| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOPICAL REVIEW |
1 Human Cortical Physiology and Stroke Neurorehabilitation Section, National Institute of Health, National Institute of Neurological Disorders and Stroke, 10 Center Drive, Bethesda, MD 20892, USA
2
Sobell Department of Motor Neuroscience, Institute of Neurology, 8–11 Queen Square, London WC1N 3BG, UK
 |
Abstract |
|---|
 Top Abstract References |
|---|
(Received 11 September 2007;
accepted after revision 30 October 2007;
first published online 1 November 2007)
Corresponding author L. G. Cohen: Human Cortical Physiology Section, National Institute of Health, National Institute of Neurological Disorders and Stroke, 10 Center Drive, Bldg 10, Rm 5 N226, Bethesda, MD 20892, USA. Email: cohenl{at}ninds.nih.gov
Introduction
Recent work on electrophysiology of non-invasive brain stimulation has contributed to identifying physiologically active interactions between different cortical regions in awake, healthy human subjects. The cortical motor output has been studied in particular detail, as the output of the primary motor cortex (M1) can be objectively measured in the form of a motor evoked potential (MEP). Using surface electromyographic (EMG) recording electrodes, most commonly positioned on the skin overlying the hand muscles, a compound MEP can be elicited in response to a single suprathreshold transcranial magnetic stimulation (TMS) pulse delivered to M1. The MEP amplitude elicited by stimulation of M1 can be modulated by a preceding conditioning pulse delivered either to the same cortical area or elsewhere, allowing the exploration of intra- and inter–regional physiological interactions in real time.
Detailed study of the corticofugal discharge in response to a motor cortical stimulus by Amassian et al. (1989) revealed multiple components of the MEP. These can be observed either by epidural recordings or by measuring single motor unit recordings with needle electrodes, and consist of a short latency direct wave (D-wave) followed by several longer latency indirect waves (I-waves). The D-wave is thought to result from direct depolarisation of the initial axon segment of the corticospinal neuron and is most effectively activated in human subjects by transcranial electrical stimulation or high intensity TMS. The I-waves following the D-wave occur sequentially with a periodicity of approximately 1.5 ms, reflecting the delay required for synaptic discharge. Thus, the first I-wave (I1) is thought to be generated through the depolarisation of an axon synapsing directly onto a corticospinal neuron (i.e. monosynaptically), while following I-waves (I2 and later) may require local polysynaptic circuits. I-waves can be elicited using relatively low TMS intensities in humans and are thus readily amenable to study.
The last few years have seen a flurry of investigations into inter-regional physiological interactions linking M1 with other ipsilateral and contralateral motor regions, parietal cortex, cerebellum and sensory afferents. A new picture of M1 is gradually emerging in which its role is considerably greater than that of the passive servant of higher order motor regions – rather, M1 may be seen as performing a complex integration of multiregional influences that result in purposeful motor behaviour. Some of the mechanisms involved in this process are now better understood, but many questions remain unanswered.
The first aim of this review is to update the reader on this rapidly changing subject, highlighting in the process the gaps in current knowledge. The second aim is to assist investigators by illustrating the electrophysiological interactions tested with TMS that may influence goal specific motor behaviour in humans.
Because of the very extensive literature on TMS, we chose to review in this paper electrophysiological interactions tested with TMS in particular detail, rather than extensively reviewing work on the use of TMS to induce a ‘virtual lesion’, only partially discussed, or on the interactions of TMS with other techniques like EEG, MEG and MRI.
Resting and activity-dependent interactions between M1 and other regions or afferents are grouped into intrahemispheric (within M1), interhemispheric (M1 to M1) and interregional (e.g. premotor cortex or cerebellum to M1). For the sake of simplicity, these interactions are separated into inhibitory and excitatory, but it should be kept in mind that they are likely to overlap to some extent, such that what is measured represents a net effect. Separating such influences often requires subtle manipulations of stimulus parameters. A summary of the net interregional influences to be considered is provided in Fig. 1.
|
The interactions within M1 that are currently known to modulate its output are illustrated in Fig. 2.
|
Intracortical facilitation (ICF) of a test MEP can be elicited at interstimulus intervals (ISIs) of 6–25 ms, using a subthreshold conditioning stimulus (CS) to influence the response to a subsequent suprathreshold test stimulus (TS). This effect was first described by Kujirai et al. (1993) in a now classic paper reporting facilitation of the test MEP at intervals of 10–15 ms. Facilitation becomes stronger with increasing CS intensity (Kujirai et al. 1993), but tends to be weaker with increasing TS intensity (Daskalakis et al. 2004). The question arises as to whether ICF may be merely a ‘rebound’ phenomenon from the robust inhibition described by these authors at shorter interstimulus intervals (see below) or whether it represents a separate phenomenon. Ziemann et al. (1996c) demonstrated that short interval intracortical inhibition (SICI) occurs at lower CS intensities than ICF, becoming stronger with increasing intensities. Furthermore the two phenomena behave differently depending on the current direction of conditioning and test pulses: while inhibition can be elicited regardless of the direction of current flow, reliable ICF requires a conditioning stimulus to be induced in a postero-anterior (PA) direction. The authors concluded that separate neuronal populations were likely to mediate intracortical inhibition and facilitation (Ziemann et al. 1996c). A cortical (rather than spinal) site of action of the CS in this context was supported by the findings that the CS intensity required to elicit this effect was below the threshold for producing an MEP and that spinal H-reflexes were unaffected. This was investigated further by studying the effect of the CS on descending volleys recorded in cervical epidural electrodes. This approach demonstrated facilitation of the late I-waves at an interstimulus interval of 25 ms, suggesting a synaptic interaction within M1 (Nakamura et al. 1997). However, a recent study by Di Lazzaro et al. (2006) examined cervical descending volleys in more detail: while facilitation of late I-waves was again seen at 25 ms, no changes in the amplitudes or number of I-waves were seen at 10 and 15 ms, despite facilitation of the compound MEP. This raises the possibility that facilitation at these shorter intervals may be mediated by subtle changes in spinal excitability. However, in the same study, test MEPs generated by delivering an electrical TS directly to cervical epidural electrodes were not facilitated by a magnetic cortical CS, making such a spinal interaction unlikely. An alternative and more likely possibility is that any additional corticospinal discharge produced in the presence of a CS is temporally dispersed, and thus not apparent in the mean I-wave traces. Thus, while ICF at 25 ms appears likely to have a cortical origin, the site of facilitation at 10–25 ms is less clear.
Excitatory glutamatergic interneurons within M1 and N-methyl-D-aspartate (NMDA) receptors appear to influence ICF (Ziemann, 2003). NMDA antagonists have been shown in two separate studies to abolish (dextromethorphan) or even reverse (memantine) ICF measured at 10 or 15 ms (Ziemann et al. 1998a; Schwenkreis et al. 1999). This issue has been clouded somewhat by the demonstration that ICF is unaffected by the non-competitive NMDA antagonist ketamine, given at a subanaesthetic dose (Di Lazzaro et al. 2003). However, while ketamine is thought to reduce transmission at NMDA receptors, it is also believed to increase glutamate release and transmission at AMPA synapses. Furthermore, the increase in unconditioned test MEP amplitude after ketamine makes the lack of effect on ICF difficult to interpret. ICF is also thought to be modulated by GABAA activity, since it is reduced by the GABAA agonist lorazepam and abolished by ethanol, which potentiates GABA-mediated currents (Ziemann et al. 1995, 1996b; Ziemann, 2004). This is consistent with the idea that the inhibition of I3 waves that is responsible for short interval inhibition (SICI – see below) may persist as late as 20 ms after the CS (Hanajima et al. 1998). Thus the phenomenon of ICF is likely to be influenced by glutamatergic facilitation tempered by persisting GABAergic inhibition.
The interactions between ICF and other physiological processes have not been explored extensively. Ziemann et al. (1996c) demonstrated in a triple pulse TMS protocol that SICI and ICF can be shown to interact in an approximately linear relation, e.g. strong SICI might abolish ICF, further supporting a different origin for these two processes. In another triple pulse TMS protocol, ICF tested in M1 in the setting of cerebello-M1 inhibition (described below) appears to be enhanced (Daskalakis et al. 2004). However, a within-group correlation analysis suggested that this is likely to be due to a reduction in the SICI component rather than an increase in the excitatory component (tested at 10 ms), making a direct interaction with the excitatory population unlikely.
A different kind of facilitatory interaction can be demonstrated within M1 over shorter interstimulus intervals. This short interval intracortical facilitation (SICF, also known as I-wave facilitation) occurs when a suprathreshold stimulus (S1, in this case considered as the test stimulus, TS) is followed by a subthreshold stimulus (S2, in this case considered as the conditioning stimulus, CS) (Ziemann et al. 1998c), or alternatively when two stimuli near motor threshold are given consecutively (Tokimura et al. 1996). Using this approach, facilitation can be demonstrated at three distinct ISIs after the first stimulus: 1.1–1.5, 2.3–2.9 and 4.1–4.4 ms. If S2 is fixed at 90% of resting motor threshold (RMT) and the intensity of S1 is gradually increased, the first facilitatory peak is observed with an S1 of 70% RMT: further increasing the intensity of S1 produces second and third peaks at approximately 90% and 100%, respectively, with latencies that shorten with increasing S1 intensity. This effect is absent if S1 precedes a transcranial electric (instead of magnetic) S2, implying a cortical site of such facilitation, and it was proposed that the three facilitatory peaks observed reflect the generation of subsequent I-waves by S1 (Tokimura et al. 1996; Ziemann et al. 1998c). This was demonstrated conclusively for the earliest such peak by showing similar effects in the descending volleys generated by such stimuli in cervical epidural electrodes (DiLazzaro et al. 1999). A study of the precise timings of these interactions and their relation to stimulus intensity shed light on the contrast between this phenomenon and that mediating inhibition at similar intervals. If S1 < S2 (with S1 subthreshold and S2 suprathreshold) inhibition occurs mainly in the I3-wave. By contrast, if S1 = S2 or S1 > S2 (with S1 suprathreshold), facilitation occurs, but this is primarily in the I2 (or even I1) wave latency range. Thus the facilitation appears to take place one I-wave cycle earlier than the inhibition. Ilic and colleagues have proposed that this is because after a suprathreshold S1 the excitatory interneurons mediating the later I-waves are still hyperexcitable at the time of the earlier I-waves resulting from S2 (Ilic et al. 2002). Thus, while SICI and ICF are mediated via a trans-synaptic action on excitatory interneurons, SICF may instead involve a direct action on the initial axon segment of these excitatory interneurons (Ilic et al. 2002).
The effects of SICF are suppressed in the period following a peripheral sensory stimulus, suggesting an inhibitory interaction between afferent inputs and the interneuron populations responsible for I-wave generation. There is also a suppression of ICF in this context, but this occurs at lower CS intensities than for SICF (Zittel et al. 2006), reinforcing the hypothesis of different mechanisms for these two phenomena.
Inhibition within M1. Two principal types of local intracortical inhibition can be studied using paired pulse TMS. Short interval intracortical inhibition (SICI) was first described by Kujirai et al. (1993) and can be elicited by a subthreshold CS followed by supra threshold TS. At interstimulus intervals (ISIs) of 1–6 ms the test motor response is inhibited by the conditioning shock. Two main phases of inhibition have been described, at ISIs of 1 ms and 2.5 ms (Fisher et al. 2002; Roshan et al. 2003). Based on indirect evidence such as the lack of change in spinal reflexes, Kujirai et al. (1993) originally suggested that SICI was the result of synaptic interactions occurring within M1. A later study used direct recordings of descending spinal cord volleys to confirm that the initial I1-wave was suppressed by the CS, indicating that SICI seems to be mediated at the cortical level (Nakamura et al. 1997). An important limitation of this study was that the intensity of the CS was relatively large, raising the possibility that the CS alone could depolarize the axon, causing subsequent refractoriness during TS delivery. Therefore, it was not until 1998 that Di Lazzaro and collaborators provided the first direct evidence that SICI originated at the cortical level (DiLazzaro et al. 1998). In their study, a subthreshold CS suppressed the size of both the descending spinal cord volleys and the MEP evoked by the suprathreshold TS. Inhibition of the descending spinal volleys was most pronounced at an ISI of 1 ms and disappeared by 5 ms. This inhibition was evident for all later I-waves but not the I1-wave. Pharmacological studies have continued to provide more detailed information about the mechanisms of SICI. It has been shown that GABAA agonists enhance SICI (Ziemann et al. 1996a; Ilic et al. 2002). However, a single dose of the GABAA antagonist flumazenil did not alter SICI, suggesting that there might be no tonic activity at the benzodiazepine binding site of the GABAA receptor in the normal human M1 (Jung et al. 2004). It has also become apparent that inhibition at the short ISI of 1 ms does not depend on GABAA, while ‘true’ SICI at an ISI of 2.5 ms is likely to be mediated by GABAergic inhibition at the intracortical level (Fisher et al. 2002; Roshan et al. 2003), supporting the point of view that they are mediated by different mechanisms. Previously, Fisher et al. (2002) proposed that SICI at an ISI of 1 ms may be due to refractoriness or changes in axonal excitability of excitatory interneurons. In this scenario the subthreshold CS would convey excitatory interneurons into the refractory state, leading to less impact of the TS reflected as inhibition. Thus, with increasing TS intensity less inhibition would be expected, as the TS would activate more non-refractory interneurons. However, the fact that SICI at 1 ms increases with TS intensity at rest and decreases with voluntary muscle contraction (Roshan et al. 2003) argues against simple axonal refractoriness. In addition, a recent study showed that SICI at an ISI of 1 and 2.5 ms decreased to a similar extent during the cortical silent period (CSP) (Ni et al. 2007). Since the CSP is unlikely to affect axonal refractoriness, a synaptic mechanism is very likely responsible for SICI at 1 ms ISI.
While SICI can be considered as a well-characterised ‘standard’ TMS parameter, much less is known about the inhibitory phenomenon occurring at longer interstimulus intervals. Long interval intracortical inhibition (LICI) is elicited by a suprathreshold CS and TS applied at ISIs of approximately 50–200 ms (Valls-Sole et al. 1992; Wassermann et al. 1996) – thus two MEPs are elicited, of which the second is smaller in amplitude. Previous evidence has suggested that LICI at ISIs longer than 50 ms is mediated within M1 rather than subcortical structures (Nakamura et al. 1997). Although this evidence supports the view that LICI is related to reduced corticofugal excitability, it still remains unclear whether the same population of neurons mediates LICI and SICI. Pharmacological studies suggest that LICI is mediated by GABAB receptors (Werhahn et al. 1999; McDonnell et al. 2006) while SICI is primarily mediated by GABAA receptors (Ziemann, 2003). Nevertheless, the involvement of different receptor subtypes does not in itself exclude the possibility of a shared neuronal population mediating these two inhibitory phenomena.
Recent studies have shown that SICI and LICI interact with each other. SICI increases with higher test MEP amplitudes, while LICI decreases with higher test MEP amplitudes (Chen & Curra, 2004). These findings suggest that motor cortical neurons recruited at low TS intensities are more susceptible to LICI than to SICI, while those recruited at higher intensities appear to be more susceptible to SICI than LICI. On this basis, it is likely that different populations of inhibitory interneurons mediate LICI and SICI. In addition, previous evidence has shown that SICI is reduced in the presence of LICI at matched size of test MEP amplitude and test stimulus intensity, suggesting an inhibitory effect of LICI on SICI (Chen & Curra, 2004). While most studies of SICI and ICF have been implemented using distal hand muscles, it has been shown that relatively similar phenomena occur also in more proximal arm representations as well (Chen et al. 1998).
Interhemispheric interactions (M1–M1)
The interhemispheric interactions between homologous M1s currently described, and their relationships to intracortical processes, are illustrated in Fig. 3.
|
Interhemispheric inhibition between primary motor cortices. In contrast to interhemispheric facilitation, interhemispheric inhibition (IHI) is more robust and occurs over a wide range of ISIs (6–50 ms) (Ferbert et al. 1992; Daskalakis et al. 2002). This form of inhibition is lacking in patients with ischaemic lesions affecting transcallosal populations, supporting the idea that this phenomenon is mediated via the corpus callosum (Boroojerdi et al. 1996). Emerging evidence suggests that IHI elicited at relatively short ISIs (e.g. 8–10 ms) is mediated by different mechanisms than that elicited at longer intervals (e.g. 40 ms). Therefore short (IHI10) and long (IHI40) interval IHI will be discussed separately. Other than the ISI, the stimulation parameters required to elicit IHI10 and IHI40 are similar. Both require a suprathreshold CS and TS intensity adequate to elicit an MEP of 0.5–1.5 mV in amplitude (Kukaswadia et al. 2005). Both are also believed to be dependent on GABAB-mediated neurotransmission in the target hemisphere (IHI10: Daskalakis et al. 2002; Kukaswadia et al. 2005; IHI40: Kukaswadia et al. 2005). This was confirmed for long latency IHI by a recent study of pharmacological modulation by GABA agonists: IHI at ISIs of up to 200 ms was strengthened after application of the GABAB agonist baclofen, suggesting that long interval IHI is most likely mediated by postsynaptic GABAB receptors (Irlbacher et al. 2007).
Interactions within the target hemisphere
Evidence for differing mechanisms of IHI at these two ISIs comes primarily from studies of their interactions with other inhibitory phenomena. A number of studies have examined the interactions of such phenomena with IHI within the ‘target’ hemisphere (i.e. that receiving the inhibition), and it has been suggested that LICI and IHI40 may be mediated by an overlapping population of inhibitory neurons. As nicely reviewed by Kukaswadia et al. (2005) the evidence for this is three-fold: (1) both parameters preferentially affect lower threshold M1 interneurons (Gerloff et al. 1998; Daskalakis et al. 2002); (2) both require a suprathreshold CS (Kujirai et al. 1993; Daskalakis et al. 2002; Chen et al. 2003); and (3) both inhibit SICI in the receiving hemisphere (Sanger et al. 2001; Chen, 2004). However, a third phenomenon, long afferent inhibition (LAI – discussed below), helps to shed more light on the relationship between IHI and LICI. It is known that LAI directly inhibits LICI (Sailer et al. 2002; Chen, 2004). Kukaswadia et al. (2005) found that LAI also directly inhibits IHI40. They therefore concluded that LICI is probably more closely related to IHI40 than to IHI10. This idea is consistent with the finding that IHI8 decreases with voluntary muscle activation (Chen et al. 2003), while IHI40 (Chen et al. 2003) and LICI (Valls-Sole et al. 1992; Wassermann et al. 1996) both show little change. Thus far, the differential effects of LICI on IHI40 versus IHI10 in the target hemisphere have not yet to our knowledge been tested directly. The relationships between SICI and IHI40 versus IHI10 have likewise not been directly compared (in fact, little is known regarding the relationship between SICI and IHI40). However, it is known that IHI10 inhibits SICI in the target hemisphere and is decreased in the presence of LAI (Kukaswadia et al. 2005). This occurs only under certain conditions and, more importantly, the amount of decrease in IHI10 is not related to the strength of LAI or IHI10. Thus, LAI probably does not inhibit IHI10 directly, but both may act on a similar neuronal population (Gilio et al. 2003). In contrast, LAI strongly inhibits IHI40, in most cases changing inhibition to facilitation, and the decrease in IHI40 is directly related to the strength of LAI (Kukaswadia et al. 2005). Therefore, LAI and IHI40 seem to show a direct inhibitory interaction, with LAI inhibiting IHI40.
Interactions within the conditioning hemisphere. The above studies describe intracortical interactions with IHI within the target hemisphere. A recent study has examined the effects of such intracortical interactions within the conditioning hemisphere on IHI targeting the contralateral hemisphere (Lee et al. 2007). IHI10 and IHI40 were elicited with the CS in the presence or absence of SICI, LICI or ICF (with stimulus intensities adjusted to maintain MEP amplitudes). Both forms of IHI were suppressed in the presence of SICI or LICI but were unaffected by ICF. This last result could be interpreted as suggesting that the corticospinal output and the transcallosal projections mediating IHI arise from different neuronal populations. However, this conclusion should be guarded in view of the recent data casting doubt on the cortical site of ICF's action (Di Lazzaro et al. 2006). Moreover, in a separate study the effect of SICF within the conditioning hemisphere on IHI40 (but not IHI10) was examined (Avanzino et al. 2007). IHI was enhanced by SICF at identical latencies to the facilitation of the contralateral MEP (1.5 ms and 3.0 ms, coinciding with the I1 and I2 waves, respectively), suggesting that circuits modulating transcallosal and corticospinal projections have at least similar properties. The fact that IHI10 and IHI40 are affected similarly by SICI, LICI and ICF (Lee et al. 2007) raises the possibility that a shared population may convey each form of inhibition across the corpus callosum, with different target populations influenced at different latencies. However, the impact of SICI on IHI might depend on the intensity of the CS for SICI (Kujirai et al. 1993), which was kept constant in the study.
Summary of interhemispheric M1–M1 interactions. As described above, the approach of paired pulse TMS between the two M1 hand areas has revealed at least three facilitatory and two inhibitory distinct interactions, depending on the parameters used (ISI, coil orientation and intensities of CS and TS). Facilitation or inhibition can even be produced at overlapping ISIs, depending on the nature of the CS and TS, suggesting that such interactions are likely to occur in parallel. With regard to the cell populations involved, it is likely that even in the case of IHI the transcallosal projections are excitatory, synapsing onto local inhibitory circuits within the target hemisphere. From such manipulation of physiological parameters as described above it is not possible to infer whether the various phenomena are mediated by distinct transcallosal populations or whether common projections are used with, for example, different coding characteristics. Consistent with these findings, it was reported that down-regulation of excitability of one motor cortex modulates cortical excitability in the opposite M1 and motor function in the ipsilateral hand in health (Schambra et al. 2003; Kobayashi et al. 2004; Johansen-Berg et al. 2007) and disease (Ward & Cohen, 2004; Talelli et al. 2006; Fregni & Pascual-Leone, 2006).
Inter-regional interactions
The inter-regional cortico–cortical interactions currently described that are known to modulate the output of M1 (excluding M1-to-M1) are illustrated in Fig. 4.
|
The second approach has employed two coils in a paired pulse protocol. Civardi et al. (2001) showed that a subthreshold CS (defined by the M1 MEP threshold) over PMd reduces the excitability of the ipsilateral M1, with a maximum effect at an ISI of 6 ms – this ipsilateral PMd–M1 inhibition requires a CS given at 90% of AMT with antero-posterior current flow. The authors argued that this interaction did indeed involve conditioning PMd rather than acting via current spread to M1, on the basis of spatial separation (conditioning at an intermediate point produces no inhibition), temporal separation (a time course that is distinct from SICI) and the effect of coil orientation (Civardi et al. 2001). However, in addition to this inhibitory interaction, facilitation could also be elicited if a higher conditioning intensity was used (120% AMT).
Mochizuki et al. (2004a) applied a similar paired pulse approach to investigate the interhemispheric interaction between PMd and the contralateral M1. At ISIs between 4 and 20 ms (with a CS to the right PMd and a TS to the left M1), they found significant inhibition of the test MEP using a CS intensity of either 90% RMT (with an ISI of 8 ms) or 110% RMT (ISI of 8–10 ms). This interhemispheric PMd–M1 inhibition is spatially specific for PMd (as not detected when stimulating 2 cm anterior, lateral or medial to the target area) but not for the hemisphere (Baumer et al. 2006). Stimulation of the left PMd and right M1 revealed the same results (Baumer et al. 2006; Koch et al. 2006). This interaction can be distinguished from M1-to-M1 IHI at the 90% CS intensity (on the basis of a lower threshold and differing effects of voluntary contraction) but this distinction is less clear at the suprathreshold intensity. Interhemispheric PMd–M1 inhibition has also been described at the longer interstimulus interval of 150 ms, using a latero-medial CS at 110% of AMT (Mochizuki et al. 2004b), but at such a long interval the effect cannot be assumed to be transmitted transcallosally. The effect of PMd stimulation on the contralateral M1 seems to depend on the stimulation intensities used, as demonstrated recently by Baumer et al. (2006). Conditioning the left PMd with low stimulus intensity (80% of AMT) and targeting the right M1 (with small test MEPs), interhemispheric PMd–M1 facilitation was described at an ISI of 8 ms (Baumer et al. 2006). This facilitation was dependent on a postero-anterior current flow for the TS, providing indirect evidence that this form of facilitation preferentially affects I1 waves in the target hemisphere.
A mechanism proposed for these interregional effects is the activation of long distance projections from the PMd to ipsi- or contralateral M1, consistent with anatomical studies showing dense connections between those areas, which are known to be both inhibitory and facilitatory (Ghosh & Porter, 1988; Tokuno & Nambu, 2000). Details of how these long-range projections interact with the intracortical circuits described above are not well known. The only study to directly address this question has been Mochizuki et al. (2004a), who showed that interhemispheric PMd–M1 inhibition was associated with a reduction in intracortical inhibition in the target hemisphere (SICI at ISI of 2 ms).
M1–SMA interactions. The supplementary motor area (SMA) is by contrast a more difficult area to target than the lateral premotor cortex (PMd and PMv), as it is located in the interhemispheric fissure, relatively unexposed on the surface of the hemisphere. TMS can reliably elicit MEPs from leg muscles by stimulating the leg area of M1 adjacent to SMA (Gerloff et al. 1997; Perez et al. 2004). Stimulation of the SMA is therefore possible if cortical elements targeted by TMS have a threshold comparable in the lower limb area of M1 and in the SMA. However, there are few electrophysiological studies of SMA stimulation in healthy subjects. Civardi et al. (2001) found that a conditioning stimulus applied over the SMA, defined as a cortical area 3 cm anterior to the M1 leg area (1–4 cm anterior to Cz), reduced the excitability of the ipsilateral M1 at an ISI of 6 ms, indicating that SMA stimulation is likely to lead to changes in activity in anatomically connected regions in a way similar to that seen after stimulation of PMd (Civardi et al. 2001). Matsunaga et al. (2005) used 5 Hz suprathreshold rTMS (110% AMT) over the SMA to investigate the effects on ipsilateral M1 excitability. Stimulation increased the MEP amplitudes (as for PMd), but in this case SICI/SICF was unchanged as were cortical silent period and H-reflexes. Thus an inhibitory interaction is suggested by the paired pulse approach (Civardi et al. 2001), whereas excitatory rTMS appears to cause facilitation (Matsunaga et al. 2005).
Posterior parietal cortex (PPC). The two coil paired-pulse approach has recently also revealed a facilitatory interaction between the posterior parietal cortex (PPC, defined as the P4 position on the 10–20 EEG system) and both the ipsilateral and contralateral M1 (Koch et al. 2007). Significant facilitation of an MEP elicited from the ipsilateral M1 was observed at ISIs of 4 ms and 15 ms when a CS of 90% RMT was used (postero-anterior current direction). This PPC–M1 facilitation was not seen at higher or lower CS intensities, or with the opposite conditioning coil orientation. Single motor unit recordings suggested that PPC stimulation enhances the I3-wave component of the test MEP, a finding that may explain the surprisingly short ISI of 4 ms (as late I-waves may take as long as 7 ms to leave the motor cortex – Day et al. 1989). In addition to the ipsilateral effect, facilitation of MEPs elicited from the contralateral M1 was also observed. This was seen at a CS intensity of 90% RMT and ISIs of 6 ms and 12 ms. The precise anatomical pathways mediating these effects are not known, and it may be that such facilitation involves parieto-premotor projections (which are known to be more numerous than direct parieto-motor projections). An interesting feature of this effect is that, unlike premotor-M1 or M1–M1 interactions inhibition was not observed at any of the CS intensities tested.
Afferent input and somatosensory cortex. Although peripheral nerve stimulation provides a strong and temporally precise afferent input, it stimulates a mixed population of nerve fibres, including muscle afferents, cutaneous afferents, joint afferents and motor efferents. Hence, it is not surprising that studies of the effects of peripheral nerve stimulation on cortical excitability provide a set of heterogeneous results whose underlying mechanisms and origin of the interactions are difficult to interpret. Two different approaches are most frequently used: (1) a paired-pulse protocol combining a peripheral and a cortical stimulus to study sensorimotor interactions, and (2) repetitive peripheral nerve stimulation as a tool to elicit changes in cortical excitability (with or without implementing the paired-pulse technique).
Activity in afferent pathways can condition motor cortical excitability in a paired-pulse protocol (summarised in Fig. 5). For example, a conditionings electrical stimulus applied to a mixed nerve (most often the median or digital nerve at the wrist) has an inhibitory effect on motor cortex excitability. These effects, more evident at ISIs of 20 ms and 200 ms, are described as short- (SAI) and long-latency afferent inhibition (LAI), respectively (Tokimura et al. 2000).
|
Indirect evidence for a cortical site of action of LAI originates from the finding that F-wave amplitudes remain unchanged at an ISI of 200 ms (Chen et al. 1999). As for SAI, it remains to be determined if this effect is mediated through direct somatosensory projections to M1 or indirectly through the primary somatosensory cortex. As mentioned above, in the presence of LAI, both LICI and IHI40 are reduced (Sailer et al. 2002; Kukaswadia et al. 2005).
Similar results have been shown for cutaneous stimulation of digital nerves. MEPs were inhibited when a TMS pulse was delivered 25–50 ms after homotopic stimulation of a digital nerve (Classen et al. 2000; Tamburin et al. 2005). However, when the TMS pulse preceded the cutaneous stimulus (ISI 16–22 ms) or if the ISI was longer than 50 ms (up to 200 ms) MEPs were facilitated. In a muscle heterotopic to the site of the cutaneous stimulus a reversed pattern of MEP size was found. In the presence of a cutaneous stimulus applied 
35 ms before a conditioning TMS pulse, a reduction in SICI was described (Ridding & Rothwell, 1999), suggesting that the afferent input provoked by the digital stimulus had a direct effect on circuits involved in intracortical inhibition, most likely as an interference with later I-waves weakening the efficacy of the cortical conditioning stimulus.
Introduced by Stefan et al. (2000) another approach combines peripheral and low frequency cortical stimulation in a repetitive, timing-specific pattern. Resembling mechanisms of associative plasticity in animal slice preparations by pairing pre- and postsynaptic action potentials (Wigstrom et al. 1986; Markram et al. 1997), this paired associative stimulation (PAS) protocol consists of a peripheral stimulus (most often electrical stimulation of the median nerve at the wrist) that is followed by a suprathreshold TMS stimulus to the contralateral M1 targeting a muscle innervated by the stimulated nerve.
The ISI is determined by the time lag in evoking an MEP from M1 via activation of the primary somatosensory cortex (S1). The shorter the ISI for facilitatory PAS (shorter than 25 ms), the more likely later inputs to corticospinal neurons are targeted by PAS, as the afferent input would arrive after firing of the initial input (I1 input) produced by the TMS pulse. Thus, if the peripheral stimulus is given approximately at the N20 latency of a somatosensory evoked potential (SEP) plus 1–4 ms for the S1 to M1 transit time (ISI of approximately 20–25 ms) an increase in the conditioned MEP can be found, if the PAS procedure is repeated for a period of 30 min (Stefan et al. 2000). Conversely, at even shorter ISIs, most often 10 ms, a decrease of cortical excitability as measured by a reduced MEP size is found after repeated stimulation (Wolters et al. 2003).
In a relaxed muscle, the intensity of the peripheral stimulus and the TMS stimulus used for PAS need to be suprathreshold to induce long-lasting changes in cortical excitability (Stefan et al. 2000; Wolters et al. 2003). The reason for this is still unclear, but it is thought that associative plasticity requires either a certain amount of synaptic activity or the neuronal population that is targeted by PAS might have high thresholds for activation. Interestingly, weak voluntary muscle contraction further enhances the after-effect of PAS compared to a resting condition (Kujirai et al. 2006). Also, the direction of the current flow in the brain following a subthreshold TMS pulse significantly alters the effectiveness of this procedure: tested during voluntary contraction, PAS using subthreshold TMS with AP current flow and 25 ms ISI is superior to PAS using subthreshold TMS with PA current flow (Kujirai et al. 2006) in eliciting excitability changes, presumably reflecting the later arrival of inputs preferentially activated by AP pulses (I3 input; Di Lazzaro et al. 2001).
The after-effects of PAS are relatively long lasting (duration up to 90 min) and have topographical specificity (Stefan et al. 2000). Furthermore they can be abolished by the application of the N-methyl-D-aspartate (NMDA) receptor antagonist dextromethorphan (Stefan et al. 2002; Wolters et al. 2003). Recently it was described that motor learning prior to PAS can also prevent induction of the LTP-like plasticity in M1 for several hours (Stefan et al. 2006) suggesting the occlusion of further plastic changes after maximized LTP following training. Conversely, Ziemann et al. (2004) found even greater reduction in cortical excitability following a PAS protocol that usually elicits LTD-like plasticity when it was preceded by motor learning, supporting the idea that PAS exerts its action via LTP/LTD-like mechanisms.
A modulation of cortical excitability can also be elicited by repetitive mixed peripheral nerve stimulation (PNS). Trains of five slightly suprathreshold pulses of 1 ms duration delivered at 10 Hz for at least 1.5 h resulted in a somatotopically specific increase of MEPs only in muscles innervated by the stimulated nerve (Ridding et al. 2000a; Ridding & Taylor, 2001; Kaelin-Lang et al. 2002), outlasting the end of the stimulation by 20 min (Kaelin-Lang et al. 2002). The somatotopy and the fact that MEPs and maximal peripheral M-waves were not altered in response to electrical brainstem stimulation suggest a cortical site of action (Kaelin-Lang et al. 2002), consistent with the lack of alteration of F-waves (Ridding et al. 2000a). No changes in motor thresholds (RMT and AMT), SICI and ICF have been found after repetitive mixed PNS, but pharmacological blockage of the effect of PNS was seen after administration of lorazepam, a positive allosteric modulator of the GABAA receptor (Kaelin-Lang et al. 2002). PNS has also been shown to increase the beneficial effects of motor training when applied for a period of 1–2 h immediately preceding the motor training period (Kaelin-Lang et al. 2005; Sawaki et al. 2006). Interestingly, proprioceptive input originating in the training motions does not appear to be sufficient to elicit substantial changes in cortical plasticity (Kaelin-Lang et al. 2005; Lotze et al. 2003).
Another form of somatosensory input is provided by low amplitude muscle vibration, which stimulates predominantly large Ia fibres and can mimic joint proprioception (Burke et al. 1976), clearly influencing excitability in somatosensory pathways (Cohen & Starr, 1985). If M1 corticospinal excitability is tested using a TMS pulse after 1 s of hand muscle vibration, MEP amplitudes are found to increase in the vibrated muscle while decreasing in adjacent non-vibrated muscles (Rosenkranz & Rothwell, 2003). Although muscle vibration certainly alters spinal excitability (Claus et al. 1988), there are associated changes in paired pulse TMS parameters which strongly suggest an effect at the level of M1: SICI targeting the vibrated muscle is reduced, while LICI is enhanced (the converse changes are seen in surrounding muscles, Rosenkranz & Rothwell, 2003). This muscle-specific surround inhibition suggests that the effects of proprioceptive input on M1 are more spatially specific at rest than those of cutaneous inputs, which give rise to less exquisitely somatotopic changes (Classen et al. 2000; Tamburin et al. 2001). Such a cortical change in response to muscle vibration is consistent with the observation in baboons that proprioceptive afferent input, unlike cutaneous input, projects directly to the motor cortex (Hore et al. 1976). Using a two-coil paired pulse approach, it has been shown that vibration of a hand muscle is also associated with stronger IHI targeting the motor cortical representation of the contralateral homologous muscle, with increased SICI and reduced MEP amplitudes in that muscle (Swayne et al. 2006).
Cutaneous anaesthesia of one hand increases MEP amplitudes in muscles immediately proximal to the deafferented hand (Ziemann et al. 1998b; Brazil-Neto et al. 1993) and in hand muscles in the unanaesthesized hand (Werhahn et al. 2002b) in the absence of excitability changes in other body part representations. This effect was blocked by a positive modulator of the GABAA receptor, lorazepam. MEPs resulting from brainstem electrical stimulation remained unchanged suggesting that the effect is probably of cortical origin. Additionally, IHI targeting the unanaesthesized hand muscles decreased during the anaesthetic procedure (Werhahn et al. 2002b). These results were interpreted as indicative that acute hand deafferentation can elicit a focal increase in cortical excitability in the hand motor representation contralateral to the deafferented cortex that is influenced by transcallosal interactions and GABAergic neurotransmission. Interestingly, these effects appeared to rebalance in the setting of chronic deafferentation following amputations. In the somatosensory domain, cutaneous anaesthesia of one hand results in focal rapid improvements in tactile spatial acuity in the opposite hand that are accompanied by increased cortical SEP amplitudes elicited by stimulation of the unanaesthesised hand (Werhahn et al. 2002b). These results are consistent with the idea that deafferentation of a cortical representation influences the homotopic representation in the opposite hemisphere; perhaps supporting the unanaesthesised hand's need to tackle enhanced environmental requirements, and is consistent with interhemispheric competition models of sensory processing. It is of relevance that these principles appear to operate also after cortical lesions like stroke, in which cutaneous anaesthesia of a healthy hand exerts beneficial effects on motor function of a paretic hand after stroke in both motor and somatosensory domains (Floel et al. 2004; Voller et al. 2005).
Cerebello-thalamo-cortical interactions. The most distant brain area over which TMS has been shown to modulate the motor cortical output is the cerebellum (summarised in Fig. 6). Cerebello-cortical (CbC) interactions in humans were originally described by Ugawa et al. (1995). They investigated how a CS over the cerebellum influences the amplitude of an MEP elicited by a subsequent TS over the contralateral M1. Either an electrical (Ugawa et al. 1991) or a magnetic cerebellar CS (Ugawa et al. 1995) resulted in a net suppression of corticomotor excitability at ISIs of 5–7 ms, as reflected in the decreased amplitude of MEPs elicited by a magnetic TS over M1 (Ugawa et al. 1995). In contrast, CbC inhibition was not observed when an electrical TS was applied over M1, suggesting an interaction upstream of the corticospinal neurons. Inhibition of TMS-induced MEP in M1 by a magnetic cerebellar CS has been consistently replicated (Pinto & Chen, 2001; Daskalakis et al. 2004). CbC inhibition can be best obtained with a double-cone coil positioned 3–5 cm lateral to the inion, with the induced current flowing upward in the cerebellar cortex (Meyer et al. 1994; Werhahn et al. 1996). The intensity of the cerebellar CS is usually set at 5–10% below AMT for direct recruitment of the corticospinal tract at the level of the foramen magnum when the double-cone coil is placed over the inion (Werhahn et al. 1996).
|
As recently suggested by studies in patients with a variety of strokes, the key cerebellar structures involved in CbC interactions elicited by a magnetic cerebellar CS are the superior cerebellum and the dentate nucleus (Liepert et al. 2004; Battaglia et al. 2006). The dentate nucleus exerts a background tonic facilitatory drive onto the contralateral M1 through synaptic relay in the ventral lateral thalamus. This dentato-thalamo-cortical pathway is one of the many cerebello-cortical loops that specifically link cerebellar and cortical areas through dedicated channels (Middleton & Strick, 2000; Dum et al. 2002; Ramnani, 2006). The activity of the dentate nucleus is under the inhibitory control of the Purkinje cells, whose axons are the exclusive output of the cerebellar cortex. It has been proposed that a magnetic cerebellar CS activates the Purkinje cells; this results in an inhibition of the dentate nucleus that leads in turn to a disfacilitation of the contralateral M1, due to a reduction in dentato-thalamo-cortical facilitatory drive (Pinto & Chen, 2001; Daskalakis et al. 2004). However, this is still under debate as the short ISIs of 5 ms to elicit inhibition of the dentate nuclei would necessitate an extremely fast inhibitory system. Moreover, data from rTMS studies over the cerebellum inconclusively showed a reduction (Fierro et al. 2007) or an increase (Oliveri et al. 2005) of ICF in M1, but no changes in SICI. However, as reviewed above, the origin of ICF at shorter interstimulus intervals may be mediated by subtle changes in spinal excitability, further supporting the possibility of peripheral effects of magnetic stimulation of the cerebellum. One should remark that the use of a flat figure-of-eight coil in these studies would promote simultaneous activation of afferent peripheral nerve fibres in the brachial plexus (Ugawa et al. 1995; Werhahn et al. 1996) and therefore might lead to this result. Thus, a specific inhibitory effect of rTMS on the dentate-thalamo-cortical pathway has yet to be proven.
CbC inhibition is more pronounced for small test MEPs (0.5 mV) elicited by slightly suprathreshold TS than for large test MEPs (2 mV) (Ugawa et al. 1995; Pinto & Chen, 2001). This may reflect either a preferential inhibition of the neuronal elements generating the I1 wave (which have a lower threshold than the D-wave and later I-waves) or the fact that the dentato-thalamo-cortical pathway projects predominantly to the core of cortical muscle representations, where the motor threshold may be lower (Pinto & Chen, 2001). The cell populations within M1 targeted by these projections are likely to include both pyramidal cells and inhibitory interneurons (Shinoda et al. 1993; Daskalakis et al. 2004).
CbC interactions with the intracortical populations mediating SICI, ICF and LICI have been tested by Daskalakis et al. (2004), who used a triple-pulse TMS protocol with small adjustments of the test MEP amplitudes. A magnetic cerebellar CS reduces SICI in the opposite M1, most likely through reduced facilitatory dentato-thalamo-cortical drive to intracortical inhibitory interneurones. This reduction in SICI may shift the intracortical balance of excitability toward excitation, leading to the observed increase in ICF. Finally, in the presence of LICI, CbC inhibition is decreased. The mechanism of this interaction is unclear and could result either from a saturation effect if LICI and CbC inhibition converge onto the same population of cortical inhibitory interneurons, or alternatively from changes in subcortical excitability.
Further evidence for the presence of a tonic facilitatory drive from the dentate-thalamo-cortical pathway onto M1 in healthy humans is provided by studies in patients with cerebellar stroke or degeneration. These have consistently demonstrated an increased RMT in the contralateral M1 (as well as increased SICI and decreased ICF) (Liepert et al. 2004; Battaglia et al. 2006).
Interrupting tonic contraction: silent periods
Contralateral silent period. In a voluntarily contracted muscle, the MEP elicited by a single suprathreshold TMS pulse is followed by a period of EMG inhibition called the contralateral silent period (CSP) (Fuhr et al. 1991). While there is evidence that the early part of the CSP is mediated by spinal mechanisms, the later part is thought to result from suppression of neural output by interneurons at the cortical level (Fuhr et al. 1991; Tergau et al. 1999). Cracco et al. (1989) have shown that cortical stimulation excites inhibitory interneurons (probably Golgi-II cells with long axons) connected to the pyramidal cells, thus decreasing corticospinal neuron firing (Cracco et al. 1989). The CSP duration is greater following an antero-posterior and biphasic stimulus than a postero-anterior stimulus, and correlates strongly with the amplitude of the evoked MEP, raising the possibility that it may depend on activity in recurrent collaterals from discharging pyramidal tract neurons (Orth & Rothwell, 2004). The CSP has been reported to be prolonged following administration of either oral tiagabine (a GABA re-uptake inhibitor) or intrathecal baclofen (a GABAB agonist), suggesting that the CSP, like LICI, may be mediated by GABAB populations (Siebner et al. 1998; Werhahn et al. 1999). However, this effect of baclofen was not replicated in studies using oral (McDonnell et al. 2006) or intravenous administration (Inghilleri et al. 1996).
The study by Werhahn et al. (1999) also provided evidence of a reciprocal relationship between CSP and SICI, in that tiagabine increased CSP duration while weakening SICI. Daskalakis et al. (2006) used rTMS in order to evaluate the effects of several different stimulation frequencies (1, 10 and 20Hz) on SICI and CSP (Daskalakis et al. 2006). They showed that the rTMS-induced change in SP was associated with a change in SICI and this inverse relationship was greatest in the highest stimulation condition (i.e. 20 Hz). Recently, this interaction was explored in a human study by Ni et al. (2007). The authors used a triple-pulse protocol, investigating SICI and ICF during different time points of the CSP. While SICI was decreased (80–140 ms following the stimulus that induced the CSP) ICF was increased, followed by normalization of both parameters after termination of the CSP. Since the lack of SICI was already present at low CS intensities (60% aMT) and the threshold of inhibitory interneurons is known to be lower than that of facilitatory interneurons (Chen et al. 1998), the decrease of SICI during the CSP is likely to be due to a decreased inhibition rather than increased facilitation. This relationship may be seen as analogous with the suppression of SICI in the presence of LICI, another GABAB-mediated phenomenon (Sanger et al. 2001), and is consistent with previous lines of evidence from both animal and human studies demonstrating that activation of presynaptic GABAB receptors inhibits further release of GABA (Deisz, 1999). Interestingly, the inhibition of SICI by LICI was also observed during the CSP (Ni et al. 2007), further supporting this hypothesis.
Ipsilateral silent period. Application of a single suprathreshold TMS pulse to the M1 ipsilateral to a tonic voluntary contraction can cause an interruption of the ongoing voluntary EMG activity known as the ipsilateral silent period (iSP), even in the absence of an ipsilateral MEP (Ferbert et al. 1992; Meyer et al. 1995). Several lines of evidence suggest that the iSP is mediated by fibres passing through the corpus callosum: iSPs were absent or delayed in patients with agenesis or surgical lesions of the corpus callosum (Meyer et al. 1995), but were preserved in patients with subcortical cerebrovascular lesions that interrupted the corticospinal tract but spared the corpus callosum (Boroojerdi et al. 1996). In addition, in young children the iSP is significantly shorter than in adults; the protracted development and myelination of the corpus callosum are paralleled by the appearance and strengthening of the iSP (Heinen et al. 1998).
Several studies have investigated the iSP's relationship to IHI (Ferbert et al. 1992; Chen et al. 2003). Chen et al. (2003) have examined the effects of different stimulus intensities and current directions on the two forms of interhemispheric inhibition. They showed that paired-pulse IHI measured with a 40 ms ISI, both at rest and during muscle activation, significantly correlated with iSP duration for some of the stimulus intensities and current directions tested, while IHI at an ISI of 8 ms and iSP did not correlate under any of the experimental conditions. These results suggest that while common neuronal populations may mediate IHI40 and iSP, the same is not true of IHI8. Thus iSP and IHI8 are separate phenomena, mediated perhaps through different sets of transcallosal fibres or, alternatively, different sets of effector neurons in the contralateral ‘target’ M1. The duration of the iSP can be modulated by a CS delivered to the stimulated hemisphere in a paired pulse protocol: the iSP can be suppressed by a subthreshold CS delivered 3 ms before the suprathreshold TS (Trompetto et al. 2004) or enhanced by a CS at motor threshold intensity delivered 1.5 or 3 ms after the TS (Avanzino et al. 2007). These are the same protocols used to elicit SICI and SICF, respectively, implying that the iSP is conditioned in a similar manner to the contralateral corticospinal output – the cell population giving rise to the transcallosal projection mediating the iSP seems therefore to be subject to similar modulation to the pyramidal output, although they do not necessarily have to be the same population.
State-dependent intra- and interhemispheric interactions
We have so far described different physiological interactions which modulate the output of M1 while the system is at rest, defined as muscle relaxation (except for the silent periods). It may be expected that the behaviour of these interactions should change depending on the behavioural state. If these parameters play a functional role in motor control then one may expect changes when subjects engage in preparation or performance of a motor task. Such movement-related changes have indeed been described for a number of these interactions, but there are still plenty of unknowns.
Changes affecting M1 during movement preparation.
Corticospinal and intracortical parameters can be assessed in the context of reaction time protocols, providing a picture of changing physiological interactions leading up to movement execution. Using single pulse TMS applied at a number of time points after a ‘Go’ cue (in a simple reaction time protocol), three studies have described a gradual increase in corticospinal excitability starting 80–120 ms prior to movement onset (Rossini et al. 1988; Leocani et al. 2000; Nikolova et al. 2006). In the study of Leocani et al., this finding was accompanied by a suppression of MEPs in the contralateral resting hand (if the dominant right hand was being moved). The role of excitability changes in the 
-motorneuron pool, which was not studied in detail in these early investigations, was evaluated more recently. It appears that the true ‘lead time’ between premovement excitability increases in the motor cortex and the spinal cord may be of the order of 10–15 ms: shorter than previously thought (MacKinnon & Rothwell, 2004; Schneider et al. 2004). This faster build-up of motor cortex excitability is perhaps not surprising when one considers that healthy volunteers may have a total reaction time of around 100 ms. Recent work suggests that motor cortical excitability is also modulated by the expectancy of the need to make a movement. In an elegant version of the simple reaction time task (SRTT), van Elswijk and colleagues manipulated the interval between a preparatory stimulus and a response stimulus in order to create four time intervals at which subjects had various expectancies of the likelihood of a cue to move. Not only were reaction times shorter with high cue expectancy (relative to intervals with a low expectancy), but MEP amplitudes to a single TMS pulse were also increased (van Elswijk et al. 2007). Thus it would seem that premovement modulation of M1 excitability is exquisitely sensitive to the precise nature of the upcoming task, and is modulated in advance of expected movements.
Paired pulse TMS can be used in a similar manner to investigate intracortical excitability changes in relation to movement. Reynolds & Ashby (1999) demonstrated that SICI begins to decrease approximately 95 ms prior to the onset of a phasic movement, and that this change is seen in the agonist but not antagonist muscle groups. As a local intracortical phenomenon, SICI would be well placed to modulate the relationship between adjacent intracortical representations via changes in horizontal connections. It was thus speculated that the reduction in SICI could contribute to the focal increase in corticospinal excitability affecting the target muscle (Reynolds & Ashby, 1999). Conversely, SICI targeting a neighbouring uninvolved hand muscle may become stronger in some subjects when tested in relation to phasic finger movements (Stinear & Byblow, 2003a). SICI also increases after a no-go signal in a go/no-go reaction task protocol (Sohn et al. 2002). These results are consistent with a role of SICI in actively suppressing execution of prepared movements. A comparison of synchronized versus syncopated externally paced finger movements has also suggested that movement-related SICI changes may be task dependent (Byblow & Stinear, 2006).
The precise timing of SICI changes has been recently investigated in a simple reaction time protocol, revealing that inhibition is in fact stronger more than 70 ms before movement onset, but is then progressively abolished relative to rest (Nikolova et al. 2006). A trend was also observed for ICF to become weaker from 150 ms before movement. This reduction in inhibition is likely in fact to occur closer to the onset of movement than described here, as this study did not test for early subtle increases in spinal excitability. With this in mind, it may be the case that the reduction in SICI occurs alongside (or later than) the increase in MEP amplitudes. If so, this would suggest that SICI modulation is unlikely to drive the corticospinal excitability increase, but may serve to focus it appropriately to the task. There is a further inherent difficulty in this kind of experiment in that premovement MEP facilitation may distort the degree to which apparent SICI in fact reflects activity in the inhibitory population. While the phasic movement experiments of Stinear & Byblow (2003b) make efforts to correct for this, it would be technically very challenging to do so across a range of time points – it is possible that such a consideration may affect the changes reported by Nikolova and colleagues. In the case of phasic movements the question arises as to whether afferent feedback, known to focally reduce SICI (Rosenkranz & Rothwell, 2003), may be responsible for the observed changes. However, reduced SICI has been observed during imagined thumb abduction movements, suggesting that afferent feedback is not necessary to produce these changes (Stinear & Byblow, 2003b). Thus it seems likely that both motor drive and afferent feedback may contribute to movement-related modulation of SICI.
Using a two coil approach it is also possible to test the activity of inter-regional interactions during movement preparation. If tested in a simple reaction time protocol, IHI targeting the moving hand is reversed to become IHF in the period immediately before movement onset (Murase et al. 2004). This effect is more prominent when tested for IHI targeting the dominant hand (Duque et al. 2007). It was suggested that this reversal of tonic inhibition may allow for accuracy of movement when the hands need to be used separately. In view of the importance of bimanual control in primate evolution it could be speculated that the role of interhemispheric interactions between the hand areas may differ between unimanual and bimanual tasks, but this has not been directly tested yet. Indirect evidence that this may be the case is provided by studying changes in MEP amplitudes (in response to a single TMS pulse) in a hand muscle before a bimanual movement. MEPs increase or decrease during this period depending on both the agonist–antagonist and kinematic relationships between the two moving fingers (Duque et al. 2005). This suggests that information describing such relationships is coded at the level of M1, but does not directly support a role for IHI/IHF in this process.
A similar approach has been employed to investigate activity in the inhibitory and facilitatory interhemispheric PMd–M1 interactions described above during a choice reaction time task (Koch et al. 2006). During movement preparation these investigators found a crucial timing dependence: the facilitatory effect on MEP amplitude was evident 75 ms after the cue if the target hand was being moved (but not the contralateral hand), while the inhibitory effect was evident 100 ms after the cue if the contralateral hand was being moved (but not the target hand). Interestingly, both the inhibitory and facilitatory influences of PMd on M1 were absent at all other time intervals before movement. The authors speculated that the reaction cue may initially cause both left and right hand movements to be specified, with the incorrect movement being eliminated at a later stage. The expectation of the need to move may thus cause the interhemispheric interactions to be suppressed, only for the relevant interaction to become active during the appropriate premovement time window. Thus, the left PMd exerts a brief facilitatory or inhibitory influence on the right M1 depending on which hand is to be selected to move, supporting a role for the left PMd in movement selection. This dependence of the interhemispheric PMd–M1 interaction on the motor state is in keeping with recent work which used the effect of a TMS input on haemodynamics in remote areas (during fMRI imaging) to assess functional connectivity. This approach also demonstrated that the PMd–M1 interaction is inhibitory at rest but facilitatory during movement preparation (Bestmann et al. 2007). A recent paper by Davare et al. (2006) supports this point of view, demonstrating that 1 Hz rTMS over the left PMd impairs movement preparation as tested in a pinch-lift task.
The brief periods of activity in the PMd–M1 interactions (Koch et al. 2006) occur considerably earlier than the excitability changes within M1 described above (although simple and choice reaction time protocols are being compared). This is consistent with the relative timings of the two regions as revealed by attempting to prolong rea
