abductor pollicis brevis
contingent negative variation
common peroneal nerve
movement execution phase
movement-related cortical potential
paired associative stimulation
- • Here we use the naturally generated brain activation when a person imagines a simple movement and combine this with the afferent inflow that would be generated had the movement been performed rather than imagined.
- • We show that the excitability of the neural projections connecting the relevant brain areas to the target muscle is increased only when the afferent inflow arrives during the highest activation phase.
- • Furthermore, the changes are specific to the task and the neural connections between the brain and muscle involved in the task.
- • This novel intervention will open up the possibilities to alter afferent-generated feedback depending on the demands of the movement to be performed.
Abstract In monkeys, the repeated activation of somatosensory afferents projecting onto the motor cortex (M1) has a pivotal role in motor skill learning. Here we investigate if sensory feedback that is artificially generated at specific times during imagination of a dorsiflexion task leads to reorganization of the human M1. The common peroneal nerve was stimulated to generate an afferent volley timed to arrive during specific phases of the cortical potential generated when a movement was imagined (50 repetitions). The change in the output of M1 was quantified by applying single transcranial magnetic stimuli to the area of M1 controlling the tibialis anterior muscle. The results demonstrated that the concomitance between the cognitive process of movement (motor imagination) and the ascending volley due to the peripheral nerve stimulation can lead to significant increases in cortical excitability. These increases were critically dependent on the timing between the peripherally generated afferent volley and the cortical potential generated during the imagined movement. Only if the afferent volley arrived during the peak negative deflection of the potential, were significant alterations in motor cortical output attained. These results demonstrate that an artificially generated signal (the peripheral afferent volley) can interact with a physiologically generated signal in humans leading to plastic changes within the M1, the final output stage for movement generation within the human brain. The results presented may have implications in systems for artificially inducing cortical plasticity in patients with motor impairments (neuromodulation).
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Long-term potentiation (LTP) is one proposed mechanism for inducing synaptic plasticity thought to underlie processes of memory and learning (Bliss & Collingridge, 1993; Cooke & Bliss, 2006; Letzkus et al. 2007). One form of LTP, associative LTP, is based on Hebbian theory, according to which synapses that experience correlated activation of two converging inputs are strengthened whereas those weakened by uncorrelated activity are lost (Hebb, 1949).
Animal experiments demonstrated that induction of LTP is enhanced in regions of increased excitability following lesions (Hagemann et al. 1998). In humans, several techniques have been proposed to induce associative LTP within the motor cortex (M1; for review see Ziemann et al. 2008). During the execution of a movement, the excitability of M1 is increased (Lazzaro et al. 1998). Evidence supports the notion that associative LTP can be more readily induced when M1 is transiently disinhibited as during a voluntary muscle contraction (Kujirai et al. 2006; Mrachacz-Kersting et al. 2007; Thabit et al. 2010).
Voluntary movements are associated to slow negative cortical potentials, the movement-related cortical potentials (MRCPs). If a warning stimulus is provided followed by a cue to start the movement the MRCP is generally termed the contingent negative variation (CNV). The CNV, first described by Walter et al. (1964), is comprised of several well-defined parts, which are known to be linked to specific neurophysiological mechanisms. The first deflection, the early CNV, typically commences immediately following the warning stimulus (Hamano et al. 1997). The late CNV appears 1 to 0.5 s prior to the cue to perform the movement (Hamano et al. 1997). It is associated with the planning of voluntary movements and is altered in neurological disease (Ikeda et al. 1997). During the movement execution and after the cue, a more complex waveform may be observed, thought to be related to somatosensory feedback or alternatively to the attention level during task execution (Hamano et al. 1997). Although the exact generators of the CNVs are not known, these potentials are modulated by the way a task is executed since their characteristics change with variations in the force and speed of the executed or imagined movement (do Nascimento et al. 2005a,b, 2006; do Nascimento & Farina, 2008; Gu et al. 2009).
Since the CNVs contain information on movement planning and execution, they may be used to trigger the artificial activation of somatosensory afferents projecting onto M1. Repetitive electrical stimulation leads to increases in excitability of the cortical projections to the target muscle that are accompanied by significant functional improvements (e.g. walking speed in stroke patients) (Everaert et al. 2010). The repeated activation of somatosensory afferents projecting onto M1 has a pivotal role in motor skill learning in monkeys (Pavlides et al. 1993). Here we describe a conditioning protocol that uses the CNV generated by a cue-driven imagined movement and a single peripheral electrical stimulus timed to arrive at specific components of the CNV to induce associative LTP. LTP induction is assessed by exploring changes in the excitability of the cortical projections to the target muscle using single-pulse transcranial magnetic stimulation.
Six experiments (one reliability experiment, one main experiment and four control experiments) were con-ducted in a total of 24 subjects (15 male, 9 female; age: 24.9 ± 4.9 years) with no prior history of neurological conditions. All procedures were in accordance with the Declaration of Helsinki and approved by the Scientific Ethics Committee of Northern Jutland (reference number: VN-20100067). Subjects gave their written consent prior to participation.
Contingent negative variation (CNV)
Monopolar EEG signals were recorded by an EEG amplifier (EEG100c, Biopac system, USA) and acquired by an acquisition card (National Instruments, NI6122) at 1024 Hz (0.1–100 Hz). Ag–AgCl scalp electrodes were placed according to the International 10–20 system in the positions C1, Cz, C2 and CPz. The right ear lobe was used as a reference and the ground electrode was placed at the nasion.
To investigate the reliability of the recorded CNVs over time, 10 subjects (5 male, 5 female; age: 25.3 ± 2.4 years) attended three separate sessions on day 1, 7 and 21, respectively (‘CNV reliability experiment’). In a separate set of experiments (‘Intervention experiment’), 9 subjects (5 male, 4 female; age: 27 ± 6.9 years) participated in one baseline session (determining the peak negativity of the CNVs) before proceeding to three intervention sessions separated by at least 2 days (range 2–9 days), as described under the section ‘Interventions’. For each subject, the CNV was recorded during the imagination of a right ankle dorsiflexion. A custom-made Matlab script provided visual information via a screen positioned 2 m in front of the subject on when to mentally prepare, execute and release the movement. Each trial consisted of five time periods, namely focus, preparation, execution, hold and rest period (Fig. 1). When the cursor reached the ramp (task period), the subjects had to imagine the kinaesthetics of a ballistic (as fast as possible) dorsiflexion and hold this for 2 s before resting prior to the subsequent trial. The time period of the focus phase was randomized between 2–3 s for each trial, while the preparation phase remained fixed at 2 s. The holding phase was set to 2 s and the inter-trial rest time was randomized between 4 to 5 s. The timing of these phases was chosen empirically. A total of 50 imaginary dorsiflexion movements were recorded in two 25-trial sets, with a 1–2 min rest period between each set.
Prior to the CNV recording, every subject was instructed to perform 25 ballistic dorsiflexion movements with the right foot following the same procedure as above. The reason of this was to train the subjects in performing ballistic dorsiflexion movements uniformly in relation to the visual cue and build up their strategy for the motor imagery task.
Feature extraction from CNV
The EEG data containing the CNVs were divided into epochs of 4.5 s (from 2 s before to 2.5 s after the visual cue) for each imaginary movement. The root-mean-square (RMS) values for all epochs were calculated. Any epoch containing activation of the tibialis anterior (TA) muscles (more than 3× the resting RMS value) were discarded, as were all epochs containing eye-movement artifacts. EEG signals were band-pass filtered from 0.05 Hz to 10 Hz, and subsequently wavelet denoising was applied (wavelet: ‘db4’) to smooth the CNV in each epoch. Next a window of 500 ms on either side of task onset was chosen. If any epoch's peak negativity was outside the selected window it was discarded. The time of occurrence for the minimum values of the remaining epochs were subsequently plotted in a histogram. Based on these remaining epochs the mean peak negativity (PN) was defined as the time of occurrence for the minimum value of the averaged CNV in relation to the visual cue. The mean PN and standard deviation (SD) were used to calculate the points in time for when to apply the peripheral stimulation in the intervention sessions described below.
Recording and stimulation
Surface electrodes (20 mm Blue Sensor Ag–AgCl, AMBU A/S, Denmark) were used to record the electromyographic (EMG) activity of the TA muscle of the right leg for all aspects of the experiments. The electrodes were placed in accordance with the recommendations of Garland et al. (1994). All data were sampled at a frequency of 4 kHz. The EMG signals were amplified and band-pass filtered at 20 Hz–2 kHz.
A Magstim 200 (Magstim Company, Dyfed, UK) with a focal figure-of-eight double-cone coil (110 mm diameter) was used to apply single pulses to elicit a motor-evoked potential (MEP) in the TA. The direction of the current was directed from posterior to anterior.
Peripheral nerve stimulation was performed during the interventions. Stimulation of the common peroneal nerve (CPN) was applied using a NoxiTest isolated peripheral stimulator (IES 230). Stimulating electrodes (32 mm, PALS Platinum, Patented Conductive Neurostimulation Electrodes, Axelgaard Manufacturing Co., Ltd, USA) were placed on the skin overlying the deep branch of the right common peroneal nerve (CPN – L4 and L5) with the cathode proximal. A suitable position for stimulation, defined as the site where a maximal M-wave was produced in the TA with no activity from the synergistic peroneal muscles and no activity from the antagonist soleus (Sol), was identified. Palpation of Sol and peroneal muscles was performed during stimulation trials to ensure that this was occurring. The stimulation site corresponded to a point just anterior to the level of the caput fibulae. The pulse width was 1 ms and the intensity 1 × motor threshold.
During all experiments described below, subjects were seated in a chair (hip 90 deg, knee 130 deg) with their right and left foot resting on separate footplates. Initially the intensity for the magnetic stimulation was set at approximately 50% of the stimulator output (SO) to find the optimal site for evoking a MEP in the TA. The best spot for stimulation (also termed the hot-spot) was defined as the coordinate where the peak-to-peak amplitudes of the MEPs were greater in the target muscle than the amplitudes of adjacent coordinates for a given stimulus intensity. For all subjects, this site was approximately 2–3 cm anterior to the vertex and a stimulation applied to this area also evoked a response in the Sol. Once the hot-spot was identified, it was marked to ensure that the coil position was maintained so that the stimulation was always applied over the same area of M1.
Subsequently, the resting threshold (RTh), defined as the highest stimulus intensity that produced no more than 5 of 10 consecutive TA MEPs with a peak-to-peak amplitude of ∼50 μV while the muscle was at rest, was identified. Next, 12 MEPs were elicited in the resting TA at five TMS intensities; 90, 100, 110, 120 and 130% of RTh. The TMS stimuli were delivered randomly every 5–7 s in a randomized order.
The intervention protocols consisted of a single electrical stimulation of the common peroneal nerve de-livered at motor threshold (MT) timed to arrive during (a) the late CNV (CPN+lateCNV), (b) the movement execution phase (CPN+ME) or (c) the holding phase (CPN+HP) (Fig. 2). The timing was calculated according to the following formulae: mean peak negativity – 50 ms (CPN+ME); mean peak negativity – 50 ms – SD (CPN+lateCNV); mean peak negativity – 50 ms + SD (CPN+HP). The 50 ms represent the average latency for the peripheral stimulus to reach the somatosensory cortex plus a cortical processing delay and is based on previous work (Mrachacz-Kersting et al. 2007). The resulting timing was subsequently rounded up/down to the nearest 50 ms due to a limitation of the computer algorithm.
A total of 50 pairings (every 10–12 s) were applied in two sets of 25 trials in nine subjects. The mean peak-to-peak TA MEP amplitude measured prior to and following each intervention was plotted against TMS intensity. This relation was fitted with the Boltzmann sigmoidal function by the Levenberg–Marquard non-linear, least-mean-squares fit, as previously described (Devanne et al. 1997).
Control experiment 1–3: imagery alone, specificity and effects of the visual input
Three control experiments were performed. In eight subjects (3 male, 5 female; age: 23.3 ± 3.5 years), the intervention performed consisted of only the imagined movement (every 10–12 s) to quantify the effects of imagination alone (‘Control experiment 1’). In nine subjects (7 male, 2 female; age: 24.2 ± 2.4 years), the tibial nerve was stimulated at MT, timed so that the afferent volley generated arrived during the movement execution phase (as for the CPN+ME intervention; 50 pairs, 10–12 s apart) (‘Control experiment 2’). This was to establish the specificity of the protocol. In the third control experiment (‘Control experiment 3’), four subjects (2 male, 2 female; age: 22 ± 2.7 years) were asked to simply look at the screen, which contained the same display of information as for the main interventions. The electrical stimulus was applied to the CPN at the same time as during the CPN+ME intervention. The purpose of this last control experiment was to exclude the effect of the visual cue as possibly altering the excitability of the cortical projections to the TA. In all the control experiment the subjects were naïve to the protocol.
Control experiment 4: stretch reflex recording
It is not possible to differentiate if alterations in the MEP are due to changes in spinal or cortical circuitry. For this purpose, TA stretch reflexes were elicited prior to and following the CPN+ME intervention in four subjects (1 male, 3 females; age: 22.5 ± 2.5 years). The right leg was affixed to a servo-controlled hydraulic actuator (MTS-systems Corporation, 215.35; Voigt et al. 1999), such that the anatomical ankle axis of rotation was aligned with the fulcrum of the actuator. The foot segment of the right leg of the subject was firmly strapped to a custom-made plate that extended from the actuator, thus producing a tight interface between the arm of the motor and the foot of the subject, ensuring that the movement of the actuator was transmitted solely to the ankle joint. The angular position of the actuator was monitored by an angular displacement transducer (Transtek, DC ADT series 600). The subjects were asked to produce three maximum voluntary contractions (MVC) of the TA, separated by 3 min of rest. The greatest of the three MVC forces was used as the reference MVC. The root mean square value of the rectified TA EMG for the MVC over a 1 s period was calculated. Subsequently, the subjects were provided with visual feedback via a computer screen displaying horizontal markings set at 5% MVC and a vertical bar displaying the subject's current level of TA activation. Subjects were asked to maintain the bar between the horizontal markings while the perturbations were applied without interfering with the imposed plantarflexion perturbations.
Thirty stretches were randomly applied at intervals ranging from 5 to 7 s (velocity: 100–200 deg s−1; amplitude: 4–6 deg; hold-time: 460 ms). The angular velocity and the amplitude of the imposed perturbations were adjusted for each subject prior to the intervention so that the amplitude of the three response peaks observed in the TA EMG trace were approximately the same and also similar to the amplitude of the TA MEP prior to the intervention. The latency of the first response peak (termed M1 or alternatively short latency reflex (SLR) in the literature) were extracted from the data both prior to and immediately following the intervention. The root mean square (RMS) value of a window extending 10 ms on either side of the SLR was calculated and used as an indication of the size of this component of the TA stretch reflex.
Since it is possible that the motor units targeted by an imposed stimulus (mechanically or electrically) differ when a muscle is relaxed compared with when it is contracted, it was necessary to record the TA MEP during a contraction of the TA. For this, the active threshold (ATh) was identified as the highest stimulus intensity that produced no more than 5 out of 10 consecutive TA MEPs with an amplitude of ∼100 μV while the muscle was contracting at 5% of its maximum voluntary contraction. Prior to and following the CPN+ME intervention, 12 MEPs were recorded during a 5% MVC contraction of the TA. The TMS intensity was chosen based on the amplitude of the rectified MEP being matched to the amplitude of the M1 component of the stretch reflex prior to the intervention.
A one-way repeated-measures ANOVA with the factor ‘day’ (1, 1; 2, 7; 3, 21) was used to establish the repeatability of the onset and occurrence of peak negativity of the MRCP. A three-way repeated-measures ANOVA with factors ‘time’ (1, pre; 2, post), ‘intervention’ (1, CPN+lateCNV; 2, CPN+ME; 3, CPN+HP) and ‘stimulus intensity’ (1, 90%; 2, 100%; 3, 110%; 4, 120%; and 5, 130% RTh) was used to investigate the effects of the three main interventions (CPN+lateCNV, CP+ME and CP+HP) on the changes in the MEP amplitude. Paired t tests with Bonferroni correction were used to assess changes in the parameters of the input–output curve (the stimulus intensity required to obtain a response that is 50% of the maximum (S50), the slope and the r2 value). A two-way repeated-measures ANOVA with factors ‘time’ (1, pre; 2, post) and ‘stimulus intensity’ (1, 90%; 2, 100%; 3, 110%; 4, 120%; and 5, 130% RTh) was used to assess the changes in MEP amplitude for the three control experiments. Each analysis was performed separately for each control experiment. The Student's t test was used to investigate possible difference in the pre- and post-stretch reflex data. For all experiments, statistical significance was set to P < 0·05.
Reliability of CNV measures
As CNVs were used to identify the instant of time to trigger peripheral stimulation, it was essential to analyse the reliability of their features when measured over different days. Figure 3 shows the average CNV for all subjects on day 1, 7 and 21. Across all subjects the average onset was −1630 ± 309 ms and the PN occurred at −15 ± 217 ms. On average seven trials were rejected for all the subjects. There was no significant difference in the onset (F(2,27)= 0.31; P = 0.60) or the time of PN (F(2,27)= 0.52; P= 0.74) of the CNV.
Interventions: dependence on timing of peripheral stimulus
The effectiveness of the intervention protocols was quantified by assessing changes in the input–output properties of the TA MEP prior to and following each intervention in a total of nine subjects. The TA MEP was elicited at five intensities of TMS, ranging from 90% to 130% of RTh. A Boltzmann fit was applied to the extracted peak-to-peak TA MEP curve. Figure 4A shows the averaged raw data as well as the fitted data for one subject prior to and following each intervention. For this subject, the peak negativity of the CNV occurred at 119 ± 229 ms (Fig. 4B). In the figure, the onset of the imagined task is also indicated (vertical dashed line) and assumed as time zero by convention. The arrows in the figure indicate when the signal from the peripheral nerve stimulation arrives in relation to the CNV based on our assumptions (see Methods). For this subject, the timing is 200 ms prior to the start of the imagined movement (CPN+lateCNV) in the first intervention, 50 ms after the start of the imagined task (CPN+ME), and 300 ms after the start of the imagined movement (CPN+HP) for the other two interventions.
Across all subjects the afferent volley from the common peroneal nerve stimulation was elicited 101 ± 110 ms prior to the start of the imagined movement (CPN+lateCNV), 134 ± 115 ms after the start of the imagined task (CPN+ME), and 368 ± 196 ms after the start of the imagined movement (CPN+HP). Figure 5 shows the averaged TA MEP size prior to and following the three interventions for all subjects. Data are expressed as a fraction of the maximum TA MEP prior to the intervention. Across all subjects, the Boltzmann fit accounted for more than 80% of the total variance in the data (r2≥ 0.8). Three variables were extracted from the fitted data: the maximum value (TA MEPmax), the slope of the steepest part of the curve (k) and S50. However, the maximum MEP varies widely across individuals due to the TA representation on M1 being buried deep in the interhemispheric fissures. We thus normalized the peak-to-peak TA MEP for each subject and all stimulation intensities to the maximum TA MEP recorded in the pre-intervention. A three-way repeated-measures ANOVA with factors ‘time’ (1, pre; 2, post), ‘intervention’ (1, CPN+lateCNV; 2, CPN+ME; 3, CPN+HP) and ‘stimulus intensity’ (1, 90%; 2, 100%; 3, 110%; 4, 120%; and 5, 130% RTh) revealed a significant effect of time (F(1,8)= 22.65; P = 0.001) and a significant time × intervention interaction (F(2,16)= 4.83; P = 0.02). Post hoc analysis of the latter interaction using paired t tests with Bonferroni correction revealed significantly greater TA MEPs following the CPN+ME intervention (P = 0.02) but not after the CPN+lateCNV (P = 0.92) or following the CPN + HP (P = 0.19) interventions.
The size of the TA MEPs in the pre-test were not significantly different across the three main interventions (F(2,16)= 0.14; P = 0.39; two-way repeated-measures ANOVA with factors ‘intervention’ (1, CPN+lateCNV; 2, CPN+ME; 3, CPN+HP) and ‘stimulus intensity’ (1, 90%; 2, 100%; 3, 110%; 4, 120%; and 5, 130% RTh).
The changes in the TA MEP sizes may be explained by alterations in the slope of the input–output relation (k), the threshold for eliciting a response or both. Paired t tests with Bonferroni correction revealed that the S50 variable of the Boltzmannn fitted data was significantly different before and after the CPN+ME intervention (P = 0.003). Across all subjects, the S50 had a value of 49.6 ± 5.1% RTh prior to and 47.8 ± 4.4% RTh post-intervention. The S50 parameter is not a direct measure of threshold, and thus it does not indicate the effectiveness of the TMS onto M1 representation of the TA. Rather, it represents the stimulation intensity required to elicit a response 50% of the maximum TA MEP. In our case, this intensity was significantly lower following the CPN+ME intervention while the slope parameter k remained unchanged (P = 0.16).
Control experiment 1: CNV – effect of imagination only
As the changes observed in cortical excitability could have been elicited by imagination only, as a control in eight subjects we recorded the TA MEP recruitment curve prior to and following an intervention where subjects were asked to only imagine the task following the same cue as for the previous interventions but without receiving any peripheral stimulation. A two-way repeated-measures ANOVA with factors ‘time’ (1, pre; 2, post) and ‘stimulus intensity’ (1, 90%; 2, 100%; 3, 110%; 4, 120%; and 5, 130% RTh) revealed a significant effect of stimulus intensity (F(4,28)= 11.36; P < 0.001). However, there was no significant effect of time (F(1,7)= 0.25; P = 0.63). This control measure indicates that imagination only was not effective in increasing cortical excitability.
Control experiment 2: specificity – effect of peripheral tibial nerve stimulation
To ensure that the effects from the main intervention were due to specific afferent inflow from the muscle involved in the imagined movement, this intervention tested if stimulation of a nerve not involved in the imagined movement would alter the TA MEP amplitude. Nine subjects received tibial nerve stimulation at motor threshold that was timed to arrive during the movement execution phase of the MRCP. A two-way repeated-measures ANOVA with factors ‘time’ (1, pre; 2, post) and ‘stimulus intensity’ (1, 90%; 2, 100%; 3, 110%; 4, 120%; and
5, 130% RTh) revealed a significant effect of stimulus intensity (F(4,32)= 19.84; P < 0.001); however, there was no significant effect of time (F(1,8)= 3.53; P = 0.097), nor significant interactions (F(4,32)= 0.68; P = 0.61). This control measure indicates that stimulation of the antagonistic nerve was not effective in increasing cortical excitability.
Control experiment 3: vision – effect of attention on the visual display combined with peripheral nerve stimulation
In an additional control experiment, we asked four subjects to simply view the visual cue and at the time where the cursor indicates the movement onset, we applied the peripheral nerve stimulus. During this control experiment the subjects were unaware of the meaning of the visual cue. The only instructions to the subjects were to pay attention to the display screen in front of them. A two-way repeated-measures ANOVA with factors ‘time’ (1, pre; 2, post) and ‘stimulus intensity’ (1, 90%; 2, 100%; 3, 110%; 4, 120%; and 5, 130% RTh) revealed a significant effect of stimulus intensity (F(4,12)= 8.44; P = 0.002); however, there was no significant effect of time (F(1,3)= 0.05; P = 0.84), nor significant interactions (F(4,12)= 0.34; P = 0.85). This control measure indicates that the visual input from seeing the cursor moving upwards combined with the peripheral stimulus was not effective in increasing cortical excitability.
CPN+ME intervention versus control experiments 1–3
In an additional test, we compared the specificity of the main finding, namely that the excitability of the cortical projections to the target muscle is only increased when the peripheral afferent volley was timed to arrive at the peak negativity of the CNV. An ANOVA with factors ‘intervention’ (1, CPN+ME; 2, TN+ME; 3, imagination only; and 4, watching only), ‘stimulus intensity’ (1, 90%; 2, 100%; 3, 110%; 4, 120%; and 5, 130% RTh) and time (1, pre and 2, post) revealed a significant effect of intervention (F(3,260)= 3.44; P = 0,02), stimulus intensity (F(4,260)= 24,14; P = <0.001) and a significant interaction between these two factors (F(12, 260)= 4.30; P < 0.001). Post hoc analysis using paired t tests with Bonferroni correction revealed significantly greater TA MEPs following the CPN+ME intervention (P < 0.001) but not after imagination only (P = 885), TN+ME (P = 0.11) or watching only (P = 86) control experiments. Figure 6 displays the changes in the TA MEPmax for all conditions across all subjects. Changes are expressed as a percentage of pre-intervention values.
Control experiment 4: spinal excitability
To identify potential changes in spinal excitability, reflex responses in the TA to sudden plantarflexion rotations were investigated prior to and following the CPN+ME intervention in four subjects. These imposed rotations had an amplitude of 4–6 deg and a velocity of 100–200 deg s−1. The changes in ankle angle as well as the EMG recording of the TA both prior to and following the CPN+ME intervention are shown for one subject in Fig. 7A and B, respectively. Each trace is the average of 30 imposed rotations. The TA responds with three discernible peaks (Fig. 7B), of which the SLR is generated exclusively through spinal circuitries. Across all subjects, the SLR component of the TA stretch reflex did not significantly change in amplitude (t= 2.9821, P = 0.06; pre: 21 ± 10 μV, post: 25 ± 13 μV). The background level of activation during the imposed plantarflexion perturbations did not differ significantly pre and post the intervention (t= 0.45, P = 0.06; pre: 17 ± 3 μV, post: 17 ± 3 μV). Figure 7C shows the average size of the SLR, MLR and LLR prior to and following the CPN+ME intervention for all four subjects.
The active MEP was also recorded prior to and following the intervention. Across all subjects, the rectified and averaged active MEP did not change significantly in amplitude (t= 1.13, P= 0.13; pre: 53 ± 16 μV, post: 46 ± 22 μV).
We presented a new conditioning technique for inducing changes in the excitability of cortical projections to the TA muscle. The pivotal proof of principle in our study is that only the CPN+ME intervention led to significant excitability changes that outlasted the stimulation period while other interventions did not have significant effects. The results thus indicate the crucial importance of the timing of peripheral stimulation in relation to the CNV components and therefore the timing between motor planning and the afferent volley to induce associative LTP. Invasive animal and human experiments employing either single-cell or subdural recordings suggest that neural elements within M1, the prefrontal area as well as the supplementary sensorimotor area contribute to the late CNV and ME phase during cued movements (Okano & Tanji, 1987; Kurata & Wise, 1988; Hamano et al. 1997).
The results also showed that afferent feedback from the periphery is necessary to induce the observed plastic changes as a disinhibition of M1 by imagery alone did not lead to a significant change in excitability. We did not assess the effects from peripheral nerve stimulation alone; however, past studies have demonstrated that any changes in MEP size from peripheral nerve stimulation alone require higher stimulation frequencies than those used in the current study (Khaslavskaia & Sinkjaer, 2005) or, alternatively, a longer period of stimulation (2 h) at lower frequencies (Kaelin-Lang et al. 2002). Further, the results demonstrate the specificity of the observed effect as the afferent volley generated by the stimulation of a nerve supplying the antagonist did not lead to any significant changes.
The proposed approach is based on the assumption that CNVs are stable over different recording sessions. In past studies it has been possible to extract parameters of the intended movement from the recorded CNV (do Nascimento et al. 2005a,b, 2006; do Nascimento & Farina, 2008; Gu et al. 2009). This implies that the CNV is stable within one recording session. Here we provide further evidence of this as well as of the CNV stability over days. In the current protocol we used the peak negativity rather than the cue provided to the subject to establish the timing of the afferent generated signal. Indeed, the cue and the peak negativity do not necessarily occur at the same time. In investigating the repeatability of the peak negativity, although we found it to be repeatable within one subject between days, the variability across subjects is in the order of a few hundred milliseconds. This is probably due to the fact that some individuals are better able to react to a cue than others, thus by timing the arrival of the afferent volley at peak negativity we were ensuring that it arrived during the subjects’ imagination of the task execution.
Jackson et al. (2006) used cortically implanted electronic circuits in freely behaving primates to record the activity from one cortical site to subsequently stimulate a different site, creating a causal relation between the neural activity at the two sites. Over a number of days the output of the stimulated site resembled that of the recorded site and this effect, which was specific to the stimulated circuitry, persisted over weeks. This elegant study demonstrated that a physiologically generated signal may be used to drive stimulation at a remote site leading to associative LTP. In the current study we used a physiological signal and paired this with an artificially generated signal to induce associative LTP at the cortical level in humans. Although we did not assess how long this effect lasted, the duration of the post-MEP measures was approximately 10 min after the cessation of each intervention. In addition, the effect was not only timing dependent but also specific since electrical stimulation of the nerve supplying the antagonist muscle did not lead to significant changes in the size of the MEPs. These effects taken together strongly suggest that the changes evoked from the CPN+ME intervention resembled associative LTP. There are several recent studies which have combined a peripheral nerve stimulation with a direct stimulation of M1 using TMS. This protocol, termed paired associative stimulation (PAS), was initially developed to target hand muscles by Stefan et al. (2000) and has since been developed to target lower limb muscles (Mrachacz-Kersting et al. 2007; Poon et al. 2008). There are several differences of the current intervention over PAS. The number of pairings required in the current study is only 50 to produce the desired changes in MEP amplitude. This is far less than those used for the conventional PAS protocols. Perhaps more importantly, here we use a physiologically generated signal within the human brain and pair it with a peripheral electrical nerve stimulation. This opens the possibility to investigate non-invasively those areas of the brain that may be involved in PAS protocols. TMS at the intensities used in PAS protocols (usually 120% RTh) will stimulate a wide area over M1. In particular for lower limb muscles, it is thus not possible to differentially affect the left leg muscles versus the right leg muscles due to the location of the representation of the leg muscles within M1. This may be one reason why the interstimulus interval for PAS when targeting lower limb muscles differs so widely across studies and subjects.
Thabit et al. (2010) modified the PAS protocols described above by using a simple reaction time task and imposed a single magnetic stimulus over M1 timed either 50 ms prior or 100 ms following movement onset. The protocol of their study may not be directly comparable to that implemented here as the TMS current will probably spread to areas other than the M1 representation of the target muscle. However, it is striking that unlike in the present study, they reported significant increases in the TMS-induced MEP of the target muscle (abductor pollicis brevis, APB) only when the TMS stimulus arrived 50 ms prior to movement onset. Several reasons may explain the discrepancy between the results in the current study and those of Thabit et al. (2010). The effects of TMS have been shown to interact with the peripheral nerve stimulus even if TMS is delivered up to 90 ms prior to the peripheral nerve stimulus (Roy et al. 2007). Further, it is specifically the later-occurring I3 waves that interact with the afferent signal from the peripheral nerve stimulation to induce the LTP-like effects following PAS (Kujirai et al. 2006). It can thus not be excluded that the summation of the two stimuli in their study occurred during the movement execution rather than prior to it.
Since there were no alterations to the spinally mediated TA stretch reflex, it is unlikely that subcortical sites contributed to the observed effects. In past studies, the H-reflex has been assessed as an indicator of spinal excitability (Mrachacz-Kersting et al. 2007). The H-reflex probes only that pathway arising from muscle spindle Ia afferents as opposed to the stretch reflex that has various components which are believed to arise from different muscle afferents (Kearney & Hunter, 1984; Petersen et al. 1998). However, it is not possible to probe the entire network of spinal pathways even with the stretch reflex and thus it cannot be excluded that some of the changes may have occurred at subcortical or spinal sites.
In the current study the subjects were presented with a warning signal prior to the presentation of the cue to imagine the movement. The subjects consequently acted on a sensory signal from the outside world and it is generally accepted that in this situation the cortical potential up to 500 ms prior to movement initiation is generated in the contralateral premotor cortex (PMd) (Jenkins et al. 2000). The PMd receives afferent input from several parietal areas and connects with several motor areas (Rizzolatti et al. 1998). Although it is part of a complex neural network involved in selecting, planning and executing of voluntary movements, the intervention CPN+RP did not result in a significant change in the excitability of the cortical projections to the TA. Using a repetitive transcranial magnetic stimulation (rTMS) protocol over the contralateral PMd to increase excitability, Lu et al. (2011) failed to demonstrate a significant change in MEP size while Rizzo et al. (2004) reported significant increases. Similar to the protocol implemented in this study, rTMS as used in the latter studies is thought to induce LTP-like effects. One important difference with respect to our protocol is that we combined peripheral afferent feedback with the natural activation of PMd. As the intensity of the CPN stimulation was at motor threshold it would have activated the lowest threshold afferents, which are those originating in receptors of the target muscle. It may be speculated that if the intensity had been elevated to also recruit pain receptors or conversely if the receptors located in the antagonist muscle had also been activated, this would have resulted in significant MEP changes following the CPN+lateCNV.
The generator of the cortical potential during movement execution is most probably the primary motor cortex (M1; Hamano et al. 1997). This region connects to and receives input from somatosensory and other non-primary cortices (Rizzolatti et al. 1998). In the CPN+ME intervention, cells within M1 are activated simultaneously via commands from other brain regions as well as from the peripherally generated sensory feedback from the target muscle. Together, this produced a significant increase in our outcome measure (the TA MEP). While data presented here do not provide direct evidence as to the locus of this modification, the lack of changes in the stretch reflex renders this to a cortical site and probably M1. As mentioned in the previous section, the intensity of the peripheral stimulus was at motor threshold. It is widely accepted that at this intensity only the low-threshold afferents originating in muscle receptors are recruited. While it is not the aim of this study to investigate the role of muscle afferents in movement generation, several studies have reported these to contribute by up to 50% to the muscle activation during normal walking (Sinkjaer et al. 2000; Nielsen & Sinkjaer, 2002). It may therefore not be surprising that the control experiment where subjects only imagined the task without any input from the periphery did not produce any change in MEP size.
The results presented may have implications in systems for artificially inducing cortical plasticity in patients (neuromodulation). In these applications, the intention to move can be detected without a cue directly from the EEG traces. By identifying the late CNV, there would be sufficient time for triggering the electrical stimulator. It is well known that the early components of the MRCP are prone to alterations in onset in various neurological conditions (Dick et al. 1989; Filipoviçet al. 2001; Bai et al. 2006). In stroke patients, the onset of the MRCP does not change; however, the size of the waveform is decreased over time (Jankelowitz & Colebatch, 2005). In the current study this has little implication; however, if movement characteristics are to be derived from the changes in the shape of the MRCPs, then this factor has to be addressed.
In conclusion, we have developed a novel intervention using self-generated physiological signals for inducing plastic changes within the intact M1. The significant coincident summation effect when the peripheral stimulus was timed to arrive during the peak negativity of the MRCP has wide implications. On the one hand, as it is believed that during cue-driven movements M1 is the generator of the cortical potential during movement execution, our study demonstrates that this activation is sufficient when combined with peripheral afferent feedback for the state of excitability in M1 to be altered beyond the intervention. It also further supports the importance of peripherally generated muscle feedback in the generation of normal ongoing movement. In this case, muscle receptors probably have a greater role beyond their involvement in the stretch reflex. It remains to be investigated how these changes in excitability are reflected in alterations of functional parameters.
All authors contributed to the concept and design of the experiments as well as to the collection, analysis and interpretation of the data. The work was performed at the Centre for Sensory–Motor Interaction, Aalborg University. N.M.-K. and D.F. drafted the manuscript and all authors approved the final version for publication.
This work was supported by the Danish Agency for Science and Technology and Innovation, by the European project BETTER (contract no. 247935) and by the Bernstein Focus Neurotechnology Göttingen.