Neural interactions between regulatory systems for rhythmic arm and leg movements are an intriguing issue in locomotor neuroscience. Amplitudes of early latency cutaneous reflexes (ELCRs) in stationary arm muscles are modulated during rhythmic leg or arm cycling but not during limb positioning or voluntary contraction. This suggests that interneurons mediating ELCRs to arm muscles integrate outputs from neural systems controlling rhythmic limb movements. Alternatively, outputs could be integrated at the motoneuron and/or supraspinal levels. We examined whether a separate effect on the ELCR pathways and cortico-motoneuronal excitability during arm and leg cycling is integrated by neural elements common to the lumbo-sacral and cervical spinal cord. The subjects performed bilateral leg cycling (LEG), contralateral arm cycling (ARM), and simultaneous contralateral arm and bilateral leg cycling (A&L), while ELCRs in the wrist flexor and shoulder flexor muscles were evoked by superficial radial (SR) nerve stimulation. ELCR amplitudes were facilitated by cycling tasks and were larger during A&L than during ARM and LEG. A low stimulus intensity during ARM or LEG generated a larger ELCR during A&L than the sum of ELCRs during ARM and LEG. We confirmed this nonlinear increase in single motor unit firing probability following SR nerve stimulation during A&L. Furthermore, motor-evoked potentials following transcranial magnetic and electrical stimulation did not show nonlinear potentiation during A&L. These findings suggest the existence of a common neural element of the ELCR reflex pathway that is active only during rhythmic arm and leg movement and receives convergent input from contralateral arms and legs.
- cutaneous reflex
- interlimb interaction
- rhythm-generating system
functional neural coupling between the arms and legs during rhythmic movement is an important issue in motor control research and rehabilitation neuroscience. This coupling has been investigated by examining the effects of remote rhythmic limb movement on the modulation of reflex amplitude in the muscles of stationary limbs (Nakajima et al. 2013a, 2014; Sasada et al. 2010; Massaad et al. 2014; Vasudevan and Zehr 2011). We recently reported that early latency cutaneous reflexes (ELCRs; peak latency 30–70 ms) in arm muscles are facilitated by rhythmic leg cycling (Sasada et al. 2010). This modulation occurs in a muscle-, task-, and cadence-dependent manner but is absent during voluntary isometric contractions at levels equivalent to those recorded during movement. These findings are thought to result from interactions between a rhythm-generating system for the legs and ELCR pathways in the arms.
A separate but integrated rhythm-generating system is thought to be employed for each limb (Grillner 1981; Zehr and Duysens 2004; Juvin et al. 2007; McCrea and Rybak 2008; Huang and Ferris et al. 2010). Thus it is challenging to investigate the interactions of rhythm-generating systems for each limb with a given reflex pathway. Previously, we found that ELCR facilitation in arm muscles is similar during unilateral and bilateral leg cycling (Sasada et al. 2010). In addition, it was shown that the excitability of cutaneous reflex pathways within and between the arms is strongly modulated during unilateral and bilateral arm cycling (Carroll et al. 2005). Despite these findings, the integration of rhythm-generating systems responsible for the excitability of ELCR pathways in remote limbs has not been fully described in humans. Integration of synaptic outputs by common spinal interneurons has been demonstrated using intracellular recordings in cats (LaBella and McCrea 1990), while common synaptic drive to synergistic muscles has been demonstrated during human gait (Hansen et al. 2001).
Interestingly, there is a measurable interaction between neural activity regulating arm and leg movements during locomotion that is specifically enhanced with cutaneous input from the hand (Zehr et al. 2007b). Therefore, we hypothesized that the effects arising from the activation of rhythm-generating systems in the leg and contralateral arm on ELCR excitability in a given muscle may be integrated by an interneuronal system promoting motor output during locomotor activity (hypothesis 1 in Fig. 1). It is also plausible that separate outputs from rhythm-generating systems project to interneurons receiving common cutaneous afferent inputs from the hand and are subsequently integrated at the level of the segmental motoneuron (hypothesis 2 in Fig. 1).
Motor-evoked potentials (MEPs) induced by transcranial magnetic stimulation (TMS) of the leg motor area are specifically facilitated by stimulating cutaneous afferents from the foot during walking (Nielsen et al. 1997; Christensen et al. 1999). In addition, MEPs in leg muscles are modulated during walking (Capaday et al. 1999), and activity in the motor cortex suggests direct monitoring of limb muscles during human walking (Petersen et al. 2001; Barthelemy and Nielsen 2010). Thus motor outputs from the cortex could activate both segmental motoneurons and rhythm-generating systems that share last-order interneurons mediating the ELCR pathway (hypothesis 3 in Fig. 1).
To assess the presence of the neuronal system described in hypothesis 1, we measured the excitability of ELCR pathways in arm muscles during arm and leg cycling performed in isolation (contralateral arm or legs only) or in combination (contralateral arm and legs together). If the outputs of rhythm-generating systems for the arm and legs are integrated by last-order interneurons mediating the ELCR pathway (hypothesis 1), motoneuronal excitability should markedly increase when these inputs are given with appropriate timing (Baldissera et al. 1981; Fournier et al. 1986; LaBella and McCrea 1990; Pierrot-Deseilligny and Burke 2005; Nakajima et al. 2013, 2014). However, a lack of marked increase suggests that these outputs might be integrated at the level of the segmental motoneuron (hypothesis 2). It is worth noting that hypotheses 1 and 2 are not mutually exclusive because the spinal motor system likely contains abundant interactive neural networks to support locomotor activities. We also investigated MEPs following TMS in arm muscles during the three cycling tasks to test hypothesis 3. If outputs from rhythm-generating systems are integrated within the cortex or cortico-motoneuronal projections, the modulation of MEPs should be similar to that of ELCRs during combined arm and leg cycling. These experiments will clarify a premotoneuronal integration of outputs from interlimb rhythm-generating systems and cutaneous reflex pathways during locomotor activity in humans.
Subjects and Experimental Setting
Twenty-six subjects (ages 21–52 yr; 22 men and 4 women) were enrolled in the present study, which was approved by the local ethics committee of the Faculty of Education at Chiba University and provided informed consent according to the Declaration of Helsinki. All subjects were physically active almost every day, and no subject had a documented neurological deficit.
ELCRs were evoked by applying trains of rectangular electrical pulses (3 pulses at 333 Hz; 1-ms duration) to the right superficial radial (SR) nerve at the wrist using a constant current stimulator (SS-100; Nihon Kohden, Tokyo, Japan) controlled by a pulse generating system (SEM7201; Nihon Kohden). Ag/AgCl disk electrodes (φ 1 cm; Ne-101; Nihon Kohden) for SR nerve stimulation were placed on the dorsal surface of the right forearm just proximal to the radial head. The stimulation intensity was set at approximately twice the minimum stimulus intensity at which the subject could perceive the stimulus [perceptual threshold (PT)]. The intensity of electrical stimulation was initially decreased to obtain a detectable ELCR amplitude that was 2 standard deviations (SDs) above background surface electromyogram (EMG) levels during isolated arm or leg cycling. This procedure was required to avoid the saturation of neural elements receiving two different inputs and therefore enable us to determine whether neural elements integrate inputs from two different sources driven by arm and leg cycling. No subjects detected any nociceptive sensation upon application of electrical stimulation. We monitored the PT and adjusted the stimulation as needed to maintain a constant intensity throughout the experiment. To elicit the maximum M wave in the right flexor carpi radialis (FCR) muscle, the median nerve on the ventromedial side of the upper arm was stimulated with a single 1-ms rectangular pulse.
Transcranial Electrical and Magnetic Stimulation
MEPs were evoked by both transcranial electrical stimulation (TES; eMEPs) and TMS (mMEPs). TES consisted of a single pulse (100 μs; DS-180; Digitimer, Wel-wyn Garden City, UK). For TES, the anode was placed on the scalp ∼4.0 cm lateral to the vertex, while the reference cathode was placed at the vertex. TMS was applied using a figure-eight coil connected to a Magstim 200 stimulator (Magstim, Whitland, Dyfed, UK). Each circle within the figure-eight was 95 mm in diameter and generated a peak magnetic field of 2.0 T. The stimulating coil was placed over the optimal site for eliciting responses in the FCR and oriented so that the current in the brain flowed in a posterior-to-anterior direction. The active motor threshold (AMT) was defined as the minimum stimulus intensity that produced detectable motor responses in ∼50% of trials with both TMS and TES. The level of EMG activity in the right FCR was ∼3–10% of maximum EMG activity (EMGmax). At least 10 MEPs were evoked during each of the three cycling tasks at random intervals between 5 and 8 s.
Surface EMG signals were recorded from the right FCR and posterior deltoid (PD) muscles. These muscles were selected because our prior study revealed significant effects in these arm muscles during leg cycling (Sasada et al. 2010). EMG signals were also recorded from the left vastus lateralis (VL) and anterior deltoid (AD) muscles to monitor their activation levels during leg and arm cycling, respectively. EMG electrodes were placed longitudinally, 3 cm apart, over each muscle belly and fixed with surgical tape. EMG signals were amplified (×1,000; model 1206; NEC Sanei, Tokyo, Japan) and band-pass filtered at 32–3000 Hz. All signals were converted to digital data via an A/D converter system at a sampling rate of 5 kHz for later offline analysis. The data were full-wave rectified and smoothed (moving average, 5-ms time interval) with built-in software (CED 1401 interface with Spike2 software; Cambridge Electronic Design, Cambridge, UK). Visual feedback for the activation level required for steady isometric contraction of these muscles was provided by analog voltmeters (AX-313TR; Sanwa, Tokyo, Japan). Analysis of MEPs elicited by TMS and TES was conducted using unrectified FCR EMG. EMGmax was determined in the FCR, PD, AD, and VL muscles before starting the experiments by averaging a 1-s window of rectified EMG during a 3-s maximum voluntary contraction.
Single motor unit (MU) activity was recorded from the FCR muscles of six subjects using surface and concentric bipolar needle electrodes. MU potentials were amplified (×10,000), band-pass filtered at 16-10 kHz, and stored on a hard disk with a sampling rate of 10 kHz (CED 1401 interface with Spike2 software).
Each subject sat in an armchair with the right elbow flexed to 120° and wrist supinated and shoulder positioned at 90°. The subject's right arm was secured firmly to the armrest with Velcro straps. In this position, each subject was able to perform bilateral leg cycling (LEG), left arm cycling (ARM), and combined leg and arm cycling (A&L) without conscious attention to postural control while performing isometric contractions of the right arm muscle. The feet and left hand of each subject were fixed to the pedals (15-cm-long cranks) of the leg (Power Max V; COMBI, Tokyo, Japan) and arm ergometers (EU6210; Matsushita, Osaka, Japan), respectively. The position of the leg ergometer axis was carefully adjusted so that the left knee was semiflexed when the crank was at the 9-o'clock position (Fig. 2). The rotation axis of the arm ergometer was set at shoulder level, and the distance between the subject and this axis was adjusted, as with the leg ergometer. The crank position during the cycling tasks was detected by a photocell (PS-102; Coco-research, Tokyo, Japan) placed on the gear wheel. The cadence was displayed by a digital meter (TDP-3301A-E; Coco-research) connected to the photocell, so that subjects were able to actively monitor cadence. In addition, LEDs on the digital meters were set to flash when the crank passed through the 6-o'clock position. Therefore, the subjects were able to visually monitor the time lag between arm and leg cycling. Positional signals at the 6-o'clock position of both the arm and leg ergometers were monitored and recorded to evaluate cycling phase synchronicity between the arm and legs.
At the beginning of the experimental session, we measured the maximal levels of background EMG activity in the VL and AD muscles during the A&L task to determine the necessary control levels during stationary contraction (see below). Subsequently, ELCRs were evoked during focused isometric contractions of the right FCR or PD muscles at 5% of EMGmax while keeping muscles in the other limbs relaxed (Control task). SR nerve stimulation was given at least 20 times at random intervals between 0.8 and 1.5 s to evoke ELCRs. This control procedure was repeated before each of the different motor tasks (see below). For all tasks, ELCRs in the FCR were recorded while exerting 5% of EMGmax.
Before starting the cycling task, subjects practiced the LEG, ARM, and A&L tasks for several minutes with visual feedback from the digital meter and verbal feedback from experimenters on the cadence or synchrony of arm and leg movement. In the A&L task, the subjects were asked to move the arm and leg cranks at the same phase (Fig. 2); the cycling cadence was set at 60 rpm for all tasks.
In the first session, to clarify the separate effects of leg and arm cycling on ELCR excitability, we investigated the discrete effects of the LEG, ARM, and A&L tasks in 18 subjects (FCR: n = 18 and PD: n = 16). The average (±SD) phase lag during A&L cycling was 18 ± 65 ms. Additionally, to determine whether ELCR modulation was related to the cycling tasks or simply to leg muscle activation, ELCR amplitude was investigated in a subset of subjects during stationary contraction tasks in addition to the three cycling tasks (FCR: n = 10 and PD: n = 9). In addition to the LEG, ARM, and A&L tasks described above, these subjects performed isolated stationary left knee extensions (LEG stationary contraction), left shoulder flexions (ARM stationary contraction), and simultaneous knee extensions and shoulder flexions (A&L stationary contraction) during ELCR elicitation. For these isolated stationary contractions, the cranks of the left leg and arm were held at the 12-o'clock position. The activity levels of the AD and/or VL muscles during stationary contraction were set at the approximate peak EMG activity measured during the cycling tasks.
In the second session, we investigated FCR muscle ELCRs during the A&L task in greater detail. SR nerve stimulation intensity was lowered to 1.1–1.3 times PT during the LEG, ARM, and A&L tasks in nine subjects. This procedure was used to avoid saturation of the last-order interneuronal component of the ELCR pathway. In this experiment, to obtain a minimum response in the FCR, the level of isometric contraction was set at ∼3–10% of EMGmax. We also investigated changes in firing probability of a single MU in the FCR following SR nerve stimulation in six subjects during all three cycling tasks. SR nerve stimulation intensity was set at 1.05–1.25 times PT. Auditory feedback of MU firing was provided to enable subjects to maintain a constant firing rate of ∼10 Hz.
In the third session, we investigated the effects of the ARM, LEG, and A&L tasks on eMEP and mMEP amplitudes in the FCR muscles of seven subjects. For each subject, the level of isometric contraction in the FCR was maintained at ∼3–10% of EMGmax in all three cycling tasks. The stimulus intensity was set at 1.05 times AMT to obtain a minimum response in the FCR.
In the fourth session, to reveal the origin of facilitation of ELCR amplitudes during the A&L task, 10 subjects performed isolated right leg cycling (LEG-ipsi), left leg cycling (LEG-contra), and bilateral leg cycling (LEG-both) tasks. They also performed combined right leg and left arm cycling (A&L-diagonal) and combined left leg and left arm cycling (A&L-contra) tasks. The task sequence was randomized for each subject.
The first and second experimental sessions tested hypotheses 1 and 2; the third session tested hypothesis 3. The fourth experimental session was necessary to determine the combination of limbs in rhythmic movement that induces neural interaction between rhythm-generating systems. Of the 26 subjects, 18 participated in 2 or 3 sessions; 7 subjects participated in only 1 session; 1 subject participated in all four sessions. Subjects were randomly assigned to each session.
The rectified and smoothed EMG signals were centered with respect to the stimulus and averaged to obtain the ELCR for each subject in each task. The resulting EMG traces were analyzed over a 400-ms time window (100 ms prestimulus and 300 ms poststimulus). ELCR amplitudes were determined by measuring the baseline-to-peak amplitude within the preset time window from 30 to 70 ms after stimulation. The background baseline value was calculated from the mean prestimulus value. ELCR amplitudes were normalized to EMGmax for each subject.
Single MU peristimulus time histograms.
To investigate the changes in firing probability of single FCR MUs after SR nerve stimulation during the LEG, ARM, and A&L tasks, peristimulus time histograms (PSTHs; bin width = 2 ms) were constructed between 50 ms before and 150 ms following stimulation. Approximately 100 stimuli were delivered to obtain PSTHs for FCR MUs during the LEG, ARM, and A&L tasks. MU discrimination was accomplished offline with a computerized spike-sorting algorithm based on amplitude, duration, and waveform shape (Spike2, version 5; Cambridge Electronic Design). Each MU was analyzed on a spike-by-spike basis; only units that were clearly identified in all tasks were included in the analysis. Subsequently, the timing of MU firing was obtained with a window discriminator using the Spike2 software and used to construct PSTHs. Stimulus artifacts were replaced with the 50-ms prestimulus mean count. The sum of bin counts from 30–70 ms following stimulation was analyzed to compare the effects of the cycling tasks on the firing probabilities of single FCR MUs.
Transcranial stimulation and MEPs.
The raw evoked EMG was averaged over at least 10 sweeps with respect to the stimulus. These averaged EMG traces were analyzed over a 120-ms time window (20 ms prestimulus and 100 ms poststimulus). MEP amplitudes were determined by measuring the peak-to-peak amplitudes of the averaged unrectified EMG traces. The peak-to-peak amplitude was normalized to the maximum M wave (Mmax). The onset latencies of eMEPs and mMEPs were detected from the averaged EMG traces.
Differences in ELCR amplitudes and latencies across tasks (Control, LEG, ARM, and A&L) were examined by one-way repeated measures (RM) ANOVA. Two-way RM ANOVA was used to compare the moving limb (LEG, ARM, and A&L) and task conditions (stationary contraction or cycling). Multiple comparisons among tasks were conducted using the Bonferroni post hoc test. Differences between the algebraic sum of ELCR amplitudes in the LEG and ARM tasks and amplitudes on the A&L tasks were tested using the χ2-test.
Differences in eMEP and mMEP amplitudes across different tasks (LEG, ARM, and A&L) were examined using one-way RM ANOVA. Differences in onset latencies between eMEPs and mMEPs were examined using the Student's t-test.
Differences in ELCR amplitudes between leg cycling laterality (ipsi vs. contra) and moving limbs (A&L vs. LEG) were tested by two-way RM ANOVA. Differences between the moving limbs during cycling tasks for each side were determined using the Student's t-test. Significant differences in amplitudes among the LEG-ipsi, LEG-contra, and LEG-both tasks were also evaluated by one-way RM ANOVA. The Student's t-test was used to determine significant differences from the Control task.
F value significance was assessed after a Greenhouse-Geisser correction, when appropriate; subsequently, the correction coefficient epsilon was determined. The level of statistical significance was set at P < 0.05.
Cutaneous Reflex Latencies
The ELCR onset latencies in the FCR and PD muscles were ∼36–38 and 33–36 ms, respectively. The ELCR peak latencies in the FCR and PD muscles were ∼52–56 and 52–59 ms, respectively. There were no significant differences in onset and peak latencies across tasks in the FCR [onset: F(3,51) = 0.929, P > 0.05; peak: F(3,51) = 1.245, P > 0.05] or PD [onset: F(3,45) = 1.993; P > 0.05]. However, a significant difference in ELCR peak latency between the LEG and A&L tasks was observed [F(3,45) = 3.399, P < 0.05; Table 1].
Task-Dependent Modulation of Cutaneous Reflexes
Figure 3A shows representative recordings of cutaneous reflexes in the FCR and PD muscles following SR nerve stimulation with two times PT during the Control, LEG, ARM, and A&L tasks. ELCRs (peak latency of ∼50 ms) were greater during all the cycling tasks than in the Control task. Although the increases in ELCR amplitudes during the LEG and ARM tasks were small, this effect was enhanced during the A&L task for both muscles. Figure 3B shows the group data for ELCR amplitudes during the LEG, ARM, and A&L tasks. One-way ANOVA showed that the averaged ELCR amplitudes during the A&L task were significantly larger than those during the Control, LEG, and ARM tasks for both the FCR and PD [FCR: F(3,51) = 22.500, P < 0.001; PD: F(3,45) = 16.755, P < 0.001]. Background EMG was well maintained and there was no significant difference across tasks [FCR: F(3,51) = 1.448, P > 0.05; PD: F(3,45) = 1.547, P > 0.05].
To clarify the importance of limb movement for ELCR modulation, we compared ELCR amplitudes during stationary isometric contractions and rhythmic arm and/or leg cycling. As shown in Figure 3C, ELCR amplitudes during the cycling tasks were markedly larger than amplitudes during stationary contraction tasks for both muscles. Two-way ANOVA showed a significant main effect of task conditions for both muscles [FCR: F(1,9) = 35.084, P < 0.001; PD: F(1,8) = 8.968, P < 0.05]. In addition, there were interactive effects of moving limbs and task conditions for both muscles [FCR: F(2,18) = 9.600, P < 0.001; PD: F(2,16) = 7.392, P < 0.01]. Furthermore, ELCR amplitudes during the A&L task were significantly larger than during the LEG or ARM tasks [two-way ANOVA, main effects of moving limbs in FCR: F(2,18) = 8.823, P < 0.01; PD: F(2,16) = 4.972, P < 0.05]. In contrast, stationary isometric contractions failed to modulate ELCR amplitudes.
To more clearly examine the combined effects of arm and leg movement on ELCR amplitude during the A&L task, we conducted additional experiments in which the intensity of SR nerve electrical stimulation was decreased to obtain a minimal response in the FCR muscle. Although small responses were observed during the LEG (Fig. 4Aa) and ARM (Fig. 4Ab) tasks, a large excitatory effect was detected during the A&L task (Fig. 4Ac). ELCRs during the A&L task (5.37 ± 5.6%) were larger than the summated EMG traces of the LEG and ARM tasks (3.2 ± 3.5%), suggesting a nonlinear integration. The group data show that ELCR amplitudes during the A&L task were significantly larger than the algebraic sum of the effects during the LEG and ARM tasks (Fig. 4B; χ2 test, P < 0.05).
A large excitatory effect on the firing probability of FCR single MUs was observed during all three cycling tasks. Figure 5 shows PSTHs of single MU activity during the LEG (Fig. 5A), ARM (Fig. 5B), and A&L tasks (Fig. 5D). The algebraic sum of PSTHs from the LEG and ARM tasks was also calculated (Fig. 5C). The firing probability during the 30–70 ms after the stimulus was greater in the A&L task than in the LEG and ARM tasks. PSTH peaks from the A&L task were larger than those of the summed PSTH from the LEG and ARM tasks (Fig. 5C). In 15 of the 27 MUs recorded from six subjects, the cumulative sum of bins during the A&L task was larger than the algebraic sum during the ARM and LEG tasks. The averaged latency of the peak counts in these 15 MUs was 49 ± 9 ms, which corresponded to the range of ELCR peak latencies recorded by surface EMG (Table 1). The level of background surface EMG activity in the FCR muscle was 2–3% of EMGmax.
Modulation of FCR MEPs During the ARM, LEG, and A&L Tasks
Figure 6 illustrates the differences in eMEP and mMEP amplitudes during the ARM, LEG, and A&L tasks in a single subject. The stimulation intensity was 1.05 × AMT in both cases. The grand average onset latencies of eMEPs and mMEPs were 12.2 ± 0.5 ms and 13.9 ± 0.9 ms, respectively (P < 0.001). In contrast to the above results for ELCR modulation, eMEP and mMEP amplitudes were maintained across the three cycling tasks. This result was confirmed by the group data, which showed no significant differences in eMEP and mMEP amplitudes or latencies across the three cycling tasks [amplitude: eMEP F(2,12) = 0.14, P > 0.05; mMEP: F(2,12) = 0.478, P > 0.05; and latency: eMEP F(2,12) = 0.29, P > 0.05; mMEP F(2,12) = 0.28, P > 0.05]. There were no significant differences in eMEP or mMEP amplitudes during the A&L task compared with the algebraic sum of amplitudes during the ARM and LEG tasks.
ELCR Laterality During Cycling
Finally, we assessed all possible combinations of limb cycling on ELCR amplitude. Figure 7 shows schematic illustrations of the cycling tasks (Fig. 7A) and the pooled ELCR amplitude data (mean ± SD) obtained during the LEG (gray bars) and A&L tasks (black bars). ELCR amplitudes in both FCR and PD muscles were significantly larger during all cycling tasks than during the Control task. However, ELCR amplitudes increased when arm cycling was performed in conjunction with cycling of either leg compared with leg cycling alone. This pattern was observed in all but one of the subjects. Two-way ANOVA revealed a main effect for moving limbs (A&L vs. LEG) in both muscles [FCR: F(1,9) = 12.692, P < 0.01; PD: F(1,9) = 6.611, P < 0.05]. There were no significant differences in amplitudes according to leg cycling laterality regardless of whether leg cycling was combined with arm cycling [FCR: F(1,9) = 0.163, P > 0.05; PD: F(1,9) = 0.355, P > 0.05]. There were also no significant differences in amplitudes among the LEG-both, LEG-ipsi, and LEG-contra tasks [FCR: F(2,18) = 0.337, P > 0.05; PD: F(2,18) = 0.371, P > 0.05].
In the present study, combined rhythmic arm and leg cycling significantly increased cutaneous reflex amplitudes in arm muscles. This facilitatation was also observed in the firing probability in 15 of 27 single MUs. Moreover, we did not find nonlinear increases in transcranial mMEP or eMEP amplitudes. These results support hypothesis 1, which predicts a neural system that integrates and amplifies inputs from low-threshold cutaneous afferents and rhythm-generating systems for both arms and legs.
Interaction of Neural Systems Responsible for Rhythmic Arm and Leg Cycling
We found that ELCR amplitudes in arm muscles were significantly higher during the A&L task than during the individual ARM or LEG tasks. This result indicates facilitated transmission of afferent feedback that is most likely to be explained by coupling between neural systems that regulate rhythmic arm and leg cycling. Interlimb neural coupling during rhythmic movements in humans has been observed in several previous studies (Wannier et al. 2001; Huang and Ferris 2004; Sakamoto et al. 2007; Mezzarane et al. 2011) and seems to depend strongly on the motor task (Dietz et al. 2001; Haridas and Zehr 2003; Sakamoto et al. 2006; Balter and Zehr 2007; Klimstra et al. 2009; Nakajima et al. 2013a, 2014). Interlimb neural coupling is an absolute necessity in quadrupedal animals for the functional coordination of all four limbs needed to cope with the physical demands of the environment during locomotion (Miller et al. 1975; Juvin et al. 2005, 2007). In humans, hand dexterity and the degrees of freedom in the arms were enhanced by the emergence of an upright stance and bipedal locomotion (Dietz 2002; Isa et al. 2007). Thus strict interlimb coupling during rhythmic arm and leg movements may not be necessary for human locomotion (Gysin et al. 2008); in fact, a “loose” coupling has been suggested (Zehr et al. 2009). Functional neural coupling for interlimb coordination may work to optimize efficiency for a variety of locomotor movements, depending on biomechanical constraints (Zehr et al. 2007a; Klimstra et al. 2009; Huang and Ferris 2010). The task dependency of interlimb coupling may reflect the functional evolution of human arm use as it relates to the acquisition of bipedal locomotion (Zehr et al. 2009).
Modulation of ELCR Amplitude During Isolated and Combined Arm and/or Leg Cycling
The neural organization of interlimb coupling can be investigated by monitoring reflex modulation during locomotor movements, as originally established in animal experiments (Burke 1999; McCrea and Rybak 2008). Cutaneous reflexes are susceptible to modulation by locomotor rhythm-generating systems, as demonstrated by strong modulation of cutaneous reflex magnitude during locomotion (Forssberg et al. 1977; Moschovakis et al. 1991; Seki and Yamaguchi 1997). Human locomotor studies also show that the spinal Hoffmann reflex (H reflex) in moving and stationary limbs is strongly modulated during rhythmic movement compared with voluntary movement (Capaday and Stein 1986; Brooke et al. 1997; Frigon et al. 2004; Nakajima et al. 2011, 2013b), demonstrating convergent effects of arm and leg activation (Mezzarane et al. 2011).
Here, we demonstrated that ELCR magnitudes in arm muscles are significantly larger during the A&L task than during the isolated ARM or LEG tasks. Importantly, ELCR amplitudes during limb movement are significantly larger than those seen during stationary isometric contractions, despite similar EMG activity (Fig. 3C). This result shows that functional interlimb neural coupling networks are activated during rhythmic limb movement.
We recently demonstrated that rhythmic leg cycling significantly enhanced ELCR amplitudes in arm muscles; this enhancement was absent during tonic activation (Sasada et al. 2010). Carroll et al. (2005) showed that cutaneous reflexes in arm muscles are modulated in a task-dependent manner during cycling of the contralateral arm. Based on these results, one can expect that, if interlimb coupling is regulated by rhythm-generating systems, and if there is a regulatory system that integrates convergent inputs arising from distinct sources, then arm muscle ELCRs should increase nonlinearly when arm and leg cycling is performed. We tested this hypothesis in the present study and found that ELCR amplitudes showed significant and nonlinear increases during combined contralateral arm and leg cycling (i.e., the A&L task; Figs. 4 and 5). Balter and Zehr (2007) demonstrated that amplitude modulation of the middle latency cutaneous reflex in the tibialis anterior muscle during arm cycling was increased during arm and leg cycling. This result confirms that the excitability of cutaneous reflexes in leg muscles is not only regulated by the neural circuitry for leg movements but also integrates inputs from arm movements (Balter and Zehr 2007). The current findings in arm muscles (Figs. 3 and 7) are in agreement with these earlier observations.
Increased ELCR amplitudes during A&L tasks were also observed during combined single-leg and single-arm cycling (A&L-diagonal and A&L-contra), a result that was absent during bilateral leg cycling (LEG-both). Thus we speculate that ELCR amplitude increases are specific to the combination of arm and leg cycling rather than the number of moving limbs. This suggests that ELCR excitability in arm muscles reflects the convergent activity of neural circuits regulating rhythmic movements (Zehr et al. 2007b).
Nonlinear Increases in Cutaneous Reflex Amplitudes During the A&L Task
The ELCR is mediated via polysynaptic pathways (Jenner and Stephens 1982), and the modulation of cutaneous reflexes during rhythmic movements is thought to be regulated at the premotoneuronal level (De Serres et al. 1995; Duysens and Van de Crommert 1998; Zehr and Duysens 2004). ELCR amplitudes increased nonlinearly during the A&L task compared with the algebraic sum of amplitudes during the ARM and LEG tasks. This observation is consistent with premotoneuronal mechanisms proposed by hypothesis 1 (see Figs. 4 and 5). Nonlinear facilitation of reflex amplitudes was previously demonstrated using a spatial facilitation technique, which delivers two distinct inputs that converge presynaptically at the motoneuronal level, in reduced animal preparations (Baldissera et al. 1981; LaBella and McCrea 1990) and humans (Fournier et al. 1986; Pierrot-Deseilligny and Burke 2005; Nakajima et al. 2013, 2014).
Alternatively, spatial or temporal summation may occur at the level of the motoneuron pool, as described in hypothesis 2. This may occur if subthreshold excitation of motoneurons is due to inputs from SR nerve stimulation and/or rhythm-generating systems responsible for arm and leg cycling. If there were two distinct pathways from rhythm-generating systems for ARM and LEG, and they were independently inserted in the ELCR pathway, their outputs would be integrated at the motoneuronal level. In this case, the amount of ELCR facilitation during A&L tasks should equate with those of the simple summation during ARM and LEG tasks (Burke et al. 1970; Burke 1981; Henneman and Mendell 1981; Fournier et al. 1986; Forget et al. 1989; Kernell and Hultborn 1990; Capaday 1997; Klimstra and Zehr 2008). To effectively test this hypothesis, we had to avoid the potential confounds of unobservable subthreshold motoneuronal excitation. Thus, we examined changes in firing probability of a single MU following SR electrical stimulation during the LEG, ARM, and A&L tasks. The spatial facilitation technique has been validated using the PSTH method, which reflects changes in the discharge rate of a single motoneuron after stimulation (Pauvert et al. 1998). As shown in Fig. 5, we found a nonlinear increase in the firing probability of a single MU following SR stimulation during the A&L tasks within a time interval corresponding to ELCR latency, while background firing rate was nearly equal across the three cycling tasks. These results suggest that the summation of locomotor inputs arising from rhythmically moving arms and legs may be integrated by common interneuronal reflex circuits (hypothesis 1 in Fig. 1). MU recordings should reflect a limited number of MUs recruited during the tasks, whereas surface EMG activity reflects compound action potentials from multiple MUs underneath the recording electrodes. These findings may represent the first indication that some recruited MUs receive inputs from common interneurons that integrate cutaneous afferent input from the hand with regulatory control of rhythmic arm and leg movements. The nonlinear increase in firing probability of single MUs during the A&L task is consistent with activity in putative common neural elements converging on the ELCR reflex pathway at the premotoneuronal level (cf. Nakajima et al. 2014).
ELCR amplitudes following low-intensity electrical stimulation clearly showed nonlinear facilitation during A&L (see Fig. 4), implying that this mechanism is highly dependent on stimulation intensity. These findings suggest that interneurons in the ELCR pathways receive strong excitatory inputs from rhythm-generating systems for both the arms and legs. Thus high-intensity conditioning stimulation could produce saturation (see Fig. 3). Because the stimulus intensity was carefully adjusted to avoid this ceiling effect, nonlinear facilitation could be detected (Fig. 4). It is also possible that lower stimulus intensities may selectively activate large-diameter afferent fibers with specific modalities and restricted connectivity. On the other hand, integration of outputs from rhythm generating systems at the level of the motoneuron (such as in hypothesis 2) could partly account for the facilitation of ELCR with high-intensity stimulation (Fig. 3 and Fig. 7). While it may be possible to mutually reconcile hypothesis 1 and 2 (as described in the Introduction), the current study placed more importance on testing hypothesis 1. Thus we cannot deny the possibility that there is a neural system such as that described in hypothesis 2 and such possibilities require further investigation.
As shown in Fig. 6, mMEP and eMEP amplitudes during the A&L task were similar to those during the LEG and ARM tasks. Neither additional facilitation nor latency modulation was observed across the three cycling tasks. eMEPs following anodal electrical stimulation to the motor cortex at the active motor threshold are reportedly induced by activation at the axon hillock or a site just proximal to the axons of pyramidal neurons. The action potential then activates motoneurons (Burke et al. 1990; Rothwell et al. 1994; Di Lazzaro et al. 1998). Thus a change in eMEP amplitude is thought to reflect subcortical excitability (Taylor 2006). In contrast, low-intensity TMS of the motor cortex indirectly activates pyramidal neurons, with the size of the evoked potential reflecting cortical excitability (Amassian et al. 1987; Day et al. 1989). The absence of amplitude modulation of eMEPs and mMEPs during the ARM, LEG, and A&L tasks suggests that the rhythm-generating systems for the arms and legs are not influenced by cortical and subcortical excitability. Thus the nonlinear summation of ELCR amplitudes cannot be explained by an integrative function of the motor cortex and the cortico-motoneuronal system (hypothesis 3 in Fig. 1).
We found no significant differences in ELCR amplitudes across the LEG-both, LEG-ipsi, and LEG-contra tasks (Fig. 7). This finding is inconsistent with the notion that rhythm-generating systems for each limb interact to modulate the excitability of reflex pathways in a stationary remote limb (Carroll et al. 2005; Balter and Zehr 2007; de Ruiter et al. 2010; Sasada et al. 2010). Strong neural coupling between the left and right legs has been reported previously (Cheng et al. 1998; Zehr and Duysens 2004; Musselman and Yang 2007). Although the current findings are consistent with previous reports regarding neural coupling between arms and legs (Balter and Zehr 2007; Sasada et al. 2010), we did not find any nonlinear convergent effects on ELCR amplitude during rhythmic right and left leg cycling. It is possible that ELCR pathway sensitivity was insufficient to detect neural coupling between the rhythm-generating systems for the right and left leg. Alternatively, the ELCR pathway may simply reflect the combined activity of rhythm-generating systems for both legs. Our data do not allow us to clearly differentiate between these possibilities.
In conclusion, the current study demonstrates that a premotoneuronal locus integrates inputs from interlimb rhythm-generating systems in human cutaneous reflex pathways during locomotor activity. This finding adds to evidence recently obtained with a similar approach using H-reflexes as the test parameter (Mezzarane et al. 2011). Together, these observations contribute to the growing body of evidence that rhythmic arm and leg movements are functionally integrated during locomotion with characteristics that are reminiscent of quadrupedal interlimb linkages.
The research was supported by Grants-in-Aid for Scientific Research from the Japan Society for the Promotion of Science No. 22500525 (to T. Komiyama) and No. 24700579 (to S. Sasada).
The funding organization had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
No conflicts of interest, financial or otherwise, are declared by the author(s).
Author contributions: S. Sasada, T.T., T.N., and T.K. conception and design of research; S. Sasada, T.T., T.N., G.F., H.O., S. Suzuki, and T.K. performed experiments; S. Sasada analyzed data; S. Sasada, T.T., T.N., E.P.Z., and T.K. interpreted results of experiments; S. Sasada prepared figures; S. Sasada and T.N. drafted manuscript; S. Sasada, T.T., T.N., H.O., S. Suzuki, E.P.Z., and T.K. edited and revised manuscript; S. Sasada, T.T., T.N., G.F., H.O., S. Suzuki, E.P.Z., and T.K. approved final version of manuscript.
- Copyright © 2016 the American Physiological Society