**The Inhibitory Tendon-Evoked Reflex Is Increased in the Torque-Enhanced State Following Active Lengthening Compared to a Purely Isometric Contraction**

#### **Vincenzo S. Contento 1, Brian H. Dalton <sup>2</sup> and Geo**ff**rey A. Power 1,\***


Received: 22 November 2019; Accepted: 18 December 2019; Published: 23 December 2019

**Abstract:** Residual torque enhancement (rTE) is a history-dependent property of muscle, which results in an increase in steady-state isometric torque production following an active lengthening contraction as compared to a purely isometric (ISO) contraction at the same muscle length and level of activation. Once thought to be only an intrinsic property of muscle, recent evidence during voluntary contractions indicates a neuromechanical coupling between motor neuron excitability and the contractile state of the muscle. However, the mechanism by which this occurs has yet to be elucidated. The purpose of this study was to investigate inhibition arising from tendon-mediated feedback (e.g., Golgi tendon organ; GTO) through tendon electrical stimulation (TStim) in the ISO and rTE states during activation-matching and torque-matching tasks. Fourteen male participants (22 ± 2 years) performed 10 activation-matching contractions at 40% of their maximum tibialis anterior electromyography amplitude (5 ISO/5 rTE) and 10 torque-matching contractions at 40% of their maximum dorsiflexion torque (5 ISO/5 rTE). During both tasks, 10 TStim were delivered during the isometric steady state of all contractions, and the resulting tendon-evoked inhibitory reflexes were averaged and analyzed. Reflex amplitude increased by ~23% in the rTE state compared to the ISO state for the activation-matching task, and no differences were detected for the torque-matching task. The current data indicate an important relationship between afferent feedback in the torque-enhanced state and voluntary control of submaximal contractions. The history-dependent properties of muscle is likely to alter motor neuron excitability through modifications in tension- or torque-mediated afferent feedback arising from the tendon.

**Keywords:** electromyography; history dependence of force; residual force enhancement; eccentric; golgi tendon organ; afferent

#### **1. Introduction**

Residual torque enhancement (rTE) is a history-dependent muscle property that results in an increase in steady-state isometric (ISO) torque production following an active lengthening contraction as compared to a purely isometric contraction at the same muscle length and level of activation [1–3]. The presence of rTE has been demonstrated in vitro, from the level of the sarcomere [4] to whole muscle preparations [5], and at the single muscle fiber level [6] to the whole muscle level, via electrical stimulation [7,8] and submaximal and maximal voluntary isometric contractions (MVCs) in humans [2,9,10]. Although numerous reports have focused on the basic mechanical mechanisms of rTE [5,11–14], the neural consequences of rTE have been largely underappreciated. Recent studies

have reported that, during voluntary contractions, rTE may be linked to excitability modifications within the corticospinal pathway [9,15]. For example, rTE led to an increase in corticospinal excitability during plantar flexion MVCs [15] and a decrease in spinal excitability during submaximal steady-state dorsiflexion contractions [15]. However, it is not clear what factors may be modulating these changes in excitability.

Sypkes et al. [15] proposed that reduced spinal excitability during a condition of enhanced muscle force production capacity may be corresponding to greater inhibition of the agonist motor neuron pool arising from tendon-mediated feedback or, more specifically, the Golgi tendon organ (GTO) [15]. The GTOs are located in series with extrafusal muscle fibers at the interface of the musculotendinous junction (MTJ) and act to sense tension produced by its corresponding activated muscle [16–18]. Rising muscle tension increases the firing of the Ib afferent neurons that project from the GTO to a variety of targets within the central nervous system [17,19], including inhibitory interneurons synapsing onto the agonist motor neuron pool [20,21]. An effective mode to assess the efficacy of inhibitory pathways onto the motor neuron pool is using tendon electrical stimulation (TStim) to elicit a short latency reflex [22,23]. Applying TStim during times of increased muscle tension (i.e., rTE state) may elicit increased Ib inhibitory reflex parameters [22,24,25]. Recently, using this technique, we have shown that the tendon-evoked inhibitory reflex is reduced in the shortening-induced residual torque depressed state [23]. The opposite may be true for a condition of enhanced torque production (e.g., rTE). For a review on residual torque depression, please see Chen et al. [26].

Residual torque enhancement can be measured as the increase in torque following active lengthening in the isometric steady-state phase compared to purely ISO contractions during activation-matched tasks (i.e., electromyography), and it has been observed during both submaximal and maximal voluntary activation [27–29]. It is proposed that rTE occurs due to a stiffening and shortening of titin's free spring length in the presence of calcium and cross-bridge cycling [27], effectively increasing the contribution of passive force to total force following active lengthening. Therefore, while matching torque, the increase in passive tension associated with rTE [5,30] requires less muscle activation as compared with a purely ISO contraction [28,29]. This activation reduction is regularly measured through a decrease in electromyography (EMG) and has been observed in both submaximally and maximally activated contractions [3,7,28,31]. These two methods of evaluating rTE—activation matching and torque matching—provide a unique approach to investigate the role of tension-mediated afferents on previously observed reductions in spinal excitability.

The purpose of this study was to determine how the tendon-evoked inhibitory reflex is modified between the rTE and ISO states and to discuss its implications in the reduction of motor neuron pool excitability in the rTE state. During activation-matched ISO and rTE contractions, muscle tension in the isometric steady state is expected to be greater, whereas muscle tension is expected to be equivalent during torque-matched ISO and rTE contractions. If previously observed reductions in agonist motor neuron pool excitability during rTE contractions are in fact modulated from tendon-mediated feedback, we hypothesize that increases in TStim-elucidated reflex parameters (such as amplitude) should be apparent in rTE contractions during isometric steady state compared to purely ISO contractions for the activation-matching trials but not the torque-matching ones.

#### **2. Materials and Methods**

#### *2.1. Participants*

Fourteen healthy male participants with a mean age of 22 ± 2 years, height of 180 ± 6 cm, and mass of 81.8 ± 13.2 kg with no prior history of neuromuscular disease or ankle joint injuries were recruited from the university population. Participants gave written informed consent prior to testing. All procedures were approved by the Human Research Ethics Board of the University of Guelph (REB: 15NV008).

#### *2.2. Experimental Set Up*

A HUMAC NORM dynamometer (CSMi Medical Solutions, Stoughton, MA, USA) was used to record torque, angular velocity, and position. Each participant sat with their right hip and knee angles set at 110◦ and 140◦ (180◦ = full extension), respectively. Joint angles were measured using a goniometer. The right knee was immobilized just proximal to the patella with the dynamometer's leg restraint preventing hip flexion and a cushion positioned beneath the distal hamstrings preventing hip extension, while movement at the torso was restricted with a four-point seatbelt harness. The right foot was fixed to the dorsi/plantar flexor adaptor with one inelastic strap secured over the ankle and another across the mid-distal portion of the metatarsals. The dynamometer's maximum ankle dorsi- and plantar flexion angles were set to 0◦ and 40◦ plantar flexion (PF; 0◦ = neutral), respectively, allowing for 40◦ of ankle excursion.

Locations for the surface EMG electrodes (Ag/AgCl, 1.5 × 1 cm: Kendall, Mansfield, MA, USA) were prepared by shaving and cleaning the skin with alcohol swabs. The active electrode was placed over the tibialis anterior (TA) approximately 7 cm inferior and 2 cm lateral to the tibial tuberosity, and a reference electrode was placed inferiorly, adjacent to the active electrode in line with the muscle fibers. To record antagonist activity, the active electrode was placed on the soleus, along the midline of the leg approximately 2 cm inferior to the border of the heads of the gastrocnemii, and a reference electrode was placed inferiorly, bordering the active electrode. A single ground electrode was positioned over the patella.

Surface EMG, torque, angular velocity, dynamometer position, and stimulus trigger data were digitized using a 12-bit analog-to-digital converter (PowerLab System 16/35, ADInstruments, Bella Vista, Australia) and analyzed with Labchart software (Labchart, Pro Modules 2014, version 8). Torque angular velocity and position as well as EMG data were recorded at a sampling rate of 1000 and 2000 Hz, respectively. The EMG data were bandpass filtered using a digital filter (10–1000 Hz).

#### *2.3. Peripheral Nerve Stimulation*

To test the voluntary activation of the dorsiflexors (see next section) and obtain compound muscle action potentials (M-waves) from the TA and soleus, peripheral nerve stimulation was delivered transcutaneously with a standard clinical bar electrode (Empi, St. Paul, MN, USA) coated in conductive gel. The deep fibular nerve, innervating the dorsiflexor muscles, was located by palpating the head of the fibula and moving posteroinferiorly until the nerve was identified. Stimulation distal to the bifurcation of the common fibular nerve was ensured in order to limit activation of the peroneal muscles. The tibial nerve was stimulated via a bar electrode positioned within the popliteal fossa to maximize the M-wave. All stimulations were delivered as a single square-wave pulse from a constant-current, high-voltage stimulator (model DS7AH, Digitimer, Welwyn Garden City, Hertfordshire, UK). Voltage and pulse width were set to a maximum of 400 V and 200 μs, respectively. The current was increased incrementally until a plateau was reached for the peak-to-peak amplitude of the resting M-wave (Mmax). To ensure consistent activation of all motor neurons throughout the experiment, the current was then increased to a supramaximal level, equivalent to 110% of that required to generate Mmax (range: 20–200 mA and 25–250 mA for deep fibular and tibial nerves, respectively). An overview of all experimental procedures is provided in Figure 1A.

**Figure 1.** Timeline and raw data traces. Schematic timeline of experimental procedures (**A**). To establish maximum compound muscle action potentials (Mmax) for the tibialis anterior (TA) and soleus, peripheral nerve stimulation was performed on the deep branch of the common fibular and tibial nerves, respectively. A 10 s maximum voluntary contraction (MVC) was followed by an initial MVC performed with the interpolated twitch technique (ITT) to assess maximum torque and voluntary activation. Tendon electrical stimulation (TStim) was applied at increasing currents until perceptual threshold (PT) was found, which was defined as the minimum current intensity that induced a tingling or tapping sensation that was detectable to the participant. Stimulation intensity was then increased to 6 × PT, a current at which participants reported a muscular sensation, including a tugging or pulling at the muscular insertion, a deep tingling near the cathode, or a muscle twitch. If a visible muscle twitch was induced, stimulation intensity was reduced to the maximum current that failed to produce a visible muscle twitch. This stimulation intensity was used for all subsequent trials. Participants then performed 10 pairs of activation-matched dorsiflexion trials (**B**) consisting of an isometric (ISO) trial (black trace) followed by a residual torque-enhanced (rTE) trial (gray trace). These contractions were performed at 40 ± 5% of the participant's maximum TA activation, and 10 TStims were applied during the isometric steady state of each contraction. Participants then performed 10 pairs of torque-matched dorsiflexion trials (**C**) consisting of an ISO trial (black trace) followed by an rTE trial (gray trace). These contractions were performed at 40% of the participants maximum dorsiflexion torque, and 10 TStims were applied during the isometric steady state of each contraction to evoke an inhibitory reflex (activation-matched trial (**D**), torque-matched trial (**E**)). Lastly, participants were instructed to perform a final MVC to assess for fatigue. A resting period of 5 min was given to participants after all contractions, and each contraction was performed with visual feedback as well as verbal encouragement.

#### *2.4. Maximum Voluntary Contraction and Voluntary Activation*

Voluntary activation of the dorsiflexors was assessed during brief MVCs (~5 s) performed twice, separated by 4 min of rest both prior to and following the experimental trials. The interpolated twitch technique was used to evaluate voluntary activation [32]. The torque resulting from peripheral nerve stimulation delivered during the plateau phase of the MVC was compared to a resting twitch evoked 1–2 s after relaxation. The level of voluntary activation was calculated as follows: voluntary activation (%) = [1 −(interpolated twitch torque/resting twitch torque)] × 100%. Participants were encouraged verbally and provided visual feedback of torque output during all MVC attempts [33]. All participants were required to reach a minimum of 95% voluntary activation in order to be included in the study and were given five minutes of rest following the qualifying MVCs before continuing with the experiment.

#### *2.5. Determining Submaximal Muscle Activation*

To determine the submaximal integrated EMG (iEMG) and torque targets, participants were instructed to perform a 10 s dorsiflexion MVC at an ankle angle of 40◦ PF. The average iEMG and torque collected between 6 and 8 s was then used to determine the 40% submaximal iEMG and torque targets [23] (Figure 1B). For activation-matched contractions, a ± 5% window was calculated about the 40% iEMG target, and participants were instructed to maintain their iEMG amplitude within set guidelines marking this target window [23].

#### *2.6. Tendon Electrical Stimulation*

Percutaneous TStim was used to induce tendon-evoked inhibitory agonist reflexes. This technique involved percutaneous square-wave electrical stimulation of the tendon near the MTJ and evoked a short-latency (<50 ms) reflexive inhibition in the agonist muscle. This reflex has been demonstrated for several upper and lower limb muscles [22–25] and is thought to be mediated via Ib spinal pathways owing to the short latency and the polarity of the response, which is consistent with Ib autogenic inhibition [22,23]. However, contributions from other sources, such as muscle or tendon type III afferents, cannot be ruled out completely [25]. Still, non-GTO origins of TStim, specifically cutaneous receptors overlying the tendon or muscle stimulation via current spread, were excluded in prior studies [24]. Previous reports [22,34] have demonstrated that indwelling electrical stimulation of the tendon evokes an inhibitory response with similar characteristics as percutaneous stimulation, indicating that the percutaneous tendon-evoked reflex technique used here is most likely tension-mediated via GTO afferents. Ag/AgCl electrodes (1.5 × 1 cm: Kendall, Mansfield, MA, USA) were used for TStim in order to generate tendon-evoked inhibitory reflexes. The cathode was placed near the MTJ of the TA, and the anode was placed over the distal tendon at the level of the malleoli [23]. Single stimuli were presented with a constant-current, high-voltage stimulator (DS7AH). Voltage was set to a maximum of 400 V and pulse width to 200 μs. The stimulation protocol was initiated with the detection of perceptual threshold (PT), which was defined as the minimum current intensity that induced a tingling or tapping sensation that was detectable to the participant (6.63 ± 2.13 mA) [23]. Stimulation intensity was then increased to 6 × PT, a current at which participants have reported a muscular sensation, including a tugging or pulling at the muscular insertion, a deep tingling near the cathode, or a muscle twitch [9,23]. If a visible muscle twitch was induced, stimulation intensity was reduced to the maximum current that failed to produce a visible muscle twitch [23]. This stimulation intensity was used for all subsequent trials (5.2 ± 0.90 × PT).

#### *2.7. Experimental Procedures*

Each rTE trial was preceded by an ISO trial, and protocol A was followed by protocol B. Five rTE trials and five ISO trials were performed for each of the two protocols for a total of 20 contractions. Participants were provided visual feedback of the iEMG and torque amplitudes on a computer monitor

and were verbally encouraged to match the target as closely as possible during all submaximal contractions. Four minutes of rest separated all submaximal contractions.

#### *2.8. Protocol "A": Activation-Matching Condition*

For each rTE trial, the protocol consisted of a 40% iEMG contraction involving a 2 s isometric phase at an ankle angle of 90◦, a 1 s isokinetic lengthening phase (angular velocity: 40◦/s) and ~20 s isometric phase at 40◦ PF (Figure 1B). A series of 10 TStim pulses were delivered at random 1–4 s intervals [9,23] during the isometric phase at 40◦ PF. During the ISO trials, an isometric dorsiflexion contraction corresponding to 40% iEMG was performed for ~23 s at an ankle angle of 40◦ PF, with a similar pattern of stimuli delivered as described for the rTE trials.

#### *2.9. Protocol "B": Torque-Matching Condition*

For each rTE and ISO trial, participants were instructed to maintain 40% MVC torque. The movement and stimulation protocols were identical to those of protocol A (Figure 1C).

#### *2.10. Data Analysis and Statistics*

Mean torque from each protocol contraction was calculated from 500 ms prior to the first stimulation to the end of each contraction [23]. Root mean squared EMG (EMGRMS) amplitude was calculated in a 500 ms window that occurred between 6 and 8 s after contraction initiation following the achievement of an isometric steady state. It was ensured that the window selected was matched between corresponding rTE and ISO trials [23]. The EMGRMS of the resting Mmax recorded at the TA and soleus was used to normalize the voluntary TA and soleus EMG, respectively. A paired *t*-test was performed to compare the torque and EMG data between rTE and ISO trials to validate the presence of rTE and activation reduction.

To obtain the reflex parameters, a stimulus-triggered average of the raw TA EMGRMS was generated using Labchart software (Labchart, Pro Modules 2014, version 8). For each protocol, separate rTE and ISO averages were constructed from all stimuli delivered in each condition; therefore, each average was composed of 50 stimuli delivered over 5 contractions (Figure 1D,E). A paired *t*-test was performed to compare reflex characteristics between the rTE and ISO states, including latency, duration, amplitude, and change in average EMGRMS from baseline, in order to characterize changes in the tendon-evoked inhibitory reflex in the rTE and AR state [23]. Baseline was measured as the average EMGRMS in a 300–500 ms window occurring before TStim, and the onset of the stimulation artifact was defined as 0 s. Latency was measured as the time from the initiation of the stimulus artifact to the sharp decrease in baseline EMGRMS occurring at the start of the reflex. Duration was measured as the time occurring from the initial decrease in EMGRMS from baseline at the start of the reflex to when baseline EMGRMS was once again reached. Amplitude was measured as the magnitude of EMGRMS from baseline to the lowest trough in the reflex [23].

Change in average EMGRMS from baseline was measured as the difference between baseline EMGRMS and the average EMGRMS from the duration of the reflex. Paired *t*-tests were also used to detect any differences in the torque produced during MVCs performed before and after the experiment in order to assess any effects of fatigue during the experimental protocol. Descriptive data found in text are reported as means ± standard deviation, while data presented in figures are reported as means ± standard error of the mean. Significance was determined based on α < 0.05.

#### **3. Results**

#### *3.1. Maximum Voluntary Contraction and Voluntary Activation*

The mean pretrial MVC torque was 27.6 ± 6.4 Nm, and all participants were capable of achieving near-maximal values for voluntary activation (98.9 ± 1.5%). Following the 20 submaximal contractions, MVC torque was not different from the pretrial values (27.1 ± 6.6 Nm).

#### *3.2. Dorsiflexion Torque and Muscle Activity*

#### 3.2.1. Activation Matching

Normalized EMGRMS of both TA (*p* = 0.8) and soleus (*p* = 0.6) were not different between rTE and ISO contractions (Figure 2B,C). Following active lengthening, steady-state isometric torque was 15.0 ± 9.7% (*p* < 0.0001) greater than that produced during the purely isometric contractions at the corresponding muscle length and level of activation (Figure 2A). Participants successfully maintained the EMG target level such that iEMG of the TA did not differ in the rTE and ISO contractions (*p* = 0.1; Figure 1B). This indicates indirectly that motor neuron output was similar in both the rTE and ISO states. For the tendon-evoked inhibitory reflex, reflex latency was not significantly different between the rTE and ISO states, with the onset of inhibition occurring at 47.8 ± 7.8 ms following TStim in rTE trials and 48.6 ± 4.8 ms following TStim in ISO trials (*p* = 0.6; Figure 2E). Further, reflex duration (*p* = 0.5; Figure 2F) and reduction in average reflex EMGRMS from baseline were not different when rTE and ISO contractions were compared (*p* = 0.2). However, inhibitory reflex amplitude differed by 22.6 ± 41.8% (*p* < 0.05; Figure 2D) in the rTE state compared to the ISO state.

**Figure 2.** Activation-matching trial. Mean values for each participant across measures (colored lines) and the group mean (black line; error bars indicate standard error of the mean) in the rTE and ISO states. For the activation-matching trial, in the rTE state as compared to the ISO state, there was a 15.0% increase in torque (**A**) and a 22.6% increase in reflex amplitude (**D**) (\* *p* < 0.05). There was no significant difference in EMGRMS collected from the tibialis anterior (**B**) or soleus (**C**) reflex duration (**F**) or reflex latency (**E**) between the two states.

#### 3.2.2. Torque Matching

Steady-state isometric torque was not different following active lengthening when the rTE and ISO states were compared (*p* = 0.2 Figure 3A). In the rTE state, however, there was a significant 14.4 ± 13.3% decrease in normalized EMGRMS (*p* < 0.01; Figure 3B) and a 26.9 ± 23.7% decrease in iEMG (*p* < 0.01) for the TA compared to ISO contractions, with no change in antagonist coactivation (Figure 3C). For the tendon-evoked inhibitory reflex, reflex latency was not significantly different between the rTE and ISO states, with the onset of inhibition occurring at 48.0 ± 6.8 ms following TStim in rTE trials and 50.6 ± 5.6 ms following TStim in ISO trials (*p* = 0.1; Figure 3E). Further, reflex duration (*p* = 0.2; Figure 3F), reflex amplitude (*p* = 0.2; Figure 3D), and reduction in average reflex EMGRMS from baseline (*p* = 0.2) were not different when rTE and ISO contractions were compared.

**Figure 3.** Torque-matching trial. Mean values for each participant across measures (colored lines) and the group mean (black line; error bars indicate standard error of the mean) in rTE and ISO states. For the torque-matching trial, in the rTE state as compared to the ISO state, as expected, there was a 14.4% decrease in TA activation (**B**) (\**p* < 0.05). There were no differences in torque (**A**), antagonist coactivation (**C**), or any reflex parameter (**D**–**F**) between the two states.

#### **4. Discussion**

The purpose of this study was to determine how the tendon-mediated inhibitory reflex is modified between the rTE and ISO states through TStim of the tibialis anterior during submaximal activation-matched and torque-matched dorsiflexions. The activation-matched task successfully elicited an ~15% increase in torque in the rTE state compared to ISO, while the torque-matching task resulted in an ~14% decrease in EMG. Our hypothesis of an increased tendon-mediated inhibitory reflex in the rTE state was supported by a 23% increase in inhibitory reflex magnitude in the rTE state compared to the ISO state during the activation-matching task but not the torque-matching task. Therefore, these results support an underlying tension-mediated factor as a plausible explanation for the previously reported decrease in agonist motor neuron pool excitability in the rTE state [15].

In the force-enhanced isomeric steady state, there is greater relative contribution of passive force to total force production, possibly owing to stiffening of the giant molecular spring titin [27]. This increase in passive tension can manifest in rTE contractions in two ways when compared to purely ISO contractions. The first is an activation reduction in torque-matching tasks in which no change in muscle tension occurs [3,29]. The second is an increase in torque, and consequently muscle tension, during activation matching [29,35]. Critical to the present study was the use of both paradigms: activation matching and torque matching. Previous reports of decreases in agonist motor neuron pool [15] and increases in corticospinal [9] excitability have been observed for submaximal and maximal rTE contractions compared to purely ISO contractions. Given the aforementioned investigations [9,15] had similar motor neuron outputs for both ISO and rTE contractions, the alteration in spinal excitability was most likely owing to peripheral sensory inputs. One source of peripheral input upon the motor neuron pool are muscle spindles. While activating their respective Ia afferents via Achilles tendon vibration, no modulatory effect on rTE was observed [35]. Therefore, a tension-mediated factor was speculated as the most likely factor driving the previously reported results [15]. Golgi tendon organs increase Ib afferent firing during periods of increased muscle tension [16] and excite inhibitory interneurons within the spinal cord [17,19]. Excitation of these inhibitory interneurons results in decreased neural output of the agonist motor neuron pool, which can be observed within the surface

EMG signal [24]. Through single pulse electrical stimulation of the TA tendon, an inhibitory reflex can be observed in the EMGRMS ~50 ms following stimulation [22,24,25]. Following active lengthening, during an activation-matching task, rTE torque increased by ~15% compared to ISO, which resulted in an ~23% increase in tendon-evoked inhibitory reflex magnitude. The torque-matching task served as a control, i.e., with no changes in torque or muscle tension between rTE and ISO states, no changes in tendon-evoked reflexes should occur. This is particularly important because it ensures any changes observed in the activation-matching trials are indeed due to GTO-mediated reflexes. While there was an ~14% reduction in TA activation (normalized EMGRMS), indicating the presence of rTE, negligible differences in muscle tension were detected, and as expected, there was no significant differences in the tendon-evoked inhibitory reflex across the rTE and ISO tasks. The lack of changes in reflex characteristics during the torque-matching task indicates our findings were due to a tension-mediated factor. Therefore, the tension-dependent GTO and the Ib afferent is most likely a key contributor in modulating agonist motor neuron excitability during voluntary control of submaximal contractions in the rTE state.

The results of our investigation are similar to previous reports on rTE and activation reduction. While we found ~15% increase in torque across all 40% activation-matched rTE trials in comparison to ISO, Pinniger and Cresswell [36] found a 12% increase in torque at 25% of maximally activated TA dorsiflexion. The 40% MVC torque-matching task induced an ~14% reduction in activation of the tibialis anterior EMGRMS. Other studies reported a similar activation reduction ranging from 5% to 20% [28,29,31]. Additionally, the aim of this study was to elucidate a measurable change in tendon-evoked reflex parameters between ISO and rTE states, where we hypothesized an increase in inhibitory reflex parameters in the rTE state compared to the ISO during the activation-matching protocol but not the torque-matching one. In support of this hypothesis, we found an ~23% increase in inhibitory reflex magnitude in the rTE state compared to the ISO state during the activation-matching task but not the torque-matching one. Under similar premises—decreased torque production in the torque-depressed state eliciting a decrease in tendon-evoked reflex parameters—Sypkes et al. [23] found a 16% reduction in reflex magnitude. The symmetry in the findings of Sypkes et al. [23] and the present study demonstrate a clear relationship between these history-dependent properties of force and the GTO mechanoreceptor. The resulting neural reflex pathway consisting of Ib afferent neurons and spinal inhibitory interneurons provide a plausible explanation to decreased motor neuron pool excitability reported previously [15].

A limitation to the present study was the lack of randomization for the order of protocols A and B. This had the potential to change perceptual threshold and reduce stimulus efficacy throughout the testing session. However, when protocol A ISO was compared to protocol B ISO, there were no differences in reflex parameters (Reflex Latency, Reflex Duration, Reflex Magnitude; Change in TA EMGRMS from Baseline). Thus, the lack of randomization does not appear to influence the results.

#### **5. Conclusions**

Residual torque enhancement, a history-dependent property of muscle, was present during submaximal activation- and torque-matching tasks. In the isometric steady state following an active lengthening contraction, there was an increase in tendon-evoked inhibitory reflex magnitude compared to purely isometric contractions for the activation-matching task but not the torque-matching task. This observation likely characterizes a tension-dependent increased contribution of tendon-mediated inhibitory feedback on the agonist motor neuron pool and as such may explain—at least partially—the documented decrease in agonist motor neuron excitability during an rTE state [15]. This study provides novel insight into the peripheral contributions of the history dependence of force and bridges the gap between the central nervous system and a property that was once thought to be purely intrinsic to muscle.

**Author Contributions:** Conceptualization, V.S.C., B.H.D., and G.A.P.; data curation, V.S.C.; formal analysis, V.S.C. and G.A.P.; funding acquisition, G.A.P.; investigation, V.S.C. and G.A.P.; methodology, V.S.C., B.H.D., and G.A.P.; supervision, B.H.D. and G.A.P.; writing—original draft, V.S.C., B.H.D., and G.A.P.; writing—review & editing, V.S.C., B.H.D., and G.A.P. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Natural Sciences and Engineering Research Council of Canada (NSERC).

**Acknowledgments:** We would like to thank all of the participants in this study.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **The Task at Hand: Fatigue-Associated Changes in Cortical Excitability During Writing**

#### **Kezia T. M. Cinelli, Lara A. Green and Jayne M. Kalmar \***

Department of Kinesiology, Wilfrid Laurier University, Waterloo, ON N2L 3C5, Canada; cine2150@mylaurier.ca (K.T.M.C.); lgreen@wlu.ca (L.A.G.)

**\*** Correspondence: jkalmar@wlu.ca

Received: 11 October 2019; Accepted: 29 November 2019; Published: 2 December 2019

**Abstract:** Measures of corticospinal excitability (CSE) made via transcranial magnetic stimulation (TMS) depend on the task performed during stimulation. Our purpose was to determine whether fatigue-induced changes in CSE made during a conventional laboratory task (isometric finger abduction) reflect the changes measured during a natural motor task (writing). We assessed single-and paired-pulse motor evoked potentials (MEPs) recorded from the first dorsal interosseous (FDI) of 19 participants before and after a fatigue protocol (submaximal isometric contractions) on two randomized days. The fatigue protocol was identical on the two days, but the tasks used to assess CSE before and after fatigue differed. Specifically, MEPs were evoked during a writing task on one day and during isometric finger abduction to a low-level target that matched muscle activation during writing on the other day. There was greater variability in MEP amplitude (F (1,18) = 13.55, *p* < 0.01) during writing compared to abduction. When participants were divided into groups according to writing style (printers, *n* = 8; cursive writers, *n* = 8), a task x fatigue x style interaction was revealed for intracortical facilitation (F (1,14) = 9.90, *p* < 0.01), which increased by 28% after fatigue in printers but did not change in cursive writers nor during the abduction task. This study is the first to assess CSE during hand-writing. Our finding that fatigue-induced changes in intracortical facilitation depend on the motor task used during TMS, highlights the need to consider the task-dependent nature of CSE when applying results to movement outside of the laboratory.

**Keywords:** corticospinal excitability; task-dependent; transcranial magnetic stimulation; fatigue; writing

#### **1. Introduction**

Single-pulse transcranial magnetic stimulation (TMS) elicits motor evoked potentials (MEPs) that are recorded from the muscle of interest using surface electromyography (EMG). MEPs are used as a measure of corticospinal excitability (CSE) that reflect the excitability of the pathway from the site of stimulation to the site of recording, such that both cortical and spinal mechanisms contribute to changes in the MEP evoked using single-pulse TMS. Intracortical mechanisms that may contribute to CSE are assessed using conditioned MEPs elicited via paired-pulse TMS. For example, paired-pulses with a brief interpulse interval (1–5 msec) provide a measure of short interval intracortical inhibition (SICI), whereas longer interpulse intervals (10–15 msec) provide a measure of intracortical facilitation (ICF) [1]. One challenge associated with using single- and paired-pulse TMS to assess CSE is the variability in MEP amplitude, both within and between participants. To minimize variability, most TMS protocols are conducted with the muscle at rest or during submaximal isometric muscle contractions to standardize levels of muscle activation and control for joint position, muscle length, movement, and other factors. Although these laboratory tasks are intended to minimize variability, they do not reflect activities of daily living to which the results may be extrapolated. This is problematic because

CSE depends on the net excitability of the targeted brain region and net excitatory and inhibitory input to spinal motor neurons at the time of stimulation, all of which vary with different motor tasks and states [2–4]. Thus, the conclusions drawn from TMS studies depend on the motor task employed during stimulation and therefore may not translate to movement outside of laboratory settings.

Muscles of the hand (e.g., first dorsal interosseous (FDI) and adductor pollicis (AP)) are frequently used in studies of CSE because of the ease with which MEPs are elicited in distal muscles of the upper extremity. In such studies, MEPs are typically evoked with the hand at rest or during submaximal isometric contractions. For example, FDI MEPs are often assessed with the hand pronated with the wrist and hand secured to allow movement only at the metacarpal phalangeal joint of the index finger. Outside the lab, however, we use intrinsic muscles of the hands in a variety of positions, including power and precision grips. Measures of CSE differ between these grips. For example, FDI MEP amplitude is greater during conventional abduction tasks compared to power, pincer, or grasping tasks [5–7], whereas APB MEPs are greater during pincer tasks compared to during power tasks or at rest [8]. More direct measures made in the monkey reveal that corticospinal neurons are more active during a precision grip compared to a power grip, despite increased EMG activity in the latter [9]. This suggests that CSE is not simply related to the force of contraction, but also to the specific task. Pearce and Kidgell (2010) demonstrate that CSE is modulated when the precision required for a given task is increased [10] Furthermore, more complex tasks will require the involvement of proximal muscles for stability during finger movements. Flament and colleagues [11] compared index finger abduction to simple manual tasks including precision grip, power grip, grasping of a petri dish, and rotation of a bottle cap. Compared to isolated finger abduction, which was restricted to FDI use, all other tasks required activation of at least one additional muscle, which was speculated to contribute to the increased CSE found during the complex tasks [11].

Conventional laboratory tasks, such as matching force output to a static target displayed on a computer monitor, require very different cognitive demands compared to more natural motor tasks such as hand writing, use of a keyboard, or object manipulation. Nonetheless, there is a tendency to attribute the TMS results solely to the changes within or downstream to the primary motor cortex, and to overlook the influence of upstream cognitive processes transmitted to the primary motor cortex [12]. For example, internally guided movements, such as writing or drawing, have been shown to generate greater activation in the pre-supplementary motor area and dorsal premotor cortex as compared to externally guided movements, such as tracing [13]. Using the hand to convey language introduces additional cognitive influences on the motor system. In fact, mere observation of letters and words can alter CSE measured in the FDI when at rest [14,15]. Interestingly, this effect is specific to handwritten text, including handwritten "non-words", and is not observed with typed text [14,15], indicating that the recognition of hand-writing represents a unique cognitive demand that is conveyed to the motor system, likely via the mirror neuron system [14]. Accordingly, it is unlikely that estimates of CSE obtained during simple, isometric laboratory tasks would translate directly to CSE during more complex tasks outside the lab. In this study, we compared a hand-writing task, which represents an internally generated and complex task that is familiar and relevant outside the laboratory, to a conventional externally guided, isometric force-matching task often used in TMS research.

We wanted to determine whether changes in CSE in response to a well-studied intervention (neuromuscular fatigue) would depend on the motor task that is used to produce background muscle activity during the delivery of TMS. TMS has been used for many years to understand the role of the central nervous system in neuromuscular fatigue (for review, see [16,17]). Such studies have revealed that CSE is briefly facilitated and then undergoes a more prolonged period of depression following fatigue when MEPs are evoked in resting muscles [18–21]. Post-exercise depression is also observed after a fatigue protocol when MEPs are evoked at rest just prior to a voluntary contraction [22]. On the other hand, MEPs recover much more quickly when assessed during a strong muscle contraction after fatigue [23,24], suggesting that reductions in CSE after fatigue are overcome with sufficient voluntary drive. Given that measures of CSE depend on limb position, muscle activation level, and other aspects

of the motor task, such as different cognitive demands, we speculate that fatigue-induced changes in CSE made during a conventional laboratory task may not represent fatigue-induced changes in CSE made during the more relevant and complex task of writing. Therefore, the purpose of this study is to elicit neuromuscular fatigue and then compare fatigue-associated changes in CSE during writing to the same measures made during a conventional isometric finger abduction task. We hypothesized that CSE would be greater during the writing task compared to the simple isometric abduction task. Furthermore, we hypothesized that the effect of fatigue on CSE would depend on the motor task employed during stimulation, even when the fatigue task itself was the same.

#### **2. Materials and Methods**

#### *2.1. Participants*

Twenty participants with a mean age of 22.6 ± 1.1 were recruited from the Wilfrid Laurier University student population. This sample size was based on a power calculation made using data from an earlier study of intracortical mechanisms of fatigue from our lab [22]. All participants were right-handed as determined by the Annett Handedness Questionnaire [25]. One participant was excluded from analysis due to an inability to complete the full testing protocol, leaving 19 participants (six male). Exclusion criteria included any neurological conditions, and orthopaedic conditions or pain of the hand, wrist, or arm. The study was approved by the Wilfrid Laurier University Research Ethics Board (REB #5381), and all participants provided written informed consent prior to participating.

#### *2.2. Experimental Design*

In this repeated-measures study design, each participant attended two testing sessions occurring no less than 48 h apart and was tested at the same time on both days to decrease between session variability [26]. On one day, participants completed a writing task and on the other day, participants completed an abduction task. The order of the two days was randomized and counterbalanced between participants. On each day participants began with a short familiarization to each task. Following familiarization, participants completed 90 pre-fatigue trials of the selected task (writing or abduction), a fatigue protocol of intermittent isometric abduction contractions, and finally 90 post-fatigue trials of the same task (writing or abduction) (Figure 1).

On each testing day participants completed a familiarization protocol including the writing task, the abduction task, and maximal voluntary contractions (MVCs). The writing task included 20 trials where the participant repetitively wrote the word "name" on a iPad tablet with a Adonit Pro3 precision disc stylus pen. The word "name" was specifically selected for the writing task because it is familiar, it is a short enough word to be written within the 5-s recording frame, and because the letters n, a, m, and e do not contain ascenders or descenders that would require greater finger movement. The tablet screen presented a blank 7 × 2 cm rectangle on a series of Powerpoint slides that were refreshed every 5 s. The rectangle served as the boundaries within which the participant was asked to write the word. These writing boundaries and the refreshing of the writing "page" on the tablet before each trial, allowed the participants to maintain a constant hand position and to minimize wrist deviations between trials. The participants were allowed to place the tablet in any position and orientation on the table in front of them. This position was marked and maintained between trials, blocks, and testing days. Participants were instructed to maintain the same self-selected grip of the pen (e.g., dynamic tripod, lateral tripod, quadrupod, etc.) for the duration of the protocol. In the abduction task, TMS was applied during isometric abduction of the right index finger against a force transducer for 3 s with a 2-s rest. The target level of contraction during the abduction task was set to the average RMS amplitude of FDI EMG activity over the 20 familiarization writing trials so that the contraction intensity was matched to activity required during writing for each participant (Figure 2). This level of activity was marked with horizontal cursors (±2.5%) to allow the participant to match it using a real-time smoothed and rectified EMG biofeedback channel. The location of the force transducer was adjusted

to be in line with the self-selected graphic tablet location, such that shoulder and elbow angle were maintained between the writing and abduction tasks (Figure 3). The hand was in a pronated position for abduction with the third finger secured by a strap and wooden dowels placed on either side of the wrist. Participants then performed three isometric MVCs from which the highest was taken as maximal finger abduction force.

**Figure 1.** Each day began with a familiarization period that included 20 repetitions of each task to record muscle activation during writing, familiarizing participants with the force-tracing task, and an assessment of baseline maximal voluntary contraction (MVC). Participants then completed the Pre-Fatigue Test, which included 90 trials (abduction or writing task), followed by a Fatigue Protocol that was comprised of submaximal (60% MVC target) abduction contractions until failure (force fell below 58% MVC force for >3 s during attempts to hold the 60% target), followed by a post-fatigue MVC. The protocol ended with the Post-Fatigue Test that was identical to the Pre-Fatigue Test followed by a recovery MVC. The 90 pre-fatigue and post-fatigue trials were separated into 3 blocks of 10 sets each. Each set contained 3 randomized trials: (1) test MEP (TS), (2) short-interval intracortical inhibition (SICI) paired-pulse, and (3) intracortical facilitation (ICF) paired-pulse stimulation, resulting in pseudo-randomization of TS, SICI, and ICF across each block.

**Figure 2.** First dorsal interosseous (FDI) electromyographic activity in two representative participants: one who printed (**A**,**B**) and one who wrote in cursive (**C**,**D**). FDI EMG activity is shown during writing (**A,C**) and abduction (**B,D**). The vertical lines denote the 500-msec window prior to TMS stimulation from which the root-mean-square activity was calculated.

**Figure 3.** Experimental set-up of participant positioning during the writing (**A**) and abduction (**B**) tasks. The table design allowed the position of the force transducer to match the self-selected writing position of the graphic tablet.

Following familiarization, the participant was set up for TMS. The pre-fatigue and post-fatigue tests included a total of 90 trials, which consisted of 30 pulses each of TMS stimulation type, including single-pulse test MEP (TS), and paired-pulse stimuli to assess short-interval intracortical inhibition (SICI), and intracortical facilitation (ICF). The 90 trials were performed in 3 blocks, with 10 sets per block and 3 trials per set to pseudo-randomize the TMS stimulation such that each set would include a single-pulse TS, SICI, and ICF evoked in random order (Figure 1). Each trial was 5 seconds in duration. On the writing day, the timing of the screen refresh was set to ensure that the stimulus was delivered as the participant wrote the letter "a" or "m". Timing between pulses (5 s) was never changed. On the abduction day, the TMS stimulation occurred 2 seconds into the 5-second trial to ensure the participant was in the plateau portion of the submaximal, isometric, finger abduction contraction.

The fatigue protocol consisted of repeated 4-second isometric finger abduction contractions at 60% of the participant's MVC, with two seconds between each contraction. Task failure was defined as the point at which force fell below 58% MVC for more than three seconds despite encouragement to meet and hold the 60% target. Immediately following task failure, participants completed a post-fatigue MVC before beginning the post-fatigue test (Figure 1). It is important to note that on both days, regardless of the condition (abduction or writing), participants performed the same isometric finger abduction fatigue protocol. In this way, we examined the task-dependent nature of TMS measures of CSE rather than the task-dependent nature of neuromuscular fatigue.

#### *2.3. Experimental Set-Up and Recordings*

#### 2.3.1. Electromyography and Force Recordings

The skin over the right FDI muscle was cleaned with isopropyl alcohol and two Ag/AgCl EMG electrodes were affixed in a bipolar configuration (0.5 cm recording surface, 1 cm interelectrode distance) over the muscle belly. A ground electrode was placed on the dorsal aspect of the right hand after being shaven and cleaned with alcohol. To allow the hand to rest comfortably on the graphic tablet, a glove covering the hand and the fourth and fifth digits was worn by participants. The skin over the extensor carpi radialis (ECR) and the flexor carpi radialis (FCR) was prepared for EMG by shaving, abrading, and cleansing with alcohol. Parallel bar surface EMG electrodes (10 × 1 mm Ag contacts, 1 cm interelectrode distance, DE-2.1 DELSYS Inc., Natick, MA, USA) were affixed over the muscle belly of the ECR and FCR, and a ground electrode was placed on the elbow. EMG and force signals were digitized at 2000 Hz using the Micro 1401-3 data acquisition unit and Signal 6.0 waveform acquisition software (Cambridge Electronics Design, Cambridge, UK). The FDI EMG signal was pre-amplified 300× and band-pass filtered from 15 to 450 Hz (Motion Lab Systems, Inc. Los Angeles, CA, USA). The

ECR and FCR EMG data was amplified 1000× and band-pass filtered from 20 Hz to 450 Hz (Bagnoli-16, DELSYS Inc., Natick, MA, USA).

#### 2.3.2. Transcranial Magnetic Stimulation

Participants were seated at a table with an adjustable headrest situated in front of the forehead to allow the body and head to be supported in a comfortable writing position (Figure 3). A figure-eight magnetic stimulating coil (D70 coil, Magstim Company Ltd., Whitland, UK) was positioned over the primary motor cortex with the handle positioned posteriorly 45◦ to the midsagittal line and the induced current in a posterior to anterior direction and held in place using a lighting support arm and clamp (Manfrotto Supports, Cassola, Italy) with additional support and position maintenance by the investigator. The TMS coil was moved in small increments in order to determine the optimal site for generating a motor evoked potential (MEP) in the FDI via suprathreshold stimulations from the BiStim2 system (Magstim Company Ltd., Whitland, UK). Once located, this spot was marked on the cap worn by participants. Stimulator output was then adjusted to find the active motor threshold (AMT), determined as the minimum intensity that elicited a 200 μV MEP in 5 out of 10 of trials while the participant maintained a target matched to the average EMG over 20 writing trials recorded during the familiarization period at the beginning of each experimental day. Pulses were delivered at intervals of at least 5 s during threshold hunting. Paired-pulse TMS was used to assess SICI and ICF with a conditioning stimulus of 80% AMT preceding the test MEP set at 120% AMT. To elicit SICI, the conditioning stimulus preceded the test MEP by 3 ms To elicit ICF, the conditioning stimulus preceded the test MEP by 12 ms [27].

#### 2.3.3. Motor Evoked Potentials (MEPs)

Motor evoked potential (MEP) peak-to-peak amplitudes were measured offline using Signal 6.0 (Cambridge Electronics Design, Cambridge, UK). Trials were excluded if (a) the participant responded incorrectly (e.g., no contraction), (b) TMS stimulation occurred between EMG bursts during writing, and (c) no MEP was elicited (MEP < 200 μV) [28]. Exclusion criteria a, b, and c were based on trial by trial inspection of the waveform data. SICI and ICF MEPs were normalized to the corresponding test MEP within the same set (each series of three randomized trials that include TS, SICI, and ICF). If a test MEP was excluded, MEPs were normalized to the test MEP in the next closest set such that the conditioned MEP was always normalized to an unconditioned MEP that was no further than three trials away. Test MEP peak-to-peak amplitudes are reported in millivolts (mV). The coefficient of variation of the test MEP amplitude was calculated for each participant to determine the consistency of the measure. Conditioned MEP peak-to-peak amplitudes (SICI and ICF) were normalized to the test MEP amplitude and are therefore reported as ratio values where <1 would indicate inhibition and >1 would indicate facilitation. Sample data from one participant are shown in Figure 4.

#### 2.3.4. Cortical Silent Period

The duration of the test MEP CSP was measured offline using a custom script in MATLAB (MathWorks, Inc., Natick, MA, USA). CSP duration was calculated from the point of the test pulse stimulus delivery to the point at which EMG activity returned to prestimulus activity (i.e., average RMS amplitude calculated over a 300-ms period prior to stimulation). CSP was calculated as the average for the 30 test pulse trials only (CSP following paired-pulses were not included).

**Figure 4.** Block averages (red line) and error (black lines) of evoked potentials for one participant. Each block included 10 single test pulses, 10 SICI paired-pulses, and 10 ICF paired-pulses. This participant completed the writing task in cursive. Pre-fatigue averages are shown above the post-fatigue averages.

#### *2.4. Statistical Analysis*

Analysis was performed using Statistica 13.2 (TIBCO Software Inc., Palo Alto, CA, USA). For TMS pulses, trial outliers were removed if they fell >2 standard deviations from the average within each individual participant. Two participants were excluded from the SICI analysis due to a large number of outliers occurring within a single block of trials. One participant was removed from the CSP analysis due to an inability return to the target level of EMG activity following stimulation. Assumptions for the analysis of variance (ANOVA) were tested for each variable using the Shapiro–Wilk test for normality. Test MEP amplitudes were not normally distributed, thus a 1/sqrt transformation was applied for statistical analysis. Preliminary analysis was performed to ensure that there was no effect of block (i.e., the 3 blocks of 30 trials pre- and post-fatigue). No block main effect was present for any variable, therefore all pre-fatigue trials and all post-fatigue trials were averaged for each participant.

To examine potential differences between testing days a dependent samples *t*-test was conducted on baseline measures (see Table 1). To determine the effect of task on dependent measures, a one-way repeated measures ANOVA was performed comparing pre-fatigue values between writing and the abduction task. To determine the effect of task on the dependent measures during fatigue, a 2 (task: abduction versus writing) × 2 (fatigue: pre versus post) repeated measures ANOVA was performed for each dependent variable. The number of participants included in each analysis is noted in Tables 2 and 3.

**Table 1.** Baseline measures made during the familiarization period on the abduction and writing days. These measures included maximal index finger abduction force and first dorsal interosseous (FDI) activity during MVCs, the level of FDI activity during 20 writing trials, the number of fatiguing sets completed, and active motor threshold (AMT) for TMS (Mean ± SD).


FDI: first dorsal interosseous; MVC: maximal voluntary contraction; AMT: active motor threshold; RMS: root mean squared; EMG: electromyography; MSO: maximum stimulator output.


×

36

§


§Calculated on the pre-fatigued values only. % Change is the (post − pre)/pre fatigue values. Cohen's *d* is calculated for the pre- to post-fatigue means. Abbreviations: SICI: short-interval intracortical inhibition; ICF: intracortical facilitation.

As the study progressed, we found that of the 19 participants, 8 wrote in cursive, 8 printed, and 3 used a combination of cursive and printing. Accordingly, we conducted an additional post hoc analysis of writing style. This separate-groups post hoc analysis included writing style (cursive vs. printing) as a categorical predictor to determine the baseline effect (task × writing style) and the effect of task and style on fatigue (task × fatigue × writing style) in a posteriori analysis. The three individuals who used a combination of printing and cursive writing strategies to complete the task were excluded from this analysis. Because this analysis of writing style was not planned a priori, it is important to note that it is underpowered.

Tukey's HSD testing was used for post hoc analysis where applicable. Partial eta-squared was calculated in Statistica for the ANOVA models. Cohen's d was calculated for the pre-fatigue to post-fatigue values where applicable [29].

#### **3. Results**

#### *3.1. Baseline Measures*

The baseline measures recorded during familiarization (force and FDI RMS amplitude during MVCs, FDI activity of writing, and AMT) were not different between the abduction and writing days. Additionally, the number of fatiguing trials performed was not different between the two days. The amount of fatigue elicited was not significantly different (Task × Style: F (1,13) = 0.88, *p* = 0.37) between printers on the writing day (−26.0% MVC force), cursive writers on the writing day (−18.6%), printers on the abduction day (−27.5%), and cursive writers on the abduction day (−22.6%).

#### *3.2. E*ff*ect of Task*

#### Motor Evoked Potentials

In the pre-fatigue trials, there was no effect of task (writing versus abduction) on the amplitude of the test MEP (F(1,18) = 0.60, *p* = 0.45, η<sup>2</sup> *<sup>p</sup>* = 0.03, Figure 5A), ICF (F(1,18) = 1.69, *p* = 0.21, η<sup>2</sup> *<sup>p</sup>*= 0.09, Figure 5C), or CSP (F(1,17) = 1.14, *p* = 0.30, η<sup>2</sup> *<sup>p</sup>*= 0.06). However, the coefficient of variation of test MEP amplitude was 23% greater in the writing task compared to the abduction task (F(1,18) = 13.55, *p* < 0.01, η2 *<sup>p</sup>* = 0.43). The coefficient of variation of the test MEP was higher in both the printing (48.4 ± 7.3, *n* = 8) and cursive writing (42.4 ± 5.5, *n* = 8) styles, as compared to abduction in printers (38.3 ± 7.9, *n* = 8) and abduction in cursive writers (35.1 ± 8.2, *n* = 8). Although there appeared to be a trend toward a greater level of inhibition (SICI) during writing compared to abduction (F(1,16) = 4.40, *p* = 0.052, η<sup>2</sup> *<sup>p</sup>*= 0.22, Table 2, Figure 5B), this analysis included only 17 participants and was not adequately powered. When writing style (printing vs. cursive writing) were factored in (task × style: F(1,13) = 8.00, *p* = 0.01, η2 *<sup>p</sup>*= 0.38, Table 3), post hoc analysis revealed a significant difference (*p* < 0.01) in SICI between printing (0.71 ± 0.13, *n* = 7) and abduction (1.03 ± 0.28, *n* = 7, Figure 6B) in printers, but no difference between cursive writing (0.86 ± 0.26, *n* = 8) and abduction (0.87 ± 0.17, *n* = 8) in cursive writers.

**Figure 5.** Pre-fatigue levels of test MEP (*n* = 19), intracortical facilitation (*n* = 19), and intracortical inhibition (*n* = 17) during finger abduction and writing (*n* = 19). There were no differences between test stimulus (**A**), short-interval intracortical inhibition (SICI; **B**), and intracortical facilitation (ICF; **C**) between the two tasks. In panels B and C, dashed lines represent no effect of the conditioning stimulus. Values below the line represent inhibition (conditioned MEP < test MEP), and values above the line represent facilitation (conditioned MEP > test MEP). Error bars represent standard deviation.

**Figure 6.** Motor evoked potentials during finger abduction and writing in printers (*n* = 8) and cursive writers (*n* = 8). There were no differences in single-pulse (test stimulus) MEPs between writers and printers or between tasks (test stimulus; **A**). Short-interval intracortical inhibition (SICI; **B**) was greater during printing than during abduction pre-fatigue (\*, *p* < 0.01). In printers, there was a fatigue-associated increase in intracortical facilitation (ICF; **C**) when assessed during printing, but not during abduction (\*, *p* < 0.01). Solid bars represent pre-fatigue values, and hatched bars represent post-fatigue values. In panels B and C, dashed lines represent no effect of the conditioning stimulus. Values below the line represent inhibition (conditioned MEP < test MEP), and values above the line represent facilitation (conditioned MEP > test MEP). Error bars represent standard deviation.

#### *3.3. Muscle Activity*

Task × fatigue repeated-measures ANOVAs were conducted for FDI, FCR, and ECR surface EMG measures of muscle activation. Level of FDI muscle activation was successfully maintained between the two tasks (F(1,18) = 0.60, *p* = 0.45, η<sup>2</sup> *<sup>p</sup>* = 0.03). This was expected because average writing RMS amplitude was used as the muscle activation target during the abduction task. The abduction task elicited greater levels of FCR activity (F(1,18) = 7.24, *p* = 0.02, η<sup>2</sup> *<sup>p</sup>* = 0.29) and lower levels of ECR activity (F(1,18) = 7.73, *p* = 0.01, η<sup>2</sup> *<sup>p</sup>*= 0.30) compared to the writing task. Analysis of writing styles (task × style) revealed a significant interaction for both FCR activity (F(1,14) = 4.70, *p* = 0.048, η<sup>2</sup> *<sup>p</sup>*= 0.25; Figure 7A) and ECR activity (F(1,14) = 5.25, *p* = 0.04, η<sup>2</sup> *<sup>p</sup>*= 0.27; Figure 7B). Post hoc analysis identified that cursive writing required significantly less FCR activity (12.58 ± 7.42 μV) than abduction (29.79 ± 17.57 μV) in the cursive writing group. There was no significant difference between FCR activity during printing (18.36 ± 9.10 μV) or abduction (20.92 ± 13.07 μV) in the printing group. In the extensor muscle, post hoc analysis identified that printing (154.22 ± 71.53 μV) required significantly more ECR activity than abduction (70.36 ± 47.40 μV) in the printers. There was no significant difference between ECR activity during cursive writing (99.07 ± 27.63 μV) or abduction (92.84 ± 39.92 μV) in the cursive group.

**Figure 7.** Pre-fatigue levels of wrist flexor (**A**) and wrist extensor (**B**) activity during the abduction and writing tasks in printers and cursive writers. Wrist flexor muscle activity was higher in the writing task than in the abduction task in cursive writers, whereas wrist extensor muscle activity was higher in the writing task than in the abduction task in the printers (\*, *p* < 0.05). This analysis included participants who completed the writing task using a pure printing (*n* = 8) or cursive (*n* = 8) strategy. Error bars represent standard deviation.

#### *3.4. E*ff*ect of Task on Fatigue*

#### 3.4.1. Motor Evoked Potentials

The task × fatigue repeated-measures ANOVA was not significant for any measure of CSE (see Table 2 for test MEP, SICI, and ICF results), including test MEP coefficient of variation (F(1,18) = 3.59, *p* = 0.07, η<sup>2</sup> *<sup>p</sup>* = 0.17) and CSP (F(1,17) = 0.29, *p* = 0.60, η<sup>2</sup> *<sup>p</sup>* = 0.02). However, when writing style was taken into consideration, a 3-way (task × fatigue × style) repeated-measures ANOVA revealed a significant 3-way interaction for ICF (F(1,14) = 9.90, *p* < 0.01, η<sup>2</sup> *<sup>p</sup>* = 0.41) (Table 3). Post hoc analysis revealed a significant increase in ICF from pre-fatigue to post-fatigue (28.1%, d = 0.94, *p* = 0.04) in the printing group. Planned comparisons were conducted using percent change.

Scores (pre- to post-fatigue) to determine the task × style effect on fatigue. The increase in facilitation in printing (28.1%) was significantly different than the decrease in facilitation of printers performing the abduction task (−12.0%, *p* < 0.01, Figure 6C). There were no significant effects of writing style on the test MEP or SICI (see Table 3), test MEP coefficient of variation (F(1,14) = 0.02, *p* = 0.88, η2 *<sup>p</sup>* = 0.002), or CSP (F(1,13) = 0.30, *p* = 0.60, η<sup>2</sup> *<sup>p</sup>* = 0.02).

#### 3.4.2. Muscle Activity

The task × fatigue repeated measures ANOVA was not significant for FDI activity (F(1,18) = 1.31, *p* = 0.27, η<sup>2</sup> *<sup>p</sup>* = 0.07), FCR activity (F(1,18) = 1.69, *p* = 0.21, η<sup>2</sup> *<sup>p</sup>* = 0.09), or ECR activity (F(1,18) = 0.86, *p* = 0.37, η<sup>2</sup> *<sup>p</sup>* = 0.05). Similarly, an analysis of writing style (task × fatigue × style) did not yield any significant interactions for FDI activity (F(1,14) = 0.57, *p* = 0.46, η<sup>2</sup> *<sup>p</sup>* = 0.04), FCR activity (F(1,14) = 0.57, *p* = 0.47, η<sup>2</sup> *<sup>p</sup>* = 0.04), or ECR activity (F(1,14) = 2.82, *p* = 0.12, η<sup>2</sup> *<sup>p</sup>* = 0.17).

#### **4. Discussion**

Our study demonstrates that fatigue-associated changes in CSE depend on the motor task completed during single- and paired-pulse TMS stimulation. Not only did we find differences in fatigue-associated changes in cortical excitability assessed during hand-writing compared to isometric finger abduction, but we also found differences between participants who completed the writing task by printing compared to those who wrote in cursive. Specifically, we found that the same fatigue protocol elicited an increase in ICF during printing, but not during cursive writing. It is important to note that the fatigue protocol itself was the same on each day, and only the motor task during pre-fatigue and post-fatigue assessments of CSE differed. Furthermore, because the level of FDI muscle activation was held constant during the two tasks, the differences in CSE between writing and abduction days were not simply due to differences in voluntary drive to that muscle.

The present study examined CSE during voluntary movement. Because the level of muscle activation during voluntary movement affects levels of CSE [30–32], it was critical in the present study to match FDI EMG between tasks. We accomplished this by measuring FDI muscle activation during the writing task in a familiarization period each day, and then using biofeedback to set a target on the abduction day that matched muscle activation during writing for each participant. This approach to matching muscle activation between the two tasks (writing and abduction) was equally effective in printers and cursive writers. Therefore, differences in intracortical measures of excitability pre- and post-fatigue cannot be attributed to differing FDI EMG activity between the tasks.

During baseline (pre-fatigue) testing, the variability of single-pulse MEP amplitudes was greater in the writing task than the abduction task. This was expected because the abduction task is a single-joint, externally cued task frequently used in TMS studies for the purpose of reducing variability. The abduction task was maintained at a stable force level using visual feedback. To make the writing task more functionally-relevant, we did not control the style, or speed of each participant's writing. Therefore, it is not surprising that the writing task resulted in greater variability than the abduction task, both within and between subjects. Similar results have been found comparing test–retest reliability of

MEP amplitudes between static and dynamic tasks in the lower limb [33], where MEPs evoked during a static task (i.e., plateau at a target force level) were less variable than those evoked during a dynamic task (i.e., continuously increasing force level). The style of writing did not impact the MEP amplitude variability, as we found higher coefficient of variation values for both printing and cursive writing compared to abduction. A future challenge in TMS research will be to utilize laboratory tasks that balance variability with the relevance of the task to motor activities outside of the lab.

We hypothesized that CSE would be greater during writing compared to abduction due to the complexity of the writing task. However, single-pulse MEP amplitude was not different between tasks. Reports of differences in CSE between precision tasks and conventional abduction tasks are inconsistent. Original research in monkeys suggested that corticospinal neurons were more active during a precision task compared to a power task [9], a finding supported by several studies that found greater MEP amplitudes during complex, precision tasks compared to simple, power tasks [4,11,34]. Alternatively, larger MEP amplitudes have been reported during conventional abduction tasks compared to power, pincer, or grasping tasks [5–7]. The discrepancy between our findings and previous research may be that other studies have used precision tasks that were visually guided and externally controlled, whereas our writing task is a dynamic, internally generated task that is retrieved and implemented from memory [35,36]. Furthermore, writing involves higher levels of activation in the dorsal premotor cortex in comparison to simple finger contraction tasks, as well as unique activation in multiple brain regions (e.g., premotor cortex and anterior putamen) not activated during tapping or "zigzagging" finger actions [37]. This association of writing with higher cognitive demands would suggest that writing is a more complex task in comparison to simple finger contractions. On the other hand, because writing is learned and practiced from a young age into adulthood, it is also associated with a degree of automaticity [38]. Accordingly, our finding that writing was not associated with increased CSE compared to abduction, may have been due to either the increased complexity or the automaticity of writing compared to unpracticed, less natural, and externally guided precision-grip tasks used in previous studies.

Despite the fact that baseline (pre-fatigue) CSE did not differ between tasks, levels of intracortical inhibition (SICI) assessed using paired-pulse TMS trended (*p* = 0.052) toward greater inhibition during writing compared to abduction. When participants were subdivided into those who printed and those who wrote in cursive (excluding those who used a combination of printing and cursive within a single word), printers had the greatest intracortical inhibition during the writing task. We hypothesize that this may be due to the increased control required during the more intermittent task of writing distinct letters during printing, compared to the continuous nature of cursive writing (Figure 2). The differences in intracortical inhibition observed between writing styles may also be explained by the activity of the proximal forearm muscles. While FDI EMG activity between abduction and writing did not differ, extensor (ECR) activity was greater and flexor (FCR) activity was lower during the writing task compared to abduction. It has been suggested that activity and position of proximal muscles can have an effect on the corticospinal pathway leading to the distal muscle of interest [39–41]. This relationship has been shown between the FDI and proximal arm muscles (including the ECR, FCR, and deltoid muscles), where proximal muscle activity resulted in facilitation of distal muscle MEPs [39–41]. Similarly, forearm position (pronation vs. semi-supinated) is also known to alter CSE [42]. In our study, however, the forearm was semi-supinated for both printing and cursive writing, and therefore forearm position does not explain the differences in intracortical excitability that we observed when we compared the two different writing styles. The role that proximal muscle activation may play in the CSE of distal muscles highlights the importance of choosing laboratory tasks that more closely resemble natural movements outside of the laboratory to which we aim to extrapolate our results.

TMS is used to assess changes in corticospinal and intracortical excitability in response to many different interventions and perturbations (e.g., fatigue, strength training, skill training, disease, injury, aging, and pharmacological agents). We sought to determine whether changes in CSE following an intervention depend on the motor task used to elicit muscle activity during the delivery of TMS. To this

end, we employed neuromuscular fatigue as an acute intervention, and measured changes in CSE during two different motor tasks before and after two identical fatigue protocols. It is well established that neuromuscular fatigue is associated with changes in CSE, specifically a reduction in unconditioned MEP (test MEP) amplitudes in the target muscle following neuromuscular fatigue [43,44]. This depression has been attributed to central mechanisms of fatigue [44,45]. Paired-pulse TMS techniques have been used to identify intracortical mechanisms of fatigue [22,46,47]. Following fatiguing contractions, SICI decreased [18,46,48]; however, the association between fatigue and ICF is not as clear. For example, ICF measured in the biceps brachii decreases during a sustained fatiguing contraction [48,49], whereas ICF measured in the FDI is elevated at the point of task failure when assessed at rest [22,46]. In other studies, ICF does not change with fatigue [47,49]. It has been suggested that decreased intracortical inhibition [18] and increased facilitation following fatigue [22] may serve as a compensatory mechanism to optimize motor output as fatigue develops.

In our study, there was no task-dependent effect of fatigue on any of the corticospinal and intracortical excitability measures between writing and abduction. However, most of our participants used one of two distinct writing styles: printing or cursive. Therefore, we completed additional a posteriori analyses to determine whether there was an effect of writing styles on fatigue-induced changes in CSE, and found that printers had a significant increase in ICF following fatigue. This is consistent with previous reports of increased ICF following fatigue that employ more conventional laboratory tasks [46,50]. However, this only occurred during printing, and was not found during cursive writing or the conventional finger abduction task. Previous research has suggested that increased ICF post-fatigue may be a mechanism to compensate for peripheral contractile failure or reduced upstream drive to the primary motor cortex [22]. However, this would not explain the increased ICF seen in printing but not in cursive writing or abduction. Possibly the differences in wrist stabilization during printing and cursive, reflected by differences in wrist extensor and wrist flexor muscle activity during the two tasks, had a greater impact on ICF than on SICI. Furthermore, it is possible that printing employs cortical circuits that are affected by isometric finger abduction fatigue task differently than the cortical circuits employed in cursive writing. Although there is very little research regarding the neural control of printing versus cursive writing, one clinical study reports impairment in cursive writing, but not printing, drawing, or the ability to draw continuous loops, following ischemic damage to the parietal lobe [51]. Another clinical case found that the ability to write in cursive was lost with the development of a large cranial tumor impinging on the left frontal lobe, while the ability to print remained intact. In this case, the patient regained the ability to write in cursive following tumor resection [52]. Case studies like these support the idea that printing and cursive writing have different cortical representations, and may therefore be associated with different inputs to the primary motor cortex. This novel finding certainly warrants further exploration, as printing and cursive writing appear to be two different tasks that further demonstrate the task-dependent nature of fatigue-associated changes in CSE.

We started this repeated-measures study with 20 participants based on a power analysis conducted using data from a previous study from our lab that also investigated intracortical mechanisms of fatigue [22]. Of the 20 participants we recruited, 1 participant was unable to complete the study, and 2 participants were excluded from the SICI analysis. Accordingly, this analysis is not robustly powered, and these data should be interpreted cautiously. The analysis of hand writing style was not planned in advance. From the start, we told participants to write as they preferred given that our aim was to assess corticospinal excitability during a natural motor task. As the study progressed, we found that many university students were unable to write in cursive, and we conducted the post hoc analysis of writing style. It should be noted that this a posteriori analysis of writing style (cursive vs. printing) is underpowered. Furthermore, in the present study, we assessed only three intracortical mechanisms (SICI, ICF, and cSP). Each of these measures has been associated with different types of neurotransmission (SICI with GABA-A receptors, ICF with glutamatergic transmission, and the cSP with GABA-B receptors) as reviewed by Reis [1]. It is possible that other intracortical (e.g., long interval

intracortical inhibition) and interhemispheric pathways (e.g., interhemispheric inhibition) may reveal fatigue-associated changes that are dependent on the motor task that is conducted during stimulation. This is an area that requires further investigation

#### **5. Conclusions**

This study is the first to assess CSE and the effect of fatigue during a writing task. Although fatigue-associated changes in CSE have been well studied over the last three decades, this study highlights the importance of considering the task used during TMS measures of corticospinal and intracortical excitability. Although controlled laboratory tasks are required to reduce variability of motor evoked potentials to allow for reproducible results, it is important to note that CSE is task-dependent. Because of this, measures of CSE made during a laboratory task may not translate to motor tasks outside of the lab. Despite the fact that the fatigue task and the level of FDI muscle activation during TMS was the same in our study, hand-writing revealed fatigue-associated changes in CSE that were not evident during abduction. Therefore, the task-dependent nature of CSE emphasizes the need for experimental paradigms that better reflect relevant motor tasks or at least acknowledge the differences between the task employed in the laboratory and the movements outside of the lab to which results may be extrapolated. With this in mind, it is essential that studies of CSE consider the "task at hand".

**Author Contributions:** K.T.M.C.: pilot work, data collection, data analysis, and writing of manuscript, L.A.G.: supervision and instruction of techniques, data analysis, writing of manuscript, and revision of manuscript, J.M.K.: study conception, supervision, instruction of techniques, data analysis, and revision of manuscript.

**Funding:** This work was funded by a NSERC Discovery (386601) grant to J.M.K. and an NSERC PDF to L.A.G.

**Acknowledgments:** The authors would like to thank Ron Daniels, Laurier Science Maker Lab, for his assistance with the design and creation of custom-built TMS table used in this experiment.

**Conflicts of Interest:** The authors declare no conflicts of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **The Impact of Glucose on Corticospinal and Intracortical Excitability**

**Stephen L. Toepp, Claudia V. Turco, Mitchell B. Locke, Chiara Nicolini, Roshni Ravi and Aimee J. Nelson \***

Department of Kinesiology, McMaster University, Hamilton, ON L8S 4K1, Canada; toeppsl@mcmaster.ca (S.L.T.); turcocv@mcmaster.ca (C.V.T.); lockemb@mcmaster.ca (M.B.L.); nicolinichiaratn@gmail.com (C.N.); roshniravi67@gmail.com (R.R.)

**\*** Correspondence: nelsonaj@mcmaster.ca

Received: 21 October 2019; Accepted: 22 November 2019; Published: 25 November 2019

**Abstract:** Neurotransmission is highly dependent on the availability of glucose-derived energy, although it is unclear how glucose availability modulates corticospinal and intracortical excitability as assessed via transcranial magnetic stimulation (TMS). In this double-blinded placebo-controlled study, we tested the effect of acute glucose intake on motor-evoked potential (MEP) recruitment curves, short-interval intracortical inhibition (SICI), short-latency afferent inhibition (SAI) and long-latency afferent inhibition (LAI). Eighteen healthy males participated in four sessions. Session 1 involved acquisition of an individualized blood glucose response curve. This allowed measurements to be time-locked to an individualized glucose peak after consuming one of three drinks during the subsequent three sessions. Participants were administered a 300 mL concealed solution containing 75 g of glucose, sucralose, or water in separate sessions. Dependent measures were assessed at baseline and twice after drinking the solution. Secondary measures included blood glucose and mean arterial pressure. Corticospinal excitability and blood pressure increased following the drink across all treatments. No changes were observed in SICI, SAI or LAI. There was no rise in corticospinal excitability that was specific to the glucose drink, suggesting that acute changes in glucose levels do not necessarily alter TMS measures of corticospinal or intracortical excitability.

**Keywords:** glucose; SICI; SAI; LAI

#### **1. Introduction**

Glucose is the brain's primary energy substrate and provides the main carbon source for de novo synthesis of large compounds required for essential ranging processes, from neurotransmission to the management of oxidative stress [1–4]. Although the substantial influence of glucose is evident in several clinical contexts [5–7], a comprehensive neurophysiological profile has yet to be compiled in healthy humans during periods of fasting versus high-circulating glucose following feeding.

A small number of studies have used non-invasive transcranial magnetic stimulation (TMS) to compare neurophysiological measures in hyperglycemic, normoglycemic or fasting conditions [8–10]. Specterman et al. [10] reported a 3-fold increase in the size of motor-evoked potentials (MEPs) 60 min after ingestion of 68 g of glucose, such that greater increases in MEPs were correlated with greater increases in blood glucose levels. Badawy and colleagues [9] observed greater long interval intracortical inhibition (LICI) in epileptic and healthy individuals when in a fed (i.e., two hours after a meal) compared to fasted state (i.e., 12 h overnight). However, not all studies have detected an effect of glucose on TMS measures. Andersen and colleagues [8] manipulated glucose levels in type 1 diabetics via an intravenous glucose pump and did not observe any change in cortical motor thresholds, a common measure of corticomotor excitability. TMS can also be used to probe the sensorimotor system with measures of short- and long-latency afferent (SAI, LAI). To date, no studies have investigated the

influence of glucose on afferent inhibition. SAI and LAI are impaired in populations with Alzheimer's and Parkinson's disease [11], both of which display altered central glucose metabolism [12,13].

Most research to date that directly tests the effects of glucose on the motor system provides information regarding a limited variety of neurophysiological measurements. For example, the study conducted by Specterman and colleagues [10] measured MEPs using a time-efficient protocol involving the delivery of only 15 suprathreshold TMS pulses to obtain a measure of average MEP size. This approach allows for relatively high temporal resolution but contrasts with more comprehensive tests which could probe underlying mechanisms. For example, acquisition of MEP recruitment curves, while more time consuming, provides information regarding cortical glutamate levels [14]. Investigating a proxy measure of glutamate levels is useful, since glutamatergic neurotransmission has been linked with glucose in in vitro [15,16] and in situ [17] neurobiological studies. It may also be worthwhile to probe gamma-aminobutyric acid (GABA)-mediated intracortical inhibition using paired-pulse TMS. Although Badawy and colleagues found no significant effect of glucose on short interval intracortical inhibition (SICI), all significant and non-significant differences between fasted and fed participants were in the direction of increased inhibition or decreased facilitation after feeding as opposed to after fasting [9]. This is counterintuitive, given the apparent increase in MEP size after glucose ingestion [10], and warrants further investigation.

The goal of this study was to test the effect of glucose ingestion on the healthy human brain using a variety of non-invasive neurophysiological measures which have relevant mechanistic underpinnings. This investigation also controlled for the influence of a sweet placebo and time-locked measurements to individually measured peak blood glucose latencies. It was hypothesized that glucose would increase corticospinal excitability and SICI. The primary observation was that glucose did not change SICI, SAI, LAI or corticospinal excitability. Furthermore, corticospinal excitability and blood pressure increased over time when data were averaged across all treatments.

#### **2. Materials and Methods**

#### *2.1. Participants*

Healthy, young (*n* = 18, 22.8 ± 2.4 years), right-handed, male non-smokers were recruited from the McMaster University student population. Participants passed a screening for TMS contraindications [18] and were identified as right-handed using a modified handedness questionnaire [19]. Inactive individuals were excluded using the International Physical Activity Questionnaire (IPAQ, <600 MET-minutes/week) to reduce the risk of influence from prediabetic impairment of glucose metabolism, which is inversely correlated with physical activity level [20]. This study was approved by the Hamilton Integrated Research Ethics Board (HiREB) and conformed to the declaration of Helsinki.

#### *2.2. Electromyography*

Motor-evoked potentials were recorded via surface electromyography (EMG) over the first dorsal interosseous (FDI) muscle. FDI was chosen because MEPs from this muscle have demonstrated good inter- and intra-session reliability [21]. Adhesive electrodes (9 mm diameter Ag-AgCl) were placed over the FDI muscle belly and the metacarpal head of the index finger. The EMG signal was amplified 1000x and sampled at 5 kHz with low and high pass signal filters of 2.5 kHz and 20 Hz, respectively. EMG data was recorded using an analog-to-digital interface (Power 1401; Cambridge Electronics Design, Cambridge, UK) in combination with Signal/CED analysis software (Signal version 6.02; Cambridge Electronics Design).

#### *2.3. Transcranial Magnetic Stimulation*

Participants sat upright in the testing chair with their palms resting supine and elbows at an approximate 45◦ angle. Single and paired-pulse TMS was delivered with a custom 50 mm figure-of-eight coil, connected to a Magstim Bistim stimulator (Magstim, Whitland, UK). The coil was held over the left primary motor cortex (M1) and the optimal stimulation location or "motor hotspot" was targeted using Brainsight neuro-navigation software (Rogue Research, Canada). The motor hotspot for FDI muscle was determined by delivering pulses at 50% of the maximum stimulator output (%MSO) over the approximate location of M1 while adjusting coil placement until the TMS pulses reliably evoked large MEPs in the FDI muscle. The angle of the coil relative to the midsagittal plane was maintained at 45◦ to induce posterior-to-anterior current in cortical tissue.

#### *2.4. Resting Motor Threshold*

Resting motor threshold (RMT) was defined as the stimulus intensity (%MSO) that evokes an MEP (i.e., peak-to-peak amplitude >50 μV) 50% of the time. This value was determined using TMS\_MTAT\_2.0 freeware (http://clinicalresearcher.org/software.htm). The starting stimulus intensity was set to 37% MSO. Twenty TMS pulses were then delivered over M1, adjusting the intensity after each pulse, as determined by the MTAT software based on the MEP occurrence (or lack thereof) in the previous trial [22].

#### *2.5. MEP Recruitment Curve*

Corticospinal excitability was measured using single-pulse TMS to obtain MEP recruitment curves. Eight TMS pulses were delivered at 90%, 100%, 110%, 120%, 130%, 140%, 150%, 160%, 170%, 180%, 190%, and 200% of RMT in a randomized order, with an inter-stimulus interval of 4 s. The MEP amplitude was plotted against stimulus intensity and data were fit with a Boltzmann sigmoidal curve. The regression line was segmented 1000 times and the area under the recruitment curve (AURC) was quantified by trapezoidal integration.

#### *2.6. Short Interval Intracortical Inhibition*

Short-interval intracortical inhibition (SICI) was measured using a paired-pulse TMS protocol. The suprathreshold test stimulus (TS) was adjusted to the intensity that evoked an MEP with a size of approximately 1 mV in peak–peak amplitude. The conditioning stimulus (CS) preceded the TS by 2 ms, with a stimulus intensity of 80% of RMT. Twelve unconditioned (MEPTS) and 12 conditioned (MEPCS-TS) pulses were delivered in a randomized order, with a 5 s inter-trial interval. The magnitude of SICI was quantified using the ratio of conditioned to unconditioned MEP size (MEPCS-TS/ MEPTS).

#### *2.7. A*ff*erent Inhibition*

Electroencephalography (EEG) electrodes were positioned over C3 (located 2 cm posterior to C3) and referenced to Fz (International 10–20 system). A bar electrode was positioned over the median nerve at the wrist (cathode proximal) to deliver square wave electrical pulses (0.2 s pulse width) using a constant current stimulator (DS7AH; Digitimer, Welwyn Garden City, UK). Nerve stimulation was delivered at the minimum intensity that evoked a visible twitch in the abductor pollicis brevis (APB) muscle. Time-locked averaging of five hundred stimuli delivered at 3 Hz was used to determine the latency of the N20 peak of the somatosensory-evoked potential (SEP).

To acquire afferent inhibition, the TMS intensity was set to evoke a MEP of ~1 mV in the right FDI muscle. Electrical stimulation was delivered to the median nerve at the wrist at the minimum intensity that evoked a visible twitch in the APB muscle. The average intensity of nerve stimulation was 11.1 ± 3.8 mA. For SAI, the interstimulus interval (ISI) between peripheral nerve stimulation and TMS was 4 ms longer than the N20 latency (i.e., N20 + 4 ms). An ISI of 200 ms was used to acquire LAI. Twelve unconditioned stimuli (MEPTS) were randomly presented among 36 conditioned stimuli (nerve stimulation followed by TMS, twelve stimuli per ISI), with a 5 s inter-trial interval. The magnitude of SAI/LAI was expressed as the ratio of the conditioned to the unconditioned MEP amplitude.

#### *2.8. Blood Glucose and Blood Pressure*

Capillary blood glucose measurements were performed via the glucose oxidase method using a hand-held diabetes monitoring device (Abbott MediSense FreeStyle Precision Neo Blood Glucose and Ketone Monitoring System, Abbott). Since previous research has indicated that blood pressure may be elevated by ingestion of a large glucose bolus [23,24], mean arterial blood pressure was measured using an automated blood pressure monitor (OMRON Blood Pressure Monitor, OMRON Healthcare). The mean arterial pressure (MAP) was calculated from the systolic (SBP) and diastolic blood pressure (DBP) as indicated below:

$$\text{MAP} = (\text{2DBP} + \text{SBP})/3 \tag{1}$$

#### *2.9. Experimental Design*

This study implemented a double-blinded, three-way crossover design in which fasted participants were assessed before and after ingestion of water, a sucralose-flavored placebo or a 75 g glucose bolus. All solutions were 300 mL. Prior to the first experimental testing session, participants completed a preliminary testing session. A schematic of the study schedule for each participant is shown in Figure 1.

**Figure 1.** Timeline for the preliminary visit (visit 1) and the three experimental visits (visits 2, 3 and 4). For visit 1, baseline blood glucose was measured (0 min) and then participants drank 75 g of glucose. Next, blood glucose was measured every 10 min for 50 min or until peak was observed (open arrows). For visits 2, 3 and 4, the transcranial magnetic stimulation (TMS) testing bouts (black boxes) at T0 and T1 were separated by a rest period equal to the glucose latency minus 5 min. Mean arterial pressure (grey arrows) and blood glucose (open arrows) were measured before, after and in between the TMS testing bouts and are labeled T0, T1, T2 and T3. TMS—transcranial magnetic stimulation; AURC—area under the recruitment curve; SICI—short-interval intracortical inhibition; SAI/LAI—short-latency afferent inhibition/long-latency afferent inhibition.

Visit 1 was used to assess a time-course for glucose metabolism, allowing the subsequent TMS measures on visits 2, 3, and 4 to be individualized. Participants arrived in the lab having fasted for a minimum of 10 h, and then ingested a 75 g glucose bolus in 300 mL of solution. Finger-prick blood samples were collected and analyzed at 10 min intervals, as indicated in Figure 1 (top). The latency at which peak blood glucose occurred was used to ensure that TMS tests are conducted during a period of high-circulating glucose for each individual.

Visits 2, 3 and 4 were scheduled at least 48 h apart. On the day of each visit, participants arrived in the lab having fasted for 10 h and then ingested a 300 mL solution containing either plain water, sucralose-sweetened placebo (5 g/300 mL Splenda® solution), or a 75 g oral glucose tolerance test bolus. TMS measures were acquired before ingestion (T0), 5 min before each participant's peak blood glucose latency (~30 min after drink ingestion) as determined in Visit 1 (T1), and ~1 h after ingestion, corresponding to the approximated peak of glucose levels in the cerebrospinal fluid (T2) occurring ~30 min after plasma glucose [25]. SAI, LAI, SICI and MEP recruitment curves were measured in a pseudorandomized order which was determined using an online Latin square generator (https://hamsterandwheel.com/grids/index2d.php). Capillary blood glucose and blood pressure were measured before and after each of the post-drink bouts, as denoted by the labels T1, T2 and T3 (see Figure 1, bottom).

The McMaster University Medical Centre (MUMC) research pharmacy provided a randomized treatment schedule. All treatment solutions were provided in uniform, shrouded bottles, with a letter code corresponding to the order of delivery. MUMC pharmacy held the drink randomization (i.e., drink identity) key until collection was complete to ensure that the experimenters were blind to the identity of the drink. Blood glucose and subjective ratings of sweetness were recorded by an unblinded researcher who did not otherwise take part in data collection or analysis. The sucralose-sweetened placebo was taste-matched with the 75 g glucose solution by MUMC pharmacy. Participants were explicitly asked not to comment on the taste of the drink to the researchers and it was made clear that this was very important to the integrity of the study. The participants were blind to the identity of the drink to the degree that they could not distinguish between the sucralose placebo and the glucose solutions (water was not masked with any taste).

#### *2.10. Statistical Analyses*

Trials were discarded if the EMG activity was >100 μV in the 100 ms preceding the stimulation artefacts, similar to previous work [26]. Normality was assessed with the Shapiro–Wilks test and a square-root or log transformation was applied in cases where data was not normally distributed.

First, to confirm that inhibition was observed for measures of SICI, LAI and SAI, two-way ANOVAs with the factors PATTERN (two levels: unconditioned MEP and conditioned MEP) and TIME (three levels: T0, T1, T2) were performed. Next, one-way ANOVAs using the within-subjects factor of TREATMENT (three levels: glucose, sucralose, water) were used to confirm that T0 data were not different between visits. Next, outlier analysis was performed using SPSS. Four outliers were removed from the SICI data and one was removed from the SAI data. Data were subsequently analyzed using repeated-measure ANOVAs with factors TREATMENT (three levels: glucose, sucralose, water) and TIME (three levels: T0, T1, T2). Post-hoc testing was performed with Bonferroni-corrected two-tailed paired t-tests. A Conover's ANOVA [27] was performed in lieu of a parametric ANOVA in cases where the data was not normally distributed (even after attempted transformations), with the Wilcoxon-signed rank test used for post-hoc testing. Supplementary measures of capillary blood glucose and MAP were assessed using two-way repeated measures ANOVA with four levels of TIME (T0, T1, T2, T3).

Wilcoxon signed-rank tests were used to assess the hypotheses that glucose would strengthen SICI and increase AURC, as well as all indicated post-hoc comparisons. Effect sizes were calculated using Cohen's d for paired t-tests and r for Wilcoxon's tests. Significance for all statistical tests was set to alpha <0.05. Data from TMS, blood glucose and blood pressure measurements are included in the supplementary materials.

#### **3. Results**

#### *3.1. Blood Glucose*

The peak plasma glucose concentration after ingestion of the glucose bolus on Visit 1 was 9.5 ± 1.0 mmol/L. This represents a 2-fold increase from the fasting glucose level of 4.8 ± 0.4 mmol/L, with a rise of 4.7 ± 0.9 mmol/L. The majority of participants had peak glucose latencies of 40 min

(*n* = 10), followed by six peaking at 30 min, and one participant peaking at 20 min and 50 min each. The observed peak glucose latency spread of 30 min emphasizes the importance of the individualized approach tested herein.

As expected, glucose levels for the experimental visits exhibited a significant effect of TREATMENT (F (2,17) = 108.261; *p* < 0.001), TIME (F (3,17) = 31.037; *p* < 0.001) and TREATMENT × TIME (F (6,17) = 29.722; *p* < 0.001). In the glucose delivery condition, plasma levels measured at T1 were 3.4 ± 1.9 mmol/L higher than fasting and remained elevated by 2.1 ± 1.2 mmol/L at T3. These data confirm that glucose levels were substantially increased during both post-drink TMS testing bouts in the glucose delivery condition. Notably, there was a slight increase in glucose level (<1 mmol/L) at T1 after the sucralose-sweetened placebo and a slight reduction at T1 following water. However, blood glucose returned to fasting levels at T2 and T3 in both cases. These data are displayed in Figure 2a and *p*-values and effect sizes for significant changes are reported in Table 1.

**Figure 2.** (**a**) Mean and standard error of capillary blood glucose levels over the course of each testing session. Measurements are plotted at each timepoint before and after ingestion of glucose, sucralose or water, which were consumed at 0 min. (**b**) Means and standard errors of AURC data entered into the ANOVA and treatment-averaged data, showing the significant effect of TIME. (**c**) Average relative increase in plasma glucose across T1, T2 and T3, plotted against the average relative increase in AURC across T1 and T2. (**d**) Treatment-averaged motor-evoked potentials (MEP) responses with mean and standard error at each TMS intensity of the recruitment curve for T0, T1 and T2. \* indicates a significant difference from T0 (*p* < 0.05).


AURCall: area under the recruitment curve averaged across all treatments, GLU: glucose data for each of the three drinks, MAPall: mean arterial pressure averaged across all treatments \* indicates that effect test statistics and effect size are derived from Wilcoxon's test (i.e., *z* and *r*, respectively).

#### *3.2. MEP Recruitment Curve*

As shown in Figure 2b, there was no increase in AURC after glucose at T1 or T2 (*p* = 0.110). There was only a significant effect of TIME, which was independent of treatment (Table 2, Figure 2b,d). No other effects in the ANOVA and no association between capillary blood glucose levels and the change in AURC at T1 or T2 were found (Table 3, Figure 2c). Further, no difference was detected between T0 data for the three treatments (Table 4).



AURC: area under the recruitment curve, SICI: short-interval intracortical inhibition, SAI: short-latency afferent inhibition, LAI: long-latency afferent inhibition. \* indicates data was square root transformed, # indicates log transformation, ¶ indicates data was ranked.


**Table 3.** Results from correlations between area under the recruitment curve (AURC) and capillary glucose (GLU) or mean arterial pressure (MAP).

Correlations were tested between relative change in AURC and GLU for the adjacent timepoint on the glucose visit only (T1glu, T2glu and T3glu). Mean arterial pressure correlations were calculated for treatment-averaged data at each timepoint (T1all, T2all and T3all). Correlations between the average change across all post-drink measures were also examined (bottom). All correlations were carried out using Pearson's r unless data were not normally distributed, in which case Spearman's rho (indicated by \*) was used. Bonferroni correction was applied to all correlations.

**Table 4.** Results from preliminary ANOVAs confirming the presence of inhibition and no differences between T0 data for each measure.


AURC: area under the recruitment curve, SICI: short-interval intracortical inhibition, SAI: short-latency afferent inhibition, LAI: long-latency afferent inhibition.

#### *3.3. SICI*

The presence of SICI was confirmed by two-way ANOVAs which showed the main effects of PATTERN for each treatment (Table 4), indicating that inhibition was observed at all timepoints. There was no increase in SICI following glucose (Figure 3a) and there was no effect of TREATMENT, TIME or TREATMENT × TIME observed in the ANOVA (Table 2).

**Figure 3.** Means and standard errors of (**a**) SICI, (**b**) SAI and (**c**) LAI data. All data are expressed as the ratio of the conditioned response (CSTS) to the unconditioned response (TS) such that the degree to which the ratio falls below 1.0 (i.e., the dotted line) reflects the magnitude of inhibition observed.

#### *3.4. A*ff*erent Inhibition*

For SAI and LAI, two-way ANOVAs confirmed the presence of significant inhibition (Table 4). There was no effect of TREATMENT, TIME or TREATMENT × TIME (Table 2, Figures 2c and 3b).

#### *3.5. Mean Arterial Pressure*

Figure 4 shows a main effect of TIME (F(2,12 = 15.119; *p* < 0.001), with significant increases in mean arterial pressure at all post-drink pressure which were observed across all treatments. Post-hoc analysis of treatment-averaged MAP data indicates that this increase from baseline (T0) was evident at T1, T2 and T3 (*p* < 0.001) (Figure 4a). There were also no correlations between the increase in mean arterial pressure across treatments and changes in AURC at T1 or T2 (Table 3).

**Figure 4.** (**a**) Means and standard errors of mean arterial pressure for each treatment and for the treatment-averaged data showing the effect of TIME. (**b**) Average increase in mean arterial pressure across T1, T2 and T3 plotted against the average relative increase in AURC across the T1 and T2 TMS testing bouts. \* indicates a significant difference from baseline (*p* < 0.05).

#### **4. Discussion**

The primary finding of this study is that glucose did not lead to an increase in corticospinal excitability. Further, there was no change in SICI, SAI or LAI. However, we did observe that all treatments contributed to an increase in corticospinal excitability and mean arterial pressure. This increase in excitability was not related to the increase in mean arterial blood pressure. We will

discuss below the lack of glucose effects on TMS measures and the possible role of hydration and prolonged brain stimulation on changes in mean arterial pressure.

The increase in AURC across all conditions contrasts with previous research reporting an increase in corticospinal excitability following glucose ingestion and not after a no-calorie placebo [10]. Notably, Specterman and colleagues [10] tested a small sample size (*n* = 4) compared to the present study. They also measured corticospinal excitability by quantifying the average size of MEPs evoked at 110% of RMT, while the present study did so across the multiple intensities of a recruitment curve. Therefore, it is possible that the contrasting results reflect a combination of differences in the study population and measurement protocol.

The finding that glucose did not alter SICI is in line with the results reported by Badawy and colleagues [9], who found that SICI was not significantly different after 12 h of fasting versus 2 h after a meal. It should be noted that the authors did report significantly greater LICI after feeding than after fasting. These findings suggest that glucose has a different effect on LICI versus SICI. Indeed, each measure is thought to reflect the activity of different neurotransmitter receptors, with GABAA receptors mediating SICI and GABAB receptors involved in LICI [28,29].

Intake of glucose, sucralose or water did not change measures of SAI or LAI. SAI and LAI are impaired in multiple neurodegenerative conditions, including Alzheimer's and Parkinson's disease [30], and have potential clinical utility as diagnostic tools or biomarkers of sensorimotor function. Therefore, it is important to establish whether factors of daily living influence the acquisition of these measures. The present study suggests that neither elevated glucose levels or hydration change the magnitude of SAI or LAI. The results also suggest that the weakening of SAI/LAI in these neurodegenerative populations [12,13] does not reflect the impairment in glucose metabolism.

Increased AURC at T1 and T2 coincided with a treatment-nonspecific rise in mean arterial pressure, suggesting influence from factors intrinsic to the testing protocol. These factors may include hydration, discomfort with the testing setup (i.e., a white-coat effect) or the delivery of a high number of TMS pulses over M1. Since participants were not asked to dehydrate themselves overnight, and the 300 mL drink is a relatively low volume of fluid, it seems unlikely that hydration was a major factor with respect to mean arterial pressure. It is more likely that changes in emotional state due to fatigue or level of interest, or the repetitive stimulation itself could have contributed to the observed TIME effects. The sensorimotor cortex has functional connections with brainstem structures involved in the regulation of vasomotor tone, which are modulated by cortical stimulation [31]. Transcranial direct-current stimulation of the sensorimotor area has been shown to acutely suppress blood pressure and adrenocorticotropic hormone levels [32]. If TMS changes blood pressure similarly to direct-current stimulation, it is possible that an association between changes in mean arterial pressure and AURC was masked by acute effects of TMS delivered at T1 and T2. In addition, Binkofski et al. [32] described longer-latency changes (1–2 h) in cerebral energy metabolism and glucose uptake following brain stimulation, suggesting that baseline TMS could have masked relationships between glucose and AURC at T1 and T2. This may provide an explanation for the contrasting results with Specterman et al. [10], who delivered only 45 TMS pulses at baseline and 15 per timepoint. They observed a positive association between glucose level and MEP size, while we delivered 180 pulses at T0, T1 and T2 and did not observe the same relationship.

In the present study, the timing of T1 was based on the latency of peak glucose levels obtained on Visit 1. The purpose of this protocol was to maximize the opportunity to observe a change in corticospinal excitability following glucose ingestion. However, the day-to-day variability in glucose metabolism was not assessed. Therefore, while we observed a significant rise in glucose levels following ingestion of the solution in the glucose session, the latency of peak glucose may have varied from day-to-day.

#### *Future Considerations*

It is important for future research to attempt to replicate previously observed effects of glucose on TMS measurements, using testing protocols which consider potential confounding factors such as the number of TMS pulses, session duration and hydration. The effect of TMS pulse load on changes in glucose levels and sympathetic tone should also be investigated to facilitate the interpretation and design of future TMS research. While glucose did not change our TMS measures, other dietary factors such as caffeine consumption [10,33], prolonged fasting [34] and ketogenic diets [35] merit further investigation as these may be important and easily modifiable factors in TMS research.

#### **5. Conclusions**

The present study found no explicit effect of glucose on corticospinal excitability, intracortical inhibition or afferent inhibition, but corticospinal excitability and mean arterial pressure increased across all treatments over the course of the experiment. These non-treatment-specific increases suggest that TMS measurements could be sensitive to various confounding factors related to repeated magnetic stimulation of the cortex, hydration, or fatigue. Further investigation of the influence of diet and acute carbohydrate consumption is warranted. However, studies should first directly examine the impact of the aforementioned confounding factors so that they can be effectively taken into account during the design and interpretation of research on this topic. Findings from such studies will work to reduce the likelihood that confounding effects arising from changing brain metabolism or autonomic modulation complicate the interpretation of TMS data.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2076-3425/9/12/339/s1, Table S1: AURC, Table S2: SICI MEP size and SICI ratio, Table S3: SAI MEP size and SAI ratio, Table S4: LAI MEP size and LAI ratio, Table S5: Glucose, Table S6: Mean Arterial Pressure.

**Author Contributions:** Conceptualization, S.L.T., C.V.T. and A.J.N.; Data curation, S.L.T. and A.J.N.; Formal analysis, S.L.T. and C.V.T.; Funding acquisition, A.J.N.; Investigation, S.L.T., C.V.T., M.B.L., C.N. and R.R.; Methodology, S.L.T., C.V.T. and A.J.N.; Project administration, S.L.T. and A.J.N.; Resources, A.J.N.; Software, A.J.N.; Supervision, S.L.T. and A.J.N.; Validation, A.J.N.; Visualization, S.L.T., C.V.T. and A.J.N.; Writing—original draft, S.L.T.; Writing—review and editing, S.L.T., C.V.T., M.B.L., C.N., R.R. and A.J.N.

**Funding:** This research was funded in part by a Natural Sciences and Engineering Research Council of Canada grant (NSERC RGPIN-2015-06309) to AJN.

**Acknowledgments:** The authors would like to thank Gita Sobhi and Megan Jutting of the McMaster University Research Pharmacy, who provided the treatment drinks and randomization schedule for this study. We would also extend our gratitude to Diana Harasym, Patrick Dans, and Jenin El-Sayes, all of whom assisted with data collection.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*
