Enhanced availability of serotonin increases activation of unfatigued muscle but exacerbates central fatigue during prolonged sustained contractions

Abstract

Key points

• Animal preparations have revealed that moderate synaptic release of serotonin (5‐HT) onto motoneurones enhances motor activity via activation of 5‐HT₂ receptors, whereas intense release of 5‐HT causes spillover of 5‐HT to extrasynaptic 5‐HT_1A receptors on the axon initial segment to reduce motoneurone activity.

• We explored if increasing extracellular concentrations of endogenously released 5‐HT (via the selective serotonin reuptake inhibitor paroxetine) influences the ability to perform unfatigued and fatigued maximal voluntary contractions in humans.

• Following the ingestion of paroxetine, voluntary muscle activation and torque generation increased during brief unfatigued maximal contractions. In contrast, the ability to generate maximal torque with increased 5‐HT availability was compromised under fatigued conditions, which was consistent with paroxetine‐induced reductions in motoneurone excitability and voluntary muscle activation.

• This is the first in vivo human study to provide evidence that 5‐HT released onto the motoneurones could play a role in central fatigue.

Abstract

Brief stimulation of the raphe–spinal pathway in the turtle spinal cord releases serotonin (5‐HT) onto motoneurones to enhance excitability. However, intense release of 5‐HT via prolonged stimulation results in 5‐HT spillover to the motoneurone axon initial segment to activate inhibitory 5‐HT_1A receptors, thus providing a potential spinal mechanism for exercise‐induced central fatigue. We examined how increased extracellular concentrations of 5‐HT affect the ability to perform brief, as well as sustained, maximal voluntary contractions (MVCs) in humans. Paroxetine was used to enhance 5‐HT concentrations by reuptake inhibition, and three studies were performed. Study 1 (n = 14) revealed that 5‐HT reuptake inhibition caused an ∼4% increase in elbow flexion MVC. However, when maximal contractions were sustained, time‐to‐task failure was reduced and self‐perceived fatigue was higher with enhanced availability of 5‐HT. Study 2 (n = 11) used twitch interpolation to reveal that 5‐HT‐based changes in motor performance had a neural basis. Enhanced 5‐HT availability increased voluntary activation for the unfatigued biceps brachii and decreased voluntary activation of the biceps brachii by 2–5% following repeated maximal elbow flexions. The final study (n = 8) investigated whether altered motoneurone excitability may contribute to 5‐HT changes in voluntary activation. F‐waves of the abductor digiti minimi (ADM) were unaffected by paroxetine for unfatigued muscle and marginally affected following a brief 2‐s MVC. However, F‐wave area and persistence were significantly decreased following a prolonged 60‐s MVC of the ADM. Overall, high serotonergic drive provides a spinal mechanism by which higher concentrations of 5‐HT may contribute to central fatigue.

Key points

• Animal preparations have revealed that moderate synaptic release of serotonin (5‐HT) onto motoneurones enhances motor activity via activation of 5‐HT₂ receptors, whereas intense release of 5‐HT causes spillover of 5‐HT to extrasynaptic 5‐HT_1A receptors on the axon initial segment to reduce motoneurone activity.

• We explored if increasing extracellular concentrations of endogenously released 5‐HT (via the selective serotonin reuptake inhibitor paroxetine) influences the ability to perform unfatigued and fatigued maximal voluntary contractions in humans.

• Following the ingestion of paroxetine, voluntary muscle activation and torque generation increased during brief unfatigued maximal contractions. In contrast, the ability to generate maximal torque with increased 5‐HT availability was compromised under fatigued conditions, which was consistent with paroxetine‐induced reductions in motoneurone excitability and voluntary muscle activation.

• This is the first in vivo human study to provide evidence that 5‐HT released onto the motoneurones could play a role in central fatigue.

Introduction

The effects of serotonergic activity in the CNS are diverse with widespread actions occurring throughout the brain and spinal cord. In the spinal cord, serotonin (5‐HT) plays a critical role in modulating motor activity, where it is commonly reported that 5‐HT facilitates motoneurone excitability. The inhibitory effects of 5‐HT on motoneurones are less frequently observed but have been occasionally described in cellular and slice preparations (Perrier & Cotel, 2015). Recent studies using adult turtle spinal cord preparations report that under conditions of moderate 5‐HT synaptic release, 5‐HT₂ receptors are activated and motoneurone excitability increases (Cotel et al. 2013). However, with greater activity of descending serotonergic fibres, and hence greater 5‐HT release, inhibitory 5‐HT_1A receptors are activated and motoneurone facilitation gives way to depression (Cotel et al. 2013; Perrier et al. 2018).

To understand how the concentration of 5‐HT can differentially affect motoneurone excitability, there must first be an appreciation for the location of 5‐HT release sites and 5‐HT receptor subtypes. The somato‐dendritic compartment of motoneurones is densely innervated by 5‐HT synaptic boutons as well as 5‐HT₂ receptors. With this arrangement, the most immediate binding site for 5‐HT is one associated with excitation. As well as influencing membrane thresholds, activation of 5‐HT₂ receptors strongly enhances the activity of persistent inward currents, which in turn promote repetitive firing of the motoneurones. In contrast, 5‐HT_1A receptors are expressed extrasynaptically on the motoneurone axon initial segment, which is devoid of 5‐HT release sites. Therefore, inhibitory motoneurone responses can only occur when large synaptic release of 5‐HT ‘spills over’ into extrasynaptic compartments (Cotel et al. 2013; Perrier et al. 2018). Activating 5‐HT_1A receptors on the axon initial segment causes inwardly rectifying K⁺ conductance via GIRK (K3) channels to create a state of tonic hyperpolarization, and inhibits Na⁺ channels that are responsible for the genesis of action potentials (Williams et al. 1988; Bayliss et al. 1997; Cotel et al. 2013; Petersen et al. 2015). This mechanism also appears relevant to humans, as ingestion of a 5‐HT_1A receptor agonist reduces the capacity to perform prolonged bouts of exercise (Marvin et al. 1997) and suppresses spinal motoneurone excitability (D'Amico et al. 2017). Therefore, a reduction in motoneurone firing due to inhibitory actions of 5‐HT provides a spinal mechanism by which higher concentrations of 5‐HT may contribute to declines in motor performance during fatigue‐inducing exercise.

The present study examined how increased extracellular concentrations of endogenously released 5‐HT influences the ability to perform brief, as well as sustained, maximal voluntary contractions (MVCs) in humans. Our hypothesis was that increased levels of 5‐HT in brief, non‐fatiguing contractions would be excitatory and hence increase motor output, whereas extra 5‐HT during sustained contractions would be inhibitory and decrease motor output. The selective serotonin reuptake inhibitor (SSRI) paroxetine was used to enhance 5‐HT concentrations by reuptake inhibition, and three studies were performed. First, we assessed whether 5‐HT reuptake inhibition affected MVC amplitude, time‐to‐task failure and self‐perceived levels of fatigue. We predicted that enhanced availability of 5‐HT would increase the ability to generate force during brief maximal contractions, but would exacerbate characteristics of fatigue during prolonged fatiguing contractions. Second, we used twitch interpolation techniques to assess whether changes in muscle activity induced by 5‐HT were due to neural mechanisms or mechanisms within the muscle. We hypothesized that 5‐HT would not affect the force‐generating mechanisms of the muscle, but would be detrimental to the ability to voluntarily activate muscle via neural processes. Third, we assessed how 5‐HT availability affected motoneuronal excitability following brief, and prolonged, MVCs. If 5‐HT plays a role in centrally mediated fatigue then it was expected that increasing the availability of 5‐HT during maximal efforts would cause a larger than normal depression of motoneurone excitability.

Methods

Experiment design

Three studies were performed. Each study was a double‐blind, placebo‐controlled, cross‐over design in human participants. For each study, participants attended the laboratory on two occasions where paroxetine, or a placebo, was ingested at 11.00 h and experimental testing occurred 4 h after ingestion.

Participants and ethical approval

Fifteen healthy individuals (age: 24 ± 7 years, 7 female) participated in three studies. No participant had neurological or musculoskeletal dysfunction, or was taking any form of CNS medication that could influence the outcome measures of the experiments. All participants were right‐handed, which was determined using the Oldfield Edinburgh Inventory. The study received approval from the Griffith University Human Research Ethics Committee, and conformed to the standards set by the Declaration of Helsinki except for registration in a database.

Drug intervention

Paroxetine 20 mg and the placebo were compounded into opaque capsules to facilitate blinding. Paroxetine was selected as the serotonin reuptake inhibitor as it is a potent and selective inhibitor of 5‐HT reuptake and has a low affinity for adrenergic, dopaminergic, histaminergic and muscarinic receptors. In addition, paroxetine is a very weak inhibitor of noradrenaline uptake, where the uptake inhibition ratio of serotonin to noradrenaline is amongst the highest of SSRIs (Bourin et al. 2001). The order of drug administration was counterbalanced whereby half of the participants received paroxetine during the first session and half were administered placebo for their first session. The first and second testing sessions for each participant were performed one week apart.

Force and EMG recordings

For Studies 1 and 2, isometric elbow flexion torque and the corresponding biceps brachii EMG were measured during a series of maximal contraction tasks. Participants sat comfortably in a chair with their dominant arm fixed in a transducer designed to measure isometric elbow flexion torque (Fig. 1 A). This device placed the shoulder in 90 degrees of flexion and ensured that the elbow remained in 90 degrees of flexion throughout testing. A precision S‐beam load cell (PT4000, PT Ltd, New Zealand) with a 1.1 kN range and full scale output of 3 mV/V measured elbow flexion force, which was converted to flexion torque for each participant. Surface EMG was recorded from biceps brachii and triceps brachii by placing two Ag/AgCl electrodes (Kendall Arbo) on each muscle with an interelectrode distance of 24 mm. EMG data were amplified (×1000, NL844, Digitimer Ltd, Welwyn Garden City, UK), filtered (3–1000 Hz, NL135 and NL144, Digitimer Ltd) and sampled at 2 kHz using Spike2 software and a Power 1401 data acquisition interface (Cambridge Electronic Design, Cambridge, UK).

Figure 1. The experimental setup for Study 1 and Study 2.

Isometric elbow flexion torque, biceps brachii EMG and triceps brachii EMG was recorded from the dominant limb following the administration of paroxetine or a placebo (A). Study 1 assessed the duration that participants could maintain a maximal isometric contraction until force decreased to 60% MVC (B). The lower panel (C) presents maximal unfatigued elbow flexion torque from Study 1. All data in this panel are for brief (∼2 s) unfatigued muscle contractions to determine baseline MVC. Group data are presented as the mean ± SD and individual data are presented as solid grey lines (n = 14). The asterisk (^*) indicates significant difference between the paroxetine and placebo condition (P < 0.05).

For Study 3, F‐waves were obtained from the abductor digiti minimi (ADM) before and after a series of maximal fifth digit abduction contractions. Participants sat comfortably in a chair with their dominant arm rested on a table. The elbow was flexed to 90 degrees and the hand was pronated and firmly secured into a hand‐plate that prevented movement of the forearm and hand. The thumb was secured in a mid‐flexed position, and the second to fourth fingers were firmly secured with velcro. The little finger was fixed into a rigid bracket to prevent movement during stimulation. Surface EMG was recorded from the ADM with Ag/AgCl electrodes (Kendall Arbo) placed over the motor point of the muscle and over the lateral aspect of the fifth metacarpo‐phalangeal joint. EMG data were amplified (×1000, NL844, Digitimer Ltd), filtered (20–1000 Hz, NL135 and NL144, Digitimer Ltd) and sampled at 5 kHz using Signal software and a Power 1401 interface (Cambridge Electronic Design).

Study 1: time‐to‐task failure and perceived levels of fatigue after elbow flexion MVCs

Fourteen individuals participated in the first study. Each participant performed five brief isometric maximal elbow flexions (∼2 s) to determine MVC torque (Fig. 1 B). Contractions were separated by periods of 3–4 min of rest, and the largest torque was taken as the individual's MVC. After a 5‐min rest period, a time‐to‐task failure protocol was performed. The protocol consisted of eight maximal elbow flexions, each maintained until torque declined to less than 60% MVC for greater than 3 s. Rest periods of 40 s were provided between fatiguing contractions and strong verbal encouragement was provided at all times. Time‐to‐task failure was calculated from the onset of elbow flexion torque to the time that torque declined to 60% MVC for greater than 3 s during each sustained contraction. From this window of data, the area under the torque curve and the area under the rectified EMG curves for biceps and triceps were calculated using Spike2 software.

To provide a psychophysical measure of fatigue, a CR‐10 Borg scale was used following each sustained contraction. Prior to testing, the participants were instructed that the lowest value on the CR‐10 represented ‘no fatigue at all’, whereas the highest value on the scale represented ‘so much fatigue that you would not be able to lift your arm’. The Stanford Sleepiness Scale was used to assess the degree of alertness of each participant at the time of capsule ingestion, immediately before the contraction protocol, and immediately following the contraction protocol.

Study 2: level of voluntary activation for biceps brachii following elbow flexion MVCs

Eleven individuals participated in the second study. Nine had also participated in Study 1. The effects of serotonin reuptake inhibition on fatiguing muscle contractions observed in Study 1 were further explored with electrical stimulation in this experiment. Voluntary activation was examined for the non‐fatigued and fatigued biceps muscle via stimulation of intramuscular fibres (motor nerve). Electrically evoked increases in torque were quantified during maximal contractions (superimposed twitch) and for the relaxed muscle (resting twitches) in this experiment.

While the arm was fixed in the force transducer, an anode was placed over the bicipital tendon and a surface cathode was placed on the muscle. Prior to the cathode being attached, a motor point ball‐pen electrode was used to identify the biceps region that (1) produced the largest twitch response with, (2) the smallest stimulation intensity and (3) the lowest triceps activation. Single electrical stimuli with a pulse width of 100 μs were delivered with a constant‐current stimulator (DS7AH, Digitimer) to the motor nerve. The stimulation intensity for each participant was set to 20% higher than the intensity required to elicit a maximal resting twitch (range: 75–260 mA).

Participants performed five brief maximal efforts (∼2 s). During each MVC a stimulus was delivered to the motor nerve to elicit a superimposed twitch, and 3 s later a stimulus was delivered to the relaxed muscle to elicit a resting twitch. A contraction protocol was then used to induce biceps fatigue, whereby four sustained maximal contractions were performed, with each contraction maintained until force declined to 60% MVC. Each fatiguing contraction was followed 3 s later by a resting twitch, and then 3 s later by a brief maximal contraction with superimposed twitch. The short period before performance of the brief MVC reduced the likelihood that central mechanisms of fatigue had recovered following the fatiguing contraction. After the brief MVC a 32 s rest period occurred until the next fatiguing contraction (which approximated the rest cycles in Study 1). All resting twitches occurred after a maximal contraction so that every resting twitch was obtained from a potentiated muscle (Todd et al. 2003).

Data were analysed using Spike2 software. Peak torque was calculated from MVCs, and the root mean square amplitude of biceps and triceps EMG (EMG rms) was calculated from a 100 ms window spanning the peak torque (i.e. 50 ms either side of peak). Peak‐to‐peak amplitude of the superimposed and resting twitches were calculated from increases in the torque signal following electrical stimulation. Time‐to‐peak and half‐relaxation time were calculated from the stimulus to the peak of the superimposed and resting twitch. Voluntary activation was calculated as [1 − (superimposed twitch/resting twitch)] × 100, where the superimposed and resting twitches followed the same fatiguing contraction.

Study 3: F‐waves in ADM after 2 s and 60 s MVC

Eight individuals participated in the third study, seven of whom had also participated in Studies 1 and 2. The third study assessed the effects of paroxetine on motoneurone excitability by evoking F‐waves in resting ADM. F‐wave data were collected with sets of 30 stimuli (pulse width of 100 μs, frequency of 0.5 Hz), where electrical stimuli were delivered to the ulnar nerve with a constant‐current stimulator (DS7AH). Stimulating electrodes were placed 3 cm apart over the ulnar nerve, with the cathode positioned distal to the anode and ∼2 cm from the wrist joint. After identifying the stimulus intensity required to elicit a maximal M‐wave of the ADM, the stimulator was set to 150% of this intensity for F‐wave testing (range: 27–52 mA).

F‐waves were examined in ADM before and after a 2‐s MVC and a 60‐s MVC (Khan et al. 2012). Two control sets of F‐waves, separated by 1 min of rest, were obtained from ADM. Following another 1‐min rest period each participant performed a 2‐s maximal abduction of the fifth digit. Participants were instructed to relax immediately following the MVC, and then six sets of F‐waves (separated by 30 s) were collected. To verify that participants were relaxed during post‐MVC sets, ADM EMG rms was measured over a 50 ms window prior to each stimulus. At the completion of data collection for the 2‐s MVC task, participants were given 30 min of recovery before the same protocol was repeated with a 60‐s MVC.

To extract F‐waves from ADM, a high‐pass, fourth‐order, Butterworth filter with a cut‐off frequency of 220 Hz was applied to the EMG signal using Signal software. This filter reduced the tail of the M‐wave to baseline so that clear deflections in EMG (F‐waves) could be observed following the M‐wave. An F‐wave was identified if the filtered EMG signal increased above baseline (≥5 μV) and a clear compound muscle action potential emerged at a latency greater than 25 ms. The persistence of F‐waves was calculated as the percentage of F‐waves identified in each set of 30 supramaximal stimuli. The area and peak‐to‐peak amplitude of F‐waves were also calculated from the filtered signal before being normalized to their corresponding filtered M‐waves. This normalization accounted for changes in muscle fibre action potential that may have occurred due to fatigue during the MVC (Khan et al. 2012). Post‐MVC data were then normalized to control data to account for any depression of motoneurone excitability that may occur with serotonin reuptake inhibition regardless of fatigue.

Statistics

Normality of data was assessed using Shapiro–Wilk tests. The amplitude of unfatigued MVCs for the paroxetine and placebo conditions was normally distributed and compared with paired t tests. However, baseline F‐wave characteristics were not normally distributed, and the paroxetine and placebo conditions were compared with Wilcoxon sign‐ranked tests. Stanford Sleepiness Scale data were examined using two‐way repeated measures ANOVA with factors of drug (paroxetine, placebo) and time (capsule ingestion, prior to testing, after contraction protocol). Perceived fatigue data were examined using two‐way repeated measures ANOVA with factors of drug (paroxetine, placebo) and contraction (1–8).

All torque‐derived measures and all EMG‐derived measures in the three studies were examined via two‐way repeated measures ANOVAs. Within‐subject factors of drug (paroxetine, placebo) and contraction set (Study 1: 1–8; Study 2: 1–4) were used in the ANOVA for Studies 1 and 2. Within‐subject factors of drug (paroxetine, placebo) and time (pre‐MVC to 8 min after MVC) were used in the ANOVA for Study 3. Mauchly's test of sphericity was applied to the data, and non‐spherical data were subjected to Greenhouse–Geisser corrections. If significant main or significant interaction effects were identified for each ANOVA, Tukey's multiple comparison post hoc tests were used to examine how the drug influenced the dependent variable at each contraction (or time point), or if differences existed between control and subsequent time points. All statistical procedures were performed using IBM SPSS Statistics (version 22) with alpha levels set at <0.05.

Results

Study 1

Unfatigued maximal voluntary contraction

The maximal torque that was produced during the unfatigued MVCs was significantly greater for the paroxetine condition (66.2 ± 26.1 N.m) compared to the placebo condition (63.2 ± 23.4 N.m, t ₁₃ = 3.237, P = 0.006). Of the 14 participants, 11 generated greater maximal torque for the paroxetine condition (mean 5.5% increase) whereas 3 generated a lower maximal torque for the paroxetine condition (mean 2.7% decrease; see Fig. 1 C).

Time‐to‐task failure, torque and EMG during repeated maximal contractions

The ingestion of paroxetine affected the ability to perform successive fatiguing contractions. A main effect of drug was detected for time‐to‐task failure (F _1,13 = 5.322, P = 0.038, Fig. 2 A), the area under the torque curve (F _1,13 = 7.595, P = 0.016, Fig. 2 B) and biceps EMG area (F _1,13 = 7.595, P = 0.016, Fig. 2 C), which were all reduced for the paroxetine condition compared to the placebo condition. Performing repeated maximal contractions with a fixed rest interval of 40 s resulted in different responses between contraction sets, as a main effect of contraction number was detected for time‐to‐task failure (F _1.5,20.3 = 12.59, P < 0.001), the area under the torque curve (F _1.5,18.7 = 52.843, P < 0.001), and the area under the biceps EMG curve (F _1.7,23.2 = 12.59, P < 0.001). No drug by contraction interaction effect occurred for time‐to‐task failure or the area under the torque curve, although an interaction effect was detected for area under the biceps EMG curve (F _2.7,35.8 = 5.043, P = 0.006). Post hoc tests indicated that EMG area for paroxetine was reduced compared to placebo from the first to the third contraction set (Fig. 2 C). Furthermore, EMG declined from the first to second contractions and from the second to third contractions during both the paroxetine and the placebo conditions.

Figure 2. Eight maximal effort sustained isometric elbow flexions were each performed until flexion torque declined to 60% MVC in Study 1.

Each contraction was separated by 40 s of rest. The time‐to‐task failure (A), area of elbow flexion torque (B) and area of biceps brachii EMG (C) were calculated for each contraction to represent fatigue‐related changes during the fatiguing contractions. Data are presented as group means ± SD (n = 14). Main effects of drug and contraction were identified for each data set. Asterisks (^*) indicate drug differences (P < 0.05) and hash symbol (^#) indicates contraction differences (P < 0.05) for the drug by contraction interaction.

Self‐perceived alertness and fatigue during repeated contractions

Paroxetine exacerbated inherent levels of sleepiness. For the Stanford Sleepiness Scale, a main effect of drug (F _1,13 = 13.76, P = 0.003) and time (F _1.8,23.9 = 15.392, P < 0.001), and a drug by time interaction (F _1.6,21.9 = 3.900, P = 0.042) was detected. Post hoc analysis indicated that sleepiness increased from pill ingestion (11.00 h) to the commencement of experimental testing (15.00 h) for both the paroxetine and the placebo conditions (Fig. 3 A), with paroxetine causing greater sleepiness than placebo by the start of experimental testing.

Figure 3. Self‐perceived ratings of alertness and fatigue in Study 1.

Stanford Sleepiness Scale assessed alertness at the time of drug ingestion, 4 h later at the commencement of testing, and immediately following the fatigue protocol (A). Participants also provided measures of fatigue via a CR‐10 Borg scale after each contraction throughout the fatigue protocol (B). Data are presented as group means ± SD (n = 14). Main effects of drug and contraction were identified for each data set. Asterisks (^*) indicate drug differences (P < 0.05) and hash symbol (^#) indicates contraction differences (P < 0.05) for the drug by contraction interaction.

Paroxetine also enhanced self‐perceived levels of fatigue. Significant effects of drug (F _1,13 = 9.147, P = 0.012) and contraction number (F _3.6,39.9 = 88.8, P < 0.001) were detected for rating of perceived fatigue, where rated fatigue was significantly greater for all sustained contractions throughout the fatigue protocol (Fig. 3 B). No drug by contraction interactions were identified for ratings of perceived fatigue. Although structured interviews were not performed, 10 out of 14 participants expressed feelings after the paroxetine testing session to the effect that, ‘no matter how hard I tried, it felt like I couldn't get my muscle to contract properly’.

Study 2

Paroxetine reduced the time‐to‐task failure, as well as the area of torque and EMG in repeated sustained MVCs in Study 1. Study 2 further examined the capacity for voluntary drive to maximally activate the biceps brachii following paroxetine ingestion, both when the muscle was fresh and after sustained fatiguing MVCs (Fig. 4 A).

Figure 4. Superimposed and resting twitches were obtained for unfatigued and fatigued maximal isometric elbow flexions in Study 2.

The contraction protocol (A) consisted of brief MVCs to establish control measures for the unfatigued biceps brachii, followed by four sustained maximal contractions until torque declined to 60% MVC. Supramaximal electrical stimulation was applied to the biceps brachii during MVCs (solid arrows) to elicit superimposed twitches and to the relaxed muscle to elicit potentiated resting twitches (open arrows). The left lower panels illustrate average traces of elbow flexion torque responses following motor nerve stimulation of the potentiated relaxed muscle (resting twitch, B) and during brief MVCs (superimposed twitch, C). Arrows indicate the moment of motor nerve stimulation, and data are offset to the voluntary torque produced at the time of stimulation. The right lower panels present the data as group means ± SD (n = 11). Paired t tests revealed that paroxetine reduced superimposed twitch for the control MVC (P < 0.05). A main effect of drug was identified using two‐way ANOVA where superimposed twitch was greater for paroxetine than placebo during the sustained contraction task (P < 0.05).

Unfatigued maximal voluntary contraction

Again, the maximal torque that was produced during the unfatigued brief MVCs was significantly greater for the paroxetine condition (64.8 ± 25.2 N.m) compared to the placebo condition (61.9 ± 21.6 N.m, t ₁₀ = 1.971, P = 0.038). No drug‐related differences were observed for resting twitch amplitude (unfatigued data in Fig. 4 B), but superimposed twitches (t ₁₀ = 2.345, P = 0.041, unfatigued data in Fig. 4 C) and voluntary activation (t ₁₀ = 2.650, P = 0.026, unfatigued data in Fig. 5 A) were significantly different between conditions. The superimposed twitch was 14.6 ± 17.3% lower for the paroxetine compared to the placebo condition, and this was reflected by a 1.5 ± 1.8% higher level of voluntary activation for the paroxetine compared to the placebo condition.

Figure 5. Level of voluntary activation calculated from superimposed twitches and resting twitches evoked by motor nerve stimulation in Study 2.

Voluntary activation in A is presented as group means ± SD. In B, voluntary activation is presented for individuals (n = 11). For clarity, different y‐axis scaling is used in each panel. Paired t tests revealed that paroxetine increased voluntary activation for the control MVC (P < 0.05), and a main effect of drug was identified using two‐way ANOVA where voluntary activation was lower for paroxetine than placebo during the sustained contraction task (P < 0.05).

Torque generation and EMG following fatiguing contractions

Following the fatiguing contractions, maximal elbow flexion torque declined (main effect of contraction: F _2.3,23.3 = 6.516, P = 0.004, Fig. 6 A). In contrast to the unfatigued MVCs, maximal torque produced during brief MVCs following fatigue was lower in the paroxetine condition compared to the placebo condition (main effect of drug: F _1,10 = 11.745, P = 0.006). No drug‐ or contraction‐related differences were identified for the associated biceps EMG rms during MVCs following the fatiguing contractions (Fig. 6 B). Time‐to‐task failure, area under the torque curve and area under the biceps EMG curve yielded the same outcomes as Study 1 and are not repeated here.

Figure 6. Maximal elbow flexion torque (A) and the corresponding biceps brachii EMG (B) following each fatigue‐inducing contraction in Study 2.

All data are from brief (∼2 s) maximal contractions performed approximately 6 s after the cessation of the fatigue‐inducing sustained contraction. Data are presented as a percentage of the corresponding measurement obtained during the unfatigued control MVCs (group means ± SD, n = 11).

Resting twitch, superimposed twitch and level of voluntary activation

Paroxetine did not influence the contractile properties of the muscle (i.e. mechanisms of peripheral fatigue), as there were no drug‐related effects for resting twitch amplitude (Fig. 4 B), time‐to‐peak torque or half‐relaxation time (Table 1). Instead, neural mechanisms associated with the ability to activate the muscle were affected by paroxetine.

Table 1.

Time‐to‐peak torque and half‐relaxation time for superimposed and resting twitches from Study 2

	Time‐to‐peak (ms)		Half‐relaxation time (ms)
	Placebo	Paroxetine	Placebo	Paroxetine
Superimposed twitch
Unfatigued	40 ± 9	41 ± 9	69 ± 18	66 ± 19
Contraction 1	47 ± 11	47 ± 10	71 ± 12	68 ± 18
Contraction 2	43 ± 10	45 ± 9	72 ± 13	71 ± 21
Contraction 3	43 ± 10	44 ± 10	72 ± 13	69 ± 23
Contraction 4	44 ± 10	45 ± 11	70 ± 13	69 ± 20
Resting twitch
Unfatigued	78 ± 7	77 ± 8	134 ± 20	135 ± 21
Contraction 1	78 ± 8	76 ± 10	138 ± 25	140 ± 31
Contraction 2	79 ± 8	78 ± 11	136 ± 29	139 ± 33
Contraction 3	79 ± 7	78 ± 9	140 ± 32	141 ± 32
Contraction 4	78 ± 5	76 ± 10	142 ± 32	142 ± 32

Latencies are relative to time of stimulation. Values are given as means ± SD.

For superimposed twitch, a main effect of drug was identified (F _1,10 = 4.51, P = 0.046). In contrast to the unfatigued control MVCs, the superimposed twitch after the fatiguing contractions was lower for the placebo condition compared to the paroxetine condition. No main effect of contraction, or drug by contraction interaction was detected for superimposed twitch. Consistent with the changes in superimposed twitch, voluntary activation was reduced for the paroxetine condition compared to the placebo condition (F _1,10 = 4.434, P = 0.048, Fig. 5). No drug‐related differences emerged in superimposed twitch time‐to‐peak torque, or half‐relaxation time (Table 1).

Study 3

A third experiment was performed to investigate whether altered motoneurone excitability may have contributed to the 5‐HT‐related changes identified in Studies 1 and 2. In particular, Study 3 examined how serotonin reuptake inhibition influences F‐waves in unfatigued muscle, following a brief 2‐s MVC, and following a prolonged 60‐s MVC (Fig. 7 A, B).

Figure 7. F‐waves were collected from the ADM to provide insight into motoneurone excitability.

Blocks of 30 supramaximal ulnar nerve stimulations were collected before and after a 2‐s MVC (A) and a 60‐s MVC (B). Two unfatigued control sets (C1, C2) and six post‐fatigue sets (PF1–PF6) were collected for both contraction tasks. Representative F‐waves for the 60 s MVC task in Study 3 are presented in the lower panels (C). These data illustrate 30 consecutive F‐waves with a rest period of 2 s between stimulations. F‐waves for post‐fatigue 1 were obtained over the 60 s immediately following the MVC. F‐waves for post‐fatigue 2 were collected after a 30 s break, i.e. from 90 to 150 s after the MVC. EMG signals were initially band pass filtered from 20 to 1000 Hz during data collection, before a high‐pass filter with a cut‐off frequency of 220 Hz was applied offline to extract F‐waves.

F‐wave area and persistence

Baseline F‐wave characteristics were obtained from all participants 4 h after ingestion of the placebo and paroxetine. No drug‐related differences were identified for ADM F‐wave parameters when the muscle was unfatigued and in a relaxed state (Table 2). Figure 7 C shows representative F‐waves for the resting ADM before and after performing a 60‐s MVC. In line with previous investigations, only F‐wave area is reported in our results, as F‐wave amplitude and area followed similar patterns following the 2‐s and 60‐s MVC (Khan et al. 2012).

Table 2.

Baseline F‐wave parameters for ADM measured 4 h after drug ingestion

	Onset latency (ms)	Amplitude (mV)	Area (mV.s)	Persistence (%)	Chronodispersion (ms)	F/M area ratio (%)
Placebo	31.9 ± 2.8	94.1 ± 51.4	0.184 ± 0.079	90.7 ± 6.6	5.4 ± 1.4	2.56 ± 1.02
Paroxetine	31.7 ± 3.3	93.0 ± 50.2	0.177 ± 0.068	87.8 ± 8.3	5.9 ± 1.7	2.39 ± 0.98

Baseline data are reported as the average of the two control sets (n = 60 stimulations per participant) obtained prior to the 2‐s MVC in Study 3. Values are given as means ± SD.

Following the 2‐s MVC, F‐wave area was initially reduced by 23.3 ± 16.2% for the placebo condition, and 27.3 ± 19.4% for the paroxetine condition. However, no main effects of drug or time, and no drug by time interaction effects, were detected for F‐wave area following the 2‐s MVC (Fig. 8 A). F‐wave persistence exhibited a similar profile to F‐wave area where no drug effect was observed. However, a significant effect of time (F _1.5,10.9 = 25.6, P < 0.001), and a drug by time interaction (F _2.1,14.5 = 4.21, P = 0.035) was identified. The occurrence of F‐waves decreased by 12.1 ± 8.2% for the placebo, and significantly decreased by 17.4 ± 12.1% for paroxetine for the first minute following the 2‐s MVC (Fig. 8 C). Within this 1‐min period following the 2‐s MVC, F‐wave persistence was reduced for the paroxetine condition compared to the placebo condition.

Figure 8. ADM F‐waves before and after a 2‐s MVC and a 60‐s MVC in Study 3.

F‐wave area (A and B) and F‐wave persistence (C and D) were measured following the ingestion of paroxetine and a placebo. As no differences existed in the two sets of control F‐waves obtained prior to each MVC, control data were averaged for the two sets for each condition. Asterisks (^*) indicate F‐waves where significant drug effects were identified (P < 0.05). A hash symbol (^#) indicates post‐MVC periods where F‐waves were significantly reduced compared to F‐waves obtained under control conditions (P < 0.05). Data are presented as group means ± SD (n = 8).

Following the 60‐s MVC, main effects of drug (F _1,7 = 7.105, P = 0.032) and time (F _2.5,18.1 = 5.260, P = 0.011) were detected for F‐wave area (Fig. 8 B). No drug by time interaction was detected for F‐wave area following the 60‐s MVC. F‐wave area was initially reduced by 35.9 ± 18.9% for the placebo condition and 54.9 ± 25.3% for the paroxetine condition, and remained reduced for 1 min following the 60‐s MVC. Main effects of drug (F _1,7 = 70.568, P < 0.001) and time (F _2.3,16.4 = 45.031, P = 0.002) were also identified for F‐wave persistence during the 60‐s MVC task (Fig. 8 D). The occurrence of F‐waves following the 60‐s MVC significantly decreased by 31.8 ± 15.6% for the placebo and 48.1 ± 22.1% for paroxetine, and remained lower than baseline for 4 min after the 60‐s MVC.

Discussion

The purpose of this study was to examine how increased extracellular concentrations of endogenously released 5‐HT influence the ability to perform maximal effort muscle contractions. The SSRI paroxetine was used to enhance 5‐HT concentrations, and three studies examined how muscle activity is altered with enhanced availability of endogenous 5‐HT. Our main findings were that enhanced 5‐HT availability increased voluntary activation of muscle, and hence strength, during brief unfatigued contractions. In contrast, enhanced 5‐HT availability decreased voluntary activation and maximal voluntary torque during fatigued conditions. We also identified that increasing the available 5‐HT enhanced the depression of motoneurone excitability that is typically observed with fatigue. Overall, these results are consistent with known facilitatory and inhibitory effects of 5‐HT on motoneurones. That is, 5‐HT facilitates motoneurone firing through enhancement of persistent inward currents via synaptic 5‐HT₂ receptors. However, 5‐HT can also depress motoneurone excitability and firing via extrasynaptic 5‐HT_1A receptors, which are activated when descending serotonergic drive is strong enough to result in spillover of serotonin from the synapses. Thus, this is the first in vivo human study to provide evidence that 5‐HT released onto the motoneurones could play a role in central fatigue.

5‐HT reuptake inhibition requires serotonergic drive to modulate muscle activation

Motoneurone excitability was unaffected in resting unfatigued muscle following the ingestion of paroxetine. Thus, any actions of the drug, including 5‐HT reuptake inhibition, do not directly affect motoneuronal responsiveness. Instead, voluntary drive to the target muscle was required before changes in motor activity were detected. When maximal effort contractions were performed during the SSRI condition there was an increase in voluntary activation which caused a concomitant increase in maximal voluntary torque. This suggests that motor output was enhanced by the combination of 5‐HT reuptake inhibition, an increase in central drive to the target muscle, and an increase in serotonergic drive associated with performing the contraction. While the first two factors are evident from the results of this study, the relationship between serotonergic drive and muscle activity must be inferred from in vitro, and human pharmacological experiments.

Converging lines of evidence indicate that the level of 5‐HT drive to the spinal cord may be associated with the level of activity in descending pathways that directly activate muscles. The primary site of 5‐HT synthesis is the raphe nuclei of the brainstem. From there, the raphe–spinal pathway descends to the ventral horn of the spinal cord to form well‐defined synapses with interneurons and motoneurones (Ridet et al. 1994; Alvarez et al. 1998; Kawashima, 2018). Electrical stimulation of the raphe–spinal pathway in integrated preparations of the adult turtle spinal cord promotes the release of endogenous 5‐HT (Perrier & Cotel, 2008), and the discharge rate of raphe neurons is known to positively correlate with motor activity in cats (Hounsgaard et al. 1988; Jacobs & Fornal, 1999). Consistent with the findings of animal preparations, a spinal locus for 5‐HT effects has also been identified in humans. It is postulated that increasing serotonergic drive, and hence increasing 5‐HT release on spinal motoneurones, serves to regulate gain control in the spinal cord. In particular, the gain of tendon vibration reflexes and stretch reflexes is increased by enhancing the availability of endogenous 5‐HT (Wei et al. 2014). Moreover, voluntary contractions in a single limb cause changes in tremor and force control for the opposite limb, which aligns with the known 5‐HT effects on spinal motoneurones (Kavanagh et al. 2013; Wei et al. 2014; Kavanagh et al. 2016). These effects are enhanced by paroxetine and reduced by cyproheptadine (a 5‐HT₂ antagonist), and their relationship with the strength of voluntary contraction is consistent with an association between the level of serotonergic release onto the motoneurones and the level of motor activity (Wei et al. 2014).

By administering the SSRI paroxetine, negative allosteric modulation of the 5‐HT transporter causes increased extracellular concentrations of endogenous 5‐HT. Of the available SSRIs, paroxetine is the most selective for 5‐HT reuptake, with limited or no effects on dopamine, histamine, adrenergic or 5‐HT receptor subtypes (Thomas et al. 1987; Finley, 1994; Hyttel, 1994; Sanchez & Hyttel, 1999). Accordingly, the increase in motor output observed in unfatigued muscle following paroxetine ingestion is probably explained by increased availability of endogenous 5‐HT. The probable site of action is excitatory 5‐HT₂ receptors in the somato‐dendritic compartment of motoneurones in the spinal cord. As well as altering the membrane threshold of the motoneurones, activation of these receptors promotes activity of voltage‐gated persistent inward currents and allows repetitive firing of motoneurones (Li et al. 2007; Harvey et al. 2006). However, it is crucial to acknowledge that the enhanced level of voluntary activation and MVC could also be attributed to excitable 5‐HT subtypes other than 5‐HT₂. There is increasing evidence to suggest that serotonergic enhancement of motor function involves 5‐HT₂, 5‐HT₃ and 5‐HT₇ receptors expressed on the soma and proximal dendrites of motoneurones (Fone et al. 1991; Noga et al. 2009; Sławińska et al. 2013; Fields et al. 2015). Therefore, caution should be used when attributing changes in motor output (in vivo) to specific receptor subtypes, or even attributing changes in motor output solely to neurons located in the spinal cord. Serotonergic innervation is widespread throughout the brain, including the motor cortex where it is recognized to influence various aspects of motor function (Ohno et al. 2013; Vitrac & Benoit‐Marand, 2017).

Increased extracellular concentration of 5‐HT enhances central fatigue

Central fatigue can be defined as a progressive exercise‐induced reduction in voluntary activation of a muscle or muscle group. It represents a decline in the ability of the nervous system to drive the muscle maximally as fatigue develops. In humans, supramaximal stimulation of a motor nerve can be used to separate the peripheral and central contributions of fatigue (Merton, 1954). If exercise‐induced changes occur distal to the stimulation site, and so manifest as a decline in resting twitch, factors in the muscle such as excitation–contraction coupling, sarcolemma membrane conductivity or substrate availability may be compromised (Fuglevand et al. 1993; Macdonald & Stephenson, 2004; Dutka et al. 2008). If changes occur proximal to the site of stimulation, which is reflected by an increase in superimposed twitch during MVCs, the ability to voluntarily activate the muscle via neural processes may have been compromised due to exercise (McKenzie et al. 1992; Allen et al. 1998; Todd et al. 2003). There are numerous factors which can lead to central fatigue (for comprehensive reviews refer to Gandevia, 2001; Taylor et al. 2006; Taylor & Gandevia, 2008). Among these factors, exercise‐induced changes in the concentration of the monoamines 5‐HT, dopamine and noradrenaline have all been linked to central fatigue (Taylor et al. 2016). The current study confirms that 5‐HT concentration does play a role in fatigue, as time‐to‐task failure and corresponding EMG and torque measurements during repeated MVCs decreased following ingestion of the SSRI. The mechanisms underlying this decline in motor performance had neural origins, as voluntary activation was reduced and measurements associated with peripheral fatigue were unaffected with paroxetine. Moreover, a substantial reduction in F‐wave persistence and F‐wave area following a 60‐s MVC points to the motoneurones as a site which contributed to added central fatigue with serotonin reuptake inhibition.

Since the 1980s, exercise‐induced changes in monoamines, including 5‐HT, have been linked to central fatigue with the idea that these act at supraspinal sites (Meeusen et al. 2006). By contrast, the paroxetine‐induced depression in motoneurone excitability in the current study aligns with the recent hypothesis that 5‐HT contributes to central fatigue via its action on spinal motoneurones (Cotel et al. 2013; Perrier et al. 2018). This hypothesis, put forward by Perrier and colleagues, has been tested in a series of in vitro experiments. This work demonstrated that sustained descending serotonergic drive increases 5‐HT concentration to the extent that it spills over from synapses on the soma and dendrites of the motoneurones to the axon initial segment, where inhibitory 5‐HT_1A receptors can be activated. Given that 5‐HT_1A receptors are located on the axon initial segment, which is the site for action potential generation, there is a suggestion that 5‐HT_1A receptors are able to act as a ‘gatekeeper’ to regulate neuronal firing (Cotel et al. 2013). Early radio ligand‐binding studies identified that 5‐HT₁ receptors have a higher affinity for 5‐HT compared to 5‐HT₂ receptors (Martin & Sanders‐Bush, 1982). Therefore, activation of 5‐HT_1A receptors results in a higher degree of occupancy for 5‐HT compared to 5‐HT₂ receptors in the soma–dendritic compartment. Collectively, these mechanisms may contribute to the ability of 5‐HT_1A activation to override 5‐HT_2B/C receptor stimulation and cause an overall reduction in motoneurone firing. In the current study, there was some evidence that paroxetine caused enhanced F‐wave depression after a 2‐s MVC, but greater evidence that paroxetine caused enhanced F‐wave depression after a 60‐s sustained MVC. Therefore, motoneurones were more depressed when inhibition of 5‐HT reuptake increased extracellular 5‐HT after a sustained period of strong voluntary drive. We postulate that the voluntary contraction was associated with strong descending serotonergic drive, and led to spill over of 5‐HT onto the axon initial segment to depress motoneurone excitability (as seen for the turtle spinal cord).

Although 5‐HT mechanisms that reduce motoneurone excitability have been identified for the turtle spinal cord, the inhibitory effect that 5‐HT_1A receptors have on motor function has also been confirmed in humans. Ingestion of the 5‐HT1_A agonist, buspirone, induces clear reductions in motoneurone excitability (D'Amico et al. 2017), which limits the capacity to perform prolonged bouts of exercise (Marvin et al. 1997). It should be noted, however, that exogenous activation of the 5‐HT_1A receptor may cause different responses to endogenously released 5‐HT. In particular, an individual's capacity to perform exercise may dictate serotonergic activity in the raphe–spinal pathway and 5‐HT receptor sensitivity on motoneurones. This observation is based on human experiments that use SSRIs to enhance 5‐HT concentration during prolonged submaximal cycling tasks. Of the six studies to investigate how 5‐HT concentration affects exercise performance, all three that reported SSRIs do not influence exercise duration used well‐trained cyclists (Meeusen et al. 2001; Strachan et al. 2004; Roelands et al. 2009), whereas two out of the three studies that found SSRIs decrease exercise performance used recreationally active participants (Wilson & Maughan, 1992; Strüder et al. 1998; Teixeira‐Coelho et al. 2014). Given that we identified SSRI‐mediated changes in central fatigue during sustained isometric contractions, and our participants have no training history involving our elbow flexion task, SSRI responses may be associated with the modality of exercise as well as the training status of the individual.

In summary, our study aligns with the hypothesis that 5‐HT release onto spinal motoneurones not only plays a role in the activation of the muscle, but also plays a role in the genesis of central fatigue. Sustained serotonergic drive that occurs with prolonged maximal contractions may increase 5‐HT concentration in the somato‐dendritic compartment of motoneurones in the spinal cord to the extent that 5‐HT spills over into extracellular compartments. This mechanism enables activation of inhibitory 5‐HT_1A receptors on the axon initial segment, which reduces activity of motoneurones, and hence muscle activity, during voluntary contractions.

Additional information

Competing interests and funding

The authors declare that no competing interests exist for this work, and no funding was received to perform this work.

Author contributions

All authors contributed to the conception and the design of this work, as well as the drafting and final approval of the manuscript. Data collection and analysis was performed by JJK at Griffith University, Gold Coast campus, Australia.

Biography

Justin Kavanagh is a researcher and lecturer at Griffith University in Australia. He leads the Neural Control of Movement laboratory in the Menzies Health Institute Queensland, where his team explores how the central nervous system controls voluntary and involuntary movement. He has particular interests in understanding how medications can be used to study mechanisms of human movement, and how neuromuscular fatigue compromises the ability to regulate muscle activity.