Motor unit discharge rate modulation during isometric contractions to failure is intensity- and modality-dependent
Handling Editors: Paul Greenhaff & Martino Franchi
The peer review history is available in the Supporting Information section of this article (https://doi.org/10.1113/JP286143#support-information-section).
This article was first published as a preprint. Valenčič T, Ansdell P, Brownstein CG, Spillane PM, Holobar A, Škarabot J. 2023. Motor unit discharge rate modulation during isometric contractions to failure is intensity and task dependent. bioRxiv. https://doi.org/10.1101/2023.12.04.569929
Abstract
Key points
- During prolonged muscle contractions a constant motor output can be maintained by recruitment of additional motor units and adjustments in their discharge rate.
- Whilst contraction-induced decrements in neuromuscular function are known to depend on task demands, task-specificity of motor unit discharge behaviour adjustments is still unclear.
- In this study, we tracked and compared discharge activity of several concurrently active motor units in the vastii muscles during different submaximal isometric knee extension tasks to failure, including intermittent vs. sustained contraction modalities performed in the same intensity domain (Experiment 1), and two sustained contractions performed at different intensities (Experiment 2).
- During each task, motor units modulated their discharge rate in a distinct, biphasic manner, with the modulation pattern depending on contraction intensity and modality.
- These results provide insight into motoneuronal adjustments during contraction tasks posing different demands on the neuromuscular system.
Introduction
Voluntary muscle force or torque production is regulated via activity of spinal motoneurons and groups of muscle fibres innervated by individual axons (i.e. motor units). When recruited, motor units (MUs) modulate their discharge rate in response to excitatory and inhibitory synaptic inputs to the motoneuron pool from descending supraspinal pathways, spinal interneurons and muscle afferents (Heckman & Enoka, 2012). With continuous activation, such as during prolonged muscle contractions, neural and contractile properties of the motor pathway exhibit alterations, leading to a progressive reduction in muscle force production capacity and, ultimately, task failure (Enoka & Duchateau, 2016; Enoka & Stuart, 1992). When the motor output is submaximal, the net torque can be maintained by the recruitment of additional MUs, and adjustments in their discharge rate via increases in the excitatory synaptic input (Enoka & Stuart, 1992; Hunter, 2018). Whilst previous literature consistently suggests progressive increases in the number of identified MUs during prolonged submaximal isometric contractions (Adam & De Luca, 2005; Carpentier et al., 2001; Castronovo et al., 2015; Contessa et al., 2016; Martinez-Valdes, Negro, Falla et al., 2020; Maton, 1981; Person & Kudina, 1972), the reported patterns of MU discharge rate modulation during prolonged contractions have been inconsistent. Most commonly a gradual decrease has been observed (Carpentier et al., 2001; Garland et al., 1994; Lowe et al., 2023; McManus et al., 2016; Vila-Chã et al., 2012), though others have also reported an increase (Bigland-Ritchie et al., 1986; Kuchinad et al., 2004), or a biphasic modulation pattern, whereby an initial decrease was followed by a progressive increase until task failure (Adam & De Luca, 2005; Griffin et al., 2000; Martinez-Valdes, Negro, Falla et al., 2020; Mettler & Griffin, 2016). These inconsistencies are thought to be at least partly attributed to the variety of exercise tasks employed in the previous studies, with different contraction durations, intensities and modalities (i.e. sustained and intermittent contractions) being studied. Considering that physiological mechanisms underpinning the progressive, contraction-induced decline in the maximal muscle force- or torque-production capacity and time to task failure depend on task demands (Enoka & Stuart, 1992; Hunter, 2018), MU discharge behaviour adjustments may also be task-specific. This hypothesis, however, remains to be tested.
Besides the contraction intensity, a contraction characteristic with the potential to influence neural control of the muscle during prolonged submaximal isometric contractions is contraction modality. Namely, the physiological responses to prolonged sustained and intermittent muscle contractions, even when performed in the same exercise intensity domain, differ due to differences in muscle perfusion and neural activation, with potential implications for MU discharge rate characteristics (Taylor et al., 2016; Hunter, 2018). Whilst prolonged sustained contractions critically limit muscle perfusion even at relatively low forces, causing muscle ischaemia and accelerated accumulation of contractile function-impairing anaerobic metabolites in the muscle (Sjøgaard et al., 1988), brief intermittent contractions are each followed by a hyperaemic response, which decelerates metabolic and ionic perturbations (Broxterman et al., 2014; Chidnok et al., 2013; Monod & Scherrer, 1965). Furthermore, differences in spinal reflex regulation and/or motoneuron activation between intermittent and sustained contraction could also influence MU discharge rate. For example, during sustained isometric contractions, motoneurons are tonically active, leading to intrinsically mediated reductions in responsiveness of motoneurons to constant synaptic input, shown as a reduction in motoneuron discharge rate (i.e. spike frequency adaptation). Such an observation has been noted in both animal preparations using intra- or extracellular current injections (Button et al., 2007; Kernell & Monster, 1982; Sawczuk et al., 1995; Spielmann et al., 1993) and in humans as reductions in the size of evoked response to direct stimulation of corticospinal axons (Brownstein et al., 2021; Finn et al., 2018). Conversely, during intermittent contractions, motoneuron activation is interrupted by periods of rest. Using extracellular stimulation of spinal motoneurons in an anaesthetised cat preparation, intermittent stimulation has been demonstrated to lead to lower levels of spike frequency adaptation compared to constant stimulation (Spielmann et al., 1993). Furthermore, the greater metabolic demands of sustained contractions probably increase group III/IV afferent discharge, leading to greater reductions in motoneuron and cortical excitability (Hilty et al., 2011; Martin et al., 2006), and thus possibly affecting MU discharge rate modulation in a task-specific manner. Indeed, greater muscle perfusion has been shown to induce smaller MU discharge rate modulation and lower MU discharge rate variability (Lowe et al., 2023). Similarly, excitatory spinal reflex input appears unaffected by intermittent contractions but reduced during sustained contractions to failure (Duchateau et al., 2002), with a probable influence on task-specific MU discharge rate modulation patterns.
In this study, we aimed to examine and compare MU discharge rate modulation of the vastii muscles during sustained and intermittent submaximal isometric knee extension tasks to failure. To assess the dependency of MU discharge rate adjustments on contraction modality (Experiment 1), the intensity of the tasks [20% and 50% of maximal voluntary torque (MVT) for the sustained and intermittent task, respectively] was chosen to induce similar declines in MVT, voluntary activation, and muscle contractile properties at task failure (i.e. above the critical torque for the respective contraction modality (Burnley, 2009; Monod & Scherrer, 1965). Because we could not exclude the possible confounding effects of contraction intensity to the observed MU discharge behaviour, we performed an additional experiment consisting of two sustained tasks to failure performed at different intensity levels (Experiment 2). For both experiments, it was hypothesised that the time-dependent changes in MU discharge behaviour will differ between the two experimental tasks (sustained vs. intermittent, or sustained contractions at different intensities).
Materials and methods
Participants and ethical approval
Twenty-two healthy volunteers (11 males and 11 females) participated in Experiment 1. In 13 participants, we were unable to identify a sufficient number of MUs from the recorded signals (i.e. we were not able to track at least one MU throughout the entire task in at least one muscle), and thus the final sample on which results are reported here included nine participants [one female and eight males; 26 (6) years, 1.75 (0.06) m, 73.9 (9.3) kg]. In Experiment 2, eight healthy adults participated [one female and seven males; 29 (4) years, 1.77 (0.08) m, 75.8 (12.0) kg]. The experiments were performed approximately 2 years apart, with one male and one female having taken part in both experiments.
Exclusion criteria for both experiments included any current injury to the lower limbs, back or spine, a history of a major traumatic injury or surgery, a musculoskeletal or neuromuscular disorder affecting the function of the major joints of the lower limb, and taking any medication known to influence neuromuscular function. To minimise the effects of the menstrual cycle on neuromuscular function (Ansdell, Brownstein, Škarabot, Hicks, Simoes et al., 2019; Piasecki et al., 2023), female participants were tested in the early follicular phase of the cycle (days 1–7) or during the period of contraceptive monophasic pill consumption. The female included in the final analysis in both experiments was naturally cycling. Before taking part in experimental procedures, participants provided written informed consent. The experiments were approved by Loughborough University Ethical Advisory Committee (2020-1733-1674 and 2023-14760-15109) and were conducted in accordance with the Declaration of Helsinki, except for registration in a database.
Experimental design and protocol
Experiment 1
Participants arrived at the laboratory on three separate occasions at a consistent time of day (±1 h), 2–10 days apart. Participants were instructed to arrive rested and hydrated, having avoided strenuous exercise, and refrained from alcohol and caffeine consumption for 24 and 12 h prior to arrival, respectively. All sessions involved voluntary and evoked isometric knee extensions of the dominant limb (as determined by the lateral preference inventory; Coren, 1993), with participants seated in a rigid, custom-built isometric dynamometer. The knee- and hip-joint angles were fixed at 115° and 126° (180° = full extension), respectively, with the testing chair adjusted to each participant's stature and kept consistent across the sessions. During the performance of voluntary and evoked isometric contractions, participants were firmly strapped across the waist and chest to minimise extraneous movement.
During the familiarisation session (visit 1), participants practised the performance of maximal and submaximal isometric knee extension contractions and were habituated to percutaneous femoral nerve stimulation that was used in experimental sessions for assessment of quadriceps voluntary activation level and evoked contractile muscle properties. Participants also practised the two contraction tasks, with the practice time standardised for all participants: 30 and 120 s for the sustained (20% of maximal voluntary contraction, MVC) and intermittent (50% MVC) contraction tasks, respectively.
During the experimental sessions (visits 2 and 3), participants performed the contraction tasks until failure in a randomised, crossover design. Following a standardised warm-up of 3 s submaximal isometric knee extensions at 50% (×3), 75% (×3) and 90% (×1) of perceived maximal effort (separated by 15–30 s of rest), participants performed five ∼3 s MVCs separated by 30–60 s of rest to assess maximal voluntary force. Quadriceps voluntary activation level and evoked contractile muscle properties were assessed by delivering an electrical pulse to the femoral nerve during and ∼2 s after the final three MVCs. To facilitate maximal force production, biofeedback was provided by displaying a real-time force trajectory on a computer monitor ∼1 m in front of the participant, with a horizontal cursor indicating the greatest force achieved. Participants were instructed to ‘push as hard as possible’ and strong verbal encouragement was provided. No instruction was provided to participants regarding the rate of force increase during maximal effort contractions. The peak instantaneous force attained during the MVCs was used as a reference to calculate the submaximal force targets. Then, participants performed a trapezoidal ramp contraction at 70% MVC, increasing the knee extensor force output to the target level at 10% MVC/s, maintaining the target force level for 10 s and decreasing the force level at 10% MVC/s to resting levels. This allowed quantification and comparison between MU discharge rate at task failure relative to MU discharge rate during a higher intensity contraction performed in a fresh state.
After a 5 min rest period, participants performed either a sustained or an intermittent isometric knee extension task to failure at 50% and 20% MVC, respectively (randomised). These two specific contraction intensities were chosen to ensure participants performed contractions above the respective critical torque of the modality; that is, at an intensity above the asymptote of the relationship between the isometric muscle torque and time to task failure, which demarcates the transition between metabolically steady state and unsustainable exercise (Burnley et al., 2012; Pethick et al., 2020). It has been demonstrated that the critical torque corresponds to ∼15% MVC for sustained contractions (Monod & Scherrer, 1965), and ∼35% MVC for intermittent contractions (Burnley et al., 2012). Consequently, the intensities used in the present study were ∼30–40% above previously reported critical torque values, to induce similar reductions in maximal voluntary torque, voluntary activation and potentiated twitch amplitude in both tasks, thus allowing the assessment of the modulation of MU discharge properties as a function of contraction modality without the confounding effect of differences in decrements of neuromuscular function. In short, had we compared sustained versus intermittent contractions at 20% MVC, the physiological demands and magnitude of neuromuscular impairments during the tasks would have differed, as 20% MVC intermittent contractions result in a target level below critical torque for most individuals (Burnley et al., 2012).
The intermittent contraction task involved a 2.75:1 (contraction:rest), 15 s duty cycle consisting of a 3 s ramp increase in force output to the target level, a 5 s hold phase at the target level and a 3 s ramp decrease to relaxation, followed by a 4 s rest. The selection of such a duty cycle, rather than, for example, a longer hold at target with shorter rest periods, also ensured that intermittent contractions were metabolically distinct from sustained contractions (Broxterman et al., 2014). Participants were instructed to match a trapezoidal waveform with a real-time force trajectory displayed on the computer monitor. During the sustained contraction tasks, the force target was indicated by a horizontal line, which the participants were instructed to match as closely as possible. Time to task failure was recorded to the nearest second after three consecutive failures to meet the target force for the entire 5 s during the intermittent contraction task, and after a sustained 3 s decline in the force output by >10% of the target in the sustained contraction task, despite strong verbal encouragement. No feedback of the elapsed time was provided. Immediately after reaching task failure, participants performed three additional MVCs with percutaneous femoral nerve stimulation ∼10 s apart. High-density surface EMG (HDsEMG) recordings were performed throughout the sessions.
Experiment 2
Because the sustained and intermittent isometric contractions to failure in Experiment 1 were performed at different relative force levels, we could not exclude the potential influence of contraction level on the differential MU discharge modulation (see Results). Additionally, considering the decomposition bias towards the identification of higher threshold MUs (Francic & Holobar, 2021), we could not exclude differences in MU recruitment threshold as a factor in the observed MU discharge behaviour. Therefore, we performed an additional experiment where we assessed MU discharge behaviour during two isometric sustained contraction tasks at different contraction levels to failure. We chose to focus only on sustained contractions due to lower critical torque levels compared to intermittent tasks (Burnley, 2009; Monod & Scherrer, 1965), and thus greater flexibility in the selection of contraction intensities that were above critical torque, but equally not too great as to risk compromising HDsEMG decomposition and identification and tracking of MU discharge behaviour. The main difference in Experiment 2 was the contraction intensity; here, participants performed sustained isometric knee extension to task failure at 30% and 50% MVC in a randomised, crossover design on two separate days (at least 48 h, but no more than 10 days apart). We selected the 30% MVC (rather than 20% MVC) task to allow a more direct comparison of MU discharge behaviour with other studies in the literature (Martinez-Valdes, Negro, Falla et al., 2020; Rossato et al., 2022). Before the contraction tasks to failure, participants also performed two trapezoidal ramp contractions at 50% and 70% MVC to quantify potential differences in MU discharge rate at task failure relative to MU discharge rate during a higher intensity contraction in a fresh state. HDsEMG signals were recorded throughout the contraction tasks, and femoral nerve stimulation was delivered during and a few seconds after each of the three maximal effort contractions before and after the sustained isometric knee extension tasks.
Experimental procedures
Force recordings
The force signal was measured using a calibrated S-beam strain gauge (linear range: 0–1.5 kN; Force Logic, Swallowfield, UK) attached perpendicularly to the participant's tibia using a bespoke reinforced non-extendable webbing strap (35 mm width) tightly fastened superior to the lateral malleolus at ∼15% of the tibial length (distance between the lateral malleolus and the knee joint centre). The analogue force signal was amplified (×370) and sampled at 2048 Hz simultaneously via an analogue-to-digital converter [Micro 1401; Cambridge Electronics Design Ltd (CED), Cambridge, UK] for recordings of maximal voluntary force, voluntary activation and contractile twitch properties in Spike2 software (version 10; CED), as well as via a 16 bit multichannel amplifier (Quattrocento, OT Bioelettronica, Torino, Italy) for synchronisation with HDsEMG signals.
Femoral nerve stimulation
Percutaneous electrical stimulation (single square-wave pulse, 1 ms, 300 V; DS7AH, Digitimer, Welwyn Garden City, UK) of the femoral nerve was delivered via circular self-adhesive gel-coated cathode and anode (32 mm2; CF3200, Nidd Valley Medical, Bordon, UK) positioned high in the femoral triangle and mid-way between the greater trochanter and iliac crest, respectively. The position of the cathode was adjusted in small steps to the position that elicited the greatest quadriceps twitch response with a constant stimulation intensity. The stimulation current was determined by increasing the pulse intensity from 20 mA in 20 mA increments until a plateau in the resting quadriceps twitch force was attained. Stimulus intensity was then increased by 30% to ensure supramaximal stimulation (Experiment 1: 355 ± 113 mA; Experiment 2: 367 ± 135 mA).
High-density surface electromyography
The HDsEMG signals were recorded from the vastus lateralis (VL) and medialis (VM) muscles in monopolar configuration using semi-disposable 64-electrode grids (5 rows × 13 columns; 1 mm electrode diameter; 8 mm interelectrode distance; OT Bioelettronica) attached to the skin surface using disposable adhesive foam interfaces (SpesMedica, Battipaglia, Italy). Following skin preparation by shaving, light abrasion and cleaning with ethanol, the adhesive foam interfaces were filled with conductive paste (SpesMedica), and the grids were positioned over the VL and VM muscles at ∼20° and ∼50° relative to the line between the superior iliac spine and the lateral and medial border of the patella, respectively. Muscle borders and the grid alignment with the orientation of fibres was confirmed by ultrasound (EUB-8500; Hitachi Medical Systems UK Ltd, Northamptonshire, UK; 92 mm linear-array transducer, EUP, L53L). Two self-adhesive reference electrodes (36 mm2; Cardinal Health, Dublin, OH, USA) were attached across the patella of both limbs and the signal was grounded via a water-soaked strap electrode placed superior to the malleolus of the non-dominant limb. The HDsEMG signals were sampled at 2048 Hz and amplified using a 16 bit multichannel amplifier (Quattrocento, OT Bioelettronica; 3 dB), band-pass filtered (10–500 Hz) and recorded in OTBioLab+ software (OT Bioelettronica).
Data analysis
Voluntary and evoked torque
Force data were low-pass filtered at 20 Hz (zero-lag, fourth-order Butterworth digital filter), gravity corrected and converted to torque with multiplication by lever length (distance between the knee-joint centre and the centre of the ankle strap). The peak torque value achieved during maximal voluntary contractions was denoted as MVT. Voluntary activation level was quantified by comparing the amplitude of the superimposed twitch (SIT) with respect to the potentiated quadriceps twitch (Qtw) using the established equation (1 − [SIT/Qtw]) × 100 (Merton, 1954). A correction equation implementing torque at SIT (TSIT) was employed if the timing of the delivered stimulus did not correspond to the proximity of peak torque (100 − SIT × [TSIT/MVT]/Qtw × 100 (Strojnik & Komi, 1998). For MVT, VA and Qtw, the values of the three contractions before and after the fatiguing tasks, respectively, were averaged, unless a trial deviated from the average by more than two standard deviations in which case it was excluded from the calculation of mean. Torque steadiness during the two tasks was quantified as a coefficient of variation of torque (the ratio of standard deviation and mean torque levels) in the same time windows used for quantifying discharge rate characteristics (see below).
HDsEMG decomposition and motor unit tracking
Offline analyses were performed using MATLAB (R2023a; Mathworks Inc., Natick, MA, USA). Monopolar HDsEMG signals were digitally band-pass filtered with a fourth-order, zero-lag Butterworth filter (20–500 Hz). We then removed channels exhibiting poor signal-to-noise ratio, noise or artefacts using a semi-automated tool in MATLAB (DEMUSE, University of Maribor, Slovenia). Specifically, channel selection was based on the outlier detection technique. Using the correlation coefficient measure, each HDsEMG channel was pairwise compared to its direct topological neighbours. Average correlation values with neighbouring channels were calculated for each channel and up to 5% of channels with minimal average correlation value were discarded. HDsEMG signals were decomposed using the extensively validated Convolution Kernel Compensation algorithm (Holobar & Zazula, 2007). This algorithm is based on the blind source separation principles whereby the EMG mixing model is inverted, and MU filters are estimated, yielding the estimation of an MU spike train (Holobar & Farina, 2021; Škarabot, Ammann et al., 2023).
The calculation of MU filters and their application to a different portion of the signal from the same contraction or another contraction allows identification of discharges from the same MUs, even in the case of filter application to higher contraction intensities (Francic & Holobar, 2021), or different levels of MU synchronisation (Škarabot, Ammann et al., 2023). These factors are important for tasks in the present study where recruitment of additional MUs is likely (and thus similar to higher contraction levels, leads to the presence of MUs with larger potentials in the signal), as are increases in MU discharge rate, which is mathematically linked to higher MU synchronisation (De La Rocha et al., 2007). Thus, to track motor units across the contraction tasks to failure we used an approach of estimation and application of MU filters in shorter, overlapping windows (Rossato et al., 2022). For the sustained task, decomposition of the HDsEMG signal was performed independently in 30 s windows at the beginning and end of the signal (Fig. 1A). From these decomposed sections of 30 s, MU spike trains were first visually inspected (Fig. 1Bi), and MU filters were iteratively optimised using standard procedures, segmenting the MU firings from noise/crosstalk from other MUs (Del Vecchio et al., 2020). After that, MU filters were iteratively applied in 15 s windows using a 50% overlap (Fig. 1Biii) to the rest of the signal upon which the MU filters were again optimised before further application (Fig. 1Biv).
A similar approach was used for the intermittent task, but here, 41 s portions of the signal (three contractions; Fig. 1C) were decomposed at the beginning and end of the task. MU spike trains were then visually inspected (Fig. 1Di), with MU filters optimised through segmentation (Fig. 1Dii), and then applied in 15 s windows (one contraction) using a 30 s (two contractions) overlap (Fig. 1Diii), followed by another optimisation of MU filters (Fig. 1Div). Because of successive independent decompositions (i.e. decomposition of the first and final portion of the tasks), several MU duplicates were inevitably identified. Duplicates were considered those that shared at least 30% of the same discharges (discharge match tolerance: 0.5 ms). Duplicates with the lower accuracy of MU spike train estimation (based on pulse-to-noise ratio; Holobar et al., 2014) were discarded. Only MU spike trains with a reliable discharge pattern and a pulse-to-noise ratio >28 dB were kept for analyses. Furthermore, only MU spike trains for which we could reliably identify the recruitment and derecruitment time were kept for analysis. Indeed, many more MUs were identified during the initial decomposition; however, probably due to changes in MU action potential shape and superimposition of MU action potentials (i.e. the presence of MUs with larger potentials with additional recruitment of MUs), some of the spike trains exhibited the presence of multiple MUs which impeded segmentation of true discharges from noise/crosstalk (Fig. 2A). Many such cases were discarded from further analysis (for example, see Fig. 2B).
Discharge rate characteristics
Motor unit discharge rate was calculated as the mean of the reciprocal of the interspike interval in 1 s epochs during the stable region of the sustained and intermittent tasks (Fig. 3). For the latter, this meant that the discharge rate was calculated in 1 s epochs during the plateau regions of the repeated trapezoidal contractions. Discharges involving interspike intervals >500 or <20 ms were excluded from analysis to minimise the confounding effect of sporadic discharges and/or potential decomposition errors on the subsequent calculations. Recruitment time was calculated as the time in the task at which a given MU started discharging (Fig. 3A). During the intermittent tasks, we could also reliably estimate recruitment and derecruitment thresholds, which were calculated as the mean torque level corresponding to the first and last three MU spikes, respectively. All variables were averaged using time-normalised intervals with length equal to 10% of the total contraction time (i.e. 10–100% of total contraction time in 10% increments). For trapezoidal ramps performed before the intermittent (Experiment 1) and sustained tasks (Experiment 2), the average discharge rate during the 10 s plateau region of the contraction was calculated. For comparative purposes, similar calculation was performed on the final 10 s of the sustained tasks, and the final two contractions of the intermittent task (2 × 5 s plateau). For this analysis, we only compared discharge rate in the final 10 s of the sustained tasks with discharge rate during a trapezoidal ramp contraction at a level 20% greater than that of a sustained task (i.e. for 30% MVT sustained task comparison was made with a 50% MVT trapezoidal ramp contraction, whereas the 50% MVT sustained and intermittent tasks was compared to a 70% MVT trapezoidal ramp). In Experiment 1, this analysis was only performed for the intermittent task because, unlike the 20% MVT sustained task, discharge rates increased above baseline (see ‘Results’).
To quantify biphasic MU discharge rate modulation during the tasks (see ‘Results’) and its relation to time to task failure, we also estimated the inflection point of MU discharge rate modulation. The inflection point was calculated as the intersection of the bilinear fit of the mean MU discharge rate in 1 s epochs with respect to torque (Fig. 3B). The computation of the bilinear fit involved an iterative process that minimised the fit error of the MU discharge rate. Note that for this analysis, only MUs recruited at the start of the task were used.
Interference EMG amplitude
To characterise the interference EMG amplitude, we computed the root mean square of a single bipolar recording derived from the HDsEMG signals. For this, we averaged the activity of two sets of six channels in the central portion of the bidimensional electrode grid (columns 2–4, rows 4–5 and 6–7), then differentiated the signals (interelectrode distance 16 mm). Root mean square values were then calculated in the same time windows used for quantifying discharge rate characteristics (Fig. 3). The interference EMG amplitude during the two exercise tasks was normalised to maximal interference EMG amplitude in a moving 1 s root mean square window during the performance of MVC.
Statistical analysis
All statistical analyses were performed in R (R Foundation for Statistical Computing, Vienna, Austria). To assess differences in time to task failure, a paired t test was performed. We used linear mixed effects models to assess whether discharge rate and interference EMG amplitude are dependent on fixed effects of contraction time (10–100%), task (intermittent, sustained), muscle (VL, VM) and their interaction, with participant used as a random intercept. Because additional MUs were recruited throughout the duration of the task, recruitment time was employed as a covariate. Modulation in recruitment and derecruitment thresholds was assessed for the intermittent task only, employing contraction time, muscle and their interaction as fixed effects in the linear mixed model. Muscle was removed as a fixed factor when comparing modulation of torque steadiness during the two tasks. Changes in MVT, VA and Qtw were assessed using time (pre, post task), task and their interaction as fixed effects. In Experiment 2, contraction level was used as a fixed factor instead of task. To compare differences in discharge rate during trapezoidal ramp contractions and the final 10 s of the intermittent (Experiment 1) and sustained tasks (Experiment 2), contraction (trapezoidal ramp, sustained), task (Experiment 2 only: 30% vs. 50% MVT sustained), muscle (VL, VM) and their interactions were considered as fixed effects, with participant taken as a random intercept. During statistical testing, further consideration was given to the potential effect of sex on outcome variables, particularly in light of potential differences in motor unit discharge behaviour (Jenz et al., 2023) and time to task failure during isometric contractions (Ansdell, Brownstein, Škarabot, Hicks, Howatson et al., 2019). However, adding sex as a random effect did not improve model fits (see Appendix). Furthermore, we also performed analyses without the inclusion of the female participant, but this did not affect the significance of the results (see Appendix for details).
The significance of all models was assessed using analysis of variance with Satterthwaite's method (lmerTest package; Kuznetsova et al., 2017). The significance level was set at an alpha level of 0.05. In cases of significant main effects or interactions, pairwise post hoc tests of estimated marginal means with Tukey adjustment for multiple comparisons were performed (emmeans package; Lenth & Lenth, 2018).
Results
Experiment 1
Time to task failure and decrements in neuromuscular function
Example torque traces of the intermittent and sustained tasks are depicted in Fig. 4A. Time to task failure was 531 (134) and 217 (82) s for the intermittent and sustained task, respectively (t8 = −7.0, P = 0.0001). The knee extensor MVT (F1,24 = 180.0, P < 0.0001; Fig. 4B), voluntary activation level (F1,21 = 22.8, P = 0.0001; Fig. 4C) and potentiated twitch amplitude (F1,21 = 32.0, P < 0.0001; Fig. 4D) decreased after fatiguing tasks. No significant interactions were found between task and contraction time for MVT, voluntary activation and potentiated twitch amplitude (P ≥ 0.1516), suggesting the tasks induced similar levels of neuromuscular function decrements.
Torque steadiness and global EMG activity
The coefficient of variation of torque was influenced by contraction time (F9,142.0 = 18.0, P < 0.0001), with it being greater than baseline at task failure for the intermittent task (P = 0.0021). During the sustained task, the coefficient of variation of torque increased above baseline at 80% of contraction time (P = 0.0136), followed by a continuing increase until task failure (P < 0.0001; Fig. 5A).
The normalised root mean square EMG amplitude was dependent on contraction time (F9,271.8 = 56.5, P < 0.0001), task (F1,275.3 = 1112.6, P < 0.0001), muscle (F1,272.4 = 9.8, P = 0.0019), contraction time by task interaction (F9,271.8 = 3.4, P = 0.0006) and task by muscle interaction (F1,272.7 = 40.3, P < 0.0001). During the intermittent task, global EMG activity was greater between 60% and 100% of contraction time in VL (P = 0.0016–P < 0.0001; Fig. 5B), and between 50% and 100% of contraction time in VM compared to baseline (P = 0.0125 to P < 0.0001; Fig. 5C). During the sustained task, global EMG activity was greater between 80% and 100% of contraction time compared to baseline in both VL (P = 0.0301 to P < 0.0001; Fig. 5B) and VM (P = 0.0272 to P < 0.0001; Fig. 5C). Global EMG activity was greater during the intermittent compared to the relative time points during the sustained task in both VL (P ≤ 0.0118) and VM (P < 0.0001).
Motor unit identification
During the sustained task, 29 and 18 MUs were identified at the beginning of the task in VL and VM, respectively (4 ± 3 and 3 ± 2 per participant, respectively). With additionally recruited MUs throughout the task, 123 and 105 MUs were identified in VL and VM at the point of task failure (14 ± 9 and 13 ± 4 per participant, respectively). The discrepancy in the number of identified MUs between the beginning and end of the task highlights the previously reported difficulty in identifying MUs with smaller potentials in the presence of MUs with bigger potentials (Francic & Holobar, 2021). During the intermittent task, 49 and 48 early recruited MUs were identified in VL and VM (6 ± 7 and 7 ± 4 per participant), respectively, with the number of identified MUs increasing throughout the task to 80 and 81 MUs at task failure (10 ± 7 and 12 ± 6 per participant), respectively.
Motor unit discharge modulation during intermittent and sustained tasks
Discharge rate was modulated during both tasks. Examples of a few selected motor units in two participants during the isometric and intermittent task can be appreciated in Fig. 6A and 6B, respectively. As can be seen, the later recruited MUs exhibited similar time-dependent changes in discharge rate to early recruited MUs.
Discharge rate was affected by contraction time (F9,3065.1 = 141.4, P < 0.0001), task (F1,3071.6 = 2273.5, P < 0.0001), muscle (F1,3067.5 = 13.1, P = 0.0003) and recruitment time (F1,3072.4 = 1027.7, P < 0.0001). There was also a significant interaction between contraction time and task (F9,3065.1 = 62.2, P < 0.0001), and task and muscle (F1,3066.7 = 16.2, P < 0.0001). During the sustained task, discharge rate initially decreased; in comparison with the start of the task, discharge rate was lower at 40% (VL, P = 0.0231; VM, P = 0.0064), 50% (VL, P = 0.0040; VM, P = 0.0058) and 60% (VL, P = 0.0181; VM, P = 0.0299) of contraction time, after which it returned to baseline (P ≥ 0.0823; Fig. 7A). During the intermittent task, discharge rate was similar to baseline up to 50% and 40% of contraction time in VL and VM, respectively (P ≥ 0.4264), at which point it increased above baseline (VL, P < 0.0001; VM, P = 0.0026), and continued to increase until task failure (P < 0.0001 for both VL and VM compared to baseline; Fig. 7B). When this increased MU discharge rate at task failure was compared to discharge rate during a 70% MVT contraction performed before the intermittent task to failure, discharge rate was not affected by contraction (F1,330.8 < 0.1, P = 0.9783), muscle (F1,329.8 = 2.5, P = 0.1140) or their interaction (F1,329.1 = 0.6, P = 0.4513). This comparison suggests that MU discharge rates during the 50% MVT intermittent task reached similar levels that are normally evident during a 10 s 70% MVT contraction (Fig. 7C and 7D). The inflection point of MU discharge rate modulation was positively associated with time to task failure during the 20% MVT sustained (R2 = 0.6935, P = 0.0001) and 50% MVT intermittent (R2 = 0.6710, P = 0.0002) tasks.
During the intermittent task, both recruitment (F9,1435.1 = 31.0, P < 0.0001) and derecruitment (F9,1435.1 = 11.4, P < 0.0001) thresholds of the identified MUs were influenced by contraction time. Recruitment threshold of the identified MUs was similar until 60% and 50% of contraction time in VL and VM, respectively, after which it decreased below baseline and continued to decrease until task failure (VL: P ≤ 0.0104; VM: P ≤ 0.0429; Fig. 7E). Derecruitment threshold of the identified MUs was unchanged until 90% and 80% of contraction time in VL and VM, respectively, after which it decreased below baseline (VL: P = 0.0191; VM: P ≤ 0.0013; Fig. 7F).
Experiment 2
Time to task failure and decrements in neuromuscular function
Time to task failure was 145 (26) and 72 (9) s for the 30% and 50% MVT sustained tasks (t7 = 9.9, P < 0.0001), respectively. The knee extensor MVT (F1,21 = 170.2, P < 0.0001; Fig. 8A), voluntary activation (F1,21 = 10.1, P = 0.0046; Fig. 8B) and quadriceps potentiated twitch torque (F1,21 = 163.6, P < 0.0001; Fig. 7C) decreased after both tasks. No interactions between contraction time and task were found for MVT, voluntary activation or potentiated twitch torque (P ≥ 0.2578).
Torque steadiness and global EMG activity
The coefficient of variation of torque was influenced by contraction time (F9,113.9 = 6.1, P < 0.0001), task (F1,117.4 = 7.8, P = 0.0060) and their interaction (F9,113.9 = 2.8, P = 0.0049). Post hoc testing showed that the coefficient of variation of torque was greater compared to baseline at 90–100% of contraction time during the 30% MVT sustained task (P ≤ 0.0008), whereas it remained unchanged throughout the 50% MVT sustained task (P ≥ 0.9968; Fig. 8D). The normalised root mean square EMG amplitude was affected by contraction time (F9,263.0 = 22.5, P < 0.0001), task (F1,95.0 = 295.1, P < 0.0001) and muscle (F1,263.3 = 11.2, P = 0.0009). During the 30% MVT sustained task, EMG amplitude was greater in the last 20% of contraction time compared to baseline in VL (P ≤ 0.0498) and VM (P ≤ 0.0151), respectively (Fig. 8E and 8F). During the 50% MVT sustained task, EMG was greater than at baseline in the final 10% (P = 0.0317) and 30% (P ≤ 0.0020) of contraction time in VL and VM, respectively.
Motor unit identification
During the beginning of the 30% MVT sustained task, 41 and 34 (6 ± 8 and 4 ± 4 per participant) MUs were identified in VL and VM, respectively. The number of identified MUs increased to 79 and 55 (11 ± 12 and 7 ± 5 per participant) in VL and VM, respectively. During the 50% MVT sustained task, 46 and 44 MUs were identified at the beginning (6 ± 6 and 6 ± 5 per participant), and 53 and 70 MUs at the end of the task (7 ± 8 and 9 ± 8 per participant) in VL and VM, respectively.
Motor unit discharge modulation during sustained tasks at different intensities
Discharge rate was influenced by contraction time (F9,2222.1 = 36.4, P < 0.0001), task (F1,2224.8 = 487.5, P < 0.0001), muscle (F1,2224.3 = 53.7, P < 0.0001) and recruitment time (F1,2227.2 = 2850.2, P < 0.0001). There was also a significant interaction between contraction time and task (F9,2222.1 = 2.8, P = 0.0026), and task and muscle (F1,2226.1 = 47.7, P < 0.0001). Compared to baseline, VL discharge rate increased in the final 10% of the 30% MVT sustained task (P = 0.0036), whereas it was greater in the final 20% contraction time of the 50% MVT sustained task (P ≤ 0.0216; Fig. 9A). Similarly, VM discharge rate was greater compared to baseline in the final 20% (P ≤ 0.0100) and 30% (P ≤ 0.0165) contraction time of the 30% and 50% MVT sustained tasks, respectively (Fig. 9B).
When comparing contractions before the tasks at a level 20% MVT greater than the sustained tasks (i.e. 50% MVT for the 30% MVT task and 70% MVT for the 50% MVT task) with the final 10 s of the tasks to failure, discharge rate was affected by contraction (sustained vs. trapezoidal ramp; F9,459.8 = 76.0, P < 0.0001), task (F1,460.0 = 128.9, P < 0.0001) and muscle (F1,461.8 = 18.5, P < 0.0001). There was also an interaction between contraction, task and muscle (F1,459.6 = 5.5, P = 0.0192). In VL, discharge rate at task failure was still greater during trapezoidal ramp contractions performed before the sustained tasks compared to that at task failure (30% MVT sustained task vs. 50% MVT trapezoidal ramp contraction: P < 0.0001; 50% MVT sustained task vs. 70% MVT trapezoidal ramp contraction: P = 0.0339; Fig. 9C). In VM, this was still the case when discharge rate at task failure during 50% MVT sustained task was compared to that during the 70% MVT trapezoidal ramp contraction (P < 0.0001), whereas this was not the case when discharge rate during 50% MVT trapezoidal ramp contraction was compared to that at task failure during the 30% MVT sustained task (P = 0.0705; Fig. 9D). Unlike in Experiment 1, the inflection point of MU discharge rate modulation was not associated with time to task failure during either of the two tasks (30% MVT: R2 = 0.1792, P = 0.1315; 50% MVT: R2 = 0.2144, P = 0.0822).
Discussion
This study examined task-specificity of MU discharge behaviour adjustments in VL and VM during prolonged isometric knee extension tasks to failure. Two independent experiments were performed. Experiment 1 compared MU discharge rate modulation pattern during a sustained and intermittent contraction task at 20% and 50% MVT, respectively, and Experiment 2 compared MU discharge rate adjustments during sustained contraction tasks at 30% and 50% MVT. In all conditions, we showed evidence of a biphasic pattern of MU discharge rate modulation. During the 20% MVT sustained contraction task, MU discharge rate initially decreased, which was followed by an increase such that MU discharge rate at task failure was similar to that at baseline. Conversely, no relative decrease in MU discharge rate was observed for the other contraction tasks (30% MVT sustained, and 50% MVT sustained and intermittent); rather, MU discharge rate remained similar for the first part of the task, followed by an increase towards task failure, with the greatest rate of increase observed during the intermittent task. Our results thus suggest that discharge rate modulation patterns during isometric muscle contractions to failure are contraction modality- and intensity-dependent.
Differences in the initial motor unit discharge rate adjustments
The key difference between the intermittent and sustained tasks in Experiment 1 was the decrease in MU discharge rate in the first half of the task. Specifically, a reduction in MU discharge rate was observed in the 20% MVT sustained task, but not during the 50% MVT intermittent task. However, when comparing MU discharge behaviour during sustained contractions at 30% and 50% MVT, there was no decrease in MU discharge rate during the initial period of any of the two tasks.
The observed MU discharge rate behaviour during the 20% MVT sustained task is similar to that of Martinez-Valdes, Negro, Falla et al. (2020) who also observed a decrease (reaching significance at ∼40% contraction time) followed by an increase in VL and VM MU discharge behaviour during a 30% MVT sustained task. Furthermore, we also observed that the inflection point of MU discharge rate was associated with time to task failure, similar to Martinez-Valdes, Negro, Falla et al. (2020). However, we could not replicate these results when performing a 30% sustained MVT task. The discrepant results may be, at least partially, attributable to subtle differences in approaches to analysis. These include differences in MU tracking methodology, and the inclusion/exclusion of later recruited MUs; for example, in our study we included later recruited MUs in the analysis, considering recruitment time as a covariate, whereas Martinez-Valdes and colleagues considered only MUs recruited from the beginning of the task (Martinez-Valdes, Negro, Falla et al., 2020). Since MUs recruited later in the task, particularly those recruited beyond a point of general increase in MU discharge rate (see Fig. 5A and 5B for example), are less likely to exhibit an initial decline in MU discharge, this phenomenon might be less likely to be statistically detected. Other possible explanations for the discrepancy between the two studies are the differences in the sample size and the degree of between-participant variability in the observed MU discharge rate patterns. Namely, compared to Martinez-Valdes, Negro, Falla et al. (2020), the present study included a smaller sample of participants and observed a greater degree of between-participant variability in the responses (see Appendix), which may have confounded the group-level findings. This variability might have also influenced our inability to detect the association between the MU discharge rate inflection point and time to task failure for the 30% MVT task. It should be noted, however, that high levels of between-participant variability are commonly reported in studies examining MU discharge rate during prolonged isometric contractions (Adam & De Luca, 2005; Contessa et al., 2016; De Ruiter et al., 2004; Garland et al., 1994; Maton, 1981; Rossato et al., 2022), with their influence on group-level findings not necessarily eradicated with larger sample sizes (Rossato et al., 2022). It is possible that differences in the discharge rate patterns and in the degree of interindividual variability between studies may be related to the fact neither study normalised the intensity of tasks to metabolic thresholds (i.e. critical torque), which would add heterogeneity to the physiological responses (Jamnick et al., 2020), though this hypothesis remains to be tested directly.
The decrease in discharge rate in the first part of the 20% MVT sustained task is unlikely to have been caused by the inhibitory afferent feedback via group III/IV muscle afferents as this effect would have presumably been greater during higher intensity tasks that will have induced greater ischaemia and thus probably have a greater intramuscular accumulation of noxious metabolites. The observed behaviour is similar to the reduced responsiveness of motoneurons to constant synaptic input as previously shown in reduced animal preparations (i.e. late spike frequency adaptation; Button et al., 2007; Kernell & Monster, 1982; Sawczuk et al., 1995; Spielmann et al., 1993) and indirectly in humans (Brownstein et al., 2021; Finn et al., 2018). However, spike frequency adaptation seems a less likely explanation for the observed reduction in MU discharge rate in the initial part of the 20% MVT sustained task since such an effect has been shown to be more pronounced in higher threshold motoneurons (Button et al., 2007; Spielmann et al., 1993). This is inconsistent with our results showing the lack of relative decline in MU discharge rates at higher contraction levels (30% and 50% MVT) and the fact our data probably reflect the behaviour of mainly higher threshold MUs due to decomposition bias (Francic & Holobar, 2021; see also Fig. 6C and 6D). Importantly, however, evidence for late spike frequency adaptation largely stems from studies in reduced animal preparations in the absence of neuromodulatory inputs. Indeed, increasing monoaminergic input that facilitates persistent inward currents (PICs) on motoneuron dendrites has been shown capable of diminishing the effects of late spike frequency adaptation (Brownstone et al., 2011; Hornby et al., 2002). Considering the probably greater PIC magnitude with greater contraction intensity (Škarabot, Beauchamp et al., 2023), we postulate that PIC magnitude is insufficient at 20% MVT to prevent the effects of late spike frequency adaptation. An alternative explanation for the lack of observed decreases in MU discharge rate during tasks performed at higher contraction intensities is purely methodological, grounded in normalising contraction time in 10% intervals to allow comparison between participants, which inevitably acts as a smoothing function, making smaller or shorter/shallower decreases in MU discharge rates less likely to be detectable.
The more likely mechanism for the observed reduction in MU discharge rate in the initial part of the 20% MVT sustained task relates to MU twitch potentiation. Assuming an increase in MU twitch amplitude and area with initial repetitive depolarisation (Bagust et al., 1974; Burke et al., 1976), a constant torque output could be maintained despite decreased MU discharge rate. Indeed, the time course of twitch potentiation has been demonstrated to coincide with the decrease in MU discharge rate at the beginning of the submaximal isometric task to failure (Martinez-Valdes, Negro, Falla et al., 2020). Eventually, MU twitch amplitude has been shown to decline (Martinez-Valdes, Negro, Falla et al., 2020), probably due to the accumulation of metabolites (McKenna et al., 2008) that render muscle contractile capacity reduced, necessitating greater synaptic input to motoneurons to maintain the same torque output resulting in an increase in MU discharge rate until task failure. Whilst evidence in cats suggests that the extent of MU twitch potentiation is similar between lower and higher threshold MUs (Bagust et al., 1974), we speculate that the shorter time course of potentiation (during the 30% and 50% MVT sustained tasks) and interruption of repetitive activation (during the 50% MVT intermittent task) lead to a significantly smaller or less detectable decrease in MU discharge rate in the initial part of the task at higher intensity (note a representative example of an early recruited MU in Fig. 5B that demonstrates a marginal trend of decrease in discharge rate). Overall, these results suggest that MU discharge rate behaviour in the first portion of the task is more related to contraction intensity rather than to the modality.
Inability to increase motor unit discharge rate coincides with task failure
During all tasks, MU discharge rate progressively increased across the second half of the task. During the sustained contraction at 20% MVT, the increase began at a below-baseline level (due to the initial decrease) and did not increase above baseline at task failure. During the sustained contractions at 30% and 50% MVT and the intermittent contractions at 50% MVT, MU discharge rate increased above baseline, with the degree of the increase being dependent on the contraction intensity, being lower in 30% compared to 50% MVT sustained tasks. The increase in MU discharge rate in the latter parts of the task, however, appeared to be greatest during the 50% MVT intermittent task to failure, where the increase above baseline was detected at the earliest point in the task (∼60% contraction time) and kept increasing until task failure. These observations suggest that the relative changes in MU discharge behaviour are dependent on both contraction modality and intensity.
The increase in MU discharge rate with proximity to task failure probably reflects an increase in the strength of the common excitatory synaptic input (Castronovo et al., 2015), which is necessary to maintain constant torque output, probably coinciding with the decrease in MU twitch potentiation (Martinez-Valdes, Negro, Falla et al., 2020). There are several lines of observation to support this interpretation. First, with increased strength of the common synaptic input, torque variability tends to increase (Castronovo et al., 2015); indeed, we demonstrated a greater coefficient of variation of torque towards the end of the tasks. Second, the greater strength of common synaptic input, which reflects the increase in net excitatory input (Castronovo et al., 2015), is probably responsible for recruitment of additional MUs that was demonstrated in our data. In addition to recruitment of additional MUs, the recruitment threshold of active MUs also decreased with contraction time during the intermittent task, contrasting with the changes in MU discharge rate. Specifically, recruitment threshold remained stable during the first half of the task, and then decreased below baseline (coinciding with the start of MU discharge rate increase) and continued to decrease until task failure. This is consistent with previous findings in an intermittent task (Carpentier et al., 2001; Farina et al., 2009) and suggests that, with repetitive motoneuron activation, MU recruitment range progressively compresses. The reduction in recruitment threshold coupled with the increase in MU discharge rate probably serves to maintain constant torque output in the presence of impaired contractile function. Along similar lines, derecruitment thresholds were also reduced, though later in the intermittent task, indicative of prolonged MU discharge as a result of greater common synaptic input. Finally, though several later recruited MUs exhibited a constant increase in MU discharge rate, this was dependent on their recruitment time within the task. For example, if additional MUs were recruited relatively early in the task, their behaviour was consistent with already recruited MUs (note the example in Fig. 5A, whereby an additionally recruited MU early in the task initially exhibits a decrease in discharge rate followed by an increase, akin to the MUs recruited at the beginning of the task). These observations support the notion that the common synaptic input strength is modulated throughout the task to allow constant torque output.
Motor unit discharge rate during the intermittent task at 50% MVT increased above baseline in the second half of the task, and during submaximal sustained contractions at 30% and 50% MVT in the final 10% and 20% contraction time, respectively. However, during the intermittent task MU discharge rate at task failure reached, but did not exceed, the rates exhibited during a short 70% MVT contraction performed before contractions to failure. Furthermore, during sustained tasks at 30% and 50% MVT discharge rate remained below the rates observed during a short, higher level (+20% MVT) contraction, which is in agreement with previous reports (Martinez-Valdes, Negro, Falla et al., 2020). Similarly, global EMG amplitude at task failure, though probably more reflective of MU recruitment and peripheral alterations (Del Vecchio et al., 2017; Martinez-Valdes et al., 2018), also did not reach EMG amplitude recorded during MVT, consistent with prior work (Fuglevand et al., 1993). These findings suggest that the increase in common synaptic input is ultimately insufficient to allow depolarisation and continued increases in MU discharge rate, coinciding with task failure. The relative increase in MU discharge rate at task failure seemed to be greater during the intermittent task, and during tasks performed at greater contraction intensity, suggesting a different mechanistic interaction depending on contraction modality and intensity leading to the observed MU discharge behaviour. Evidence from reduced animal preparations shows that an intense or prolonged release of serotonin may cause a spillover onto the extrasynaptic receptors on the axon initial segment and thus reduce motoneuron activity (Cotel et al., 2013; Perrier et al., 2018). However, whilst human studies have indicated the potential for ingestion of a serotonin reuptake inhibitor to shorten time to task failure (Kavanagh et al., 2019), this seems only to be the case for maximal rather than submaximal sustained contractions (Henderson et al., 2022), making it an unlikely mechanism to explain the observed differences in relative increase in MU discharge rate between contraction modalities and intensities in our study. Furthermore, evidence from both reduced animal preparations (isolated rat diaphragm) and human studies suggests that prolonged supramaximal electrical stimulation of the peripheral nerve (>15 Hz) can lead to suppressed muscle fibre discharge activity due to neuromuscular transmission failure (Aldrich et al., 1986; Bigland-Ritchie et al., 1979; Cupido et al., 1992; Krnjevic & Miledi, 1958; Kuei et al., 1990), with greater functional impairments of the neuromuscular junction found at higher stimulation frequencies (Kuei et al., 1990) and in higher threshold MUs (Sandercock et al., 1985). It has been speculated that failure in neuromuscular propagation occurs due to alterations in action potential propagation along the terminal nerve branches, reduced acetylcholine release from the terminal endings and/or reduced excitability of the motor endplate (Krnjevic & Miledi, 1958; Thesleff, 1959). During maximal prolonged voluntary muscle activation, however, this phenomenon has not been detected (Bigland-Ritchie et al., 1979). Whilst neural demands of maximal isometric voluntary contractions differ from that during submaximal contractions, it seems unlikely that neuromuscular transmission failure could meaningfully contribute to muscle fibre discharge rate saturation at task failure or explain differences in the rate of increase between contractions tasks.
The task-dependency of MU discharge behaviour is perhaps unsurprising; the greatest increase in MU discharge rate during the intermittent task is probably due to lower levels of ischaemia, with the post-contraction hyperaemic response restoring muscle perfusion and clearing accumulated metabolites (Adreani & Kaufman, 1998; De Ruiter et al., 2007), thus reducing the inhibitory inputs of group III/IV afferents to motoneurons (Martinez-Valdes, Negro, Farina et al., 2020) and the motor cortex (Kennedy et al., 2016) compared to that during the sustained contractions. Similarly, H-reflex has been shown to reduce during sustained compared to intermittent contractions (Duchateau & Hainaut, 1993), indicating reduced Ia afferent input to motoneurons, probably further contributing to smaller increases in MU discharge rate during sustained contractions. Though not observed in our study, the smaller increases in MU discharge rate during lower intensity sustained contractions are consistent with the notion that reductions in voluntary activation are typically greater during lower relative torque outputs (Burnley et al., 2012), indicating a greater impairment in the central drive. During submaximal sustained contractions, corticospinal excitability and inhibition have been shown to progressively increase (Klass et al., 2008; Lévénez et al., 2008; Søgaard et al., 2006), whereas such changes appear intensity-dependent during intermittent contractions, with corticospinal excitability increased and decreased during contraction above and below critical torque, respectively (Ansdell, Brownstein, Škarabot, Hicks, Howatson et al., 2019). Future studies should consider employing such measures concurrently with HDsEMG to attempt to explain the apparent contraction intensity-dependent alterations in MU discharge rate in isometric tasks to failure.
Further considerations
Previous studies investigating MU discharge behaviour during prolonged contractions have produced mixed findings. Although the consensus sometimes appears to suggest a progressive decline in MU discharge rate should be expected with sustained contractions (Enoka & Duchateau, 2008), comparisons must be interpreted through the lens of methodological differences, the nature of the task, the motor pool investigated and possibly the recruitment threshold of the identified MUs, among others. The studies that did not perform contraction tasks to failure seem less likely to be comparable; the progressive decrease of MU discharge rate in those studies (Mottram et al., 2005; Pascoe et al., 2014; Riley et al., 2008) might reflect task cessation before the MU twitch potentiation had reached its peak, not necessitating the need for an increase in synaptic input to motoneurons to maintain the constant torque output. Our demonstration of a biphasic behaviour in MU discharge, with an increase towards task failure, seems to agree with a large number of studies that performed tasks to failure (Adam & De Luca, 2005; Bigland-Ritchie et al., 1986; Garland et al., 1997; Griffin et al., 2000; Kuchinad et al., 2004; Mettler & Griffin, 2016), though not with some others (Carpentier et al., 2001; Garland et al., 1994; McManus et al., 2016).
It is also worth considering methodological differences between studies, with many prior studies utilising intramuscular EMG and template matching algorithms that carry a higher risk of erroneous MU identification during prolonged contraction where MU action potential waveforms are likely to change due to modulation in conduction velocity (Farina et al., 2009). In this study, we utilised an approach of estimating and applying MU filters in successive, short windows, which has been shown to be a robust approach when attempting to accommodate small changes in MU waveforms (Francic & Holobar, 2021; Kramberger & Holobar, 2021; Škarabot, Ammann et al., 2023). From this perspective, our results broadly agree with a study that utilised multichannel surface EMG decomposition, and a time-domain MU tracking approach (Martinez-Valdes, Negro, Falla et al., 2020). Furthermore, the number of concurrently active MUs identified with intramuscular EMG is limited, which may underestimate the modulation in MU discharge rate at the population level (Enoka, 2019). Finally, many of the prior studies pooled the identified MU discharges, violating the principle of co-dependence of observations (Tenan et al., 2014). Conversely, we used linear mixed statistical modelling, accounting for nesting of observations within individual participants (Wilkinson et al., 2023; Yu et al., 2022). Finally, in most of the studies showing a progressive decrease in MU discharge rate to task failure, an intrinsic hand muscle was used (Carpentier et al., 2001; McManus et al., 2016). Considering previous suggestions that the behaviour of MU discharge rate adjustments with repetitive motoneuron activation might depend on the upper limit of recruitment of the motor pool (Enoka & Stuart, 1992), we allow for the possibility that the demonstrated contraction intensity- and modality-dependent MU behaviour in our study might be constrained to the vastii motor pools, or at least motor pools with similar upper limits of recruitment. It remains unclear whether the results can be replicated in a similar design using an intrinsic hand muscle with a comparatively lower upper limit of MU recruitment (Kukulka & Clamann, 1981).
It is important to note that owing to decomposition bias, the majority of the identified MUs were of higher threshold, and thus we cannot exclude the possibility that the observed behaviour is related to MU recruitment threshold, rather than intensity of the tasks per se. For example, during intermittent contractions where recruitment thresholds could be reliably determined, most of the identified MUs at the beginning of the task had recruitment thresholds >35% MVT on average. It is conceivable that MU behaviour exhibited during repetitive activation of motoneurons is non-homogeneous across the motor pool, with previous reports showing a decrease in discharge rate of early recruited, but an increase in later recruited (higher threshold) MUs (Carpentier et al., 2001). We provide some evidence of this insofar as we observed later recruited MUs (that will have higher thresholds) to have a tendency to increase discharge rate. Future studies identifying a pool of MUs with a sufficient range of recruitment thresholds, perhaps by combining intramuscular and surface EMG recordings, or increasing the density and/or area of surface recordings (Caillet et al., 2023), might be able to answer this question more directly.
Conclusion
We showed that during isometric contractions to failure, vastus lateralis and medialis MU discharge rate alterations exhibit distinct, biphasic patterns depending on contraction modality and intensity. During a low-intensity (20% MVT) contraction, MU discharge rate initially decreased, before increasing to baseline levels. During higher intensity contractions, either sustained at 30% and 50% MVT, or intermittent at 50% MVT, we showed that MU discharge rate remained stable, after which it continually increases until task failure, though the relative increase depends on contraction intensity, with the greatest increase demonstrated by the intermittent task. In all tasks, modulations in MU discharge rate pattern were also accompanied by recruitment of additional MUs. The observed behaviour is probably related to changes in the strength of the common excitatory synaptic input that is modulated in response to MU twitch potentiation and attenuation throughout the isometric task to failure. Notably, the increased MU discharge rate at task failure for the 30% and 50% MVT sustained tasks was still lower than during short trapezoidal contractions performed at 50% and 70% MVT, respectively, suggesting insufficient strength of the excitatory synaptic input to motoneurons to allow further discharge rate increases and prolongation of task. Overall, the data presented herein provide strong evidence that MU discharge rate modulation during isometric contractions to failure is both intensity- and modality-dependent.
Appendix
Statistical considerations
Due to difficulty in identifying motor unit spike trains in female participants (see ‘Participant and ethical approval’ section of ‘Materials and methods’) the final sample on which analyses were performed was biased towards males (eight vs. one in Experiment 1, and seven vs. one in Experiment 2). Considering the potential for different modulation of motor unit discharge rate in female participants (Jenz et al., 2023) and sex differences in time to task failure (Ansdell et al., 2019), further considerations need to be given to the effect sex might have on the observed findings.
First, we considered whether the inclusion of sex in the statistical model would improve model fit. We thus compared the fit of the model with and without the inclusion of participant sex as a random effect. As demonstrated in Table A1, the inclusion of sex as a random effect had a marginal effect on the model fit. Specifically, this effect on model fit was negative in most of the outcome variables.
| AIC | BIC | |||
|---|---|---|---|---|
| Outcome variable | (1 | Sex) | (1 | Sex) | ||
| Experiment 1 | ||||
| EMG amplitude | 2089.9 | 2091.9 | 2248.2 | 2253.9 |
| CV force | 366.4 | 367.9 | 435.4 | 440.0 |
| Discharge rate | 12,561.9 | 12,563.9 | 12,821.8 | 12,829.8 |
| Recruitment threshold | 10,198.1 | 10,195.9 | 10,314.4 | 10,317.5 |
| Derecruitment threshold | 9981.4 | 9981.3 | 10,097.8 | 10,102.9 |
| Experiment 2 | ||||
| EMG amplitude | 1997.9 | 1999.8 | 2154.9 | 2160.5 |
| CV force | 230.7 | 232.7 | 295.4 | 300.3 |
| Discharge rate | 8848.1 | 8838.8 | 9094.4 | 9090.8 |
- CV = coefficient of variation; AIC = Akaike Information Criterion; BIC = Bayesian Information Criterion.
Second, we considered whether the inclusion of a single female participant influenced the outcomes of our experiments. For this purpose, we excluded the female participant from our analysis and compared the significance of the model with the model that included all participants. As demonstrated in Table A2 (Experiment 1) and Table A3 (Experiment 2), despite the smaller sample size (n − 1) the significance of these two models was similar, indicating that the inclusion of a single female participant in both experiments did not have a significant effect on our findings.
| All participants (n = 9) | |||||
|---|---|---|---|---|---|
| EMG amplitude | CV force | Discharge rate | RecT | Derec T | |
| Contraction time | F9,271.8 = 56.5, P < 0.0001 | F9,142.0 = 18.0, P < 0.0001 | F9,3065.1 = 141.4, P < 0.0001 | F9,1435.1 = 31.0, P < 0.0001 | F9,1435.1 = 11.4, P < 0.0001 |
| Task | F1,275.3 = 1112.6, P < 0.0001 | F1,145.4 = 2.5, P = 0.1170 | F1,3071.6 = 2273.5, P < 0.0001 | – | |
| Muscle | F1,272.4 = 9.8, P = 0.0019 | – | F1,3067.5 = 13.1, P = 0.0003 | F1,1441.2 = 0.7, P = 0.4061 | F1,1441.4 = 0.1, P = 0.7667 |
| Recruitment time | – | – | F1,3072.4 = 1027.7, P < 0.0001 | – | – |
| Time × Task | F9,271.8 = 3.4, P = 0.0006 | F9,142.0 = 1.8, P = 0.0708 | F9,3065.1 = 62.2, P < 0.0001 | – | – |
| Time × Muscle | F9,271.8 = 0.5, P = 0.8668 | – | F9,3065.1 = 1.1, P = 0.3084 | F9,1435.0 = 0.1, P = 0.9999 | F9,1435.0 = 0.2, P = 0.9958 |
| Task × Muscle | F1,272.7 = 40.3, P < 0.0001 | – | F1,3066.7 = 16.2, P < 0.0001 | – | – |
| Time × Task × Muscle | F9,271.8 = 0.1, P = 0.9990 | – | F9,3065.1 = 1.0, P = 0.4519 | – | – |
| Males only (n = 8) | |||||
| Contraction time | F9,242.9 = 49.0, P < 0.0001 | F9,122.9 = 16.5, P < 0.0001 | F9,2932.1 = 129.0, P < 0.0001 | F9,1321.3 = 26.5, P < 0.0001 | F9,1321.2 = 9.4, P < 0.0001 |
| Task | F1,245.7 = 913.2, P < 0.0001 | F1,126.2 = 4.4, P = 0.0376 | F1,2936.6 = 2251.2, P < 0.0001 | – | |
| Muscle | F1,243.3 = 12.4, P = 0.0005 | – | F1,2934.4 = 19.4, P < 0.0001 | F1,1323.9 = 0.6, P = 0.9999 | F1,1326.5 = 0.1, P = 0.8342 |
| Recruitment time | – | – | F1,2938.9 = 960.2, P < 0.0001 | – | – |
| Time × Task | F9,242.9 = 2.8, P = 0.0033 | F9,122.9 = 1.3, P = 0.2689 | F9,2932.1 = 56.7, P < 0.0001 | – | – |
| Time × Muscle | F9,242.9 = 0.4, P = 0.9533 | – | F9,2932.1 = 1.2, P = 0.3138 | F9,1321.2 = 0.7, P = 0.9999 | F9,1321.2 = 0.1, P = 0.9991 |
| Task × Muscle | F1,243.3 = 30.6, P < 0.0001 | – | F1,2932.9 = 22.4, P < 0.0001 | – | – |
| Time × Task × Muscle | F9,242.9 = 0.1, P = 0.9991 | – | F9,2932.1 = 0.7, P = 0.6875 | – | – |
| All participants (n = 8) | |||
|---|---|---|---|
| EMG | CV force | Discharge rate | |
| Contraction time | F9,263.0 = 22.6, P < 0.0001 | F9,113.9 = 6.1, P < 0.0001 | F9,2222.1 = 36.4, P < 0.0001 |
| Task | F1,95.0 = 273.4, P < 0.0001 | F1,117.4 = 7.8, P = 0.0060 | F1,2224.8 = 487.5, P < 0.0001 |
| Muscle | F1,263.0 = 11.1, P = 0.0009 | – | F1,2224.3 = 53.7, P < 0.0001 |
| Recruitment time | – | – | F1,2222.1 = 2.8, P < 0.0001 |
| Time × Task | F9,263.0 = 0.2, P = 0.9950 | F9,113.9 = 2.8, P = 0.0049 | F9,2222.1 = 2.8, P = 0.0026 |
| Time × Muscle | F9,263.0 = 0.3, P = 0.9631 | – | F9,2222.1 = 0.5, P = 0.8415 |
| Task × Muscle | F1,263.0 = 0.1, P = 0.7506 | – | F1,2226.1 = 47.7, P < 0.0001 |
| Time × Task × Muscle | F9,263.0 = 0.2, P = 0.9956 | – | F9,2222.1 = 0.4, P = 0.9509 |
| Males only (n = 7) | |||
| Contraction time | F9,224.0 = 17.9, P < 0.0001 | F9,95.0 = 5.9, P < 0.0001 | F9,2013.1 = 26.4, P < 0.0001 |
| Task | F1,224.3 = 217.6, P < 0.0001 | F1,98.1 = 7.2, P = 0.0085 | F1,2015.0 = 360.9, P < 0.0001 |
| Muscle | F1,224.3.0 = 9.9, P = 0.0018 | – | F1,2014.4 = 66.8, P < 0.0001 |
| Recruitment time | – | – | F1,2016.9 = 2728.0, P < 0.0001 |
| Time × Task | F9,224.0 = 0.1, P = 0.9989 | F9,95.0 = 2.0, P = 0.0428 | F9,2013.1 = 3.0, P = 0.0014 |
| Time × Muscle | F9,224.0 = 0.1, P = 0.9998 | – | F9,2013.1 = 0.5, P = 0.8965 |
| Task × Muscle | F1,224.3 = 0.1, P = 0.8795 | – | F1,2016.5 = 41.4, P < 0.0001 |
| Time × Task × Muscle | F9,224.0 = 0.1, P = 0.9999 | – | F9,2013.1 = 0.5, P = 0.8801 |
Variability of responses
To characterise the variability of motor unit discharge behaviour we performed two additional analyses. For the first analysis, we divided individual motor unit discharge behaviour into four bins as a proportion of contraction time (i.e. every 25% of contraction time). From there, we modelled the slope of the behaviour separately for each bin as follows: MU discharge rate ∼ contraction time + (contraction time | PID: MUID), where PID is the individual participant ID and MUID represents IDs for individual MUs that were nested within a participant. Individual PID slopes and the average slope are depicted for all tasks and muscles in Fig. A1. The largest inter-individual variability was exhibited for the 50% sustained task.
Additionally, we computed a normalised cumulative discharge rate in 1 s epochs and performed second-order polynomial fits to depict the individual participant cumulative trends of motor unit discharge behaviour. The results of this approach are depicted in Fig. A2.
Biographies
-
Tamara Valenčič is a PhD candidate in Neuromuscular Physiology at the School of Sport, Exercise and Health Sciences, Loughborough University (UK). She obtained her undergraduate degree in Kinesiology from University of Ljubljana (Slovenia) and MSc degree in Exercise Physiology from Loughborough University (UK). Her PhD research focuses on the long-term effects of ACL reconstruction with a hamstring tendon autograft on neural control of knee extensor and knee flexor muscles.
-
Jakob Škarabot is a Senior Lecturer in Neuromuscular Physiology at the School of Sport, Exercise and Health Sciences, Loughborough University (UK). He obtained his BSc degree in Kinesiology from University of Ljubljana (Slovenia), MSc degree in Biomechanics from University of Jyväskylä (Finland) and PhD in Neurophysiology from Northumbria University (UK). His research focuses on motor neuron physiology in health and disease.
Open Research
Data availability statement
Data are available from the corresponding author upon reasonable request.
No conflicts of interest, financial or otherwise, are declared by the authors.
The experiments were performed at Loughborough University. T.V., P.A., C.G.B., A.H. and J.Š. conceived and designed research; T.V. and J.Š. performed experiments; T.V., P.S., A.H. and J.Š. analysed data; T.V. and J.Š. drafted the manuscript; all authors edited and revised the manuscript. All authors approved the final version of the manuscript. All authors agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. All persons designated as authors qualify for authorship, and all those who qualify for authorship are listed.
J.Š. is supported by Versus Arthritis Foundation Fellowship (reference: 22569). P.A. is supported by the UK Office for Veteran's Affairs (G2-SCH-2022-11-12245). A.H. is supported by the Slovenian Research Agency (J2-1731 and P2-0041) and Horizon Europe Research and Innovation Programme (No. 101079392).
References
Additional information
