Loss of neuromuscular strength is caused by peripheral fatigue produced by changes at or away from the neuromuscular junction and by central fatigue caused by a loss of voluntary activation (defined as the level of neural drive) to the neuromuscular junction (Gandevia, 2001). Peripheral fatigue can develop due to impairments at the neuromuscular junction, changes in the electrical properties of the muscle, development of metabolites interfering with metabolic processes required to maintain Adenosine Triphosphate levels, or the depletion of glycogen available to the muscles (Bigland-Ritchie et al., 1978; Kirkendall, 1990; Amann, 2011). However, the focus of this review is not on peripheral fatigue and therefore, the discussion on peripheral fatigue has been kept brief. Central fatigue is caused due to factors at the supraspinal sites, which control the descending drive to motoneurons or at the spinal sites which form the pathway for the descending drive to reach the neuromuscular junction (Gandevia, 2001; Taylor et al., 2016). Supraspinal factors include changes in the excitability of the motor cortex, or the strength of cortical connections with motoneurons. Spinal factors include intrinsic motoneuron behavior, recurrent inhibition, presynaptic modulation of α- and γ-motoneurons or other influences to the spinal circuitry.

Central fatigue can be defined as an exercise induced loss of voluntary activation. Voluntary activation is the level of neural drive required to activate the muscle received by the neuromuscular junction from the primary motor cortex. This level of neural drive determines the level of force output that will be generated by the muscle. As evidenced by several studies, voluntary activation during exercise is submaximal even during maximal efforts (Gandevia, 2001). This means that even during maximal effort, the neuromuscular system is unable to generate the maximum evocable force that a muscle group is capable of producing. If continuous or intermittent effort is applied over time, voluntary activation further decreases indicating the progression of central fatigue (Todd et al., 2003; Taylor and Gandevia, 2008). During maximal efforts, voluntary activation declines and becomes suboptimal (Bigland-Ritchie et al., 1978; Gandevia et al., 1996) and this decline can be attributed to an increase in recurrent inhibition of the motoneuron pool, reflex inputs from type III and IV muscle afferents (spinal factors) and suboptimal descending drive (supraspinal factors) from the motor cortex (Taylor and Gandevia, 2008). During submaximal contractions, the neural drive to the muscle first increases with time to compensate for peripheral impairments but intermittent maximal efforts with nerve or brain stimulation reveal that superimposed twitch also increases indicating central fatigue (Søgaard et al., 2006). During submaximal contractions the contribution of central fatigue is higher for lower intensity contractions (Ljubisavljević et al., 1996; Eichelberger and Bilodeau, 2007). Supraspinal fatigue has a larger contribution in central fatigue development during submaximal efforts as compared to maximal efforts (Taylor and Gandevia, 2008). This indicates that during submaximal efforts a larger proportion of task failure is caused by a suboptimal descending drive from the motor regions of the brain. This supraspinal fatigue offers scope for improvement in motor performance as fatigue is not caused due to reaching the limits of the neuromuscular system but by a suboptimal drive to begin with.

The rate of recovery of voluntary activation after central fatigue depends on contraction intensity. While, recovery of voluntary activation from maximal or high intensity submaximal contractions takes 2–3 min, recovery from prolonged, low intensity exercise can take up to 60 min (Taylor and Gandevia, 2008). Excitability of the primary motor cortex also reduces with the onset of fatigue and increases during recovery (Benwell et al., 2006; Teo et al., 2012). The changes in cortical inhibition and activation during and after exercise is indicative of the adaptive neural strategies employed by the central nervous system to maintained the drive required for the contracting muscle.

To better understand how the descending motor drive is regulated, several models of brain regulation of exercise performance have been proposed. Blain et al. (2016) found the central nervous system inhibits intramuscular metabolic perturbation via group III/IV muscle afferents to regulate exercise performance. These results further show that exercise termination due to fatigue is determined by the brain by interpreting the afferent feedback from the muscles. The Central Governor model (CGM) proposes a subconscious “central governor” that regulates the voluntary drive to muscles based on afferent sensory input with the goal of maintaining body homeostasis (Noakes, 2010). According to CGM, physical activity is controlled by a pacing strategy where the “central governor” calculates the intensity of exercise to be maintained and when the exercise would be terminated in order to maintain homeostasis (Gibson and Noakes, 2004). However, according to Marcora (2008), CGM conflicts with conscious effort regulation and does not account for the effect of external motivation on exercise performance. The psychobiological model based on the motivational intensity theory (Wright, 1996) accounts for this conscious control and proposes that task failure occurs when the effort involved in task performance reaches the maximum effort that the participant had calculated to be needed for successful task performance. Although, there is disagreement between the two models on whether exercise regulation is controlled consciously or subconsciously, both models suggest that exercise is regulated by the brain to a calculated safe exertion limit.

As brain regulates effort during exercise, perturbation to certain cortical regions during exercise can also potentially lead to the development of central fatigue before the neuromuscular system reaches capacity. Such perturbations could be concurrent mental fatigue, chronic conditions, or even acute cognitive stress. Studies have shown that high mental workload or concurrent mental fatigue during submaximal isometric contractions are accompanied by a reduction in endurance time and the motor unit firing rate (Mehta and Agnew, 2012; Kowalski and Anita, 2020) and also by a stunted prefrontal cortex (PFC) (Shortz et al., 2015; Mehta, 2016). Studies have shown that changes in PFC activation are associated with changes in motor regions and muscle oxygenation during fatigue development (Rupp and Perrey, 2008; Rupp et al., 2013). However, these studies have not been able to conclusively establish a causal relationship between PFC activation, activation of the motor regions, and neuromuscular fatigue.

TMS is a cortical stimulation technique that uses high intensity transient magnetic field to induce an electrical field in the cortical regions (Carroll et al., 2001). Supramaximal stimulation of the motor cortex or the peripheral motor nerve is used to measure voluntary activation by measuring the “twitch” generated by the muscles while performing a voluntary contraction (Todd et al., 2003). The twitch generated by the supramaximal stimulation of the motor cortex or motor nerve represents the extra force the human was not able to generate voluntarily. Traditionally, voluntary activation is calculated using the following formula:

As discussed earlier, the superimposed twitch can be measured through cortical or nerve stimulation during a voluntary contraction (Gandevia, 2001). In case of nerve stimulation, the resting twitch is measured by the stimulation of muscles at rest, however the same technique cannot be used for cortical stimulation as the excitability of the motor cortical and corticospinal neurons is far less when the muscle is relaxed (Todd et al., 2016). Todd et al. (2003) developed a method of estimation of the resting twitch by calculating the y-intercept of a linear regression of the amplitude of the superimposed twitch against the contraction levels from 50 to 100% of maximum voluntary contractions. Therefore, twitch interpolation technique required twitch to be given at different levels of effort above 50% MVC to have enough data for linear regression.

Twitch interpolation by nerve stimulation has been traditionally used in measuring changes in voluntary activation but is limited in interpretation of causal mechanisms that may be behind these changes (Todd et al., 2016). Measuring voluntary activation via nerve stimulation can tell us whether the voluntary drive to the neuromuscular junction is submaximal (Allen et al., 1995) and this submaximal drive could be due to inability of motoneurons to discharge continuously, or a diminished excitatory drive to the motoneurons. This does not indicate whether the submaximal drive is due to a suboptimal descending drive from the motor cortex or due to inhibitions in the corticospinal pathways. Voluntary activation obtained from the stimulation of the motor cortex via TMS, however, reveals different information than nerve stimulation. If superimposed twitch is generated by motor cortical stimulation, the cortical output was not maximal and not sufficient to drive motoneurons maximally (Taylor et al., 2006). Thus, voluntary activation calculated from cortical stimulation by the TMS represents the level of descending drive from the motor cortex and its loss indicates supraspinal fatigue (Gandevia et al., 1996; Taylor et al., 2000, 2006).

While TMS is better suited to answer questions about mechanisms of fatigue as compared to nerve stimulation, there are some challenges associated with it. While nerve stimulation can be confined to activation of only the agonist muscle group, stimulation via TMS is less precise in stimulating only the target muscles (Taylor et al., 2006). Thus, the choice of muscle groups should be such that the relative stimulation of the agonist muscles is minimized (Todd et al., 2016). For example, studies have shown that this effect is minimal for isometric contraction for the elbow flexors where the EMG recordings of the triceps show a very small motor evoked potential (MEP) as compared to the biceps (Todd et al., 2003, 2004). This is because elbow flexors have stronger excitatory connections and are twice as strong as compared to the antagonists (Taylor et al., 2006). Therefore, the use of TMS for measuring voluntary activation is feasible for only certain muscle groups. Additionally, there are also methodological challenges with measuring voluntary activation using TMS during recovery from a fatiguing exercise (Dekerle et al., 2019). Traditionally, the resting twitch using TMS is estimated by consecutively measuring superimposed twitch at three intensities of contractions above 50% MVC. However, Dekerle et al. (2019) found that resting twitch is not estimated reliably using this method and resting twitch estimation requires at least nine superimposed twitches for reliable measurement of Voluntary activation. However, the accuracy of this “nine-point” method can also be compromised during recovery. Since, recovery of the muscle from fatigue is quick, the twitches obtained at different contraction intensities will be at different points of recovery. Therefore, the interpretations that can be made from TMS for measuring Voluntary activation during recovery are limited.

Surface EMG is often used in studies to monitor neuromuscular fatigue in real time, and analysis of EMG signals gives insight into the physiological mechanisms of fatigue. Studies in the past have shown that surface EMG activity increases linearly with force and that the increase in low frequency signals is characterized by fatigue (Gandevia, 2001). However, surface EMG signals measure compound muscle action potential at the muscle site and do not directly indicate the amount of “neural drive” to the muscles. This is because the muscle action potential (MAP) measured at the site is the result of superimposed MAPs of several muscle fibers and intracellular muscle fiber action potentials. Furthermore, these signals also depend on the conductivity of different muscle fiber types and local blood supply to the site. Therefore, EMG by itself is not a reliable indicator of upstream behavior and is used with nerve or cortical stimulation to measure spinal or supraspinal factors involved in neuromuscular fatigue (Carroll et al., 2001; Hunter et al., 2008; Sundberg et al., 2018). The following features of the EMG response along with cranial or nerve stimulation are used to quantify the mechanisms of fatigue:

While the features above have been extensively studied to understand the mechanisms of fatigue, sex differences in these features have also been studied. Sex differences have been found EMG signals of the lower extremity muscles during fatigue inducing exercise (Cioni et al., 1994; Clark et al., 2005). Pitcher et al. (2003) also found females to have a higher variability in MEP amplitude than males for hand muscles. However, Hunter et al. (2006) did not find sex differences in voluntary activation during exercise. No studies were found to have investigated Sex differences in SICI. Therefore, since there is some evidence of overall sex differences in EMG activity, sex differences for the above EMG features should also be studied.

While the techniques discussed before are efficient in identifying mechanisms of fatigue downstream of the descending voluntary drive from the motor cortex, they are not able to identify mechanisms upstream of this descending drive. Functional brain imaging techniques like electroencephalography (EEG), functional magnetic resonance imaging (fMRI), and functional near-infrared spectroscopy (fNIRS) have made it possible to study brain activity during neuromuscular performance and fatigue manifestation to bridge this gap (Liu et al., 2003; Alexandre et al., 2015; Rhee and Mehta, 2019). Just like EMG provides valuable information about the electrical activity in the muscle while performing a physical task and therefore, indirectly measures the level of neural drive reaching the muscle, brain imaging can capture the activity of the cortical regions involved in sending and regulating the neural drive descending from the primary motor cortex (M1). Brain imaging techniques can monitor cortical activity concurrently with neuromuscular performance which can provide a valuable insight into the central mechanisms of fatigue that TMS cannot provide as it cannot be used concurrently with physical tasks that are at a submaximal level. Of the brain imaging techniques, commonly used in studies on neuromuscular performance, fMRI allows for the greatest spatial resolution and is also capable of measuring cortical regions located deep into the skull (Rao et al., 1993). However, fMRI does not allow for ambulatory examinations or high temporal resolution as the participants need to be supine during fMRI scans (Zhu et al., 2019). While fNIRS is only able to measure hemodynamic activity in cortical regions closest to the brain, it allows for the greatest freedom of movement (Bunce et al., 2006; Zhu et al., 2019). Conversely, while EEG has a better temporal resolution than fNIRS, localization of source of activity is harder for EEG leading to poor spatial resolution (Blinowska and Durka, 2006). However, several source localization algorithms, like LORETA, BPNN, FOCUSS, and more widely used Bayesian techniques, have helped improving the spatial resolution of EEG (Asadzadeh et al., 2020). There are still some methodological challenges with these algorithms with respect to spatial accuracy, and susceptibility to noise. Additionally, EEG is also more susceptible to motion artifacts as compared to fNIRS and requires stronger artifact removal methods (Mehta and Parasuraman, 2013; Zhu et al., 2019). Therefore, taking into account, the advantages of fNIRS over other imaging techniques for functional studies involving movement, further review was conducted on the features that are extracted from hemodynamic activity of the cortical regions (that fNIRS measures). Figure 1C illustrates the method of measuring hemodynamic signal using fNIRS over the brain. The following are the relevant features that can be obtained from fNIRS:

Cognitive stress and high cognitive demands are associated with higher rates of perceived exertion (Marcora et al., 2009), a decrease in force steadiness (Pereira et al., 2015) and enhanced EMG activity in muscles performing the physical task (Van Galen et al., 2002). Even in the absence of physical demand, cognitive stressors contribute to keeping low threshold motor units active which causes lack of muscle rest and can contribute to muscle pain (Lundberg et al., 2002). The effect of stress on force steadiness is dependent on the type of stressor and whether exercise is performed concurrently with stress or after stress. Stressors concurrently applied with exercise (like mental math) reduce force steadiness indicating poorer muscle control (Yoon et al., 2009), whereas social stressors like the Trier-Social Stress test have no impact on force variability (Shortz and Mehta, 2017). To understand the mechanisms behind fatigue under stress, the relative contributions of peripheral and central mechanisms need to be investigated. Increased sympathetic activity caused by stress induces electromyographic activity in low threshold motoneurons (Lundberg et al., 2002) and modulates muscle contractile properties and motor unit discharge rate (Roatta et al., 2008). This mechanism can potentially be responsible for increase in both peripheral or central fatigue.

Another potential cause for exacerbated neuromuscular fatigue under stress can be the effect of stress on brain activity. Cognitive stress impairs the functioning of the PFC (Arnsten, 2009), which is involved in neuromuscular performance as evidenced by several studies (Liu et al., 2003; Muthalib et al., 2013; Rhee and Mehta, 2018). According to Robertson and Marino (2016), the PFC may be involved in regulating and terminating exercise by interpreting and integrating the afferent feedback from the motor nerves and muscles through the anterior cingulate cortex, the orbitofrontal cortex, or the premotor areas. The claim of potential involvement of the PFC in regulating exercise tolerance is also supported by the fact that cognitive and emotional factors like self-motivation impact exercise performance (Barwood et al., 2015). Cognitive declines like mild cognitive impairments are also associated with low dexterity, impaired motor skills and even greater fatigue (Roalf et al., 2018; Suzumura et al., 2018; Kukla et al., 2021). Therefore, it is possible that the increased fatigability under stress is related to the temporary cognitive perturbation and stunted PFC activity caused by stress. However, no studies have investigated whether the stunted PFC activity during acute stress is directly or indirectly responsible for shorter endurance time.

Keller-Ross et al. (2014a) found that during sustained low intensity submaximal contractions, the loss in voluntary activation due to fatigue was the same with and without stress, but the time to task failure was shorter under stress. This indicates that while the relative contributions of central fatigue to neuromuscular performance were similar with and without stressor, central fatigue development was accelerated by the stressor. Furthermore, as voluntary activation was measured by the transcranial magnetic stimulation (TMS) of the motor regions, supraspinal factors were involved in central fatigue development. While the findings allude to supraspinal mechanisms being invoked in accelerated time to fatigue under stress, no studies have investigated if the same holds true for intermittent submaximal contractions that allow the progression of fatigue over a longer period of time. Also, the mechanisms behind the accelerated decline of the descending drive to the motoneurons are not established.

Neuromuscular performance and fatigability are highly variable between the sexes due to physiological differences. Males exhibit higher strength than females while females have exhibited higher endurance for submaximal contractions at levels relative to the strength of various muscle groups (Hicks et al., 2001; Hunter and Enoka, 2001; Hunter, 2014). Hunter and Enoka (2001) however have attributed higher endurance for females to lower absolute strength. Hunter (2014) has attributed sex differences in fatigability to not only muscle specific factors like muscle mass, blood perfusion to muscles, muscle fiber type, and metabolism but also to central mechanisms. These central mechanisms arise due to sex differences in both structural and functional aspects of the brain physiology (Ingalhalikar et al., 2014; Duan et al., 2018; Rhee and Mehta, 2018). Studies have already shown sex differences in neuromuscular fatigue under acute cognitive stress (Yoon et al., 2009) but it is not yet identified if these differences are caused by different mechanisms of fatigue for males and females. Therefore, any investigation on the mechanisms of fatigue under acute cognitive stress also needs to study potential sex differences in these mechanisms.

Conditions like old age and obesity have also shown to exacerbate the negative influence of concurrent stressors on neuromuscular performance where older and obese individuals had even shorter endurance time under acute stress (Mehta, 2015; Shortz and Mehta, 2017). Along with reduction in endurance time, age also exacerbates decrease in force steadiness at higher mental workload and anxiety levels (Vanden Noven et al., 2014). Studies have shown that the impairment of neuromuscular performance depends on the intensity of the stressor applied and that strength is a primary predictor of neuromuscular performance under stress which is characterized by a shorter endurance time (Yoon et al., 2009; Keller-Ross et al., 2014a). This means that a greater reduction in endurance time is observed in weaker individuals under acute stress, which is not observed under low intensity stress and observed to a lesser extent in stronger individuals. Apart from strength, sex, age, and obesity, other individual specific factors like physical activity and other chronic conditions are also known to influence neuromuscular performance (Allman and Rice, 2002; Almeida et al., 2008; Keller-Ross et al., 2014b; Urquhart et al., 2019). Therefore, the interaction of these individual specific factors with stress during neuromuscular performance should also be investigated.