The effects of transcranial magnetic stimulation (TMS) on post-discharge histograms of single motor units in the first dorsal interosseous have been tested to estimate the input–output properties of cortical network-mediating short-interval intracortical inhibition (SICI) to pyramidal cells of the human primary motor cortex. SICI was studied using the paired pulse paradigm (2-ms interval): test TMS intensity was varied to evoke peaks of different size in post-discharge histograms, reflecting the corticospinal excitatory post-synaptic potential in the relevant spinal motoneuron, and conditioning TMS intensity was constant (0.6 × the resting motor threshold). Navigated brain stimulation was used to monitor the coil position. A linear relationship was observed between test peak size and test TMS intensity, reflecting linear summation of excitatory inputs induced by TMS. SICI was estimated using the difference between conditioned (produced by the paired pulses) and test peaks (produced by the isolated test pulse). Although the conditioning intensity (activating cortical inhibitory interneurons mediating SICI) was kept constant throughout the experiments, the level of SICI changed with the test peak size, in a non-linear fashion, suggesting that low-threshold cortical neurons (excitatory interneurons/pyramidal cells) are less sensitive to SICI than those of higher threshold. These findings provide the first experimental evidence, under physiological conditions, for non-linear input/output properties of a complex cortical network. Consequently, changes in the recruitment gain of cortical inhibitory interneurons can greatly modify the excitability of pyramidal cells and their response to afferent inputs.
Recent advances in transcranial magnetic stimulation (TMS) have provided an indirect electrophysiological approach to human cortical networks (Hallett, 2007). In the paired pulse paradigms (Kujirai et al., 1993), a first (conditioning) TMS pulse modifies cortex excitability and influences the pyramidal cell transynaptic response to a second (test) pulse. The motor-evoked potential (MEP), commonly used to evaluate cortical excitability, is influenced by the conditions of electromyographic (EMG) recording and the spinal motoneurons participating in its amplitude (Lackmy & Marchand-Pauvert, 2010). In the same way, short-interval intracortical inhibition (SICI) depends on the size of the MEP evoked by an isolated test pulse, partly due to the origin of the TMS-induced corticospinal volleys (direct D-wave vs. indirect late I-waves; Garry & Thomson, 2009). The relationship between SICI and MEP size was also attributed to the spinal motoneuron properties, and probably to non-linear summation at cortical level, but the latter was difficult to estimate using variations in MEP amplitude (Lackmy & Marchand-Pauvert, 2010). Given the heterogeneous motoneuron pool properties and the different sensitivity of the corticospinal volleys to SICI, it is difficult to distinguish the effects at cortical and spinal level. A method testing SICI on a single motoneuron and a single corticospinal volley, to avoid the effect due to their own properties, would be required to clarify summation at cortical level.
Complex neural networks mediate the information in the cerebral cortex to pyramidal cells, whose intrinsic properties (Spruston, 2008) and synaptic input characteristics (DeFelipe & Fariñas, 1992) influence their input–output properties. Both electrophysiological (Oviedo & Reyes, 2005; Williams, 2005) and computational (Poirazi et al., 2003a,b; Li & Ascoli, 2006) studies have revealed linear and non-linear summation of synaptic inputs, but the results were obtained under unnatural and limited conditions (with only one pyramidal cell), and did not take into account the greater neural network complexity. This raises questions about the input–output properties of cortical neural networks in intact individuals, a crucial issue in understanding the synaptic integrations at cortical level and the mechanisms underlying plasticity.
Synaptic integration at the cortical level is far from clear and, except that early and late corticospinal volleys are differentially affected by SICI (see Reis et al., 2008), TMS studies do not provide further insight. Investigations on single motor units allow the TMS-induced corticospinal volleys to be distinguished in the post-stimulus time histogram (PSTH; Day et al., 1989). This makes it possible to analyse a single corticospinal volley, and to avoid non-linear summation of multiple corticospinal waves at spinal level. We assumed that investigating SICI on a single volley using PSTHs could give an estimate of the synaptic integrations at the level of the cortical network underlying this volley. The paired pulse paradigm was tested on single motor units from an intrinsic hand muscle during voluntary contraction. The conditioning intensity was kept constant throughout the experiment, so that the cortical networks mediating SICI would be the same. The test intensity was varied to activate different fractions of cortical neurons (interneurons and pyramidal cells discharging in the corticospinal volleys), to investigate the summation of inhibitory and excitatory inputs to pyramidal cells in the primary motor cortex. We found a non-linear relationship between the level of SICI and the strength of the corticospinal volley, suggesting non-linear summations at the cortical level. This study constitutes the first approach to characterize the input–output properties of cortical neural networks under physiological conditions.
Experiments were carried out in 12 healthy volunteers (mean age 33.6 ± 5.1 years; seven women), all of whom gave written informed consent to the experimental procedures. The study was performed according to the Code of Ethics of the World Medical Association (Declaration of Helsinki), and was approved by the local ethics committees of the Pitié-Salpêtrière Hospital (Paris, France).
The subjects were sitting in a comfortable reclining armchair, with head support. EMG activity was recorded from right first dorsal interosseous (FDI), using bipolar surface electrodes (DE-2.3; Delsys Inc., Boston, MA, USA) positioned over the muscle belly. EMG activity was filtered (0.3 Hz to 1 kHz), amplified (× 10 000–50 000, AM502; Tektronix Inc., Beaverton, OR, USA) and converted into standard pulses, which were collected using software programmed in Labview (National Instruments, Austin, TX, USA). The subjects were first trained to perform tonic and voluntary FDI contractions at < 5% of a maximal voluntary contraction (MVC). At the beginning of the experiment, the EMG electrode location was determined during contraction to isolate one motor unit in the recorded activity. Surface EMG recordings are non-invasive, and do not cause any damage to muscle tissue, which in turn influences the motor unit potential (De Luca et al., 2006). Nevertheless, only the largest motor units with the lowest firing thresholds could be investigated. Visual feedback (EMG activity was displayed on an oscilloscope) and auditory feedback (a sound was triggered each time the motor unit potential occurred in the EMG) helped the subjects to maintain a constant motor unit discharge of ∼10 Hz for studying the effects of TMS on the motor unit firing rate (Bawa & Lemon, 1993).
Transcranial magnetic stimulation
TMS was delivered through a figure-of-eight coil (70 mm), generating postero-anterior (PA) currents in the primary motor cortex (with the handle orientated in an anterior, antero-medial or antero-lateral axis depending on the subject), at the optimal site (hot spot) for evoking an MEP in the contra-lateral FDI EMG. The coil was connected to a Bistim module combining two stimulators (Magstim 200; Magstim Company Ltd, Whitland, UK), to provide paired pulses at a 2-ms interval through the same coil. The 2-ms interval, known to evoke strong SICI (Fisher et al., 2002; Roshan et al., 2003), was kept constant throughout the experiment. The optimal coil position was marked on the scalp and, for protocol 2, TMS was assisted by the navigated brain stimulation (NBS) system (Nexstim, Helsinki, Finland), using a standard magnetic resonance imaging brain scan of each individual (http://www.nexstim.com). The NBS system uses a sophisticated algorithm to predict the actual location of the stimulating electric fields in the cortex, and to keep the coil location constant throughout the experiment. TMS intensity was adjusted in relation to the resting motor threshold (RMT), which was the lowest intensity for evoking an MEP of ∼50 μV in at least 50% of trials.
Post-stimulus time histograms
The method is explained fully in Pierrot-Deseilligny & Burke (2005). Briefly, EMG activity was displayed on an oscilloscope to monitor the shape of the investigated motor unit during the experiment. In parallel, the EMG signal was conveyed to a window discriminator with variable trigger levels, which converted the motor unit potential into a standard pulse (3-ms duration, 5-V amplitude); the trigger level position was constant throughout the experiment (Kirkwood & Sears, 1978). Each time the motor unit potential occurred in the EMG activity, the window discriminator delivered a pulse. The pulses were conveyed to a computer, which generated a histogram of the discharge (0.5-ms bin width), according to the latency after a delay R1 relative to the previous pulse; 0 ms in the PSTH thus corresponds to the delay R1. Without conditioning stimulus, the PSTH reflects the motor unit firing rate, and this fits a Gaussian curve (Butler et al., 2007). For example, in Fig. 1A, the delay R1 was ∼40 ms and the Gaussian curve peaked at ∼60 ms, thus ∼100 ms after the previous motor unit discharge, i.e. a discharge rate of ∼10 Hz (Bawa & Lemon, 1993). The delay R1 was adjusted according to the motor unit firing rate, so that TMS was delivered within the recovery phase of the after-hyperpolarization. Thus, when the computer triggered a single TMS pulse at delay R1, the effects on the membrane potential of the motoneuron are optimized, and peak(s) appeared in the PSTH (from an FDI unit) 20–35 ms after TMS (Fig. 1B). The(se) peak(s) reflect(s) the arrival of corticospinal input(s) at motoneuron level, and indicate(s) that the resulting corticospinal excitatory post-synaptic potential(s) [EPSP(s)] were sufficient to advance the discharge of the motoneuron, as compared with its firing rate during voluntary contraction, by shortening the after-hyperpolarisation duration (Fig. 1A and B). The peak in the PSTH is correlated to the ascending phase of the underlying EPSP at motoneuron level (Kirkwood & Sears, 1978; Ashby & Zilm, 1982). Therefore, the TMS-induced peak in PSTH can be used to estimate the corticospinal EPSP produced at the motoneuron level.
The hot spot for FDI and the RMT were determined at the beginning of the experiment. The intensity of both test and conditioning pulses influences the level of SICI (Chen et al., 1998; Sanger et al., 2001; Orth et al., 2003; Roshan et al., 2003; Garry & Thomson, 2009; Lackmy & Marchand-Pauvert, 2010). Therefore, the test pulse intensity was changed so as to evoke a peak in the PSTH of different size (normalized to the number of stimuli, see PSTH analysis below), reflecting corticospinal EPSPs of different size. It was necessary to adjust the intensity of the conditioning pulse to produce SICI without evoking a peak in the PSTH, to prevent possible summation of corticospinal volleys (induced by the test and conditioning pulses) at the motoneuron level. As a consequence, the conditioning pulse could only be set to 0.6 RMT, an intensity at which TMS did not produce a peak in the PSTH (Fig. 1C) but was sufficient to activate SICI (Fisher et al., 2002). At 0.65 RMT, a peak occurred in the PSTH of some motor units (see Results).
A recording session consisted of sequential alternation (0.3 Hz) of isolated test and paired pulses (conditioning + test pulses with a 2-ms interval), to deliver as many test pulses (test peak) as paired pulses (conditioned peak). To avoid muscular fatigue (which can develop rapidly in FDI), 30–50 single and 30–50 paired pulses were delivered during each recording session; the session was stopped when the subjects developed fatigue or had difficulty in maintaining a steady motor unit discharge. Care was taken to ensure that the same motor unit was studied in each session, based on the shape of the potential, its firing rate, the hand position and the movement performed by the subject, and the peak latency in the PSTH.
In four motor units (three subjects), the conditioning–test interval was set at 15 ms, to determine if intracortical facilitation (ICF) could be detected in PSTHs of single motor units with conditioning pulses at 0.6 RMT (Kujirai et al., 1993). As observed previously (Ziemann et al., 1996), no ICF could be evoked with such low conditioning stimuli, and increasing its intensity to the threshold for ICF (0.8 RMT) was not possible because the conditioning pulse by itself could evoke a peak in the PSTH (see below, Protocol 1).
Based on our previous study (Lackmy & Marchand-Pauvert, 2010), Protocol 1 was first elaborated to test the influence of the test peak on SICI. Experiments were performed on 27 motor units from ten subjects. The test pulse intensity was changed in a range defined by the threshold intensity for evoking a significant peak in the PSTH (0.75 ± 0.02 RMT), and an intensity corresponding to RMT minus 5% the maximal stimulator output (MSO; Fisher et al., 2002); for example, RMT was 51% MSO in the subject illustrated in Fig. 2, and the maximal test intensity was 46%, i.e. about 0.90 RMT. Only test intensities below the motor threshold were investigated, because if an MEP occurred in the EMG activity, it could interfere with the recording of the motor unit discharge due to superimposition of MEP and motor unit potential. The test intensity was randomly changed from one recording to another and, at the end of the experiment, we ensured that each TMS intensity (in 1% steps), between peak threshold and RMT minus 5% MSO, had been tested. The intensity of the test pulse was then normalized to RMT for inter-individual comparisons (Fig. 2A,D and G). About 10–12 recording sessions were made for each motor unit, one recording session for each intensity investigated. Each recording session lasted 4–7 min.
The hot spot for FDI was determined at the beginning of the experiment, and marked on the scalp in Protocol 1. Although the conditioning pulse intensity was kept constant throughout the experiment (0.6 RMT), a minimal change of the coil orientation might have influenced the stimulating conditions, and therefore the level of SICI: something that can be controlled only by monitoring stimulus intensity and stimulation site because the conditioning pulse (0.6 RMT) did not produce any significant change in the PSTH (Fig. 1C). This would also have influenced the effect of the test pulse, and thus the test peak size. To stabilize the stimulating conditions, a second protocol was developed with the NBS system to monitor the coil position and the TMS-induced electrical field in the brain. Based on the results of Protocol 1, we adjusted the TMS test intensity to 0.75 and 0.85 RMT to evoke small and medium peaks in the PSTH (∼10 and 20–30% the number of stimuli, respectively), and we increased TMS intensity to 0.95 RMT to explore SICI on larger test peaks than in Protocol 1 (> 30% the number of stimuli). Experiments were performed on 18 motor units from nine subjects (including seven subjects of Protocol 1). Two recording sequences were made at each test intensity, and for each motor unit investigated, but the same test pulse intensity was not tested in two successive sequences.
Because changing the test intensity affects the test peak size (Devanne et al., 1997), the influence of both parameters was tested with the two protocols. However, to make the distinction between the two protocols, the results of Protocol 1 were grouped according to the test peak size, and those of Protocol 2 to the test intensity. This corresponds to the two methods commonly used to set the test pulse in the paired pulse paradigm, either the amplitude of the test response or the intensity of the test pulse.
PSTHs were constructed for 40 ms (acquisition window) starting 15 ms after test TMS (15–55 ms), i.e. for a window larger than the duration of TMS-induced peaks in FDI PSTH (20–35 ms; Day et al., 1989). Peaks were first identified visually in the control PSTH (single test pulse; see Figs 2A,D and G, and 4A,D and G), and the analysis was limited to the first three adjacent bins in the peak (i.e. the first 1.5 ms). These three bins were then tested using a χ2 test to ensure that the increase in motor unit firing rate at this latency was significant (e.g. 25, 25.5, 26 ms in Fig. 2; the first bin in the peak is indicated by the dotted vertical arrow). Given an interval between the component waves in the corticospinal volley of 1.5 ms (see Hallett, 2007; Reis et al., 2008), the analysis was thus limited to a single corticospinal EPSP. It is worth noting that the first three bins corresponded to the peak rising phase, and included the largest bin in the peak (Figs 2 and 4). Sometimes, at low test intensity, it was difficult to determine visually the beginning of the peak (Fig. 4A). In such a case, the analysis window was determined for higher intensities, which evoked larger peaks in the PSTH (Fig. 4G). The conditioned PSTH (after paired pulse TMS) was analysed within the same window as the test peak, to compare the peak size after SICI (conditioned peak) with that evoked by single test pulse (test peak).
In Protocol 1, we grouped the data according to the size of the test peak, and for close TMS intensities, the size of the peak could be similar. For inter-individual comparisons, the recording sequences giving rise to test peaks < 30, 30–60 and > 60% of the maximal test peak size were summed for each motor unit. We thus compared three sizes of test peak for each motor unit. The number of stimuli was about 100 for each test peak size (Fig. 2J).
In Protocol 2, the two recording sessions performed at similar test intensity were added, and the number of stimuli was 100 for each intensity. However, when the test intensity was 0.95 RMT, TMS evoked an MEP in FDI EMG on ∼25% of occasions. The corresponding counts were thus deleted, and this was taken into account in the PSTH normalization.
For each protocol, the number of counts (motor unit discharges) in each bin was normalized to the number of stimuli to allow inter-individual comparisons. The number of counts in the three adjacent bins (percentage number of stimuli) was used to evaluate the test peak size. The level of SICI was estimated using the difference between the conditioned and test peak (percentage number of stimuli).
For each motor unit, χ2 tests were performed at each TMS intensity investigated, to determine if the three consecutive bins in the test peak were significantly different from the equivalent three bins in the control PSTH, and to compare the distribution in the test (test TMS alone) and conditioned peaks (paired pulse).
Analysis of the grouped data
Because the size of the test peak (Protocol 1) and the TMS intensity (Protocol 2) were the parameters retained to characterize the test pulse in each protocol, their influence on SICI was tested using one-way anova, taking into account the test peak size for the grouped data in Protocol 1, and the TMS intensity for those in Protocol 2. If a significant P value was obtained, post-hoc Fisher LSD tests were performed for comparisons of two means. The relationships between TMS intensity and test peak size (Protocol 1), and between test peak size and SICI were tested using Pearson’s correlation with repeated measures (Poon’s treatment to take into account the within- and between-subjects variances; Poon, 1988). To determine if the level of SICI was significantly different from 0, one-sample t-tests were performed for each category of test peak size (Protocol 1), and for each test pulse intensity (Protocol 2). Tests were performed using StatEL software (http://www.adscience.eu), and the significance level was 0.05. Mean data are given ± 1 standard error of the mean (SEM).
TMS-induced corticospinal peak(s) in PSTHs
In Protocol 1, the TMS test pulse enhanced significantly the firing rate of a single FDI motor unit at 25 ms (Fig. 2, dotted vertical arrow). The resulting peak in the PSTH increased with TMS intensity: 10.0% the number of stimuli when test TMS was 0.76 RMT (χ2 = 7.3, P <0.01; Fig. 2A), 25.5% at 0.83 RMT (χ2 = 25.3, P <0.001; Fig. 2D) and 36.6% at 0.90 RMT (χ2 = 14.5, P <0.001; Fig. 2G). The peak was limited to three bins (25–26 ms) at 0.90 RMT (Fig. 2G). In the 27 motor units investigated (Protocol 1), a significant linear relationship was found between TMS intensity and peak size (Fig. 3; Pearson’s correlation with repeated measures, P <0.00001, R2 = 0.87).
In Protocol 1, the mean threshold intensity for a significant peak in the PSTH was 0.75 ± 0.02 RMT (range 0.65–0.80 RMT). These values were used to determine the test intensities investigated in Protocol 2: 0.75 (peak threshold intensity), 0.85 (intermediary intensity) and 0.95 RMT (maximal intensity usable in a PSTH). Figure 4 illustrates the results on a single motor unit of Protocol 2. The test TMS increased significantly the motor unit firing rate at 27 ms (dotted vertical arrow), and the peak (27–28 ms) reached 10.7% the number of stimuli at 0.75 RMT (χ2 = 5.7, P <0.05; Fig. 4A), 19.2% at 0.85 RMT (χ2 = 6.9, P <0.01; Fig. 4D) and 24.5% at 0.95 RMT (χ2 = 22.6, P <0.001; Fig. 4G). In the 18 motor units investigated (Protocol 2), the test peak increased significantly with TMS intensity (15.3 ± 2.4% at 0.75 RMT, 28.1 ± 2.9% at 0.85 RMT and 42.6 ± 3.9% at 0.95 RMT; anova, P <0.0001).
The PSTHs of a single motor unit in Fig. 4 illustrate a 3-ms duration peak (27–30 ms), with largest bins at 27 and at 28.5–29 ms, suggesting a contribution of different corticospinal waves. In the 45 motor units investigated (Protocols 1 and 2), the mean latency of the earliest peak (P1) evoked in the PSTH was 27.1 ± 0.3 ms (range 22.5–30.5 ms). In 16/45 motor units (ten in Protocol 1 and six in Protocol 2), a second peak (P2) followed P1, and the mean time difference between P1 and P2 was 1.6 ± 0.1 ms (range 1.5–3 ms). These peaks are likely to represent motor unit discharge to separate components of a complex corticospinal volley, 1.6 ms corresponding to the interval between successive corticospinal waves (Day et al., 1989; Hallett, 2007; Reis et al., 2008). In such a case, the analysis was limited to P1, specifically to the three-first significant bins, to evaluate SICI on the first component of the corticospinal volley.
Influence of test peak size on SICI
In Protocol 1, the intensity of the test pulse was randomly changed to produce test peaks of different size, and to evaluate the resulting SICI evoked by a paired pulse using the difference between conditioned (paired pulse) and test (isolated test pulse) peaks in the PSTHs. For inter-individual comparisons, the results of each motor unit were grouped into three categories of test peak size, according to the maximal size of the test peak (peakmax), and the intensity of the test pulse was normalized to RMT. Concerning the motor unit illustrated in Fig. 2, the test peak < 30% the maximal peak, within the three-first bins (25–25.5–26 ms), was evoked at 0.76 RMT (Fig. 2A). The test peak between 30 and 60% the maximal peak was evoked at 0.83 RMT (Fig. 2D), and the test > 60% was evoked at 0.90 RMT (Fig. 2G). In the 27 motor units investigated, the peaks < 30% were evoked with test stimuli at 0.77 ± 0.01 RMT, the peaks between 30 and 60% were evoked with test stimuli at 0.84 ± 0.02 RMT, and the peaks > 60% were evoked with test stimuli at 0.90 ± 0.01 RMT (Fig. 2J).
In each motor unit, the test (isolated test pulse) and conditioned PSTHs (paired pulses) were compared within the three-first bins in the peak. In the motor unit of Fig. 2, there was no significant change in peak size after paired pulses, between 25 and 26 ms, when the test peak was < 30% of the peakmax (the difference was 2% the number of stimuli, χ2 = 0.07; Fig. 2A–C). When the test peak was 30–60% of the peakmax (Fig. 2D), the conditioned peak was significantly smaller with the paired pulses (Fig. 2E), reflecting SICI (−14.4%, χ2 = 9.9, P <0.05; Fig. 2F). When the test peak was > 60% of the peakmax (Fig. 2G), the conditioned peak was again smaller (Fig. 2H), but the difference was not statistically significant (−6.6%, χ2 = 0.09; Fig. 2I). In the 27 motor units investigated (Protocol 1), anova revealed a significant influence of the size of the test peak on SICI (P <0.0001), with significant differences between peaks < 30% and peaks between 30 and 60% (Fisher’s LSD test, P <0.001), and between peaks < 30% and peaks > 60% (P <0.001; Fig. 2J). For peaks < 30% the peakmax, the mean difference was 0.1 ± 1.2% the number of stimuli (one-sample t-test, P =0.94), revealing no SICI. For peaks between 30 and 60%, the mean SICI was −5.6 ± 1.0% (P <0.0001), and for peaks > 60% it was −5.4 ± 1.4% (P <0.001).
Correlation analyses were performed to determine the relationship between the test peak size (percentage number of stimuli) and the level of SICI. The scatter plot in Fig. 2K shows less SICI when test peak size was between 3 and 14% than when test peak size was > 14%, but no significant linear relationship was observed between test peak size and SICI (Pearson’s correlation with repeated measures, P =0.38). Given the significant influence of the test peak size on SICI, further analyses were performed using the reciprocal function of the test peak size (1/peak), and its natural logarithm [ln(peak)]. No significant correlation was found between ln(peak) and SICI (P =0.15), but there was a significant linear relationship between 1/peak and SICI (P <0.00001, R2 = 0.45; Fig. 2L). This result indicates that the level of SICI increased with the size of the test peak in a non-linear fashion (SICI depends on 1/peak).
In five of 27 units, the peak was not depressed after SICI, and when the group analysis was repeated omitting these units, the results were similar to the whole sample of 27 motor units.
Influence of stimulating conditions on SICI
To control for the possibility that the modification of SICI in Protocol 1 was not due to a change in coil position, Protocol 2 was undertaken using the NBS system to monitor the stimulating conditions.
In Fig. 4, illustrating the PSTHs from a single motor unit, when the test pulse was 0.75 RMT, the peak (27–28 ms) was not depressed after the paired pulses (difference was −0.8% the number of stimuli, χ2 = 0.36; Fig. 4B and C). At 0.85 RMT, the peak was significantly depressed after the paired pulses (Fig. 4E), producing SICI of −10.3% (χ2 = 4.18, P <0.05; Fig. 3F). Increasing the test pulse to 0.95 RMT caused the SICI to disappear (6.89%, χ2 = 2.21; Fig. 4H and I). In the 18 motor units investigated (Protocol 2), anova revealed a significant influence of the test pulse intensity on SICI (P <0.02; Fig. 4J), with larger SICI at 0.85 RMT than at both 0.75 RMT (Fisher’s LSD test, P <0.01) and 0.95 RMT (P <0.03). Indeed, the mean SICI was significant at 0.85 RMT (−7.5 ± 1.6%; one-sample t-test, P <0.001), but not at 0.75 RMT (−0.9 ± 2.0%, P =0.66) or at 0.95 RMT (−1.8 ± 1.8%, P =0.33).
Similar correlation analyses as in Protocol 1 were performed considering the test peak size, and no significant linear relationship was found between peak size and SICI (Pearson’s correlation on repeated measures, P =0.43; Fig. 4K). Again, SICI was significantly correlated to the reciprocal function of the peak size (1/peak, P <0.00001, R2 = 0.35; Fig. 4L) but not to its logarithm (P =0.8).
In two of 18 units, the peak was not depressed after SICI, and when the group analysis was repeated omitting these units, the results were similar to the whole sample of 18 motor units.
Protocols 1 and 2 revealed a significant influence of the test pulse on SICI, with significant correlation between SICI and 1/peak. Table 1 shows the mean data from the two protocols. In both, SICI was hardly evoked when the test peak was < 10–15% the number of stimuli (Figs 2K and 4K). In Protocol 2, stronger test pulses evoking larger test peaks, as compared with Protocol 1, were investigated revealing a decreased in SICI when test peak size was > 30% the number of stimuli, and with test TMS > 0.90 RMT (compare Figs 2K and 4K).
Table 1. Comparison of the mean data from the two protocols
Test TMS: test pulse intensity × RMT; Peak: test peak size (percentage number of stimuli); SICI: difference between conditioned (paired pulse) and test (isolated test pulse) PSTHs (percentage number of stimuli). *P < 0.05, **P < 0.01, ***P < 0.001. Data are shown as means ± 1 SEM.
0.77 ± 0.01
9.4 ± 0.6
0.1 ± 1.2
15.3 ± 2.4
−0.9 ± 2.0
0.84 ± 0.02
18.4 ± 1.1
−5.6 ± 1.0***
28.1 ± 2.9
−7.5 ± 1.6**
0.90 ± 0.01
28.9 ± 1.7
−5.4 ± 0.9**
42.6 ± 3.9
−1.8 ± 1.8
This study has shown that, while the test peak produced by single TMS in the PSTH increases linearly with TMS intensity, SICI in a paired pulse paradigm depends on test peak size and test TMS intensity in non-linear fashion. Small peaks (< 15% the number of stimuli) evoked at low TMS intensities < 0.80 RMT are not sensitive to SICI. The paired pulse inhibition became apparent when test peaks were larger (15–30%) with test TMS between 0.80 and 0.90 RMT. Finally, SICI was hardly evoked when the test peak was > 40%, and test pulse at 0.95 RMT.
Origin of the peak(s) evoked by the test TMS
TMS can evoke multiple corticospinal volleys, distinguishable in epidural recordings (Burke et al., 1993; Di Lazzaro et al., 1998a) and in the PSTH of single motor units (Day et al., 1989), with minimal periodicity of 1.5 ms, as in the 16 motor units exhibiting multiple peaks in the PSTH, in the present study. Each volley has a different sensitivity to SICI: the D-wave (activation of pyramidal axons) and the first I-wave (I1: transynaptic response of pyramidal cells) are less affected by SICI than late I-waves (Nakamura et al., 1997; Hanajima et al., 1998; Di Lazzaro et al., 1998b; Fig. 5). Given only the latency of a peak in a PSTH, it is difficult to be certain which wave in the corticospinal volley underlies the peak without transcranial electrical stimulation, which can be used to identify the D-wave latency (Day et al., 1989). However, I-waves are elicited at a lower threshold intensity than the D-wave under the stimulating conditions in this study (Sakai et al., 1997; Di Lazzaro et al., 2002), and because SICI was evoked in 38 of 45 motor units, we assume that the peaks we investigated were mediated by I-waves in mostly units.
The peak in a PSTH is directly related to the rising phase of the underlying EPSP at motoneuron level (Ashby & Zilm, 1982). The largest peaks in the PSTH reached 50–90% the number of stimuli, which further supports that the TMS-induced corticospinal waves produce fairly large EPSPs in FDI motoneurons, mediated by the direct cortico-motoneuronal projections (de Noordhout et al., 1999). Because the analysis was limited to the first 1.5 ms of the peak, corresponding to the periodicity between each TMS-induced corticospinal volley (Hallett, 2007; Reis et al., 2008), the peak size probably reflects the EPSP evoked by a single corticospinal volley at motoneuron level. Increasing the TMS intensity leads to larger corticospinal volleys and to additional volleys (Burke et al., 1993; Di Lazzaro et al., 1998a); earlier or later volleys would have induced earlier or later peaks in the PSTH. When test TMS intensity was increased, the latency of the earliest peak evoked in the PSTH did not change in all the motor units investigated; in 16 of the 45 motor units, a second peak could be evoked but the analysis was limited to the first peak.
Linear relationship between TMS intensity and test peak size
As observed in a previous study (Devanne et al., 1997), the peak size increased linearly with TMS intensity. This suggests a linear increase in the underlying corticospinal EPSP. This EPSP depends on the membrane properties of the spinal motoneuron innervating the motor unit investigated (Hultborn, 2002), and on the corticospinal input induced by TMS. This input depends on the summation of the effects evoked by TMS at the cortical level: the stimulating electric fields activate neural network in the primary motor cortex, including inhibitory and excitatory interneurons, and pyramidal cells (Fig. 5). The resulting corticospinal volley depends on the balance between TMS-induced inhibitory and excitatory inputs to pyramidal cells. When TMS was suprathreshold for a peak in the PSTH, and when its intensity was increased, the excitation counterbalanced the inhibition at the cortical level, which made the pyramidal cells discharge: the greater the cortical excitation, the stronger the cortical outflow (more pyramidal cells discharge) and this leads to larger peaks in the PSTH (spatial summation of corticospinal EPSPs at motoneuron level; Fig. 5). The linear relationship between TMS intensity and peak size thus reflects the input/output properties of the cortico-motoneuronal network (cortical network and spinal motoneuron). Note that this conclusion is limited to the cortical networks with the lowest thresholds, activated with very low TMS intensities that we could investigate with PSTHs. At higher intensities, MEPs are evoked in EMG activity, and the sigmoid recruitment curve could then be due to non-linear summation at both cortical and spinal levels (several motoneurons discharge, not just one; Devanne et al., 1997; Lackmy & Marchand-Pauvert, 2010).
Influence of test peak size on the measurement of SICI
In the paired pulse paradigm, the conditioning pulse was subthreshold for a peak in the PSTH but suprathreshold for SICI, the threshold intensity for inhibitory interneurons being lower than for excitatory ones (Fig. 5; Ziemann et al., 1996). The conditioning pulse was delivered in advance of the test pulse so as to investigate the integration of inhibitory inputs in the neural network activated by the test pulse; the conditioning pulse changed the excitability of the cortical neural network (excitatory neurons being hyperpolarized) and its response to the test pulse. The peak evoked by a paired pulse was thus the result of the summation at cortical level of inhibitory inputs produced by the conditioning pulse and those (excitation + inhibition) produced by the test pulse.
In the present study, the conditioning intensity was constant throughout the experiments (and stimulation site was controlled using the NBS system in Protocol 2), but SICI changed according to the test pulse. Summation of inhibitory inputs produced by the conditioning and test pulses seems unlikely because this would mean that increasing test intensity gave rise to stronger inhibition. The most parsimonious explanation is that cortical excitation increased with test pulse intensity, and the excitatory cortical neurons have different sensitivity to inhibition. Indeed, if these neurons had the same sensitivity to the conditioning-induced inhibition (considered to be constant), SICI would have been equal whatever the test peak size. Another explanation would be that the summation of corticospinal inputs of different strengths (due to SICI) could be non-linear at motoneuron level due to its membrane properties (Hultborn et al., 2004). However, this seems unlikely given the linear relationship between TMS intensity and test peak size in PSTHs (Devanne et al., 1997). Our results thus suggest a cortical mechanism, and that low-threshold neurons (excitatory interneurons and pyramidal cells) in the neural network mediating TMS-induced corticospinal waves are less sensitive to inhibitory inputs than excitatory neurons with higher threshold.
When the test peak was > 30% (the number of stimuli), SICI was less, and it was hardly seen when the peak was > 40%. This could suggest that the cortical neurons with high threshold are not sensitive to SICI, but this seems unlikely because paired pulses depressed MEPs evoked at even higher test pulse intensity (Garry & Thomson, 2009; Lackmy & Marchand-Pauvert, 2010). Increasing TMS intensity strengthened the corticospinal input, giving rise to a large EPSP at spinal level, which can greatly exceed the threshold for motoneuron discharge. SICI evoked at 0.6 RMT was probably not sufficient to depress enough the corticospinal outflow produced by the test pulse at 0.95 RMT. Although the corticospinal volley was depressed by SICI, it was still sufficient to make the motoneuron discharge, and the conditioning peak was not different from the test peak (saturation of the corticospinal input). Therefore, the level of SICI evaluated with the difference between conditioning and test peak was apparently less, but this was due to the PSTH method, which is not sensitive enough to reveal a small depression of large corticospinal EPSPs.
Using PSTHs, the effect of paired pulses can only be explored during voluntary contraction, which influences the level of SICI (Ridding et al., 1995). The non-linear relationship between test peak size and SICI could be due to opposite effects at cortical level with, on the one hand, inhibition of SICI due to voluntary motor activation and, on the other, activation of SICI by conditioning TMS. However, this seems unlikely as the subject performed very weak contraction < 5% MVC, a level at which SICI is not depressed (Zoghi & Nordstrom, 2007).
Summation at cortical level
The non-linear relationship between SICI and test peak may thus reflect non-linear input–output properties of the cortical neural networks activated by TMS. Using PSTHs, the non-invasive electrophysiological investigation of cortical networks in humans is limited to the range of TMS intensities usable for conditioning and test pulses. Indeed, it would have been interesting to assess the effects of different conditioning pulses (Chen et al., 1998; Orth et al., 2003), but this was not possible: we could use only 0.6 RMT (just above SICI threshold; Fisher et al., 2002), because TMS at 0.75 RMT regularly produced a peak in PSTHs, and did so in some motor units at 0.65 RMT. Regarding the test pulse, MEPs could occur in the EMG activity at 0.95 RMT (with one of four stimuli). The range of TMS (0.75–0.95 RMT) evoked corticospinal peaks covering a narrow range of sizes. Our conclusions are thus limited to cortical networks with low thresholds. Wider ranges of stimulus intensity can be tested only with the MEP. However, conclusions based on MEP studies are limited by the fact that the corticospinal inputs are non-linearly distributed in the motoneuron pool, making it difficult to distinguish between non-linear summation at spinal level and non-linear summation at cortical level (Lackmy & Marchand-Pauvert, 2010). Investigations on single motor units with PSTHs allow such a distinction. Comparison of the results obtained with PSTHs and MEPs would be desirable to understand synaptic integration at both cortical and spinal level, and especially the distribution of SICI in cortical networks. A non-linear relationship was also found between test MEP and SICI (Garry & Thomson, 2009; Lackmy & Marchand-Pauvert, 2010), and when varying the conditioning pulse, the larger the MEP, the greater the difference between the SICI evoked at 0.7 and 0.8 RMT (Lackmy & Marchand-Pauvert, 2010):
1 When the test MEP was small (< 10% the maximal compound action muscle potential), SICI was weak when the conditioning TMS was 0.7 or 0.8 RMT, and there was no difference between the two intensities of conditioning. This result fits with those on single motor units (small peaks in the PSTHs were hardly depressed), and supports the suggestion that the cortical neural networks with the lowest threshold are not sensitive to SICI. When the test TMS was low and the resulting corticospinal inputs weak, SICI was hardly evoked whatever the conditioning intensity (0.6 RMT in PSTH studies and 0.7–0.8 RMT in MEP studies).
2 In both PSTH and MEP studies, SICI increased with increasing test TMS intensity. However, the largest MEPs (> 30%) were hardly depressed by conditioning pulse at 0.7 RMT, as observed for a large peak in the PSTH at 0.6 RMT (saturation of corticospinal inputs). When the conditioning pulse was set at 0.8 RMT, SICI still increased with MEP size, and the largest MEPs were strongly depressed.
Results from MEP studies suggest that neural networks with high threshold are still sensitive to SICI. The decrease in SICI observed with large peaks in the PTSH and large MEPs suggests that when the conditioning stimulus is weak (0.6–0.7 RMT), the depression of the corticospinal volley has little effect on motoneuron discharge: the weakly depressed corticospinal inputs after SICI are still sufficient to make the motoneurons discharge. Therefore, the combination of the two methods (PSTH and MEP studies) suggests that (1) low-threshold cortical neurons activated at low TMS intensities are not sensitive to SICI, (2) the distribution of inhibitory inputs in cortical networks is non-linear and (3) the conditioning stimulus has to be > 0.7 RMT to depress the corticospinal volleys sufficiently and to avoid the saturation at motoneuron level that would prevent the evaluation of SICI using the difference between conditioned and test responses.
Although the non-invasive techniques used in humans can only provide indirect electrophysiological data, it has been possible (1) to give further evidence for linear input–output properties of cortico-motoneuronal networks (Devanne et al., 1997) and (2) to give the first evidence for non-linear summation of inhibitory inputs in neural networks controlling pyramidal cell discharge in the human primary motor cortex. To our knowledge, this is the first time that the input–output properties of cortical networks have been studied under physiological conditions (both in humans and in animals). These results are important for the understanding of synaptic integration at cortical level and summation at motoneuron level in studies using TMS: synaptic integration at cortical and spinal levels should be taken into account in interpreting the effects of TMS. In addition, this study provides further insights into the neural mechanisms underlying plasticity in awake humans. Intrinsic plasticity in layer V cortical neurons has been demonstrated in animal preparations in vivo (Paz et al., 2009), but a change in the relative recruitment gain of inhibitory interneurons has been proposed to participate in long-term potentiation of pyramidal cells only using a computational model (Marder & Buonomano, 2004). Given the non-linear summation of inhibitory inputs at cortical level, a change in the recruitment gain of inhibitory interneurons can strongly influence pyramidal cell excitability. Such a mechanism should be taken into account in studies using techniques recently developed to investigate TMS-induced plasticity in humans (e.g. repetitive TMS and paired-associative stimulation).
We thank Prof. David Burke (Sydney University) for reading and commenting upon the manuscript. The study was supported by UPMC Université Paris 6, Assistance Publique-Hôpitaux de Paris (AP-HP), Institut pour la Recherche sur la Moelle Epinière (IRME), and INSERM. L.S.G. was supported by a grant from UPMC Université Paris 6 (Ministère de l’Enseignement Supérieur et de la Recherche).