Weakening of Corticomuscular Signal Coupling During Voluntary Motor Action in Aging

Abstract

Background.

Aging is associated with muscle weakness and impairment in performing skilled motor tasks. Still, little is known about whether the link or functional coupling or connection between the central and peripheral systems during voluntary motor performance is compromised in the elderly subjects. The purposes of this study were to estimate functional corticomuscular connection (CMC) strength in the elderly subjects by calculating EEG–EMG coherence during voluntary motor performance, determine the relationship between the CMC and voluntary muscle force, and compare these between the old and the young subjects.

Methods.

Maximal voluntary contraction (MVC) of elbow flexion (EF) and EFs at three submaximal (20%, 50%, and 80% MVC) levels were performed in 28 healthy older (74.96±1.32 years) and 20 young (22.60±0.90 years) individuals, while EEG and EMG from biceps brachii, brachioradialis, and triceps brachii muscles were recorded simultaneously.

Results.

Compared with the young, older individuals exhibited significantly weakened CMC at all force levels tested. There was a proportional relationship between the CMC and EF force and high-positive correlation between the CMC and EF strength in both groups.

Conclusions.

Weakened CMC in aging may be a major factor contributing to age-related muscle weakness, and the linear relationship between the CMC and voluntary muscle force suggests dependence of force output on translation of the descending command to muscle electrical signal.

According to the United Nations (1), world population is getting older rapidly, with an estimation of 2 billion people being 65 years or older in 2050—up from 841 million in 2013. Aging is inevitable and is associated with body system degeneration in which neural, muscular, and musculoskeletal system deteriorations have a direct effect on motor function and quality of life. Older adults in general have difficulties to accomplish muscular activities in the same degree, compared with their younger counterparts. Weakness in aging is a consequence of neural and muscular impairments. For example, people with advanced age exhibit neural deficits resulting in increased difficulties to maximally activate their muscles (2–4). Decline in muscle mass can be as much as one third of the total mass in humans between the ages of 30 and 80 (5). Older adults experience significant deterioration in voluntary muscle strength ranging from 20% to 50% or even higher (6).

Numerous studies have reported evidence of age-related structural and functional degenerations in the central nervous and muscular systems that contribute to motor function impairment. Although the most commonly implicated mechanism involves muscle atrophy (6,7), peripheral degeneration following death of spinal motor neurons (7) only partially explains aging-related weakness (6,8). Changes within the central nervous system, both microscopic and macroscopic, exaggerate the deficit. Degenerative processes include reduction in gray matter volume (9), fewer motor cortical neurons (10), decrease in synaptic density (11), abnormal intra- and inter-cortical inhibition (12), loss of white matter integrity (13), lowered neurotransmitter levels (14), and greater variability in motor unit discharges (15). Still, little is known about whether functional connection between the two (central [brain] and peripheral [muscle]) systems during voluntary motor activities is intact in individuals with advanced ages.

Techniques that generate real-time information of central–peripheral communication in motor control may best address this question. Procedures that detect functional connection form such a method. Functional connection between the brain and muscle can be estimated by calculating coherence. Coherence is a quantitative method that estimates the level of synchronization in the frequency content between two (in this case, brain and muscle) signals, with values ranging from 0 (no synchronization or relationship) to 1 (perfect synchronization or relationship). This coherence measure has been suggested to solve the “binding” problem, acting as a mechanism to link information related to the same function but processed in different neural populations (16,17). Coherence between EEG and EMG signals at particular frequency bands has been reported by many studies (18–20) to demonstrate functional connection between the brain and muscle. For instance, Brown and colleagues (21) suggest that EEG–EMG coherence at lower gamma band (35–60 Hz) plays a role in controlling high-demanding hand muscle contractions. Others state that although the coherence at beta band (15–30 Hz) is significant, it is observed with static force output, whereas dynamic (movement) paradigms reduce beta band coherence significantly in comparison to gamma band (22). It has been previously shown that the level of brain–muscle or corticomuscular coherence (CMC) increases with the level of EMG (23) in healthy individuals, indicating a dependence of muscle output on strength of brain–muscle functional connection or CMC. Because EEG–EMG coherence or CMC is a quantitative assessment of the quality of functional connection between the brain and muscle, reduction in its value means diminished connection quality, which is expected to have a detrimental effect on peripheral motor performance (because of lower quality of descending command received) as well as on brain activation (because of difficulties to have its output signal translated into muscle activities).

Aging effects on functional connection have been explored preliminarily. Few recent studies have attempted to investigate CMC in aging. One study (24) showed shifting of the CMC to lower frequencies in people with more advanced ages compared with young individuals, but the authors did not address the relationship between CMC and voluntary force or muscle activation. James and colleagues (25) compared CMC among subjects in ages from infancy to elderly and reported prominent CMC differences during motor development in children compared to adults. Another study (26) reported a strong link between muscular (EMG–EMG) coherence and force output in the elderly subjects during abnormal walking. Although CMC changes with aging have been evidenced, direct comparisons of CMC strength at multiple force levels and the relationship between CMC and force between young and elderly subjects are missing. It is unknown if a positive relationship between the CMC and muscle force is maintained in the elderly subjects. Addressing these questions will help understand whether changes in functional connection with increasing force output can help explain muscle weakness in older adults.

The purposes of this study are to (i) estimate strength of CMC in older adults by calculating EEG–EMG coherence during voluntary motor performance and compare the CMC values between elderly and young subjects, and (ii) determine the relationship between the CMC and voluntary muscle force and to compare the relation between the two groups. It was hypothesized that the CMC would be weakened, but its relation with force was maintained in the old compared with the young. As EMG signals were recorded from synergists and antagonists, CMC levels between signals of the brain and these muscles were also estimated in the two groups and their values compared. Findings of this research are likely to contribute to a better understanding of neural mechanisms underlying motor function deficit (especially weakness) in aging. The CMC measurement may potentially serve as a tool to evaluate recovery/repair of functional brain–muscle connection as a result of neuromuscular training or therapy.

Methods

Subjects

Twenty-eight healthy older (74.96 ± 1.32 years, 20 women) and 20 young (22.60 ± 0.90 years, 10 women) individuals participated in the study. All recruited subjects were right handed (27) and screened by a physician for any ongoing neurological disorders/impairments. Those with neurological or musculoskeletal disorders impairing their sensorimotor function were excluded. All enrolled subjects did not have any systematic upper extremity motor training (eg, weekly weightlifting for at least a month) in the last 5 years. Elderly subjects had a minimum score of 25 or more on the Mini-Mental State Examination (28). The research study and its procedures were approved by the local institutional review board and all subjects signed an informed consent prior to their participation.

Elbow Flexion Strength and Submaximal EF Contractions

Maximal isometric elbow flexion (EF) force of the left arm was measured at the beginning of the experiment, indicating the subject’s maximal voluntary contraction (MVC) force at the joint. This was used as a baseline for subsequent calculation of submaximal isometric EF force levels. All the EF contractions were performed with the forearm in a neutral position (between pronation and supination) and hand open. The forces (maximal and sub-maximal) were measured by a force transducer (0-310N, JR3 Universal Force-Moment Sensor System, Woodland, CA), acquired by a Micro-1401 MK ii data acquisition system (Cambridge Electronic Design, Ltd., Cambridge, UK), digitized at 100 Hz, and shown on a computer screen using a horizontal cursor as online force feedback. For the MVC force, subjects were verbally encouraged to produce the maximal EF force for approximately 8 s, with a 45-s rest between trials (a total of five MVC trials). For the submaximal EF, subjects were asked to match their 20%, 50%, or 80% MVC force and maintain the target force as steadily as possible for 10 s. Three trials were performed at each force level and contractions at the three submaximal force levels were randomized. The subjects were asked to avoid eye blinks, teeth biting, tensing facial and neck muscles, and head movements during each contraction to minimize artifact signals produced by these activities. Practice runs were allowed for all the subjects to correctly perform the EF contractions at different intensity levels until they were comfortable to do them. The experiment lasted less than 60min.

EMG

Bipolar electrodes (Ag–AgCl, 8-mm recording diameter; In Vivo Metric, Healdsburg, CA) were placed on the skin overlying belly of the biceps brachii (BB), brachioradialis (BR), and triceps brachii (TB) muscles, with an interelectrode placement of approximately 3cm. A reference electrode was placed over the acromion process of the shoulder. EMG data for all MVC and submaximal level contractions were amplified (×1000), band-pass filtered (3- to 1000-Hz bandwidth), rectified, and digitized at a sampling rate of 2000 Hz using the Micro-1401 system.

EEG

Scalp EEG signals (with Cz being the referential electrode) were recorded continuously during the MVC and submaximal level contractions using a 128-channel dense-array EEG data acquisition system (Electrical Geodesics, Inc., Eugene, OR), along with the force and EMG signals. The EEG data recording did not start until the impedance for the whole net was below 10 kΩ. Additionally, four auxiliary channels were used to record BB, TB, and BR EMG and EF force to allow synchronization of the EEG and peripheral (EMG and force) signals for CMC analysis. EEG signals were amplified (×20,000), band-pass filtered (0.1–100 Hz), and digitized at a sampling rate of 250 Hz.

EEG–EMG (Corticomuscular) Coherence

The EEG and EMG data were visually inspected and signals with artifacts related to head movement, eye blink, or facial/neck muscle contraction were removed. EMG signals of the three muscles were re-sampled at 250 Hz to match sampling frequency of the EEG. Coherence between EEG of each of the 124 channels (four EEG channels were used for EMG and force recordings) and EMG of each of the three muscles was computed (29) (for detailed CMC calculation, see Supplementary Material).

Hanning windows were applied during the coherence calculation, with each window length of 512 sample points and a 128-sample overlap when moving the window rightward till the entire trial (~8 s with removal of the first and last second data to ensure steadiness of the data) was analyzed. An average coherence value over the three trials at each force level was obtained in each subject. Coherence was considered significant when the value was over 95% confidence level. EEGLAB software package and its SIFT toolbox (30) were used to calculate the coherence across frequency ranges with its values between 0 and 1. The coherence analysis excluded 57 out of a total of 392 trials in the elderly and 12 out of 280 in young groups, which contained artifacts mainly in EEG recordings. After computing coherence for all the EEG channels with EMG of each of the three muscles, coherence maps were generated to demonstrate channels on the scalp exhibiting significant coherence at one or more frequency bands in a given muscle. CMC values of EEG electrodes above the significance level with EMG of each of the three muscles (BB, BR, and TB) were identified for further process, which included EEG electrode grouping (clustering) and statistical analysis (see Statistical Analyses section). Clusters of EEG electrodes in the right sensorimotor (surrounding C4 location; force was exerted by left arm) and central (Cz location) areas with EMG of each of the three muscles was specifically analyzed due to close relevance of the underlying brain regions in motor control and the large volume of information with all 124 electrodes. Hence, results associated with the C4 and Cz reported hereafter represent data from CMC averaged across the clusters of EEG electrodes surrounding them (see Supplementary Figure 5 for the EEG electrodes in each cluster).

Statistical Analyses

Although 28 old and 20 young subjects participated in the study, not all their CMC data were included for the statistical analysis. Specifically, 19 out of 20 (young, 9 women) and 26 out of 28 (old, 18 women) CMC data points during submaximal force performances were analyzed. For the MVC task, only 16 out of 20 (young, 8 women) and 22 out of 28 (old, 15 women) CMC data points were analyzed. When the MVC was performed, a number of subjects in each group had co-contractions of facial or neck muscles despite repeated requests of keeping these muscles silent. Activities of facial or neck muscles significantly contaminated EEG signals in these subjects and the data could not be used for accurate EEG–EMG coherence analysis.

IBM SPSS Statistics v21.0 was used to perform the statistical analyses. Gaussian distribution of all dependent variables was confirmed with nonparametric Kolmogorov–Smirnov test. Subsequently, a repeated-measures two-way analysis of variance (ANOVA) was employed to examine the effect of age and force level on EEG–EMG coherence for each of the two EEG electrode (C4 and Cz) clusters for each muscle. A one-way ANOVA was applied for analyzing the EF strength (MVC force) and corresponding CMC during MVC between the two groups, and EEG–EMG coherence between the agonist (BB and BR) and antagonist (TB) muscles for each cluster in each group. Bonferroni corrections were applied to multiple comparisons when appropriate. Significance level was set at p ≤ .05.

Results

Strength and Corresponding CMC

One-way ANOVA analysis revealed that compared with young group, the left EF strength (MVC force) was marginally significantly lower in the group of healthy elders (old: 93.40±6.11 N vs young: 116.76±12.44 N, F(1,47) = 3.37, p = .07). Younger subjects had significantly higher CMC during MVC at beta (15–35 Hz) band for both C4 and Cz locations with the BB and BR muscles. Corresponding CMC values for young and old groups for the C4 (EEG)–BB (EMG) pair were 0.49±0.05 (young) and 0.24±0.02 (old), for Cz–BB pair 0.44±0.05 (young) and 0.21±0.02 (old), for C4–BR pair 0.46±0.05 (young) and 0.23±0.02 (old), and for Cz–BR pair 0.41±0.04 (young) and 0.20±0.02 (old). TB, as the antagonist muscle to the EF task, had the corresponding CMC values of 0.30±0.12 (young) and 0.26±0.19 (old) for the C4–TB pair and 0.24±0.09 (young) and 0.20±0.11 (old) for the Cz–TB pair. The outcome of one-way ANOVA showed that the differences in CMC between the two groups were highly significant for C4–BB (F(1,37) = 22.27, p < .001), Cz–BB (F(1,37) = 22.08, p < .001), C4–BR (F(1,37) = 21.43, p < .001), and Cz–BR (F(1,37) = 20.44, p < .001) pairs. However, the between-group difference in CMC values for EEG–TB EMG during the EF MVC was insignificant for the C4–TB (old: 0.12 ± 0.02 vs young: 0.22±0.12, F(1,37) = 2.56, p = .46) and Cz–TB (old: 0.11 ± 0.02 versus young: 0.20±0.10, F(1,37) = 3.51, p = .40) pairs. Comparing the agonist (BB + BR) with the antagonist (TB) during the EF MVC, the CMC for TB muscle was significantly lower at C4 (F(1,37) = 29.05, p < .001 for young and F(1,37) = 18.63, p < .001 for old) and Cz (F(1,37) = 20.44, p < .001 for young and F(1,37) = 17.39, p < .001 for old) locations.

CMC at Multiple Force Levels

The CMC (EEG–EMG coherence) data at multiforce levels were analyzed by two-way ANOVA with group as a between and force level as a within factor. The analysis showed a significant main group effect (F(1,44) = 25.12, p < .001) and a significant main force level effect (F(1,44) = 23.84, p < .001) for main flexors BB and BR at C4 location. The interaction between the two factors was also significant (F(1,44) = 19.36, p < .001). Post hoc analysis revealed a significantly larger difference (p < .001) in CMC between the two groups at higher versus lower force levels, with an exception of 80% MVC force level (p = .035). Figure 1 shows the CMC data at all force levels for both groups, and Figure 2 gives an example of the coherence difference between one young and one older person. The corresponding CMC analysis for the TB muscle at the three force levels between groups revealed no significant main group effect (F(1,44) = 1.875, p = .18) and no significant main force level effect (F(1,44) = 2.022, p = .16). The interaction between the two factors was also not significant (F(1,44) = 1.451, p = .23). Additionally, one-way ANOVA showed that the differences in CMC between the antagonist TB and agonists BB/BR muscles were highly significant (p < .01) at both EEG (C4 and Cz) locations and all three force levels for both groups.

Figure 1.

Group results indicate corticomuscular connection (EEG at C4 with pooled EMG data of biceps brachii and brachioradialis) was (i) weaker in the elderly vs young subjects and (ii) proportional to the exerted force (*p < .05 and **p < .001).

Figure 2.

An example of weaker corticomuscular connection within beta band especially at about 30 Hz in an older compared with a young subject.

Figure 3 shows Z-transformed (for easier visualization of) coherence maps based on all EEG channels with EMG of the BB, BR, and TB muscles (first, second, and third columns, respectively) for the elderly (upper row) and young (lower row) subjects at the beta band (15–35 Hz). The bar on the right shows color-coded coherence with red indicating higher coherence and blue lower coherence. The strongest coherence with both flexor (BB and BR) muscles can be roughly seen in the right frontal–parietal areas in young subjects but is shifted predominantly to the parietal region in the elderly subjects.

Figure 3.

Normalized (Z-transformed) corticomuscular connection (CMC) map based on EEG signals of 124 EEG channels and EMG of the biceps brachii, brachioradialis, and triceps brachii muscles in beta (15–35 Hz) band for maximal voluntary contraction. The CMC was averaged across all subjects in each group.

There was a strong association between the CMC (C4 EEG–BB EMG coherence) at MVC and MVC force (EF strength) for both groups with adjusted R ² values of 0.811 and 0.784, for the young and older subjects, respectively (Figure 4). Similar association was also observed for the C4–BR pair with the corresponding adjusted R ² values of 0.754 and 0.763, for the young and older subjects, respectively. Strong positive correlation between CMC and strength suggests that the maximal force output is highly dependent on the level of CMC.

Figure 4.

High-positive correlation between corticomuscular connection (CMC; EEG at C4 with EMG of biceps brachii) and elbow flexion strength (maximal voluntary contraction [MVC] force) in both groups. The figure included 16 data points for young (8 women) and 22 data points for old (15 women) subjects. A number of subjects’ EEG data in both groups were removed from the CMC analysis due to significant artifacts generated during the MVC performance.

Discussion

This study is the first to investigate the relationship between brain–muscle functional connection (measured by CMC) and voluntary muscle force in the elderly subjects and the difference in the relationship between young and older populations, by systematically varying force levels. The major findings are (i) the CMC was significantly lower in older compared with young subjects at all four levels of EF force, (ii) there was a proportional relationship between the CMC and force in both groups, and (iii) significant positive correlation between CMC and EF strength for both groups.

The CMC was significantly lower in older than young individuals, suggesting weakened or impaired functional connection between signals of the brain and target muscle. This impairment may partially explain age-related weakness and reduced ability to perform motor skills. A number of factors may have contributed to weakening of the CMC in aging. One such factor is reduced descending command or brain-to-muscle signal for voluntary muscle activation and this has been shown by the twitch interpolation in which a supramaximal electrical stimulation is applied to a muscle undergoing MVC and the evoked force over the MVC force measured. It has been demonstrated that the evoked (by twitch interpolation) force was significantly greater in older than young subjects, suggesting a greater deficit in the descending command in the elderly subjects (2). Such a deficit is expected to have a negative influence on the CMC. A second factor that may have influenced the CMC in aging is propagation of the descending command across neuromuscular junction (NMJ) and this is widely evaluated by assessing the amplitude of compound action potential or M wave triggered by a supramaximal electrical stimulation applied to the motor nerve proximal to the NMJ and recording the M wave distal to the NMJ. Apparently if the propagation function over the NMJ is impaired, the size of M wave would be smaller and CMC weakened. Few studies have reported a decline in amplitude of M wave in older than young individuals (31,32). However, the observation of impairment in NMJ propagation in the elderly subjects is not conclusive and needs further investigation (33). Finally, there are structural and physiological changes in the brain and skeletal muscle in aging, which may alter cortical and muscular signals and their connection. Indeed, numerous studies have reported these changes (34–36), yet their effects on the level of CMC have never been elucidated.

The finding of proportional relationship between the CMC and voluntary muscle force confirmed a previous report by Kilner and colleagues (23) that the level of CMC in the 15–30 Hz range increases with the level of voluntary EMG. Although they have demonstrated a positive relation between CMC and muscle electrical signal (EMG), which is highly correlated with voluntary muscle force, this study was the first to show such a relationship in the elderly subjects and to compare it with their younger counterparts. The outcome of the comparison suggests that although the CMC is weakened in older adults at any levels of muscle force, the CMC–force relation is maintained in the older population. This observation could be explained by dependence of voluntary muscle force on the descending command (37,38). As long as the force maintains to be a product of the descending command, the CMC–force relation needs to be maintained to ensure proper translation of the neural signal into final force output. Based on the proportional CMC–voluntary EMG relationship, Kilner and colleagues (23) hypothesized that voluntary muscle force is represented in the brain. Our results suggest that the force or EMG output is at least linearly scaled by descending command from the cortical motor control network.

The CMC is linearly correlated to EF MVC force or strength in both groups. This result was expected based on the linear relationship between the CMC and voluntary EF force observed in this study and positive association between the CMC and EMG reported (23). This finding suggests that CMC, a measurement of the level of functional connection between the brain and muscle, may serve as a biomarker of the central nervous system for strength prediction and evaluation of motor function recovery (especially strength) after intervention. Our group has previously demonstrated functional magnetic resonance imaging signal (37) and motor-related cortical potential derived from scalp EEG recordings (38) as potential predictors of voluntary muscle strength. However, functional magnetic resonance imaging does not measure neural signal in the brain directly, and obtaining a reasonable motor-related cortical potential profile usually needs averaging a large number of trials (at least 30 trials or more depending on the intensity of muscle contraction). In contrast with Brown and colleagues (21) who saw significant CMC on multiple frequency bands, we only found significant CMC at beta band in both groups. On the other hand, our findings agreed with those of Omlor and colleagues (22) under a similar isometric force condition.

Other Observations

CMC for both the BB and BR muscles were apparently higher than CMC of TB muscle (eg, the CMC was between 0.4 and 0.5 for BB and BR during MVC but that for TB was around 0.2). This is expected as the BB and BR were agonists for the EF contraction, but the TB was the antagonist (39). Our study paradigm required a high accuracy and co-ordination between the agonist and antagonist muscles for submaximal EF contractions by matching the force at each target level and TB might be active as a stabilizer of the joint, but its role in making a precise contraction may not be as great as the BB and BR. In addition, although CMC seemed to be higher for BB than BR in both groups, the differences were not statistically significant. This is an interesting observation and has been argued that BB is a greater contributor to EF than BR (40).

The majority of the elderly subjects were women (20 out of a total of 28 or 72% total older subjects), but the young group had equal numbers of men and women (50%). A larger proportion of female subjects in the aging group could influence the group CMC difference. However, as the difference in the proportion was not huge (72% vs 50%), the gender contribution, if any, should be relatively small. No studies have reported significant CMC differences between men and women. EEG and EMG signals were in general noisier in the elderly than young subjects. This may be due to increased skin impedance and reduced ability to maintain the head and body steady during data collection. We did take precautions to minimize the noise during the experiments.

Conclusions

Functional CMC during voluntary EF is reduced in older persons. This impairment could potentially be one of the mechanisms contributing to age-related muscle weakness, loss of motor skills, and co-ordination. Although the CMC is weakened in aging, its relation with voluntary force output is not affected, suggesting that older adults, similar to their younger counterparts, depend on the central or descending command and translation of the command to local signal to scale their muscle force output. The linear relationship between the CMC and voluntary strength suggests that the CMC measure may serve as an objective index or predictor of maximal muscle force output for old and young populations. However, because it is difficult to capture central neural activities below the cortical level, the role of the subcortical motor fields in controlling the muscle actions could not be delineated by this study. Future studies are recommended to focus on identifying therapies or training regimes to restore the level of CMC and cortical and muscular mechanisms contributing to impairment in CMC in aging.

Supplementary Material

Supplementary material can be found at: http://biomedgerontology.oxfordjournals.org/

Funding

This work was supported by the National Institutes of Health (R01 NS035130 to G.H.Y.).