Passive muscle stretching reduces estimates of persistent inward current strength in soleus motor units

Prolonged (usually ≥60 s; Behm et al., 2016; Kay and Blazevich, 2012) passive muscle stretching reduces maximal force production and neural drive to the muscle. For example, passive stretching of the human plantar flexors reduces electromyogram (EMG) amplitude during maximal voluntary isometric contractions and both the reduction and recovery of EMG tend to correlate strongly with the loss and recovery of maximal muscle force (Kay and Blazevich, 2009; Trajano, et al., 2013a; Pulverenti et al., 2019; Ryan et al., 2014). These changes in EMG are strongly associated with changes in other measures of neurological function, including estimates of voluntary activation and V-wave amplitude (Trajano et al., 2013a; Ryan et al., 2014), leading to the conclusion that muscle force loss largely results from a decrease in neural drive to the muscle.

The origin of this loss of neural drive is still unknown. However, there is evidence that passive muscle stretching might cause prolonged reduction in excitability of the α-motoneurons (Trajano et al., 2017). This conclusion was based on the observation in humans that passive muscle stretching could reduce the strength of the involuntary plantar flexor contractions that normally persist after cessation of simultaneous tendon vibration and electrical muscle stimulation (Trajano et al., 2014). Tendon vibration can elicit muscular contractions via the Ia reflex loop, and these contractions display characteristics that are consistent with the initiation of persistent inward currents (PICs) in α-motoneurons, including self-sustained motor unit firing (after vibration cessation), joint angle dependence and a warm-up effect (Trajano et al., 2014). Thus, the reduction in strength of involuntary contractions that persist after tendon vibration provides evidence of a stretch-induced disfacilitation of the α-motoneurons, possibly via a reduction in the strength of PICs.

PICs are depolarizing currents generated by voltage-sensitive sodium and calcium channels predominantly located on the motoneurons' dendrites (Heckman et al., 2004). They amplify and prolong synaptic input, producing sustained depolarization of the cell, causing motoneurons to continue to fire without the need for additional input (Heckman et al., 2004; Lee and Heckman, 2000). PICs can also change the input–output gain of motoneurons by ∼5-fold or more, depending on the level of monoaminergic drive to the spinal cord, allowing motoneurons to fire at much higher frequencies than when PICs are not triggered (Lee and Heckman, 2000). In fact, the neuromodulatory effect exerted by monoamines (i.e. serotonin and noradrenaline in motoneurons) on PICs is so important for normal motor behaviour that maximal activation of excitatory ionotropic inputs would produce less than 40% of the maximal motor output in the absence of neuromodulation (Heckman, 1994; Johnson et al., 2017). Thus, PICs are a fundamental component of normal motor output observed in humans, and reductions in PIC amplitude can significantly affect the muscle's ability to produce force.

In humans, PIC amplitudes are typically estimated using the paired-motor unit technique (Gorassini et al., 2002; Johnson et al., 2017; Wilson et al., 2015). This method measures the discharge rate of a lower-threshold motor unit (control unit), as a surrogate for the level of excitatory drive, at the time of recruitment and de-recruitment of a higher-threshold unit (test unit). The difference between recruitment and de-recruitment frequencies is referred to as ΔF. ΔF is considered to be proportional to the PIC amplitude and has been validated by comparison with animal data (Powers et al., 2008) as well as computer simulations (Powers and Heckman, 2015). Hence, if passive muscle stretching reduces the ability to develop PICs in humans, as suggested by previous research, then reductions in ΔF should be observed immediately after the stretching has been completed.

Given the above, the main purpose of the present study was to determine whether passive stretching can reduce ΔF, i.e. estimates of PIC amplitude, in the human plantar flexors. To do this, we measured ΔF in soleus (SOL) and gastrocnemius medialis (GM) before and after five sets of 1 min plantar flexor stretching or a control condition (5 min rest). We hypothesized that ΔF would be lower after passive stretching, indicating a loss of PIC strength.

Eighteen recreationally active participants (10 males and 8 females) without any neuromuscular limitations volunteered to participate in the study (32±4 years, 73±19 kg, 174±11 cm). Participants were instructed to avoid vigorous exercise and alcohol consumption for 24 h, and caffeine use for at least 6 h, prior to testing. All participants read and signed the informed consent document, and the Queensland University of Technology Human Research Ethics Committee approved this study (approval number: 1800000550).

All data collection was performed in a single session lasting approximately 1 h and 30 min. Participants were seated upright in the chair of an isokinetic dynamometer (Biodex System 4, Biodex Medical system, Shirley, NY, USA) with the knee fully extended (0 deg) and ankle at neutral position (0 deg). They were then instructed to practise six 3-s voluntary submaximal isometric plantar flexion contractions (2×40%, 2×60% and 2×80% of perceived maximal effort) with a 3-s passive rest interval between contractions. Two minutes after this warm-up, two maximal voluntary contractions (MVC) were performed with a 1-min passive rest interval, and maximum torque was recorded. Then, participants practised isometric triangular ramped contractions, with the instruction to contract their plantar flexors in order to increase torque from 0 to 20% of MVC in 10 s at a rate of 2% of MVC s⁻¹, and then decrease that force linearly at the same rate back to the baseline value. A contraction intensity of 20% of MVC was chosen as pilot testing revealed this contraction level to allow the greatest number of motor units to be accurately identified by the decomposition algorithm. Once the participants were familiarized with these triangular contractions, data were recorded as they performed, in the following order, three triangular contractions interspaced by 25 s immediately before (Control 1) and after (Control 2) a 5-min rest period as well as immediately after five sets of 1 min passive plantar flexor stretching interspaced by 15-s intervals (Post-stretch).

Surface electromyogram (EMG) was recorded during the contractions using two semi-disposable 32-channel electrode grids with a 10-mm inter-electrode distance (ELSCH032NM6, OTBioelettronica, Torino, Italy) placed over the muscle bellies of both SOL and GM. A strap electrode was dampened and positioned around the ankle joint as a ground electrode. The EMG signals were recorded in monopolar mode and converted to digital data by a 16-bit wireless amplifier (Sessantaquattro, OTBioelettronica). EMG signals were amplified (256×), sampled at 2000 Hz and bandpass filtered (10–500 Hz) before being stored for offline analysis.

Surface EMG signals were decomposed into single motor unit discharge events using a convolutive blind source separation algorithm using DEMUSE software (Holobar and Zazula, 2007). This algorithm has been extensively validated against the standard intramuscular fine wire methods across a variety of muscles and contractions (Holobar et al., 2014; Negro et al., 2016; Muceli et al., 2015). All decomposed motor units were visually inspected and all erroneous identified motor units and discharge times were excluded. Only motor units with a pulse-to-noise ratio above 30 dB were kept for further analysis. After that, discharge events with interspike intervals below 0.025 s and above 0.4 s were excluded. Subsequently, discharge events were converted into instantaneous discharge rates and fitted into a 5th-order polynomial function. This polynomial fit was then used for the paired motor unit analysis. ΔF was calculated as the change in discharge rate of a lower-threshold (control) motor unit from the moment of recruitment to the moment of de-recruitment of a higher-threshold (test) unit (Gorassini et al., 2002) (see Fig. 1). ΔF values were calculated for pairs of motor units that fitted the following criteria: (1) rate–rate correlations of ≥0.7; (2) test unit recruited at least 0.5 s after the control unit; and (3) no saturation of discharge rates of the control unit (discharge rate increased by at least 0.5 pulses s⁻¹ after the recruitment of the test unit) (Stephenson and Maluf, 2011). ΔF values were calculated for individual test units as the average value obtained when the units were paired with multiple control units. Subsequently, a single ΔF value was obtained per participant per muscle (SOL and GM) by averaging the values obtained for all the units from each muscle. Motor units were tracked throughout the conditions using the decomposition filter and the same motor unit pairs were used to calculate the ΔF value for each contraction within each participant (Glaser and Holobar, 2018). Peak discharge rates were measured as the highest value in the polynomial function for each motor unit and averaged across all the units obtained for each muscle. In three participants (2 females and 1 male), it was not possible to identify usable motor units.

The stretching procedures were performed on an isokinetic dynamometer with the muscles relaxed. The plantar flexors were stretched 5 times, with 15-s rest intervals, by rotating the ankle into dorsiflexion at 2 deg s⁻¹ until maximal tolerable stretch was attained and then holding at the stretched position for 1 min. This 5 min stretch protocol is the same as used in previous studies reporting reductions in maximal voluntary torque, neural drive to the muscle and tendon vibration reflexes (Trajano et al., 2014).

Separate one-way repeated-measures ANOVA were used to compare changes in ΔF and peak discharge rates over time (Control 1, Control 2 and Post-stretch). Tukey post hoc comparisons were performed as follow-up tests. Comparisons between Control 1 and Control 2 were used to investigate changes in the control condition while comparisons between Control 2 and Post-stretch were used to determine changes after the stretching condition. Test–retest reliability (Control 1 and Control 2) of ΔF values and peak discharge rates were evaluated by intraclass correlation coefficients (ICC) via linear regression (Hopkins, 2017). Statistical significance was set at an α-level of 0.05. All data are presented as means and 95% confidence interval.

The total number of motor units tracked across the three time points was 82 for SOL and 153 for GM. On average, 5.5 (3.9–7.1) SOL and 10.2 (7.5–12.9) GM motor units were tracked per participant. A total of 73 SOL and 328 GM pairs of motor units were identified across the three time points and used to compute ΔF values. On average, 4.9 (1.3–8.4) SOL and 21.9 (9.5–34.2) GM pairs of motor units were used per participant.

ICC values between Control 1 and Control 2 were high for SOL (0.94, 0.82–0.98) and GM (0.97, 0.90–0.99) ΔF. Similarly, high ICC values were observed for SOL (0.95, 0.87–0.98) and GM (0.93, 0.79–0.97) peak discharge rates.

There was a significant reduction in SOL ΔF (F_2,28=7.63; P=0.002) over time [Control 1, 2.91 pulses s⁻¹ (2.02–3.79 pulses s⁻¹); Control 2, 2.97 pulses s⁻¹ (2.03 to –3.92 pulses s⁻¹); Post-stretch, 2.23 pulses s⁻¹ (1.27–3.18 pulses s⁻¹)], with ΔF significantly reduced at Post-stretch compared with Control 2 [−25.6% (−42.1% to −9.1%); P=0.004; Fig. 1]. No changes were detected in GM ΔF (F_2,28=1.16; P=0.33) over time [Control 1, 2.89 pulses s⁻¹ (2.19–3.60 pulses s⁻¹); Control 2, 2.86 pulses s⁻¹ (2.18–3.54 pulses s⁻¹); and Post-stretch 2.62 pulses s⁻¹ (1.79–3.44 pulses s⁻¹)] (Fig. 2).

There was a significant increase in peak discharge rate in SOL (F_2,28=5.14; P=0.013) over time (Control 1, 9.94 pulses s⁻¹ (9.13–10.75 pulses s⁻¹); Control 2, 10.38 pulses s⁻¹ (9.49–11.27 pulses s⁻¹); Post-stretch, 10.49 pulses s⁻¹ (9.61–11.38 pulses s⁻¹); Fig. 3A]. However, while peak discharge rate was higher at Post-stretch than in Control 1 (P=0.014), a significant difference was not observed between Control 2 and Post-stretch (P=0.79). No changes were detected in GM (F_2,28=2.01, P=0.15) motor units over time (Fig. 3B).

The main finding of the present study was that passive stretching of the plantar flexors reduced ΔF in SOL but not GM, suggesting that stretch-induced PIC inhibition occurred only in SOL motor units. These data are consistent with the hypothesis that a stretch-induced decrease in PIC strength might underpin post-stretch force reductions.

Motoneuronal disfacilitation has been suggested as a mechanism involved in the reduction in neural drive after passive muscle stretching (Trajano et al., 2017). Preliminary evidence for this hypothesis was gathered in an experiment in which passive stretching reduced the strength of ongoing muscle contractions (i.e. self-sustained motor unit firing) that were elicited via Ia afferent input during tendon vibration and amplified by simultaneous muscle electrical stimulation but which persisted after both tendon vibration and electrical stimulation ceased (Trajano et al., 2014). These data are indicative of a reduction in motoneuron gain, possibly through reductions in PIC amplitude (Trajano et al., 2014). The present study used a more robust and validated method (the paired motor unit technique) to test this hypothesis. We observed a ∼26% reduction in ΔF in SOL but not in GM (see Fig. 2), suggesting the change caused by passive stretching was very significant, but specific to SOL. In the previous study using tendon vibration, it was speculated that inhibition of Ia afferents might be a factor influencing the loss of motoneuron facilitation, as these afferents were the main source of synaptic input in that study (Trajano et al., 2014). However, PIC amplitude was estimated during voluntary contractions in the present study, where corticospinal projections should be the main source of synaptic input. Therefore, PIC amplitude appears to be reduced regardless of whether it is estimated as the response to tendon vibration (predominantly Ia input) or using the paired motor unit technique (voluntary contraction).

It is possible that SOL ΔF reduction has a non-localized, ubiquitous, origin. Considerable evidence for this comes from experiments demonstrating that passive stretching of the ipsilateral limb can reduce the motor output of a non-stretched contralateral limb (Caldwell et al., 2019; da Silva et al., 2015; Cè et al., 2020). With respect to a non-localized mechanism, it is interesting to note that passive muscle stretching affects autonomic regulation, changing the sympathetic–parasympathetic balance. More specifically, an increase in parasympathetic and/or a decrease in sympathetic drive may be triggered by passive stretching, leading to a post-stretch reduction in noradrenergic activity (Kruse and Scheuermann, 2017; Mueck-Weymann et al., 2004; Farinatti et al., 2011; Inami et al., 2014). Importantly, PICs are strongly facilitated in the presence of both serotonin and noradrenaline, and PIC amplitude (and therefore ΔF) has been shown to be directly proportional to the level of monoaminergic input from the brainstem (Johnson et al., 2017). For instance, Udina et al. (2010) found that amphetamine ingestion leading to increased presynaptic release of noradrenaline in humans caused a 62% increase in ΔF without changing motor unit initial or average discharge rates (i.e. changes in ΔF resulted predominantly from changes in de-recruitment rates). These results (Udina et al., 2010) are consistent with the results of the present study, in which changes in ΔF were also observed without changes in peak discharge rate, i.e. no difference between Control 2 and Post-stretch (Fig. 3). It was notable that there was a small increase (0.5 pulses s⁻¹) in peak discharge rate in SOL (although not GM) from Control 1 to Post-stretch. However, there was no significant increment between Control 1 and Control 2 (coefficient of variation=3.8%) and no increment from Control 2 to Post-stretch. Moreover, the magnitude of this change is unlikely to be physiologically meaningful. In fact, we would have expected increases in ΔF alongside increases in peak discharge rate, so the observation of a decrease in ΔF from Control to Post-stretch despite an increase in peak discharge rate reinforces the effect of stretching on SOL PICs. The clear effect of amphetamine ingestion on ΔF suggests that, similar to what has been observed in animal preparations (Rank et al., 2007; Lee and Heckman, 1999), the activation of α1 adrenergic receptors in humans strongly contributes to PIC amplitude. Therefore, reductions in noradrenergic input from the locus coeruleus after passive stretching could speculatively play a role in the reduction in PIC amplitude. However, more targeted mechanistic experiments are necessary to explicitly test this hypothesis.

The specific reduction that was observed in SOL, but not GM, ΔF cannot be explained only by the noradrenergic hypothesis and requires further explanation. Several minutes of plantar flexor stretching typically decreases maximal muscle excitation capacity, as measured by EMG amplitude (Trajano et al., 2017). A number of studies have reported this reduction to occur specifically in SOL but not the medial or lateral gastrocnemii, suggesting a potential muscle-specific effect of stretching (Pulverenti et al., 2020, 2019; Trajano et al., 2013b). The reasons for this muscle-specific effect are still unknown but some possibilities deserve consideration. First, it could be speculated that SOL motoneurons might contain a greater number of PIC-amplifying monoaminergic receptors or a greater density of PIC-producing L-type voltage-gated calcium channels (Ca_V1.2 and Ca_V1.3) and/or voltage-gated sodium channels (Na_V1.1 and Na_V1.6) (Wilson et al., 2015; Binder et al., 2020). However, little is currently known about these possible differences. Moreover, the findings of the present study do not lend support to this assertion as ΔF values were similar between muscles (SOL and GM), and lower-threshold motor units are likely to have been sampled from both muscles because of the low intensity of the contractions used (20% MVC). Second, PICs tend to be more pronounced in slow-type motor units (Heckman et al., 2008), which are abundant in human SOL (∼80–70%) but less abundant in GM (∼55–60%) (Gollnick et al., 1974; Houmard et al., 1998; Harridge et al., 1996), and have been suggested to play an important role in the tonic firing of postural muscles (e.g. soleus) by decreasing the amount of descending drive required to maintain sustained contractions (Heckman et al., 2008). Evidence to support this hypothesis comes from several studies reporting the typical pattern of self-sustained firing observed specifically in SOL muscles, consistent with the occurrence of plateau-potentials that are a hallmark of PICs (Eken and Kiehn, 1989; Eken, 1998; Collins et al., 2001, 2002). Interestingly, depletion of spinal monoamines seems to decrease SOL tonic firing in rats (Kiehn et al., 1996), suggesting that SOL motoneurons might be particularly affected by reductions in monoaminergic input. In addition, SOL shows greater EMG amplitude in relation to GM during reflexive self-sustained contractions induced by tendon vibration in humans (Trajano et al., 2014). Furthermore, the magnitude of this SOL, but not GM, self-sustained firing increases with decreases in antagonist muscle length, suggesting that SOL motoneuron PICs might be more easily modulated compared with those in GM (Trajano et al., 2014). Third, muscle spindle Ia afferent input, which favours slow-twitch motor units and is an important source of PIC initiation, is greater in SOL than in GM (Tucker and Türker, 2004; Eccles et al., 1957) and could be negatively affected by stretching (Trajano et al., 2017). However, a problem with this hypothesis is that a lack of change in Ia afferent pathway excitability is commonly shown after passive stretching (Opplert et al., 2020; Budini et al., 2018; Pulverenti et al., 2020). In particular, one recent study has reported reductions in SOL EMG amplitude without detectible changes in the Ia afferent pathway (no reduction in H-reflex amplitude) after passive stretching, suggesting that reductions in this pathway are unlikely to contribute to a reduction in SOL excitation (Pulverenti et al., 2020). Fourth, an alternative explanation could be that SOL and GM might have received different magnitudes of stretch, which might then impact on the afferent signal acting on the motoneuron pools, and thus the capacity for PIC initiation. However, further research is required to test this hypothesis. Finally, further research is also required to determine how PIC strength might be differentially modulated between muscles within a single synergistic group.

In conclusion, passive muscle stretching detectibly reduced PIC strength in SOL but not GM motor units, suggesting not only that a loss of PIC strength might contribute to the loss of muscle force production after passive (static) muscle stretching but also that the muscle-specific decrease in muscle activity (EMG) observed previously (Pulverenti et al., 2019) might be explained by a specificity of PIC inhibition. Thus, the present data support the supposition that prolonged periods (e.g. several minutes) of muscle stretching can acutely affect the operation of persistent inward currents in spinal motoneurons, and that this reduces motoneuron (and thus muscle) activation in vivo in humans. The development of methods that allow the testing of associations between post-stretch alterations in PIC strength and changes in motor unit firing characteristics during maximal contraction are needed to more explicitly test this hypothesis.