfield et al. 2000; DeVita and Hortobagyi 2000; Monaco et al. 2009). Neuromuscular adaptation may also be led by pathologies such that elderly people affected by knee osteoarthritis showed increased hip extension during the late stance to gain advantage from the elastic proprieties of hip flexor muscles when pulling the thigh forward for the swing phase (McGibbon 2003; McGibbon et al. 2001).
All these findings seem to show that gait features in elderly people represent the results of the ontogeny of locomotion where the neuromuscular adaptation could have a key role in compensating the age-related modifications of the neuro-musculo-skeletal system. These hypotheses have been mainly argued by analyzing kinetics at leg joints during locomotion, which describe the entire effect of all muscles crossing each leg joint. However, there is no evidence concerning the role of the integration of motor programs and peripheral feedback on lower limb control during locomotion in elderly people. In particular, it is not clear whether aging influences the synergistic control of muscle activations and whether this is reflected in the activities of the CNS.
Previous experiments have suggested that, during locomotion, neural circuits located in the spinal cord drive muscular activations while intermediating between descending signals and peripheral sensory information (Bizzi et al. 2000; Ivanenko et al. 2006a; Patla 1985). Recent findings have shown that, during human walking, the activity of several muscles belonging to leg, trunk, and neck could be decomposed into several components (Davis and Vaughan 1993; Ivanenko et al. 2004; Patla 1985; Shiavi and Griffin 1981), also named “primitive signals,” which are generated by central structures independently from the peripheral feedback and represent the basic central features of the motor programs (Patla 1985).
The goal of this study was to verify whether the aging process could modify motor primitives and related spinal cord activities, which reflect the integration of descending signals with peripheral afference in the CNS. We hypothesized that aging may involve modifications of the motoneuronal (MN) activity even though the main features of muscular activation would remain similar in both young and elders. The rationale behind this hypothesis is that, despite slight differences in the kinetics and kinematics of locomotion, the biomechanical demand of walking in young and elderly people is similar. Therefore it is reasonable to believe that motor programs do not significantly depend on aging, as already shown in a previous and preliminary study (Monaco et al. 2008). On the other hand, CNS activity, in particular spinal cord activity, may differently integrate descending signals and peripheral afference, reflecting some aspects of the age-related neural plasticity.
METHODS
Subjects
Seven young (4 males and 3 females; 27.3 ± 4.9 yr old; height, 1.78 ± 0.02 m; weight, 75.0 ± 7.7 kg) and seven elderly (4 males and 3 females; 69.0 ± 1.4 yr old; height, 1.71 ± 0.05 m; weight, 77.0 ± 1.4 kg) healthy subjects, without any evidence or known history of postural, skeletal, or neurological diseases, were enrolled for the study after signing informed consent. Protocol was designed in accordance with the Local Ethical Committee.
Experimental setup
Experiments were carried out in a 15-m-long room where subjects walked over ground. EMG data of 12 leg muscles belonging to the right leg [peroneus longus (PERL), gastrocnemius lateralis (GL), soleus (SOL), rectus femoris (RF), vastus medialis (VM), tibialis anterior (TA), biceps femoris (BF), semitendinosus (ST), adductor longus (ADD), tensor fascia latae (TFL), gluteus maximus (GM), gluteus medius (Gmed)] were recorded using surface EMG electrodes (NORAXON, Telemyo 2400T, V2) placed on previously shaved skin. Sampling rate was 1,000 Hz, and gain of the amplifier was 1,000. Electrode placements were tested through suitable movements to verify cross-talk among muscle signals and bad contacts. Subjects donned shoes provided with foot switches to record heel strike and toe off related to the right leg.
Protocol
Subjects were asked to walk at the beat of a metronome set at the following cadences (steps/min): 40, 60, 80, 100, 120, and 140. Cadence was adopted as a metric to make data comparable between the groups, because the elderly usually walk with faster step frequency than young people, even if at the same speed (DeVita and Hortobagyi 2000; Silder et al. 2008). Therefore muscle activation was not affected by any different temporal recruitment across groups. Trials were carried out in a randomized order.
Preprocessing
For each subject and trial, EMG data concerning the first and the last three ipsilateral (right side) strides were discarded in both the young and in elderly people (because of the transitory periods), in agreement with the previous literature (Bus and de Lange 2005; Mbourou et al. 2003; Miller and Verstraete 1996; Sparrow and Tirosh 2005; Wearing et al. 1999). From the remaining strides (between 4 and 12, depending on the speed), EMG data were full rectified, low-pass filtered (Butterworth, 15 Hz, 4th order), averaged over one gait cycle, and time-interpolated over 200 points (Ivanenko et al. 2004). This preprocessing provided a 12 × 200-sized matrix for each subject and trial, in which rows and columns were, respectively, related to muscles and fraction of the gait cycle. CVs of preprocessed EMG signals were estimated according to the literature (Winter 1991) and compared between groups at all cadences via t-test (P < 0.05).
Factorization
After preprocessing, Bartlett's test of sphericity was carried out, and the Kaiser-Meyer-Olkin (KMO) measure was calculated to verify whether each matrix was adequate for factorization (Ivanenko et al. 2004). All data sets were factorized using factor analysis with varimax rotation (FA) and, in each case, five muscle synergies were retained according to the results shown by Ivanenko et al. (2004). Each synergy represents the coordinated activity of a subset of the 12 muscles and is described by factors or primitive signals, which are the basic activation patterns underlying muscle activity, and a weight coefficient matrix highlighting which muscles are involved in each synergy. Actually, the number of synergies and related factors extracted from a data set needs to be carefully chosen because it represents the threshold between capturing random fluctuations and systematic behavior (Cheung et al. 2005). Usually, the authors who used multivariate statistic tools, such as principal component analysis or factor analysis, adopted the criteria of the eigenvalue > 1 or a threshold on the cumulative variance explained by the extracted factors (Davis and Vaughan 1993; Ivanenko et al. 2004; Merkle et al. 1998; Olree and Vaughan 1995). Nevertheless, in a recent study, Ivanenko et al. (2004) showed that 5 basic muscle activation patterns significantly described the variability of a set of 12–18 ipsilateral muscles recorded from the leg, trunk, and neck. Although one of the main differences between the mentioned study and this work is a different set of accounted muscles, we decided to extract five factors to reproduce results already available in the literature to verify whether, in the framework of our protocol, our results were comparable with those shown by previous authors. We discuss the consistency of the choice concerning the number of extracted factors after data analysis.
Synergies grouping
Previous studies have shown that synergies extracted by means of multivariate statistic tools were characterized by common features, in terms of shape of the factors and muscles loading each primitive signal, which allowed the authors to group them over several subjects. In this study, for each cadence and within each group, synergies were grouped across subjects after estimating the best-matching scalar products of their related weight coefficients to guarantee the best similarity also for those which would have shown greater variability (Cheung et al. 2005). The averaged weight coefficients related to each synergy were calculated for both groups and matched between young and elderly people by means of the best scalar product between them. The ordering process was necessary to enable comparison of both factors and weight coefficients between groups and among walking speeds.
Effects of cadence and aging on synergies
To assess the effect of cadence and aging on muscle synergies, three metrics have been adopted: the scalar product of the weight coefficient vectors after normalizing with respect to their own norms (Cheung et al. 2005); the Pearson correlation coefficient (r); and the phase lag between factors, which represents the temporal offset between them (Ivanenko et al. 2004). This last parameter has been calculated for each cadence after averaging primitive signals across all subjects within each group (Ivanenko et al. 2004).
The effect of cadence was evaluated, for each subject, by comparing muscle synergies extracted at all speeds with those extracted at 100 steps/min, which is close to the step frequency during walking at self-selected speed in both groups (Silder et al. 2008; Winter 1991). The effect of aging was evaluated comparing muscle synergies of all young subjects with those of all the elderly, for each walking speed.
Descriptive statistics included means and SD.
Spatiotemporal patterns of MN activity in the spinal cord
EMG signals were used to map the MN activity related to the lumbosacral enlargement (segments L2–S2), approximated along the rostrocaudal direction. This methodology has been recently proposed by several authors who estimated MN activity in cats (Yakovenko et al. 2002), spinal cord–injured patients (Grasso et al. 2004), and healthy subjects during locomotion with different gaits (Cappellini et al. 2006, 2010; Ivanenko et al. 2006b, 2008). Briefly, because each spinal segment innervates several muscles, MN activity can be estimated by a weighted summation of the EMG signals of all muscles innervated from that spinal segment. The weight coefficients, reported in Kendall's chart (Ivanenko et al. 2006b), are 0.5 or 1, depending on the number of sources mentioning the anatomical innervations of each muscle.
The main assumption underlying such representation is that EMG signals, after full-wave rectifying and filtering, can provide an indirect measure of the net MN firing rate. This hypothesis has been assumed in accordance with earlier studies showing that EMG activity increases fairly linearly with the net motor unit firing rate (Day and Hulliger 2001; Zhou and Rymer 2004). Moreover, Masakado et al. (1994) showed that motor unit action potential is linearly correlated with macro-EMG both in young and in elderly people, such that the assumption can be considered reasonable also for the elderly, as already assumed in literature (Grasso et al. 2004).
To compare spinal maps between groups and among walking speeds, we adopted the two-dimensional (2D) correlation coefficient (r, see Matlab function corr2.m) that is an alternative representation of the Pearson correlation coefficient suitable for the comparison between two images (James 1988). In particular, for digital images in which the color represents the scale of the observed variable, the Pearson correlation coefficient is defined as
For each subject, the effect of the cadence was evaluated by comparing MN activity estimated at all cadences with that related to 100 steps/min. Furthermore, at all cadences, MN activity of each young subject was compared with the MN activity of every elderly subject.
Statistical significance of the 2D correlation coefficient, which was expected to be positive because of the features of the maps, was verified by means of the tables reporting critical values of r with respect to the degrees of freedom of each comparison. The goodness of the similarity between two maps was assessed with respect of the proportion of variance (r2) shared between the two variables, such that r < 0.45, that is, r2 < 0.2 corresponded to poor correlation; 0.45 ≤ r < 0.7, that is, 0.2 ≤ r2 < 0.5, corresponded to fairly to good correlation; and r ≥ 0.7, that is, r2 ≥ 0.5, corresponded to good correlation.
Within each group, one-way ANOVA was used to verify whether the similarity between maps extracted at all speeds and those at reference pace were statistically comparable across cadences. Moreover, the trend of the similarity between maps related to young and elderly people versus the speed was analyzed by means of the Pearson correlation coefficient (r). Significance was set at α = 0.05.
Preprocessing, FA, MN activity reconstruction, and statistics were carried out by using custom written MATLAB (The MathWorks, Natick, MA) scripts.
RESULTS
Gait parameters and EMG signals
Subjects involved in the study had comparable anthropometric features, although the young were slightly taller than the elderly. They all kept their cadence close to the beat set by the metronome (Fig. 1B) without any significant difference between the groups (P > 0.05). In accordance with the literature (Silder et al. 2008; Winter 1991), stance time decreased when the cadence increased, and in the elderly, it was significantly longer than in young subjects (between 4.3 and 7.2% of the gait cycle; Fig. 1B).
Fig. 1.Main features of EMG signals and spatio-temporal parameters. A: representative raw EMG patterns of an elderly and a young subject, while walking at 100 steps/min. Vertical grid shows the instants in which the heel strikes the ground. Data adopted for the analysis after discarding the 1st and the last 3 strides are reported in dark gray over the whole record (light gray). B: averaged and SD of the stance time expressed as percentage of the gait cycle (top) and cadence (bottom) for young and elderly people. SD is represented as a 1-side error bar to make reading easy. C: representative EMG patterns, after full wave rectification, filtering, resampling, and averaging over the accounted strides of an elderly and a young subject walking at all cadences. Horizontal axis refers to the percentage of the gait cycle. D: average and SD of the cumulative variance explained by the extracted factors. Note that extracted factors in this figure are not yet ordered.
EMG signals related to the selected steps (Fig. 1A) were not affected by the transitory phases (start and stop of the trials), and raw EMG patterns were generally in agreement with previous authors (Shiavi et al. 1987a,b; Winter 1991). Moreover, at slower cadences (40, 60, and 80 steps/min), some muscles, particularly those crossing the hip joint,were quite silent (Fig. 1C), as also reported by previous authors (Cappellini et al. 2006; Ivanenko et al. 2004, 2006b; Shiavi et al. 1987a).
Because of aging, EMG data were characterized by different features concerning both amplitude and variability, such that CVs related to the myoelectric signals of young subjects were, almost in all cases, higher than in the elderly, even if not significantly. The few cases in which CVs were statistically different between groups (P < 0.05) occurred at faster cadences, and, in particular, they concerned RF, ST, and Gmed at 100 steps/min, GL and ADD at 120 steps/min, and GL and Gmed at 140 steps/min.
Factor analysis
Data were adequate for factorizing (sphericity P < 0.001; KMO > 0.55), and five synergies were extracted according to Ivanenko et al. (2004), accounting for >90% of the cumulative variance (Fig. 1D). However, we usually observed that three synergies were already in accordance with the criteria of the eigenvalue > 1, and they accounted for >75% of the variability of data sets (Fig. 1D).
Factors and weight coefficients related to synergies extracted from both groups were comparable with those shown in previous works at all cadences (Cappellini et al. 2006; Davis and Vaughan 1993; Ivanenko et al. 2004; Merkle et al. 1998; Olree and Vaughan 1995). Nevertheless, as expected, extracted factors did not seem to be ordered, given the position of their own main peak along the gait cycle. Consequently, synergies were not directly comparable between the groups and within the walking speeds. After groupings based on the intra- and intergroup similarities of the synergies, both factors and weight coefficients seemed comparable between young and elderly subjects at all cadences (see Fig. 2 for data related to 100 steps/min).
Fig. 2.Factors (A) and weight coefficients (B), after ordering, related to all subjects while walking at 100 steps/min. A: factors of young (light gray) and elderly (dark gray) subjects are represented with respect to the gait cycle. PF, propulsion factor; LRF, loading response factor; HF, heel strike factor. F4 and F5 are the less significant primitive signals. For both groups, the averaged factor for each synergy is represented in black. B: weight coefficients referring to the 5 synergies. Similarly as for factors, PW, LRW, and HW represent the weight coefficients related to the 1st 3 synergies, wheras W4 and W5 represent the weight coefficients for the less significant ones.
Three of the five factors were always similar to those named by Davis and Vaughan (1993) as “loading response factor” (LRF), “propulsion factor” (PF), and “heel strike factor” (HF; Fig. 2). Because of their systematic behavior, LRFs, PFs, and HFs were usually extracted as first factors, accounting on average for 75% of the variability of the data sets. The remaining factors, herein called F4 and F5, seemed to capture less systematic information (∼15% of the explained variance) and were generally characterized by variable shape above all at slower cadences. However, it was possible to notice that, despite the wide intrasubject variability of F4 and F5, their related weight coefficients, W4 and W5, quite consistently accounted for the same muscle groups above all at faster speeds (Fig. 2).
Despite certain variability across subjects and speeds, in both groups, it was possible to observe a systematic distribution of the muscles loading each synergy (Fig. 2). In particular, PF was mainly loaded by calf muscles (PERL, LG, and SOL); LRF was mainly loaded by the knee extensor muscles (RF and VM) and, although not very systematically, by the monoarticular hip extensors (GM and Gmed); HF was mainly loaded by the biarticular hip extensor muscles (ST and BF) and sometimes by TA; F4 was mainly loaded by ADD and TA even if it could be also loaded by glutei and knee extensors; and F5 was mainly loaded by glutei (GM and Gmed) and TFL, although sometimes it was characterized by the activity of knee extensors (VM and RF).
Effects of cadence on synergies
In both groups, LRF and PF at faster cadences (from 80 to 140 steps/min) were well correlated (r > 0.7 for PF and r > 0.63 for LRF) to those related to the reference speed (100 steps/min), whereas HF showed lower correlation in the young (0.4 < r < 0.6) and in the elderly (0.6 < r < 0.7). Conversely, F4 and F5 in both groups showed poor correlation (r < 0.4) and greater variability at all cadences. All factors extracted at 40 steps/min correlated poorly to those extracted at the reference speed (100 steps/min), apart from PF in young subjects.
Factors extracted at slower cadences (40 and 60 steps/min) showed variable phase lags when they were compared with those extracted at 100 steps/min. At all other paces, at least for the first three factors, phase lags appeared to be reduced to zero with increasing speed. Moreover, the phase lag related to HF became close to zero in both groups, even though it was more variable in elderly subjects.
Regarding weight coefficients, PW, LRW, and HW at almost all speeds showed good similarity with those at reference cadence (scalar product > 0.7). The similarity between muscles loading W4 and W5 extracted at all paces with that at the reference velocity increased from 0.4 to 0.8 at faster cadences (80, 120, and 140 steps/min). At slowest cadences, similarity dropped or did not show any systematic trends for both these synergies.
Effects of aging on synergies
Figure 3 shows all results concerning the comparisons between synergies of young and elderly people at all cadences.
Fig. 3.Effects of aging on synergies. A: the similarity, expressed by the scalar product, between the weight coefficients of young and elderly people. As the cadence increases, the synergies occurring during the propulsive and the heel strike phases became more similar between groups. W4 and W5 were poorly correlated between the groups. B: the averaged CVs of factors for both groups, thin lines for elderly and thick lines for young, at all cadences showing that factor variability between groups was comparable, above all for the 1st 3 synergies. C: the correlation between the factors related to young and elderly subjects. At faster cadences (from 100 to 140 steps/min), averaged factors related to the 1st 3 synergies were similar between groups, whereas at the slower pace, only PF showed relevant correlation between groups. D: the phase lag between averaged factors of young and elderly subjects. Concerning the 1st 3 synergies, at faster cadences (from 80 to 140 steps/min), factors related to both groups were not affected by significant time shift, whereas at the slowest pace, elderly factors were slightly delayed with respect to the young ones. F4 and F5 did not show any consistent behavior because of their intersubject and intergroup variability.
Concerning factor variability (Fig. 3B), CVs were similar between groups at all cadences, especially when referring to the first three primitive signals. For these factors, CVs tended to increase and their explained variance decreased in both groups. CVs related to F4 and F5 did not show any systematic behavior across speeds (Fig. 3B).
PF, LRF, and HF averaged across all subjects did not show any significant time shift between the groups at faster cadences (from 80 to 140 steps/min), whereas at slower speeds, in elderly subjects, these were slightly delayed with respect to the young subjects (Fig. 3D). At slowest paces, the phase lag between the young and elderly for these factors was within 15% of the gait cycle and was affected by significant variability. F4 and F5 did not behave in a significant fashion because they were characterized by an inconsistent phase shift between the groups, wide variability, and no systematic trend.
Correlations of PF, LRF, and HF between the groups were significantly higher (r > 0.7) at faster cadences, whereas when cadence decreased, primitive signals related to the young and elderly were less correlated with each other (Fig. 3C). F4 and F5 never showed good correlation between the groups (r < 0.7), but at faster cadences, r assumed higher values (0.45 < r < 0.7).
The weight coefficients showing greater similarity between young and elderly people were those related to the first three synergies (PW, LRW, and HW), especially at faster cadences (from 80 to 140 steps/min; Fig. 3A). Conversely, weight coefficients related to the last muscle synergies showed low similarity between the groups.
Spatiotemporal patterns of MN activity
Figures 4 and 5 show the representative MN activity of a young and an elderly subject, respectively, while walking at all cadences and the comparisons carried out between groups and among the walking speeds.
Fig. 4.MN activity of 2 representative subjects while walking at all cadences. The color scale refers to millivolts of activity.
Fig. 5.Average and SD (positive error band) of the 2-dimensional (2D) correlation coefficients obtained by comparing maps estimated at all speeds with those referring to the reference pace for both groups (A and B); average and SD (positive error band) of the 2D correlation coefficients obtained by comparing MN maps between young and elderly subjects for each cadence (C). In A and B, the P value concerning the ANOVA of the 2D correlation coefficients across the speeds is reported for both groups. In C, the linear regression fitting the 2D correlation coefficients (red line) and its significance are reported.
In young subjects at the slower cadences (40 and 60 steps/min), MN activity was mainly distributed along the whole stance time in the most caudal sections of the spinal cord. Nevertheless, bursts could also occur during the late swing at rostral levels. As the cadence increased, peaks occurred in three main periods, at ∼10, 45, and 95% of the gait cycle. Bursts occurring during the early stance and the late swing phases were distributed on the whole lumbosacral enlargement with foci in the rostral sections, whereas the activity recorded at 45% of the gait cycle mainly involved the most caudal segments. At faster cadences (from 100 to 140 steps/min), a further peak of activity occurred at ∼60% of the gait cycle, mainly in the most rostral sections, even if it was characterized by lower amplitude (Fig. 4).
In elderly people, bursts of MN activity occurred likely than in young even though they were characterized by higher amplitude, and, at all cadences, they were more spread. As the cadence increased, the main bursts of activity became more distinguishable (Fig. 4).
According to previous results (Cappellini et al. 2006; Grasso et al. 2004; Ivanenko et al. 2006a,b), bursts of MN activity mainly occurred in conjunction with the main peaks related to LRF, PF, and HF along the gait cycle and appeared to be better defined with increasing speed (Figs. 2 and 4). Although F4 and F5 were not characterized by consistent behavior, spinal maps highlighted that, at faster speeds, another burst of MN activity systematically occurred during the preswing phase across sections L4–L5 in elderly people and in the rostral section in young subjects. This burst was characterized by lower amplitude than the others in both groups (Fig. 4).
All the comparisons between spinal maps were characterized by a 2D correlation coefficient significantly different from zero (P < 0.05).
Regarding the comparisons between MN activity for each subject over the cadences, results (Fig. 5, A and B) showed that, at slower speeds (40 and 60 steps/min), spinal activity was less correlated (on average, 0.52 < r < 0.64) with that at reference pace (100 steps/min) in both groups, whereas when the step frequency increased (from 80 to 140 steps/min), the similarity between the maps was greater (on average, 0.67 < r < 0.75). In this regard, the one-way ANOVA highlighted that mean r differed significantly across cadences only in young subjects (P = 0.04 for the young; P = 0.49 for the elderly; see Fig. 5, A and B).
On the other hand, when MN patterning referring to young subjects was compared with that of elderly people, results (Fig. 5C) showed that maps correlated poorly (0.41 < r < 0.67) between groups. Moreover, r significantly decreased (P < 0.001) along with an increased cadence.
DISCUSSION
The aim of this work was to verify whether aging involves the reorganization of MN activity in accordance with modifications of the motor primitives undergoing muscle activation while healthy subjects walked overground in a wide range of cadences.
Consistency of the accounted synergies
For each subject, a matrix representing the averaged EMG patterns of 12 ipsilateral leg muscles over one gait cycle was algebraically factorized in five components, in agreement with previous authors (Ivanenko et al. 2004, 2006a). These five components accounted for ∼90% of the variability of each data set in both groups, even though three of them described ∼75% of the whole variance, in accordance with the criteria of the eigenvalue > 1 (Fig. 1D).
Factors related to the first three synergies, labeled LRF, PF, and HF, were well shaped in both groups at all cadences and were similar to those first named by Davis and Vaughan (1993) because of the occurrence of their main peaks during the gait cycle (Fig. 2). Results seemed to confirm a certain relationship between muscular groups and primitives signals because LRF, PF, and HF were regularly loaded by muscles that were active during the same periods of the gait cycle (Fig. 2). Moreover, when the cadence increased, these factors behaved more consistently and were loaded by their related muscle groups more systematically. On the whole, these three synergies, reflecting the coordinated activity of the leg muscles during the loading response, the propulsion, and the heel strike phases at all cadences and in both groups (Fig. 2), were definitively in agreement with the previous literature (Cappellini et al. 2006; Davis and Vaughan 1993; Grasso et al. 2004; Ivanenko et al. 2004, 2006a,b; Merkle et al. 1998; Olree and Vaughan 1995).
Conversely, factors related to the lasting two synergies, named F4 and F5, showed greater intersubject variability despite the consistency of their own loading muscles (W4 and W5), such that it was not possible to identify any univocal functional description.
Actually, results concerning F4 and F5 are also contrasting in literature. For instance, some authors observed that the fourth factor was characterized by a bi-phasic behavior that was supposed to maintain the phase shift between the legs (Davis and Vaughan 1993; Olree and Vaughan 1995). Conversely, other authors highlighted that four of their five mono-phasic factors were coupled and shifted each other for ∼50% of the gait cycle, whereas the last factor probably accounted for features related to the speed (Cappellini et al. 2006; Ivanenko et al. 2006a). Moreover, the scheme based on five primitive components was common across speeds (Cappellini et al. 2006; Ivanenko et al. 2004) and modes of locomotion (Cappellini et al. 2006, 2010; Ivanenko et al. 2007). Finally, other authors observed that, despite the consistency they found in muscle loading each synergy, timing patterns could show greater variability (Clark et al. 2009).
The discrepant results among the authors can be ascribed to the different set of recorded muscles. For instance, trunk and hip flexor muscles (e.g., erector spinae, adductor, trapezious group, external obliquae, iliopsoas, adductor magnus) have presumably contributed, in a significant way, to make more consistent the shape of their corresponding primitive signals in the studies of Ivanenko's group (see factor 4 and factor 5 in Ivanenko et al. 2004), such that the authors concluded that five main components account for the muscle activity during locomotion with different gaits and in a wide range of speeds (Cappellini et al. 2006; Ivanenko et al. 2004, 2006a, 2008). Davis and Vaughan (1993) reported that their biphasic factors was mainly led by sartorius, adductor longus, and erector spinae, which were not recorded during our experimental sessions. Moreover, when primitive signals are extracted only from leg muscles, between two and four were shown to be adequate to reconstruct EMG signals, such that Clark et al. (2009) already acknowledged that a larger set of muscles may have helped to identify additional “modules.”
With respect to the weight coefficients, we observed a certain consistency of muscles loading the fourth and the fifth synergies (W4 and W5) above all at faster cadences (see Fig. 2). In particular, muscles loading the fourth synergy, which mainly were ADD in the elderly and TA in the young, did not significantly load other ones, whereas W5 presumably accounted for residual information already described by the synergy related to the loading response phase, even though explaining a lower fraction of data variance. Concerning W4, we would like to remark that ADD and TA are both characterized by two main peaks over a gait cycle: during the early stance and the early swing (Winter 1991). Because F4 was affected by great intersubject variability, it did not highlight to which phasic activity it referred. Therefore from our results, the fourth synergy may account for features related to both the control of the foot/leg during grounding the floor and the control of the clearance during the stance-to swing transition.
In conclusion, in the framework of our protocol, the first three synergies were related to the loading response, the body support/propulsion, and the heel strike phases, whereas the fourth and fifth ones did not report further distinctive and unequivocal information. In this regard, the most significant synergies seemed to represent the corresponding biomechanical functions during locomotion (e.g., weight acceptance, body support and forward progression, leg deceleration) according to the role of the accounted muscles (Neptune et al. 2001, 2004). Therefore in agreement with previous authors, muscle synergies reflect the relationship between the central control and the output of the motor tasks (Cheung et al. 2005; Clark et al. 2009; Drew et al. 2008; Ivanenko et al. 2006a; Kargo and Giszter 2008).
Effects of cadence and aging on synergies
Factors and weight coefficients regarding the three most significant synergies showed similar features across the faster speeds (from 80 to 140 steps/min) in both groups. It was interesting to notice that, in young subjects, HFs extracted at all cadences were less correlated (0.4 < r < 0.6) to those extracted at reference speed (100 steps/min) than in elderly people (0.6 < r < 0.8). This occurred even though the similarity between muscles loading HF at all cadences with those related to 100 steps/min was significantly higher (scalar product > 0.7) in both groups. This result supports the hypothesis that aging could involve more rigid control of the activation of the biarticular hip extensors during heel strike in a wider range of cadences, whereas young subjects would adopt more flexible motor patterns depending on walking speed. Because the consistency of the EMG signals of elderly people is usually associated with a lack of neural plasticity (Schmitz et al. 2008; Winter 1991), it is possible to speculate that this may also be reflected in the motor primitives undergoing muscle activation.
At the slowest cadences (40 and 60 steps/min), the comparisons carried out across walking speeds and between the groups (Fig. 3) did not show consistent behavior, probably because of the variability of EMG signals, both in terms of amplitude and timing of the bursts. In particular, it is well known that muscle activation is affected by a redistribution of phasic activity along the gait cycle depending on the walking speed (Hof et al. 2002; Ivanenko et al. 2004), and very slow speeds could involve shuffling or pausing in motion (Shiavi et al. 1987a), lower amplitude in kinematics, and EMG signals with slight disruption of muscle activation (Nymark et al. 2005). Therefore the slower cadences involved greater variability in muscle recruitment, which was also reflected in the consistency of the synergies undergoing muscle activation in both groups.
Although a slight discrepancy between young and elderly people concerned HF, the most significant synergies appeared to be very similar between the groups. In particular, at faster cadences, primitive signals were characterized by high correlation and little time shift, and muscles loading these synergies were significantly comparable between young and elders (Fig. 3C). Moreover, the different timing of stance and swing phases in young and elderly people (Fig. 1B) appeared to weakly affect the occurrence of factor peaks along the gait cycle (Fig. 3D).
The likelihood between the synergies related to the two groups was expected because it confirms that the synergistic activation of leg muscles during locomotion is functionally coordinated to carry out such motor tasks in a similar fashion in both groups.
Modifications of spinal activity caused by aging
Even if there were no substantial differences between young and elderly people in muscle synergies, the spinal maps were sensitive to the age of the subjects, suggesting that aging involves significant modifications of the oscillating activity of the MN pools (Figs. 4 and 5).
MN maps related to young people systematically showed three main bursts in accordance with LRF, PF, and HF peaks, which became better defined as the cadence increased (Figs. 4 and 5). These were, respectively, located around L2–L3, the most caudal sections, and L5–S2, in accordance with the spinal segments innervating muscle accounted by each synergy and in agreement with the previous literature (Ivanenko et al. 2006a,b).
As far as our MN maps are concerned, we expected another burst of activity after the toe off, across the sections L2–L4. Nevertheless, it appeared systematically only at faster speeds, representing the main discrepancy between our results and previous findings (Ivanenko et al. 2006a,b). Again, this could be caused by the different set of muscles recorded during the experimental sessions. In particular, the fourth peak would be related to the activation of hip flexor/adductor muscles (iliacus, sartorius, adductor longus) and the tibial anterior, which would, respectively, flex the leg and increase the foot clearance during the early swing phase (Shiavi 1985; Winter 1991). In this study, we only recorded ADD and TA, such that our maps lack the hip flexor muscles (iliacus and sartorious), which significantly contribute to flex of the thigh during the stance to the mid-swing period (Shiavi 1985; Winter 1991) and would have increased the amplitude of such burst.
In elderly people, MN activity also systematically showed three main bursts in accordance with LRF, PF, and HF, and at the faster cadences, another peak of activity occurred after the toe-off (Fig. 4). Moreover, MN patterning became more consistent as the cadence increased, even though, because of intersubject variability, differences across cadences were not statistically significant (Fig. 5). Nevertheless, despite the similarity of the burst occurrences between young and elderly people, aging involved spread MN activity (Fig. 4). This different pattern was more evident when maps became more consistent within each group because of the faster cadences. In particular, although with the increasing pace, maps were better correlated within each group, when MN patterning of young people was compared with those of the elderly, the similarity decreased significantly (P < 0.001; Fig. 5).
To identify reasons leading to the age-related modifications of spinal patterning, it is important to remark that MN activity reflects both descending signals and peripheral afference integrated into the spinal cord. According to the hypothesis that the central pattern generator (CPG) is located in the spinal cord and because factors extracted from the EMG did not show significant differences between the groups (Fig. 3), primitive signals generated by the CPG and undergoing muscular synchronization were comparable between young and elderly people. Conversely, because of aging, both descending information and modifications of the peripheral sensory motor system seemed to influence the spinal activity (Figs. 4 and 5).
Spinal maps, as estimated in this study, are in agreement with the more intense muscle activity of the elderly, both in terms of amplitude (see CV of preprocessed EMG data in results) and power spectral density (Monaco et al. 2009, Bologna, Italy), as already shown in the previous literature (Nielsen et al. 1994; Schmitz et al. 2008). These findings could be caused both by physiological modifications of the nervous system and by different strategies adopted by elderly people to manage the age-related modifications of the musculo-skeletal system.
Concerning the physiological modification of the CNS, because of aging, the size of MNs becomes wider (Doherty et al. 1993a,b), the living MNs innervate muscles fibers that lose their own MNs (Cederna et al. 2001; Kadhiresan et al. 1996; Sugiura and Kanda 2004), the recruitment of the motor unit increases (Ling et al. 2007), and, because of the loss of synchronization of MN activity, EMG signals could seem spread (Pitcher et al. 2003). These reasons are potentially able to involve higher and spread muscle activation and, consequently, could partially explain the different patterning of the motoneuronal pools.
On the other hand, modifications of motor strategies, in term of greater variability of gait kinematics (Kang and Dingwell 2008), redistribution of the work load between proximal and distal extensors (DeVita and Hortobagyi 2000; McGibbon 2003; Monaco et al. 2009), and stiffening of the limb during the single support (Schmitz et al. 2008), could contribute to generate different MN patterning in the elderly, mainly affecting spinal activity both during the mid-stance (from 10 to 50% of the gait cycle) and across the heel strike (from 90 to 10% of the gait cycle), respectively, at the most rostral and caudal segments of the lumbosacral enlargement.
Conclusion
Muscle synergies during walking did not seem to be affected by aging but depended on the biomechanical demand during locomotion. These results may further support the hypothesis that motor primitives are hard-wired into the motoneuronal network involved in locomotion control (Drew et al. 2008). On the contrary, spinal activities seemed to be affected by aging probably because of both physiological modifications of the CNS and age-related modifications of motor control.
In this regard, muscle synergies may be more suitable to reflect the effects of treatments aimed at inducing reorganization of muscle activity, through a longitudinal period of observation. Nonetheless, we also acknowledge that further efforts are necessary to standardize the procedures to extract significant features from muscle synergies, in accordance with others authors (Clark et al. 2009). On the other hand, spinal maps seemed to be potentially able to provide an integrate view of the information accounted in EMG signals and weakly described by muscle synergies.
GRANTS
This study was partly supported by project Progettazione e realizzazione di un sistema biomeccatronico per la riabilitazione dell'arto inferiore in soggetti emiparetici in fase acuta funded by the Fondazione Cassa di Risparmio di Pisa, project Metodi innovativi per la neuroriabilitazione robot-mediata funded by the Fondazione Monte dei Paschi di Siena, and the European Project CLONS (Closed-loop vestibular neural prosthesis; FP7-ICT-2007.8.0 FET Open, 225929).
DISCLOSURES
No conflicts of interest, financial or otherwise, are declared by the authors.
ACKNOWLEDGMENTS
We thank Prof. Ferdinando Sartucci for valuable and useful suggestions. Moreover, we thank the authors of the “Università della Terza età” of Pontedera (Pisa, Italy) and all the other subjects involved in the experiments.
REFERENCES
- . New perspectives on spinal motor systems. Nat Rev 1: 101–108, 2000.
Crossref | Google Scholar - . The influence of lower extremity joint torque on gait characteristics in elderly men. Arch Phys Med Rehab 81: 1153–1157, 2000.
Crossref | PubMed | ISI | Google Scholar - . A comparison of the 1-step, 2-step, and 3-step protocols for obtaining barefoot plantar pressure data in the diabetic neuropathic foot. Clin Biomechanics (Bristol, Avon) 20: 892–899, 2005.
Crossref | PubMed | ISI | Google Scholar - . Motor patterns during walking on a slipperly walkway. J Neurophysiol 103: 746–760, 2010.
Link | ISI | Google Scholar - . Motor patterns in human walking and running. J Neurophysiol 95: 3426–3437, 2006.
Link | ISI | Google Scholar - . Motor unit properties of nerve-intact extensor digitorum longus muscle grafts in young and old rats. J Gerontol 56: B254–B258, 2001.
Crossref | Google Scholar - . Central and sensory contributions to the activation and organization of muscle synergies during natural motor behaviors. J Neurosci 25: 6419–6434, 2005.
Crossref | PubMed | ISI | Google Scholar - . Merging of healthy motor modules predicts reduced locomotor performance and muscle coordination complexity post-stroke. J Neurophysiol 103: 844–857, 2010.
Link | ISI | Google Scholar - . Phasic behavior of EMG signals during gait: use of multivariate-statistics. J Electromyogr Kines 3: 51–60, 1993.
Crossref | PubMed | ISI | Google Scholar - . Experimental simulation of cat electromyogram: evidence for algebraic summation of motor-unit action-potential trains. J Neurophysiol 86: 2144–2158, 2001.
Link | ISI | Google Scholar - . Age causes a redistribution of joint torques and powers during gait. J Appl Physiol 88: 1804–1811, 2000.
Link | ISI | Google Scholar - . Determinants of mean motor unit size: impact on estimates of motor unit number. Muscle Nerve 16: 1326–1331, 1993a.
Crossref | PubMed | ISI | Google Scholar - . Effects of motor unit losses on strength in older men and women. J Appl Physiol 74: 868–874, 1993b.
Link | ISI | Google Scholar - . Muscle synergies during locomotion in the cat: a model for motor cortex control. J Physiol 586: 1239–1245, 2008.
Crossref | PubMed | ISI | Google Scholar - . Distributed plasticity of locomotor pattern generators in spinal cord injured patients. Brain 127: 1019–1034, 2004.
Crossref | PubMed | ISI | Google Scholar - . Gait variability: methods, modeling and meaning. J Neuroengineer Rehab 2: 19, 2005.
Crossref | PubMed | Google Scholar - . Speed dependence of averaged EMG profiles in walking. Gait Posture 16: 78–86, 2002.
Crossref | PubMed | ISI | Google Scholar - . Modular control of limb movements during human locomotion. J Neurosci 27: 11149–11161, 2007.
Crossref | PubMed | ISI | Google Scholar - . Spatiotemporal organization of alpha-motoneuron activity in the human spinal cord during different gaits and gait transitions. Eur J Neurosci 27: 3351–3368, 2008.
Crossref | PubMed | ISI | Google Scholar - . Five basic muscle activation patterns account for muscle activity during human locomotion. J Physiol 556: 267–282, 2004.
Crossref | PubMed | ISI | Google Scholar - . Motor control programs and walking. Neuroscientist 12: 339–348, 2006a.
Crossref | PubMed | ISI | Google Scholar - . Spinal cord maps of spatiotemporal alpha-motoneuron activation in humans walking at different speeds. J Neurophysiol 95: 602–618, 2006b.
Link | ISI | Google Scholar - . Similarity and correlation—template matching. In: Pattern Recognition. New York: Wiley, 1988, p 36–41.
Google Scholar - . Step length reductions in advanced age: the role of ankle and hip kinetics. J Gerontol 51: M303–M312, 1996.
Crossref | Google Scholar - . Properties of single motor units in medial gastrocnemius muscles of adult and old rats. J Physiol 493: 543–552, 1996.
Crossref | PubMed | ISI | Google Scholar - . Separating the effects of age and walking speed on gait variability. Gait Posture 27: 572–577, 2008.
Crossref | PubMed | ISI | Google Scholar - . Individual premotor drive pulses, not time-varying synergies, are the units of adjustment for limb trajectories constructed in spinal cord. J Neurosci 28: 2409–2425, 2008.
Crossref | PubMed | ISI | Google Scholar - . Biomechanical gait alterations independent of speed in the healthy elderly: evidence for specific limiting impairments. Arch Phys Med Rehab 79: 317–322, 1998.
Crossref | PubMed | ISI | Google Scholar - . Electromyographic patterns suggest changes in motor unit physiology associated with early osteoarthritis of the knee. Osteoarthrit Cartilage 15: 1134–1140, 2007.
Crossref | PubMed | ISI | Google Scholar - . Macro-EMG and motor unit recruitment threshold: differences between the young and the aged. Neurosci Lett 179: 1–4, 1994.
Crossref | PubMed | ISI | Google Scholar - . Step length variability at gait initiation in elderly fallers and non-fallers, and young adults. Gerontology 49: 21–26, 2003.
Crossref | PubMed | ISI | Google Scholar - . Toward a better understanding of gait changes with age and disablement: neuromuscular adaptation. Exercise Sport Sci Rev 31: 102–108, 2003.
Crossref | PubMed | ISI | Google Scholar - . Mechanical energy transfer during gait in relation to strength impairment and pathology in elderly women. Clin Biomech (Bristol, Avon) 16: 324–333, 2001.
Crossref | PubMed | ISI | Google Scholar - . Using factor analysis to identify neuromuscular synergies during treadmill walking. J Neurosci Methods 82: 207–214, 1998.
Crossref | PubMed | ISI | Google Scholar - . Determination of the step duration of gait initiation using a mechanical energy analysis. J Biomech 29: 1195–1199, 1996.
Crossref | PubMed | ISI | Google Scholar - . Muscle synergies during walking: Comparison between young and elderly people. Preliminary results. Conf Proc IEEE Eng Med Biol Soc 2008: 5370–5373, 2008.
PubMed | Google Scholar - . Ageing involves wider spectral bandwidth and greater amplitude of EMG signals during walking.
Proceedings of International Society for Posture and Gait Research , 2009, p. 271.
Google Scholar - . During walking elders increase efforts at proximal joints and keep low kinetics at the ankle. Clin Biomech 24: 493–498, 2009.
Crossref | PubMed | ISI | Google Scholar - . Contributions of the individual ankle plantar flexors to support, forward progression and swing initiation during walking. J Biomech 34: 1387–1398, 2001.
Crossref | PubMed | ISI | Google Scholar - . Muscle force redistributes segmental power for body progression during walking. Gait Posture 19: 194–205, 2004.
Crossref | PubMed | ISI | Google Scholar - . Power spectrum analysis of the rectified electromyogram during gait for normals and patients. J Electromyogr Kinesiol 4: 105–115, 1994.
Crossref | PubMed | ISI | Google Scholar - . Electromyographic and kinematic nondisabled gait differences at extremely slow overground and treadmill walking speeds. J Rehab Res Develop 42: 523–534, 2005.
Crossref | PubMed | ISI | Google Scholar - . Fundamental patterns of bilateral muscle activity in human locomotion. Biol Cybern 73: 409–414, 1995.
Crossref | PubMed | ISI | Google Scholar - . Some characteristics of EMG patterns during locomotion: implications for the locomotor control process. J Motor Behav 17: 443–461, 1985.
Crossref | PubMed | ISI | Google Scholar - . Age and sex differences in human motor cortex input-output characteristics. J Physiol 546: 605–613, 2003.
Crossref | PubMed | ISI | Google Scholar - . Differences in lower-extremity muscular activation during walking between healthy older and young adults. J Electromyogr Kinesiol 19: 1085–1091, 2008.
Crossref | PubMed | ISI | Google Scholar - . Electromyographic patterns in adult locomotion: a comprehensive review. J Rehab Res Develop 22: 85–98, 1985.
Crossref | PubMed | Google Scholar - . Electromyographic gait assessment, Part 1: adult EMG profiles and walking speed. J Rehab Res Develop 24: 13–23, 1987a.
PubMed | Google Scholar - . Electromyographic gait assessment, Part 2: preliminary assessment of hemiparetic synergy patterns. J Rehab Res Develop 24: 24–30, 1987b.
PubMed | Google Scholar - . Representing and clustering electromyographic gait patterns with multivariate techniques. Med Biol Engineer Comput 19: 605–611, 1981.
Crossref | PubMed | ISI | Google Scholar - . Active and passive contributions to joint kinetics during walking in older adults. J Biomech 41: 1520–1527, 2008.
Crossref | PubMed | ISI | Google Scholar - . Gait termination: a review of experimental methods and the effects of ageing and gait pathologies. Gait Posture 22: 362–371, 2005.
Crossref | PubMed | ISI | Google Scholar - . Progress of age-related changes in properties of motor units in the gastrocnemius muscle of rats. J Neurophysiol 92: 1357–1365, 2004.
Link | ISI | Google Scholar - . Effects of age on rapid ankle torque development. J Gerontol 51: M226–M232, 1996.
Crossref | Google Scholar - . A comparison of gait initiation and termination methods for obtaining plantar foot pressures. Gait Posture 10: 255–263, 1999.
Crossref | PubMed | ISI | Google Scholar - . The Biomechanics and Motor Control of Human Gait: Normal, Elderly and Pathological. Waterloo, Canada: University of Waterloo Press, 1991.
Google Scholar - . Biomechanical walking pattern changes in the fit and healthy elderly. Phys Ther 70: 340–347, 1990.
Crossref | PubMed | ISI | Google Scholar - . Spatiotemporal activation of lumbosacral motoneurons in the locomotor step cycle. J Neurophysiol 87: 1542–1553, 2002.
Link | ISI | Google Scholar - . Factors governing the form of the relation between muscle force and the EMG: a simulation study. J Neurophysiol 92: 2878–2886, 2004.
Link | ISI | Google Scholar
