Estimates of persistent inward currents in tibialis anterior motor units during standing ramped contraction tasks in humans

Abstract

Persistent inward currents (PICs) play an essential role in setting motor neuron gain and shaping motor unit firing patterns. Estimates of PICs in humans can be made using the paired motor unit analysis technique, which quantifies the difference in discharge rate of a lower threshold motor unit at the recruitment onset and offset of a higher threshold motor unit (ΔF). Because PICs are highly dependent on the level of neuromodulatory drive, ΔF represents an estimate of level of neuromodulation at the level of the spinal cord. Most of the estimates of ΔF are performed under constrained, isometric, seated conditions. In the present study, we used high-density surface EMG arrays to discriminate motor unit firing patterns during isometric seated conditions with torque or EMG visual feedback and during unconstrained standing anterior-to-posterior movements with root mean square EMG visual feedback. We were able to apply the paired motor unit analysis technique to the decomposed motor units in each of the three conditions. We hypothesized that ΔF would be higher during unconstrained standing anterior-to-posterior movements compared with the seated conditions, reflecting an increase in the synaptic input to motoneurons drive while standing. In agreement with previous work, we found that there was no evidence of a difference in ΔF between the seated and standing postures, although slight differences in the initial and peak discharge rates were observed. Taken together, our results suggest that both the standing and seated postures are likely not sufficiently different, both being “upright” postures, to result in large changes in neuromodulatory drive.

NEW & NOTEWORTHY In the present study, we show that the discharge rate of a lower threshold motor unit at the recruitment onset and offset of a higher threshold motor unit (ΔF) is similar between standing and seated conditions in human tibialis anterior motor units, suggesting that at least for these two upright postures neuromodulatory drive is similar. We also highlight a proposed technological development in using high-density EMG arrays for real-time muscle activity feedback to accomplish standing ramped contraction tasks and demonstrate the validity of the paired motor unit analysis technique during these conditions.

INTRODUCTION

The discharge of motor units, each of which comprises a single motor neuron (MN) and the set of muscle fibers it innervates, is critical for all motor behaviors ranging from basic life-sustaining functions to the diverse repertoire of human movement. Motor output is driven by supraspinal commands and/or spinal reflex pathways that recruit and subsequently modulate the discharge rate of MNs. For the better part of the 20th century, MNs were thought to be relatively simple, linear integrators of these supraspinal inputs. However, research in the past four decades has highlighted the central role of persistent inward currents (PICs) in changing the intrinsic cellular properties and consequently the input-output functions, of MNs (1). These PICs, which are facilitated by voltage-sensitive Na (2, 3) and Ca channels (4, 5), are highly reliant on the level of monoaminergic neuromodulatory drive (6–8). The monoaminergic system is itself highly dependent on arousal state and activity level (9–11), suggesting a functional role for these currents. Indeed, one of the first proposed functional implications of the initial cellular work was that PICs might serve a role in posture maintenance, where prolonged activation of muscles would be beneficial (7, 12–14).

Most activities of daily living in humans are anisometric, involving changes in joint angle. Intracellular recordings in cats suggest that changes in joint angle have a substantial effect on PICs (12), which facilitate self-sustained firing in MNs. A few investigations have studied the role of PICs in standing, one of which relied on a constrained standing setup (15) and the other using force feedback and indwelling wire EMG (16). However, the majority of assessments of motor unit discharge patterns, including those concerned with estimates of PICs in MNs, have been performed under constrained isometric conditions in the arms and the legs (17–21). In humans, the relationship between a standing versus seated posture and the functional role of PICs remains nebulous. Revealing the nature of this relationship is important for understanding postural control in healthy and diseased states.

A well-accepted method to estimate PICs noninvasively is the paired motor unit analysis technique, which uses the discharge rate profile of a lower threshold motor unit as an indicator of drive to the MN pool and measures the difference in the discharge rate of a lower threshold motor unit at the recruitment onset and offset of a higher threshold motor unit (discharge rate hysteresis or ΔF) (22, 23). It is worthwhile to consider whether the basic principles underlying the paired motor unit analysis technique are still valid during more functional behaviors. Under isometric conditions, muscle fascicle length changes in the tibialis anterior (TA) are almost nonexistent (24). By contrast, the TA muscle fascicle length changes orders of magnitude more during the gait cycle (25). During standing anterior-posterior postural adjustments, as are performed in the present study, these changes in muscle length are likely to lie somewhere in between these two extremes. Given that the shape of motor unit action potentials can change throughout a dynamic contraction, it is not obvious that motor unit action potentials can be decomposed from high-density surface EMG (HDsEMG) arrays during such conditions. The present investigation shows that we can in fact reliably discriminate the firing patterns of tens of motor units during standing dynamic contractions of the TA muscle.

In the present investigation, we used HDsEMG arrays to decompose the activity of multiple motor units in the TA during seated isometric conditions and during standing anterior-to-posterior adjustments. In particular, we were interested in whether ΔF was higher when participants assumed a standing as opposed to a seated stance. We reasoned that monoaminergic drive would be higher when participants were standing as opposed to when they were seated and accordingly hypothesized that ΔF would be increased during the standing anterior-to-posterior movement task. The present study is not the first to test this hypothesis; previously, Foley and Kalmar (16) used fine wire EMG to detect a handful of motor units in the soleus muscle and performed the paired motor unit analysis technique during standing and seated conditions. We extend these findings by measuring activity from the TA muscle with HDsEMG arrays to discriminate the firing patterns of tens of motor units. Furthermore, we also highlight the clinical and scientific benefits of using real-time feedback from the HDsEMG arrays for studies attempting to estimate PIC magnitudes. In the present study, this was necessary to allow comparisons between the seated and standing conditions (where a torque measurement was not feasible). To date, EMG has not been utilized as a source of visual feedback in studies measuring ΔF, with all the published studies at the time of this writing having utilized torque feedback. Arguably, under certain conditions, a well-isolated EMG signal from a given muscle may be a better representation of that muscle’s activity level than the torque, which can be influenced by the contraction of other muscles (e.g., the soleus and medial gastrocnemius muscles working in concert during ankle plantarflexion). Our overall findings suggest that neuromodulatory input between the standing and seated (both upright) postures is at similar levels, but several avenues of future research can illuminate this relationship in more extreme postural changes.

METHODS

Participants

A total of 15 participants (12 males and 3 females) with no known neuromuscular impairments were recruited for this study. Participants were asked to abstain from caffeine for the day of the experiment, and none reported taking any medication for attention deficit hyperactivity disorder, anxiety, or depression. All participants provided written informed consent before the experiment, and all procedures were approved by the Institutional Review Board of Temple University (Protocol #No. 23971).

Experimental Setup and Protocol: Seated and Standing

The skin over the right tibialis anterior (TA) muscle was shaved, abraded with high-grit sandpaper, and cleansed with a damp towel. Conductive cream (AC Cream, OT Bioelettronica, Turin, Italy) was applied to a 64-channel HDsEMG array (ELSCH064NM2, 8-mm interelectrode distance; OT Bioelettronica, Turin, Italy), centered over the muscle belly of the TA, and secured with medical tape (3M Transpore; St. Paul, MN). Participants were seated in a dynamometer (Biodex, Shirley, NY) and secured with straps across the torso and upper leg in a position of 90° hip flexion and 15° knee flexion. The ankle was secured to a footplate at an angle of 15° plantarflexion, with the footplate affixed to a six degrees of freedom load cell (JR3; Woodland, CA). Participants performed three maximal volitional dorsiflexion contractions (MVC) separated by 2 min of rest under isometric conditions while secured in the dynamometer. The greatest of these contractions was chosen as the MVC. If the most recent contraction was greater than the greatest previous contraction, an additional contraction was performed after 2 min of rest to ensure detection of an accurate and reliable MVC.

Participants then performed three sets of three isometric dorsiflexion contractions up to 20% MVC, with a 10-s increase and 10-s declining phase. Real-time feedback of either torque or a 500-ms root mean square (RMS) EMG output from a single differential of two EMG channels (channels 27 and 28; see highlighted channels in Fig. 1A, inset) of the HDsEMG array was displayed to the participant on a monitor as shown in the setup in Fig. 1. This differential was chosen due to its central position in the array and its larger than-standard interelectrode distance (17.9 mm as opposed to 8 mm), allowing a more global representation of muscle activity. Two different conditions were tested while participants were secured in the seated setup: isometric ramping contractions with visual feedback provided as the dynamometer torque response and isometric ramping contractions with the visual feedback provided as the RMS EMG from a single differential channel of the HDsEMG array. Subsequently, a third condition was performed. Keeping the electrode in place, participants were asked to stand and perform anterior-to-posterior postural adjustments (Fig. 1B), mimicking the pattern and level of activity of the slowly ramping isometric contractions they performed while seated. The goal of these standing anterior-to-posterior adjustments was to linearly increase the RMS EMG to allow for conditions under which ΔF calculations could theoretically be performed. Thus RMS EMG feedback was provided in this condition, and participants were instructed to increase their activity over 10 s and then reduce it back to baseline at the same time scale. In doing so, the participants tended to slightly lift their toes and we visually observed that participants shifted their weight to their heels.

Figure 1.

Experimental setup in Biodex System 4 with OT Bioelettronica amplifier and electrode array. Participants followed real-time torque and/or root mean square (RMS) EMG feedback tracings on a monitor while performing ramp tasks while seated (A) and while standing (B). Participants performed dorsiflexion ramps [20% maximal volitional dorsiflexion contractions (MVC)] while seated and anterior-to-posterior adjustments to generate linearly increasing and sure decreasing RMS EMG activity (up to 20% MVC) while standing. Inset: magnified electrode array with the channels across which differential RMS EMG is calculated in near-real-time for visual feedback purposes highlighted in blue. TA, tibialis anterior.

Across all conditions, each of the 63 differential EMG signals from the 64-channel array were filtered between 10 and 500 Hz and collected at 2,048 Hz using a 16-bit A/D converter (Quattrocento, OT Bioelettronica; Turin, Italy). The EMG signals were visualized throughout muscle contraction to ensure acceptable signal quality (i.e., when there was a clear indication of increasing electrical activity visible to the naked eye that coincided with muscle contraction in real-time).

Torque and EMG Signal Analysis

Task performance was assessed based on the feedback signals (torque and RMS EMG). All post hoc analyses were performed in Matlab (Version 2019a, Mathworks, Natick, MA). The torque was smoothed with a 10 Hz low-pass third order Butterworth filter. In the post hoc analysis, each signal was time-aligned with the peak of the target ramp and residuals were calculated as the difference between the feedback signal generated by the participant and the target ramp. There is no evidence of a change in ΔF across contraction intensities in the 10–30% MVC range with ramping contractions (15, 17), although we cannot exclude the possibility that these effects may be more apparent during the standing conditions. Thus we did not exclude ramps as long as they were within 5% MVC of 20% MVC target. A total of five trials across participants were excluded based on this criterion. Since our primary concern was a smooth linear increase and decrease in force output, the “target” ramp was adjusted to reach the same %MVC peak that the participant reached in each trial post facto, effectively allowing calculation of a detrended residual representing the error associated only with failing to maintain a continuous trajectory. We reasoned that changes in ΔF are unlikely to be evident for such a small change in torque or RMS EMG and did not wish to inflate values of the root mean square error as would be expected if we did not calculate the detrended residuals. Thus a high detrended residual for any particular trial penalized a failure to generate a steadily increasing and steadily decreasing ramp, not a failure to reach 20% MVC. The sum of the root mean squared error of the residual was calculated for all torque and RMS EMG signals and averaged across trials for each participant.

When presented with only RMS EMG feedback, some participants failed to generate approximately linear increases or decreases in the feedback signal. More specifically, if the participants had fast (1- to 2-s duration) high-amplitude (>5% MVC) shifts in their torque or RMS EMG output during the ramping task, the offending trials were excluded from all subsequent analyses. A total of four trials were excluded from analysis in this way, representing all the standing trials from one participant and one trial from a separate participant. In accordance with these criteria, the root mean square error was also higher in these trials compared with the acceptable trials. In the case of one participant during the seated RMS EMG feedback trials and two participants during the standing RMS EMG feedback trials, there were no acceptable trials; hence these participants had no data to be analyzed for those conditions.

Motor Unit Decomposition

High-density array surface EMG recordings were converted from the OT Bioelettronica format into Matlab-compatible (Version 2019a, MathWorks, Natick, MA) data files for post hoc analyses. Visual inspection was performed to remove noisy EMG channels before decomposition. The remaining EMG channels were input into a Matlab script utilizing the convolutive blind source separation algorithm (26). This methodology and similar approaches have been extensively used and have shown good reliability in several experimental conditions (26–33). In general terms, the algorithm iteratively optimizes a contrast function that maximizes the non-Gaussianity, or sparsity, of each source signal to accurately detect individual motor unit discharge patterns. The basic principle underlying this decomposition stems from the fact that the firing instances of any one motor unit are further from a normal distribution than the combined firing instances of several independent motor units. For all participants, motor unit discharge patterns were discriminated from the TA muscle during each condition (i.e., seated with torque feedback, seated with RMS EMG feedback, or standing with RMS EMG feedback). The TA was chosen because of its role in postural adjustment while standing.

The output of the decomposition algorithm is the interval pulse train, which represents the likelihood of a motor unit discharge occurring at a particular instance in time for each of the discriminated putative motor unit firing patterns. Only motor unit spike trains in which all of the motor unit spikes could be fully identified and with a Silhouette value >0.87 (26) were used for subsequent analysis. After the initial fully automated decomposition, each of the interval pulse trains were edited in a semiautomatic fashion by an experienced operator. Briefly, this editing process involved assessments of the interval pulse trains to optimize the source separation procedure, allowing identification and inclusion of previously missed spikes and exclusion of incorrectly identified spikes in the interval pulse train. Particular attention was paid to the threshold of spike detection in the interval pulse train, especially at the onset and offset of the ramp contraction.

Analysis of motor unit discharge patterns.

Motor unit firing instances were converted to a binary matrix representing motor unit spikes and quiescent periods. This binary matrix was convolved with a 1-s Hanning window and snipped at the first and last instance of motor unit firing to produce a smoothed instantaneous discharge rate, similar to previous publications (15, 18, 34, 35). The smoothed instantaneous discharge rate was used to acquire all discharge rate values (e.g., initial, maximum, final) and perform further analyses.

Discharge rate hysteresis (ΔF).

Estimates of PIC amplitudes were made using the paired motor unit analysis technique initially proposed by Kiehn and Eken (36) and formally developed for ramp contractions by Gorassini and colleagues (22, 37, 38). This technique allows a noninvasive estimate of MN PIC amplitude by quantifying the difference in the discharge frequency of a lower threshold motor unit at the onset and offset of activity of a higher threshold motor unit (ΔF = F_Recruitment – F_{Derecruitment}; Fig. 2). To ensure that each motor unit within a pair received similar levels of neural drive and processed the drive in a similar fashion, we excluded motor unit pairs that whose rate-rate coefficient of determination (r²) was <0.7 (23). Additionally, to ensure that the PIC was fully activated, we excluded all motor unit pairs in which the recruitment time was within 1 s of each other (34). After filtering ΔF values using these criteria based on previous publications, ΔF values correspond to changes in the amplitude of the PIC in the higher threshold motor unit. In the present study, we are using a unit-wise approach to calculating ΔF, as shown in Fig. 2. Thus, for each of the higher threshold (test) units, we identified all paired lower threshold (control or reporter) units that fell within the bounds of the selection criteria for ΔF calculations enumerated in detail above and in previous publications (21, 23). The ΔF values we report are based on an average of all acceptable pairs of lower threshold motor units with each higher threshold motor unit. In Fig. 2, the example shows the ΔF of motor unit k being the mean of the ΔF values from three different acceptable pairs. Using this “unit-wise” approach (in contrast to the traditional “pairwise” approach) has the benefit of reducing intra-participant variability of ΔF values and weighting ΔF values evenly across motor units regardless of the number of possible permutations for each motor unit. By contrast, the ΔF values obtained by averaging across all pairs are weighted toward higher threshold motor units since these units have a greater chance to generate more distinct permutations with many lower threshold motor units. From a theoretical perspective, a meaningful ΔF calculation requires similar levels of synaptic input onto the two motor units used in the calculation (23). This must be the case for all of the lower threshold (reporter) units paired with each of the higher threshold (test) motor units. Thus it makes abundant sense from a theoretical perspective to view each test unit as having a single ΔF value, which is represented by the average of ΔF values across all acceptable test-reporter motor unit pairs. The paired motor unit analysis technique is currently the preferred method in the field for estimating PIC amplitudes in human MNs; it has been subjected to rigorous experimental and computational evaluations of its effectiveness as a means of estimating PICs and of the factors affecting the accuracy of these estimations (17, 20, 37, 39, 40).

Figure 2.

Torque trace (black) and root mean square EMG trace (blue) are shown during a 10-s increasing, 10-s decreasing ramp contraction protocol up to 20% maximal volitional dorsiflexion contractions (MVC). The discrete and smoothed (1-s Hanning window) instantaneous discharge rate of a subset of motor units (MUs) are shown to demonstrate the unit-wise paired motor unit analysis approach. Briefly, ΔF is the change in discharge frequency of a lower threshold motor unit at the recruitment onset and offset of a higher threshold unit. In this case, motor unit k forms acceptable pairs with 3 other motor units. Thus ΔF_k is the average of those 3 pairs. Note that the permutations involving motor units 3 and 2 would also generate their own individual ΔF values as emphasized with the gray vertical lines and desaturated squares.

Statistics

All statistical analyses were performed using R. The Shapiro-Wilk test revealed that participant ΔF values were distributed normally. To perform the most direct comparisons to test our hypotheses in the present study, we constructed two linear mixed models with fixed effects as either feedback source or posture (i.e., model 1: seated comparing torque and RMS EMG feedback or model 2: RMS EMG Feedback comparing seated and standing posture); participant, trial, and ramp number were included as random effects to ensure that the independent observations assumption was not violated. This is the most straightforward way to deal with the partially crossed design of the present study and clearly indicates the comparisons that are of most interest, namely, when feedback (torque vs. RMS EMG) is the independent variable and when posture (standing vs. seated) is the independent variable. Post-hoc analyses were performed using a t test using Satterthwaite’s method to account for unequal sample sizes. Significance was set at the α = 0.05 level. Results are all presented as means ± 95% confidence interval unless otherwise stated.

RESULTS

Participants

A total of 15 participants were included for analysis in the present study. Of these, 12 were male, and 3 were female. The mean participant age was 26.6 ± 5.5 (SD) yr. The average participant weight was 78.0 ± 11.4 (SD) kg (average female weight: 62.9 kg; average male weight: 81.8 kg). The participant height was 173.9 ± 11.5 (SD) cm (average female height: 156.0 cm; average male weight: 178.4 cm). Across all participants, we discriminated 1,920 MU discharge patterns during torque feedback trials, 1,808 MU discharge patterns during seated RMS EMG feedback trials, and 1,585 MU discharge patterns during standing RMS EMG feedback trials. The average yield per participant across all three conditions was ∼125 MU discharge patterns across five to six trials. Thus the per participant motor unit yield was ∼20 MU per contraction, although approximately a third of these motor units did not form any acceptable pairs with a lower threshold unit and thus did not generate a ΔF value. Data from a single participant across three different trials are shown in Fig. 3.

Figure 3.

Bottom: individual decomposed motor unit firing patterns (gray dots) fitted by convolution of the spike train with a 1-s Hanning window and truncated at the onset and offset of each motor unit (green trace). Top: 3 identical targets for the 2 distinct types of trials and the torque (black trace) and root mean square (RMS) EMG (orange for seated or blue for standing) signals. All data are from the same participant.

Motor Unit Discharge Rates and Hysteresis

The average and variability of ΔF values for the tibialis anterior across all three conditions were similar to those reported in previous publications (15, 17). There was no evidence for a difference in ΔF in any condition (P = 0.603 and Cohen’s d = 0.107 for the feedback model and P = 0.156 and Cohen’s d = 0.080 for the posture model) as shown in Fig. 4A. Similarly, a mixed effects analysis of the average ΔF coefficient of variation for each participant revealed no evidence for a difference across seated and standing trials, regardless of feedback source (P = 0.949 and Cohen’s d = 0.019 for the feedback model and P = 0.237 and Cohen’s d = 0.393 for the posture model; Fig. 4B).

Figure 4.

The paired motor unit analysis technique was used to calculate ΔF across motor units in all 3 conditions. There was no evidence for a difference in change in discharge frequency of a lower threshold motor unit at the recruitment onset and offset of a higher threshold unit (ΔF; A) or in the coefficient of variation (CV; B) across the conditions The horizontal red bar represents the mean across participants and is bounded by the 95% confidence interval. One participant’s CV is ∼125% during the standing condition (value not shown due to ordinate scaling).

During the seated motor task, initial discharge rates were ∼0.8 Hz greater when participants were presented with RMS EMG visual feedback compared with torque visual feedback (P = 0.012, Cohen’s d = 0.424). In addition, initial discharge rates were greater on average during the standing motor task as opposed to the seated motor task (P = 0.019, Cohen’s d = 0.572), as shown in Fig. 5A. There was no evidence of a difference in the peak discharge rate (P = 0.116, Cohen’s d = 0.139) with a change in visual feedback source (torque or RMS EMG). However, the peak discharge rate was ∼0.7 Hz greater during the standing anterior-to-posterior movements (P = 0.045, Cohen’s d = 0.514; see Fig. 5B). The mixed linear models showed no evidence for a difference in the final discharge rate during either a change in the visual feedback source (P = 0.087, Cohen’s d = 0.275) or in the posture (P = 0.276, Cohen’s d = 0.397).

Figure 5.

The initial (A), peak (B), and final (C) average discharge rates are shown for all participants for each trial type. Both the initial and the peak discharge rates were slightly higher in the standing trials. The horizontal red bar represents the mean across participants and is bounded by the 95% confidence interval. *Significantly different from seated-torque; †significantly different from seated-root mean square (RMS) EMG (with RMS EMG visual feedback).

Task Performance

The torque and RMS EMG signals generated during 20% MVC ramps were analyzed in each trial to calculate the residuals. The average traces across conditions and trials across all participants are shown in Fig. 6. Overall, the average response matched the target reasonably well (see residuals in Fig. 6), although the peak of the ramp did not reach the same height in certain cases, likely due to the instruction to emphasize smoothness of contraction over precisely reaching the peak value. Despite this, the RMS EMG peak amplitude (% MVC) was 16% with torque visual feedback (RMS EMG feedback not visible; Fig. 6, top row, orange trace), 18% with RMS EMG visual feedback during the seated posture (Fig. 6, bottom row, orange trace), and ∼20% with RMS EMG feedback during the standing posture (Fig. 6, bottom row, blue trace). By contrast, the torque peak amplitude (% MVC) was ∼18% with torque visual feedback (Fig. 6, top row, black trace) and ∼21% with RMS EMG visual feedback (torque feedback not visible; Fig. 6, bottom row, black trace) during the seated posture. None of these differences in peak torque or RMS EMG were statistically significant. It is clear from these traces that the relationship between torque and RMS EMG is not one-to-one, but at these lower %MVC values the relationship is relatively linear.

Figure 6.

The average responses of all participants across 3 different trial types (seated-torque feedback, seated-root mean square (RMS) EMG feedback, and standing-RMS EMG feedback). Note that the residuals are comparatively lower for signals that were visible during the performance of the task. To allow comparisons across these different conditions, and because these are average traces from within and then across participants, standard residuals are shown as opposed to the detrended residuals. This representation emphasizes the error at the peak. Root mean square error values were calculated from the detrended residuals. Note that the torque vs. RMS EMG relationships are not one-to-one even during these 20% maximal volitional dorsiflexion contraction tasks, although the most important factors in determine ΔF values seems to be a linear increase and decrease in effort. The signal traces represent the means across the participants and shaded areas represent the 95% confidence interval.

The root mean squared error, shown in Fig. 7, represents the error between the participant-generated ramp and the target ramp for each trial. Unsurprisingly, the mixed linear models revealed that the root mean squared error was greater for the torque signal when the feedback type was the RMS EMG (P = 0.019) and vice versa (P < 0.001 and Cohen’s d ≈ 0.75 for both). This can largely be attributed to the fact that any corrections to the trajectory are made to optimize the visible signal, but the discordant relationship between RMS EMG and torque (shown in Fig. 6) may also play a role. Furthermore, the root mean square error for the RMS EMG signal was higher during the standing postural adjustments compared with the seated isometric ramp tasks, reflecting the lower precision apparent during the standing task (Fig. 7A; also see blue averaged trace with shaded 95% CI in Fig. 6; P = 0.027 and Cohen’s d = 0.75). None of these slight differences point to major differences between contractions during the distinct posture/feedback conditions.

Figure 7.

The root mean square error of the detrended residuals was greater for the signals that were not visible during the task performance. A: root mean square error for the torque and root mean square (RMS) EMG signals during the seated trials with torque feedback. B: root mean square error for the torque and RMS EMG signals during the seated and standing trials with RMS EMG feedback. There is no torque measured during the standing trials. The dots represent individual participants. The horizontal red bar represents the mean across participants and is bounded by the 95% confidence interval. *Significantly different from seated-torque; †significantly different from seated-RMS EMG (with RMS EMG visual feedback).

DISCUSSION

No Evidence for a Change in ΔF between Seated and Standing Posture

In the present study, we found no evidence for a difference in ΔF when participants performed standing anterior-to-posterior movements to linearly increase and decrease their TA RMS EMG activity, thus falsifying our hypothesis that ΔF would be greater in standing compared with seated conditions. The two main explanations of these results can take either of the following forms: following a change in posture from seated to standing, 1) the amplitude of the PIC truly was not altered, and thus ΔF did not change, or 2) the level of neuromodulation changed, but other factors (e.g., reciprocal inhibition due to activation of the antagonist muscle for stabilization) caused ΔF to be similar between conditions. Since PICs are assumed to have the functional goal of posture maintenance (7, 14, 41), it would seem that the level of neuromodulatory input (as estimated by ΔF) should increase to accommodate a standing posture. If the first form of the argument is true, then we must grapple with the result that changes in posture of the participant do not seem to change our estimates of the PIC amplitude. In this sense, our results may seem counterintuitive and seem to contradict qualitative assessments, which have suggested that rats have less erect posture following serotonin depletion (14). Additionally, studies in cats have suggested that neuromodulatory drive seems to be directly proportional to the arousal state and intensity of activity. For example, the drive increases when moving from a baseline seated posture to walking (10, 11). Increasing the speed of treadmill walking or the depth of respiration (11) both increase the magnitude of medullary serotonergic neuron activation. An important caveat here is that, at least in the Veasey et al. (11) study, their seated and standing postures were in the quiet waking state, which suggests that they also find no evidence of a difference between the two conditions. In light of this, our findings do not contradict but rather match observations from the nonhuman animal data. A study of the soleus muscle with a similar study design to the present investigation reported similar findings (16), further bolstering the claim that there is no evidence of an effect on ΔF of a seated versus a standing posture. In contrast to the HDsEMG used in the present study, Foley and Kalmer (16) obtained single motor unit recordings using intramuscular wire EMGs. Although this reduced their motor unit yield substantially, the benefit of this approach was that they were able to track the same motor units across conditions. Despite these differences in technique, the results of the present study are similar to those of Foley and Kalmar (16).

Another explanation of our results could be that that the level of neuromodulation was increased, but ΔF remained similar between the seated and standing postures due to various sources of inhibition counteracting the expected increase in the amplitude of the PIC. The PIC is extremely sensitive to inhibition (42), likely due to the magnitude of dendritic depolarization (43–45). Coactivation of the antagonist muscles in the lower limb could, in theory, increase the inhibition to the TA. There is evidence that reflex pathways may be differentially regulated with posture (46–50). Despite the relatively short time constant of these reflex pathways, changes in posture (e.g., sitting vs. standing) likely cause overall systemic effects. These changes likely reflect the types of changes that could result in inhibition of PICs. Most of these data in the legs are from soleus muscle, owing to the difficulty of eliciting reflexes in the TA (see Table 1 in Ref. 50). Although the H reflex is commonly assumed to reflect changes in the excitability of a MN pool, changes in the size of the H reflex can also result from changes in transmitter release and in presynaptic inhibition (50). Thus any evidence of a change in the H-reflex should not be interpreted merely as a change in MN excitability. It seems relatively uncontroversial to say that the H-reflex amplitude decreases, and the level of presynaptic inhibition increases with increasing intensity of activity (46, 47, 50, 51). The synopsis of the H-reflex data in the leg muscle, including the TA (49), seems to be that there is either a minimal difference or no difference in the normalized H-reflex amplitude between sitting and standing, though these responses are variable and inconsistent across participants (48, 52, 53). During standing postural sways, there is some evidence of increased corticospinal drive to TA motor units in the TA muscle (54), which may suggest that ΔF should increase in the standing condition. However, the relation between transient reflexes and PICs is not necessarily direct and certainly not well understood; after all, the time constant of the calcium PIC is around an order of magnitude greater than that of the H-reflex.

On the other hand, the magnitude of the reciprocal inhibition to various leg muscles may change when participants are standing compared with when they are seated, thus counteracting any apparent increase in ΔF. There is some evidence suggesting a much greater strength of inhibitory pathways to the TA compared with the soleus (55) and this could theoretically explain our apparent lack of a difference in ΔF. However, for ΔF to be counteracted by inhibitory mechanisms in the TA in the present study and the soleus in the study of Foley and Kalmar (16), there would have to be complex fine tuning that elicits the same result in both muscles, i.e., a counteracted ΔF between the seated and standing postures. It is true that we cannot exclude this possibility and there may in fact be this level of complex control, although Foley and Kalmar (16) found that the level reciprocal inhibition to the soleus was not different between the standing and seated postures. Although we cannot exclude the possibility that this is not the case in the TA, the most parsimonious explanation seems to be that there is no change in ΔF between standing and seated postures in the TA. Additionally, the postural differences in the present study may also change the overall maximal force generating capacity of the TA muscle (e.g., in the standing case both legs are active during the task whereas in the seated case only one leg is). Although these potential differences in MVC seem unlikely to affect ΔF (15), they may explain the higher discharge rates during the standing condition. Oscillatory changes in the TA fascicle length during the standing anterior-to-posterior movements could also affect ΔF (13, 41, 56). Oscillatory states in motor neurons do not seem to rely heavily on the Ca PIC but rather a self-inactivating NMDA channel (57), which would effectively lower ΔF despite the increased postural demands of the standing condition. These are all complications worth considering in interpreting the data in the present study.

One aspect of postural differences that seems to be underemphasized in the literature is that both the seated and standing conditions are, after all, still upright postures. It may be the case that, similar to the trends in the H reflex (48, 50), ΔF may be different between a lying down posture versus a standing/seated posture and likely elevated during walking. These types of predictions should be tested in future work to understand the role of neuromodulation in functional postural and pattern generator-influenced behaviors.

Practical Applications

In the present study, we demonstrate the utility of RMS EMG as a visual feedback source during motor tasks where torque or force measurements are more difficult (e.g., during functional behaviors and/or in a clinical setting). We have shown that HDsEMG arrays can be used to discriminate up to 30 motor units from the TA muscle during an unconstrained nonisometric task, namely, a standing anterior-to-posterior ramped contraction task. These findings highlight the expansive possibilities of HDsEMG in performing assessments during functional behaviors. Despite the obvious benefits of studying participants performing contrived motor tasks such as slowly ramping contractions in constrained isometric settings, these tasks may be difficult to relate to real-life, often nonisometric, behaviors. In addition, regardless of the signal type (torque or RMS EMG) provided as visual feedback during the ramping contraction task, participants were able to control their contractions well enough to generate an approximately linear increase and decrease in force.

Our findings suggest that in both clinical settings as well as in cases where torque measurements are a limiting factor, studies on estimates of PICs can still be performed relatively easily using RMS EMG feedback. Additionally, in the present study, we demonstrate that not only can large numbers of motor units be decomposed during free standing anterior-to-posterior movements, but a sufficient number of motor unit pairs meet the discharge rate criteria to allow for ΔF to be calculated. Thus studies on PIC activity can potentially be performed during more functional behaviors that are approximately linearly increasing and decreasing. It is important to interpret these types of data in light of the changes in joint angles, MVC values, and task specificity; thus considerable work needs to be done to more fully describe the role of spinal motoneuron excitability during functional tasks. Interestingly, the coefficient of variation for ΔF (Fig. 6) was also similar across the different conditions. Thus the differences in root mean squared error (Fig. 5) with different types of visual feedback and/or posture did not correspond to increased variability in the ΔF. This suggests that the paired motor unit analysis technique is robust as long as motor units remain activated once recruited and the activity is increasing and decreasing in a relatively linear fashion. This makes some sense given that the paired motor unit analysis technique relies primarily on the recruitment onset and offset of the test unit.

It is true, as is evident from the residuals in Fig. 6, the root mean square error (Fig. 7), and subjective reporting from participants during the experiment, that these ramping tasks were more difficult to perform when participants received RMS EMG visual feedback. Regardless, although there was a slight difference in the torque (%MVC) and RMS EMG (%MVC) both within and across feedback sources, the general linearly increasing and decreasing pattern was conserved. Furthermore, at least two studies with ramping effort protocols similar to ours have now demonstrated that there is no effect on ΔF of increasing contraction intensities between 10 and 30% MVC (15, 17). On the other hand, a recent study that clamped the rate of rise between the different contraction intensities, detected a difference of ∼0.5 pps in ΔF between across contraction intensities ranging between 10–30% MVC (58). Given that the maximum difference between the torque and RMS EMG (both %MVC) responses is ∼5% in the present study, differences in contraction intensity alone are unlikely to explain any differences (or lack thereof) in ΔF. Additionally, if the small increase in contraction intensity that we observed when participants received RMS EMG feedback was liable to alter ΔF, one would expect ΔF to increase (including in the seated RMS EMG case); thus the feedback source is unlikely to play a counteracting role to the increase expected in the standing condition. In the present study, ΔF is almost identical regardless of the feedback source during the seated isometric experiments (see Fig. 4; Torque visual feedback, ΔF = 4.73 imp/s; RMS EMG visual feedback, ΔF = 4.55 imp/s).

Conclusions

We show that ΔF for the TA is similar between standing and seated postures, which matches the findings of a previous study in the soleus muscle (16). However, we acknowledge the possibility that the neuromodulatory input likely changes with more drastic postural changes, such as lying down, and activity changes, such as walking or running. Therefore, future studies should investigate the role of more drastic postural changes on these PICs. Additionally, our findings also augur hitherto unexplored potentials for the use of HDsEMG arrays. We show that real-time EMG visual feedback as opposed to force or torque feedback is sufficient for obtaining ΔF measurements, which suggests that these techniques may be used in muscles where mechanical feedback measures are untenable. This is important lest the wellspring of human motor output is inundated with data from the limbs; after all, some of the most important muscles for sustaining life in mammals are near the core (i.e., respiratory muscles). Finally, our findings suggest the potential for using measures such as ΔF clinically without the complication of precise force measurements under strict isometric conditions.

GRANTS

This research was supported by National Institutes of Health Grants R01NS098509 (CKT, CJH) and F32HL151251 (OUK).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

O.U.K. and C.K.T. conceived and designed research; C.K.T. performed experiments; O.U.K., F.N., and C.K.T. analyzed data; O.U.K., F.N., C.J.H., and C.K.T. interpreted results of experiments; O.U.K. prepared figures; O.U.K. drafted manuscript; O.U.K., F.N., C.J.H., and C.K.T. edited and revised manuscript; O.U.K., F.N., C.J.H., and C.K.T. approved final version of manuscript.