Excitability changes in human peripheral nerve axons in a paradigm mimicking paired-pulse transcranial magnetic stimulation

Authors

  • Jane H. L. Chan,

    1. Prince of Wales Medical Research Institute, University of New South Wales and College of Health Sciences, University of Sydney, Sydney, Australia
    Search for more papers by this author
  • Cindy S.-Y. Lin,

    1. Prince of Wales Medical Research Institute, University of New South Wales and College of Health Sciences, University of Sydney, Sydney, Australia
    Search for more papers by this author
  • Emmanuel Pierrot-Deseilligny,

    1. Neurophysiologie Clinique, Rééducation, Hôpital de la Salpêtrière, 75651 Paris Cedex 13, France
    Search for more papers by this author
  • David Burke

    Corresponding author
    1. Prince of Wales Medical Research Institute, University of New South Wales and College of Health Sciences, University of Sydney, Sydney, Australia
    • Corresponding author D. Burke: Prince of Wales Medical Research Institute, Barker Street, Randwick, Sydney, NSW 2031, Australia. Email: D.Burke@unsw.edu.au

    Search for more papers by this author

Abstract

A peripheral nerve model was developed to determine whether changes in axonal excitability could affect the findings in studies of cortical processes using paired-pulse transcranial magnetic stimulation (TMS). The recovery of axonal excitability from a conditioning stimulus smaller than the test stimulus was qualitatively similar to that with suprathreshold conditioning stimuli. There was an initial decrease in excitability, equivalent to refractoriness at conditioning-test intervals < 4 ms, an increase in excitability, equivalent to supernormality, at intervals of 5–20 ms and a second phase of decreased excitability, equivalent to late subnormality at intervals > 30 ms. H reflex studies using conditioning stimuli below threshold for the H reflex established that these excitability changes could be faithfully translated across an excitatory synapse. Changing membrane potential by injecting polarising current altered axonal excitability in a predictable way, and produced results similar to those reported for many disease states using paired-pulse TMS. Specifically, axonal hyperpolarisation produced a smaller decrease in excitability followed by a greater increase in excitability. This study supports the view that changes in excitability of the stimulated axons should be considered before synaptic mechanisms are invoked in the interpretation of findings from paired-pulse TMS studies.

The activity of intracortical inhibitory and excitatory synaptic processes is often investigated in human subjects using paired-pulse transcranial magnetic stimulation (TMS) of the motor cortex, documenting changes in the motor evoked potentials (MEPs) recorded from a peripheral muscle. TMS is thought to activate the corticospinal system predominantly by a trans-synaptic mechanism, a view supported by direct recordings from the corticospinal tract documenting proportionately more I waves than D waves with TMS (Rothwell et al. 1991; Burke et al. 1993; Nakamura et al. 1997; Di Lazzaro et al. 1998).

Using paired-pulse TMS, Kujirai et al. (1993) demonstrated an initial period of inhibition lasting ≈6 ms and a subsequent period of facilitation maximal at ≈15 ms, and attributed these to inhibitory and facilitatory synaptic processes, respectively. Nakamura et al. (1997) and Di Lazzaro et al. (1998) provided further evidence to support the presence of intracortical inhibition in the cortex by showing a reduction of directly recorded descending corticospinal waves at short interstimulus intervals with paired TMS. In the pioneering experiments of Kujirai et al. (1993), a suprathreshold test stimulus was conditioned by a stimulus subthreshold for eliciting a motor response in the first dorsal interosseous (FDI) muscle. This conditioning protocol has been termed the ‘conventional’ paired TMS paradigm (Ziemann, 1999). Other investigators have used a conditioning stimulus intensity that was equal to or greater than that of the test stimulus (e.g. Ziemann et al. 1998b), particularly when investigating longer interstimulus intervals (Valls-Soléet al. 1992; Wassermann et al. 1996). Using the ‘conventional paradigm’, Ridding et al. (1995c) demonstrated reduced short-latency inhibition and facilitation during voluntary contraction of the target muscle.

The scientific literature contains an abundance of studies demonstrating changes in intracortical inhibition and facilitation in many neurological and psychiatric diseases or following drug administration (for review, see Ziemann, 1999). For example, there is reduced intracortical inhibition in patients with Parkinson's disease (Ridding et al. 1995a) and dystonia (Ridding et al. 1995b). However, a similar reduction with or without changes in intracortical facilitation has been described in patients with different forms of myoclonus (Brown et al. 1996; Caramia et al. 1996, Hanajima et al. 1996; Inghilleri et al. 1998), stroke (Sakai et al. 1998), Tourette's disorder (Ziemann et al. 1997a), obsessive-compulsive disorder (Greenberg et al. 2000), amyotrophic lateral sclerosis (Yokota et al. 1996; Ziemann et al. 1997c), stiff-man syndrome (Sandbrink et al. 2000) and even lower-limb amputees (Chen et al. 1998). Acute administration of various agents with neurological effects have generally demonstrated reciprocal effects on intracortical inhibition and intracortical facilitation, e.g. increased intracortical inhibition and decreased intracortical facilitation with apomorphine (Pierantozzi et al. 2001), memantine (Schwenkreis et al. 1999), dextromethorphan (Ziemann et al. 1998a) and alcohol (Ziemann et al. 1995) and the reverse pattern with haloperidol (Ziemann et al. 1997b) and tiagabine (Werhahn et al. 1999). Two issues are worth noting: there seems to be a common pattern of ‘abnormality’ of intracortical synaptic mechanisms in virtually all diseases, regardless of pathophysiology, and the changes produced by drugs often involve opposite changes in so-called intracortical inhibitory mechanisms and intracortical facilitatory mechanisms. As considered in Discussion, this raises the question whether the changes in inhibition and facilitation truly reflect changes in synaptic mechanisms, or do so exclusively.

An alternative explanation is that the results seen with paired-pulse TMS are complicated by the recovery of excitability of the stimulated axons of cortical interneurones, such that the recorded changes represent the resultant of independent presynaptic and synaptic mechanisms. To test this hypothesis, the present study was undertaken to document the changes in axonal excitability produced by conditioning stimuli that were weaker than the test stimuli, and then to see whether these changes in axonal excitability were faithfully reproduced across an intervening synapse. The findings indicate that the pattern of change in axonal excitability is similar to that seen with the ‘conventional’ paired-pulse TMS paradigm.

Methods

Twenty-four experiments were conducted on 11 normal subjects (aged 23–66 years, five male, six female), without clinical evidence of a peripheral nerve disorder. All subjects gave informed consent to the experimental procedures, which conformed to the Declaration of Helsinki and had the approval of the Research Ethics Committee of the South Eastern Sydney Area Health Service (Eastern Division). One or more of three studies were performed on each subject.

Two methods of stimulation were used. Test stimuli were either varied to maintain a constant compound nerve or muscle action potential (CNAP or CMAP) using ‘threshold tracking’ or kept constant while a changing CNAP/CMAP was monitored (‘amplitude tracking’). The experiments were performed using a computerised threshold-tracking program (QTRAC, Professor Hugh Bostock, Institute of Neurology, Queen Square, London, UK; see Bostock & Baker, 1988; Bostock et al. 1998). With threshold tracking, the current required to produce the target CNAP or CMAP is referred to as the threshold for the CNAP or CMAP. ‘Proportional tracking’ was used, such that the extent to which the stimulus current increased or decreased was proportional to the difference between the target and the measured response (Bostock et al. 1998). With amplitude tracking, the test stimulus remained constant during the experiment such that changes in excitability produced a CNAP or CMAP of variable size.

Surface electrodes were used to stimulate the median nerve at the wrist, and the resulting CNAP was recorded from the median nerve at the cubital fossa using surface electrodes 4 cm apart. The amplitude of CNAP was measured peak-to-peak. The soleus H reflex was used in experiments to determine the effects of an intervening synapse on the neural volleys produced in the conditioning-test paradigm. The tibial nerve was stimulated at the popliteal fossa, and the CMAP was recorded using surface electrodes 4 cm apart, in the midline over the lower third of the soleus muscle.

A conditioning-test experimental design was used. Stimuli were delivered at 2 Hz for CNAPs or 0.3-0.5 Hz for H reflex studies, cycling through a sequence of test stimuli, alone or in combination with a conditioning pulse. The conditioning-test stimulus interval was systematically increased between 1 and 150 ms. The changes in threshold current for a CNAP/CMAP of a fixed size or in the amplitude of the CNAP/CMAP to a fixed test stimulus were measured. Three sets of studies were performed as outlined below.

Recovery of axonal excitability following a conditioning stimulus that was weaker than the test stimulus

The recovery of axonal excitability was studied following a conditioning stimulus, the intensity of which was adjusted to produce a volley that was 80 % of that of the unconditioned test stimulus. Studies were performed using threshold and amplitude tracking in six subjects. Measurements were made in response to nine test stimuli delivered in a regularly repeating sequence at 0.5 s intervals, as illustrated in Fig. 1. A fixed supramaximal stimulus of duration 0.2 ms was delivered on stimulus channel 1 to produce a CNAP of maximal amplitude. In each set of two following channels (i.e. channels 2 and 3, 4 and 5, 6 and 7, 8 and 9), the duration of the test and conditioning stimuli were both 0.1 ms or both 1.0 ms on alternate channels. On six channels (channels 2-7), the stimulus current was adjusted by the computer to produce a CNAP that was a percentage of the maximal CNAP (recorded in response to stimulus 1); on channels 8 and 9 the test stimulus was fixed (amplitude tracking). Channels 2 and 3 were used to determine the current required to produce an unconditioned CNAP 50 % of maximum (i.e. the unconditioned test response). Channels 4 and 5 tracked a response that was 40 % of the maximal CNAP to set the intensity of the conditioning stimulus. On channels 6 and 7, conditioning and test stimuli were delivered, the former identical to those on channels 4 and 5 (i.e. stimuli that produced a 40 % potential), the latter tracking the current needed to maintain a test CNAP 50 % of maximum. On channels 8 and 9, conditioning and test stimuli were delivered. On these amplitude-tracking channels, the intensities of the conditioning stimuli were equal to those on channels 4 and 5 (i.e. stimuli that produced a 40 % CNAP) and the test stimulus intensity was fixed at the current necessary to produce an unconditioned CNAP that was 50 % of maximum. Accordingly, on these channels, changes in axonal excitability would produce changes in amplitude of the test CNAP. Note that the responses to channels 1–5 were used to set the stimuli on other channels and/or to provide unconditioned test responses and, but for noise, should have remained constant throughout the experiments. To measure test potentials uncontaminated by the conditioning volley, the potentials on channels 4 and 5 were subtracted on-line from the responses to the paired stimuli on channels 6 and 7 and channels 8 and 9, and measurements were made on the differences.

Figure 1.

Stimulus channels

Schematic representation of the stimulus channels, which were delivered in the displayed sequence at regular intervals of 0.5 s for studies on the CNAP. On channel 1, the stimulus was supramaximal, and the resulting CNAP was used to calibrate the response on the other channels. On channels 2, 4, 6 and 8 stimulus duration was 0.1 ms and on channels 3, 5, 7 and 9 it was 1.0 ms. Channels 2 and 3 provided the unconditioned test responses, and channels 4 and 5 the conditioning stimuli. On channels 6 and 7, the conditioned potential was measured using threshold tracking and on channels 8 and 9 it was measured using amplitude tracking.

In six subjects, using the technique of amplitude tracking, the effects of four different conditioning stimuli were determined. Conditioning stimuli were set at ≈70, ≈80, 100 and ≈110 % of the test stimuli. The duration of the test and conditioning stimuli was 1 ms.

Recovery of H reflex following subthreshold conditioning stimuli

To determine whether the changes in axonal excitability would be reproduced across an intervening synapse, the recovery of the H reflex following a subthreshold conditioning stimulus was measured using threshold and amplitude tracking. The soleus H reflex was chosen because it is the most easily elicited H reflex. The conditioning stimuli were subthreshold for the H reflex.

Stimulus channels were similar to those in the first series of experiments with the following exceptions. The test and conditioning stimuli were of 1 ms width only. An extra channel was used to monitor changes in the M wave to a fixed submaximal stimulus during the experiment to ensure the stability of stimulating conditions. H reflex size was set to ≈50 % of the maximal H reflex (≈10 % of the maximal M wave dependent on subject). The conditioning stimulus was subthreshold for the H reflex, and was of fixed intensity. Stimulus frequency was 0.5 or 0.33 Hz.

In these experiments, the recovery of the H reflex was documented during a weak voluntary contraction and subsequently at rest. This sequence was chosen because a conditioning stimulus that was just subthreshold during a contraction would also be subthreshold at rest, but not vice versa. Subjects performed a maximal voluntary plantar-flexion and subsequently maintained a contraction that was 5–10 % of maximum with visual feedback of the level of integrated EMG on an oscilloscope and auditory feedback of EMG noise. They were allowed rest breaks during the contraction sequence if they felt fatigued.

Changes in nerve recovery during depolarising and hyperpolarising currents

To determine the effects of changes in membrane potential on the recovery of axonal excitability, polarising DC was injected at the stimulus site through a non-polarisable cathodal electrode (Red Dot, 3M Canada, London, Ontario, Canada) over the median nerve at the wrist, with a similar electrode as anode, ≈10 cm away over muscle. The effects of the resulting changes in membrane potential on the recovery of axonal excitability were measured in six subjects using threshold tracking.

Test and conditioning stimuli were of 1 ms duration. Conditioned thresholds were measured for three selected conditioning-test intervals: corresponding to ‘refractoriness’ (2-3 ms), ‘supernormality’ (7-8 ms) and ‘late subnormality’ (40 ms). For each of the selected conditioning-test intervals, the median nerve was subjected to polarising currents of −50 to +50 %, changed in 10 % steps and the changes in the threshold of the test potential were measured. The intensity of the polarising current was set to a percentage of the threshold current needed to produce the unconditioned test potential. Negative values denote hyperpolarising current and positive values denote depolarising current. The duration of the polarising currents was 200 ms, and the conditioning stimulus was delivered 50 ms after its onset, with the test stimulus at the appropriate delay after the conditioning stimulus.

In all experiments, skin sensors monitored temperature continuously over the forearm or the calf. It was kept above 32 °C using blankets and by applying radiant heat if necessary.

Results

The recovery of excitability of peripheral nerve axons was found to follow an oscillating pattern, dependent on the size of the preceding conditioning stimulus (Fig. 2). The recovery of the H reflex following a subthreshold conditioning stimulus was then measured to determine whether these changes in axonal excitability would be reproduced across an intervening synapse. Finally, polarising DC was injected at the stimulus site to determine whether changes in membrane potential of the stimulated axons would alter the recovery of axonal excitability in a manner reminiscent of the findings in patients with paired-pulse TMS.

Figure 2.

Recovery curves of median nerve for four different conditioning intensities

Mean data for six subjects, with s.e.m. only for data with 70 and 110 % conditioning stimuli. The intensity of the conditioning stimulus is expressed as a percentage of the intensity of the test stimulus, as indicated on the y-axis using arrows. Note that smaller conditioning stimuli resulted in smaller changes in the test CNAP, but that the conditioned CNAP exceeded both the conditioning and test potentials at conditioning-test intervals of 5–16 ms. (Cond: amplitude of conditioning CNAP as a percentage of the unconditioned test CNAP.)

When the conditioning stimulus is less than the test stimulus, the changes in excitability may not result solely from the mechanisms responsible for refractoriness, supernormality and late subnormality and accordingly, these terms are used below in quotation marks. Indeed, as seen in Fig. 2 this may also be the case when the conditioning stimulus is equal to or just greater than the test stimulus (filled circles and triangles, respectively).

Recovery of axonal excitability following conditioning stimuli of different size

The effects of conditioning stimuli that were, on average, 70, 82, 100 and 110.3 % of the test stimulus on a test CNAP 50 % of maximum are shown in Fig. 2 for six subjects. At conditioning test intervals of 2–4 ms, the conditioned potential was smaller than the unconditioned potential, i.e. the conditioning stimulus decreased the ability of axons to respond to a subsequent test stimulus given a few milliseconds later. At conditioning-test intervals of 4–25 ms, the conditioning stimulus potentiated the response elicited by a subsequent test stimulus, with maximal potentiation at 8 ms. At conditioning-test intervals of 25–140 ms, conditioned responses were smaller than control. The conditioned test response was suppressed maximally at 60 ms and returned to control size at 140 ms.

These changes in axonal excitability correspond to the relative refractory, supernormal and late subnormal periods of recovery of axonal excitability, as described using supramaximal conditioning stimuli (Gilliatt & Willison, 1963; Stöhr, 1981; Ng et al. 1987; Kiernan et al. 1996). Although the changes in excitability for conditioning stimuli of different strengths followed a similar time course, the same was not true for the extent of the changes, which was dependent on the strength of the conditioning stimulus, i.e. greater changes in excitability were seen with stronger conditioning stimuli. Interestingly, during the ‘supernormal’ period, conditioned potentials were greater than both conditioning and test potentials; for example, when the conditioning potential was only slightly larger (110 %) than the unconditioned test potential, the conditioned response at conditioning-test intervals of 6–10 ms was much greater (125 %).

The changes in excitability produced by the 70 % conditioning stimulus during the periods corresponding to supernormality and late subnormality were small and variable, with consistent ‘supernormality’ only at the 4 and 5 ms intervals. Accordingly, to investigate further the effects of conditioning stimuli smaller than test stimuli, the recovery of excitability of the median nerve axons was measured in six subjects using conditioning stimulus intensities that produced volleys that were ≈80 % of the test volley (Fig. 1). Stimulus widths of 0.1 and 1.0 ms were used to determine whether the recovery curves were affected by stimulus width. (These durations were chosen because, while the duration of magnetic pulse from the Magstim 2000 (Magstim Company Ltd, Dyfed, Wales, UK) is 1.0 ms, this duration may not be that of the induced current stimulating the axons of cortical interneurones due to the filtering properties of intervening scalp, skull, meninges and CSF.) Again, as illustrated in Fig. 3, there were three distinct phases corresponding to refractoriness, supernormality and late subnormality. The curves obtained with threshold and amplitude tracking showed reciprocal changes.

Figure 3.

The recovery of axonal excitability following a single conditioning stimulus

Mean data for six subjects (±s.e.m.) for two stimulus durations (open circles: 0.1 ms; filled circles: 1.0 ms). A, results with threshold tracking, with data plotted as the change from the unconditioned threshold. An increase in threshold (i.e. a decrease in excitability) is plotted upwards. B, results with amplitude tracking showing reciprocal trends, with data expressed as a percentage of the maximal CNAP. The test CNAP was 50 % of maximum. Note the larger error bars in B. (TW: width of test stimulus.)

Recovery of H reflex at rest and during contraction

The recovery of the H reflex following a conditioning stimulus subthreshold for the H reflex was measured to determine whether the changes in axonal excitability would be reproduced across an intervening synapse. The conditioning stimulus was quite weak relative to the unconditioned test stimulus (65.7 ± 4.8 %, mean ±s.e.m.) because it was necessary to have a conditioning volley that was subthreshold for the H reflex even during contraction. Figure 4 shows the average recovery cycles of the soleus H reflex at rest and during contraction in six subjects using both threshold and amplitude tracking (A and B, respectively). Despite the weak conditioning stimulus the two recovery curves demonstrate three distinct phases, strikingly similar to the recovery of excitability of peripheral nerve axons (Fig. 3). Indeed, there was a strong correlation between the changes in the H reflex and the changes in axonal excitability (P/ 0.0008 for the amplitude tracking data illustrated in Fig. 5A, and P= 0.010 for the threshold tracking data). The changes in the H reflex were significant at the 2 and 7 ms intervals for amplitude tracking (Fig. 4B; P= 0.029 and 0.034, respectively), but only at 2 ms for threshold tracking (Fig. 4A; P= 0.018 and 0.056, respectively).

Figure 4.

The recovery cycle of the H reflex following a single conditioning stimulus

Data using threshold tracking (A) and amplitude tracking (B) at rest (open circles) and during voluntary contraction (filled circles). Mean data ±s.e.m. for six subjects. Note the smaller error bars during contraction with threshold tracking. In A and B the data represent the deviation from the unconditioned value.

Figure 5.

Changes in the H reflex in a paired-pulse paradigm

A, correlation between the change in the H reflex (data from Fig. 4B, filled circles) and the change in axonal excitability (data from Fig. 3B, open circles). B, relationship between the results of threshold and amplitude tracking of the H reflex. Data from Fig. 4A and B for the recovery cycles of the H reflex at rest.

There were differences between the recovery curves recorded at rest and during contraction. The initial period corresponding to refractoriness started at 2 ms and ended at 3 ms at rest and during contraction. The curves then started to follow separate time courses, with less ‘supernormality’ and a curtailment of the late period of decreased excitability during contraction. The oscillations in excitability in the three phases of the recovery cycle were more prominent at rest than during voluntary contraction. This is reminiscent of the findings of Ridding et al. (1995c) with paired-pulse TMS at rest and during contraction.

The H reflex recovery curves suggest that the recovery cycle for the reflex response is influenced by the recovery of excitability of the stimulated afferent axons. The presence of an intervening synapse did not distort the recovery curve which was correlated with that of peripheral nerve axons (as illustrated in Fig. 5A). As with all experiments involving H reflexes, the reflex response was sensitive to central and peripheral factors, and this probably explains the greater variability in the H reflex recovery curve than that of peripheral nerve (compare error bars in Fig. 3 and Fig. 4).

This is the first study to use the technique of threshold tracking to document changes in excitability of the motoneurone pool due to conditioning stimuli (see Discussion). It is important that there was a strong correlation between the two methods of assessing changes in reflex function. Specifically, there was a reciprocal relationship between changes in threshold and changes in amplitude (Fig. 5B). However, the threshold tracking technique provided data that were less variable (Fig. 4).

Effects of polarising currents

Polarising DC was injected at the stimulus site and its effect on recovery of axonal excitability was measured to determine how changes in membrane potential affected axonal excitability. Conditioning-test intervals of 2-3, 7–8 and 40 ms were studied to document changes during the ‘refractory’, ‘supernormal’ and ‘late subnormal’ periods, respectively. Although the conditioning stimulus was less than the test stimulus (75 %), polarising current produced changes in excitability expected from previous studies (Burke et al. 1998; Kiernan & Bostock, 2000). Figure 6A shows the effect of polarisation on unconditioned threshold, an indirect measure of membrane potential.

Figure 6.

Effects of changing membrane potential

A, effect of polarising current on the threshold for the unconditioned CNAP. The intensity of the polarising current is expressed as a percentage of the current needed to produce the unconditioned test potential. Negative values denote hyperpolarising current and positive values denote depolarising current. The threshold data (y-axis) were normalised so that the value at rest (i.e. without polarising current) was 1.0. The large open circles show unpolarised thresholds (equivalent to the resting state). B, C and D, the relationships between ‘refractoriness’ (B), ‘supernormality’ (C) and ‘late subnormality’ (D) and normalised threshold, respectively. These data were calculated as the percentage change in threshold produced by the conditioning stimulus at the appropriate conditioning-test interval (B, 2–3 ms; C, 7–8 ms; D, 40 ms; see text). Data are means ±s.e.m. Note that the error bars are often smaller than the symbols for the data, particularly in A. The labels for the y-axes in B-D appear in quotation marks because the underlying mechanisms are complicated when the conditioning stimulus is weaker than the test (see Discussion).

In panels B-D of Fig. 6, the change in the threshold current needed to elicit a conditioned potential is plotted as a percentage of the unconditioned value against unconditioned threshold (an indirect measure of membrane potential). Open circles indicate thresholds at rest (i.e. when there was no polarising current). At the 2–3 ms conditioning-test interval (Fig. 6B), more current (67.5 %) was needed at rest to elicit the conditioned response than the unconditioned response, i.e. axons were ‘refractory’. At this conditioning- test interval, there was a graded reduction in the current needed to elicit the conditioned response with increasing hyperpolarising currents, i.e. threshold decreased with increasing hyperpolarising currents. With depolarising current, the opposite occurred with greater depolarisation causing an increase in ‘refractoriness’.

At conditioning-test intervals of 8 ms (Fig. 6C) and 40 ms (Fig. 6D), the mean threshold difference at rest was −3.5 and +1.8 %, respectively. At the 7–8 ms conditioning-test interval, ‘supernormality’ increased with hyperpolarising currents and decreased with depolarising currents, though the changes were significant only with the more extreme degrees of polarisation. At the 40 ms conditioning-test interval, ‘late subnormality’ decreased with hyperpolarising currents and increased with depolarising currents.

Discussion

This study was undertaken to determine how a conditioning volley that is weaker than the test volley affects the response to the test stimulus, and whether any resultant changes in axonal excitability would be transmitted across a synapse. It was the thesis of the study that any such effects would complicate the interpretation of paired-pulse TMS studies. In experiments performed on peripheral nerve axons, the changes in excitability were qualitatively similar to those previously documented in experiments with supramaximal conditioning stimuli, and had a temporal pattern similar to the changes in the MEP produced by ‘conventional’ paired-pulse TMS. In other words, in these paired-pulse TMS studies, conditioning stimuli that were weaker than the test stimuli should also have induced excitability changes in the stimulated cortical axons. Furthermore, the present study has documented that the excitability changes induced in axons can be faithfully translated across a synapse, by demonstrating that the recovery cycle of the H reflex following a subthreshold conditioning stimulus was remarkably similar to that of peripheral nerve axons. Finally, the study documents, for the first time, the use of threshold tracking to study changes in the H reflex. The results suggest that threshold tracking may have advantages over conventional amplitude tracking for some indications, as discussed later.

It must be stressed that while the present data are qualitatively appropriate to explain the ‘conventional’ paired-pulse TMS data, they may not be quantitatively sufficient. Indeed there is good evidence that the conditioning TMS by itself can produce appropriate changes in voluntary EMG without a test stimulus (Petersen et al. 2001). A recent threshold-tracking study suggests that ‘intracortical inhibition’ can be dissociated into at least two processes, the first of which could represent refractoriness (Fisher et al. 2002). These results indicate that the inhibition must involve more than just refractoriness. In addition the results with paired-pulse TMS vary dependent on the strength of the conditioning stimulus (Ziemann et al. 1998b; Hanajima et al. 2002), but do so in a manner not readily explained by the data in Fig. 2. A further argument is that the changes in I waves in paired-pulse TMS paradigms affect only late I waves, sparing the wave of lowest threshold, I1 (Nakamura et al. 1997; Di Lazzaro et al. 1998; Hanajima et al. 1998). However, this begs the question how individual I waves are generated and, in this respect, it is relevant that the I wave depression produced by volatile anaesthetics also affects only late I waves: I1 is unchanged, even enhanced (Hicks et al. 1992). Either way, it is not maintained that the results of paired-pulse TMS paradigms can be explained by changes in axonal excitability, but it is maintained that all such paradigms will produce changes in axonal excitability. If the MEP in a paired-pulse paradigm is the result of a number of interacting processes, changes induced by drugs or disease need to be interpreted with caution. It may be invalid to quantify changes in the inhibitory or facilitatory phases as if they were specific measures of a single process.

There is some precedent for the suggestion that the responses to paired-pulse TMS involve presynaptic (axonal) mechanisms in addition to synaptic mechanisms. Short-term synaptic depression in cultured hippocampal neurones involves presynaptic changes in action potentials and can be enhanced by tetrodotoxin (Brody & Yue, 2000). In such neurones, presynaptic Na+ channel inactivation can contribute to paired-pulse depression and this is more likely ‘under depolarising physiological or pathological conditions’ (He et al. 2002). Recovery from Na+ channel inactivation is responsible for the relatively refractory phase of the recovery cycle of axonal excitability following a conditioning discharge (Hodgkin & Huxley, 1952), and the influences of Na+ channel blockade and membrane potential on short-term synaptic depression is relevant to ‘abnormalities’ found using paired-pulse TMS in patients (see Introduction and below).

Mechanisms underlying the changes in axonal excitability

The mechanisms involved in the recovery of axonal excitability when test stimuli are conditioned by weaker conditioning stimuli have not been investigated previously and are potentially complex. During the ‘refractory’ period, more current was required to achieve the target response when threshold tracking was used, but the change in stimulus intensity does not reflect just the refractoriness of axons activated by the conditioning stimulus. Conceivably, the test response could also include axons that were recruited by the test stimulus (but not by the smaller conditioning stimulus) and some axons of higher threshold recruited by the stronger conditioned stimulus (but not by the conditioning or unconditioned test stimuli). Hence, the increase in threshold does not reflect solely the ‘refractoriness’ of axons that fired in response to the conditioning stimulus. However, with amplitude tracking, the conditioned stimulus produced a response that was smaller than that evoked by unconditioned test stimulus. This was presumably due to the inability of refractory axons to generate an action potential in response to the test stimulus. Given this, the similarity of the curves in Fig. 3A and B is reassuring, even if not completely predictable.

During the period of ‘supernormality’, axons that discharged in response to the conditioning stimulus would have become super-excitable, thereby requiring a test stimulus of lower intensity. However, some axons of higher threshold in the test volley would not have discharged in response to the conditioning stimulus. If these axons were truly unaffected by the conditioning stimuli, threshold would not decrease during this period. Accordingly, the decrease in threshold suggests that there can be long-lasting effects on axonal excitability following subthreshold conditioning stimuli. This is well-documented with subthreshold conditioning stimuli of long duration (i.e. those that produce threshold electrotonus, see Bostock & Baker, 1988; Bostock et al. 1998), but not with brief conditioning stimuli (i.e. those that produce ‘latent addition’, see Bostock & Rothwell, 1997). There were similar findings with amplitude tracking: the test stimulus produced a larger test potential, and this could have occurred only if more axons had been recruited into the conditioned potential. This interesting finding warrants further study, but is consistent with studies on single motor axons by Shefner et al. (1996). The extent of the changes in Fig. 2 indicates that these findings cannot be explained by variable activation of axons of similar threshold on the stimulus- response curve.

Relevance of a peripheral nerve model to paired-pulse TMS

TMS stimulates axons of cortical interneurones, not the neurones themselves, and it would not be unreasonable to expect that results with paired-pulse TMS are influenced by excitability changes in the stimulated axons (as documented here) in addition to any subsequent synaptic effects on cortical neurones. Indeed, the present study demonstrates that the changes in excitability of peripheral nerve axons stimulated in a paired-pulse experimental paradigm are remarkably similar to those seen with the ‘conventional’ paired-pulse TMS paradigm.

There are potentially three sources of discrepancy between paired-pulse TMS and the present experiments on peripheral nerves. First, the size of the stimulated axons is probably different. Peripheral nerve axons studied here are likely to be larger than the axons of cortical interneurones, and one might therefore have expected a shorter relatively refractory period in the peripheral nerve model. The duration of ‘refractoriness’ in the peripheral nerve studies proved to be similar to that of short-latency intracortical inhibition with paired-pulse TMS though, if anything, it was shorter (as expected). Secondly, with paired-pulse TMS, there are at least two intervening synapses before the evoked volleys are translated into an EMG potential, at least one cortical (at the corticospinal neurone) and one spinal (at the cortico-motoneuronal synapse). The present experiments on the H reflex involve only a single excitatory synapse. However, the studies involving H reflexes were intended to determine whether the excitability changes could be translated faithfully across synapses and, if so, it is likely that this would occur, whether across one, two or a few synapses. Thirdly, electrical stimulation of the peripheral nerve elicits a single volley, whereas TMS may produce multiple volleys in corticospinal axons. With paired-pulse TMS, this would presumably affect the test volley (which was above threshold for the MEP) more than the weaker conditioning volley.

The responses to both paired-pulse TMS and peripheral nerve stimulation were reduced during muscle contraction. One suggested explanation for reduced responses to paired-pulse TMS during contraction is that voluntary drive reduces the excitability of inhibitory circuits in cortical areas that project to the active muscle (Ridding et al. 1995c). An explanation for the same phenomenon with the peripheral nerve model remains to be elucidated, though a comparable mechanism could involve disynaptic Ib inhibition from the afferents in the test volley for the H reflex (Burke et al. 1984; V. Marchand-Pauvert, G. Nicolas, D. Burke & E. Pierrot-Deseilligny, unpublished observations). Regardless of the mechanism, it is notable that contraction produced similar results with the conditioned H reflex and paired-pulse TMS, and this suggests that, while the changes with paired-pulse TMS may be cortical, they are not specific for cortical phenomena.

Paired-pulse TMS in neurological disorders

As noted in the Introduction, a decrease in short-latency intracortical inhibition with or without a change in intracortical facilitation has been described in many neurological diseases, seemingly irrespective of pathophysiology. This non-specificity suggests that the mechanism involved in producing these changes may be equally non-specific. The present study has replicated the trends seen with paired-pulse TMS in disease states by injecting polarising current into peripheral nerve axons. Axonal hyperpolarisation decreased the initial period of inhibition (i.e. it decreased ‘refractoriness’) and it tended to increase the subsequent period of facilitation (i.e. it increased ‘supernormality’). Based on these findings, hyperpolarisation of axons of cortical interneurones in patients with neurological disease would be expected to decrease intracortical inhibition and possibly to increase intracortical facilitation when tested with paired-pulse TMS.

It is conceivable that the disease states mentioned in the Introduction are associated with a change in membrane potential, perhaps due to increased cortical activity. Heightened background activity could lead to activity-dependent hyperpolarisation of the active axons (e.g. Vagg et al. 1998). It should be noted that activity is likely to increase threshold by less than 40 % (see Vagg et al. 1998; Kuwabara et al. 2002) and that there was little change in ‘supernormality’ in Fig. 6 when the threshold change was less than 40 %. This is, perhaps, consistent with the findings in disease states of clear changes in intracortical inhibition but small or variable changes in intracortical facilitation. Thus, it is necessary to exclude excitability changes associated with differences in membrane potential before changes with paired-pulse TMS in disease states can be safely attributed to a change in cortical inhibitory and facilitatory synaptic mechanisms.

Threshold tracking of H reflex

Threshold tracking was developed specifically to follow changes in excitability of peripheral nerves, but has also been used to study the excitability of motor cortex using TMS (Awiszus et al. 1999) and intracortical inhibition produced by paired-pulse TMS (Fisher et al. 2002). This is the first study to document the use of threshold tracking to study changes in excitability of the H reflex. Conventional motor control studies using the H reflex are based on measuring the changes in amplitude of the reflex response to a fixed stimulus (i.e. the amplitude tracking paradigm).

There are advantages of threshold tracking over amplitude tracking for H reflex studies, much as was reported for tracking the threshold for the MEP (Fisher et al. 2002). First, the results obtained with threshold tracking are less variable than with amplitude tracking. Indeed, in the H reflex recovery curves (Fig. 4), the variability with amplitude tracking was so large that the differences between recovery curves at rest and during contraction would not be statistically significant. Secondly, with threshold tracking the recorded response consists of a constant population of axons. The susceptibility of the H reflex to facilitation and inhibition varies with the size of the reflex, such that, at low amplitudes, the sensitivity of the H reflex to facilitation increases with increasing size of the reflex (Crone et al. 1990). This means that, when a manoeuvre changes the size of the unconditioned test H reflex, stimulus intensity may need to be changed to keep the reflex within the linear range of the input-output relationship (Crone et al. 1990). An advantage of threshold tracking is that it avoids the problem of non-linearity in the input-output relationship of the motoneurone pool by clamping the reflex response to a fixed size. Lastly, the dynamic range of threshold tracking is wide, enabling threshold changes of 200 % or more to be tracked. In contrast, amplitude tracking suffers from ‘floor’ and ‘ceiling’ effects, dependent on the size of the test stimulus. There is a limited range within which the size of test response can reflect changes in excitability and, once the response reaches maximum or is reduced to zero, further increases or decreases in excitability can no longer be reflected in the size of the test response.

Acknowledgements

This study was supported by the National Health and Medical Research Council of Australia and Institut pour la Recherche sur la Moelle Épiniére. The authors are grateful to Dr Matthew Kiernan and Dr Sabine Meunier for comments on the manuscript.

Ancillary