Role of spinal premotoneurones in mediating corticospinal input to forearm motoneurones in man

Authors

  • V. Pauvert,

    1. Laboratoire de Neurophysiologie Clinique, Rééducation, Hôpital de la Salpêtrière, 47 boulevard de l'Hôpital, F-75651 Paris cedex 13, France and UK
    Search for more papers by this author
  • E. Pierrot-Deseilligny,

    1. Laboratoire de Neurophysiologie Clinique, Rééducation, Hôpital de la Salpêtrière, 47 boulevard de l'Hôpital, F-75651 Paris cedex 13, France and UK
    Search for more papers by this author
  • J. C. Rothwell

    1. MRC Human Movement and Balance Unit, The Institute of Neurology, Queen Square, London WC1N 3BG, UK
    Search for more papers by this author

Corresponding author E. Pierrot-Deseilligny: Laboratoire de Neurophysiologie Clinique, Rééducation, Hôpital de la Salpêtrière, 47 boulevard de l'Hôpital, F-75651 Paris cedex 13, France. Email: emmanuel.pierrot-deseilligny@psl.ap-hop-paris.fr

Abstract

  • 1Evidence was sought to support the suggestion that corticospinal input can be relayed to motoneurones (MNs) via a population of interneurones (premotoneurones) in the cervical cord, and that this pathway operates in parallel with the direct monosynaptic pathway.
  • 2Single motor units were recorded in forearm muscles and post-stimulus time histograms (PSTHs) of their firing pattern were constructed during voluntary activation. Weak transcranial magnetic stimulation of the contralateral motor cortex was used to produce a small facilitation in the PSTH. We then examined how the size of this peak was affected by low threshold electrical stimulation of either the homonymous muscle nerve or the musculo-cutaneous nerve at various interstimulus intervals (ISIs).
  • 3Homonymous nerve stimulation had the following characteristics: (a) the cortical peak was facilitated when stimuli were timed so that both inputs arrived simultaneously at the MN; (b) the amount of facilitation was only slightly greater than the sum of the effects of each stimulus given alone; and (c) facilitation affected even the earliest bins of the cortically evoked peak. These three features are consistent with a monosynaptic input onto the MN from both sources.
  • 4Stimulation of the musculo-cutaneous nerve (which has no monosynaptic connections with forearm MNs) had no effect at similar timings. It (a) produced facilitation only at longer intervals corresponding to an extra central delay of 4-6 ms; (b) always gave a significantly larger facilitation than expected from the algebraic sum of the effects of each stimulus given alone; and (c) never affected the earliest bins of the cortical peak. These features are compatible with interaction of peripheral and cortical inputs at a population of premotoneurones.
  • 5These results confirm the suggestion that premotoneurones mediate part of the cortical command to MNs innervating forearm muscles.
  • 6Excitation is followed by an inhibition which may almost completely suppress the cortical peak. It is suggested that cortical and musculo-cutaneous volleys also converge onto inhibitory interneurones projecting to the premotoneuronal pool.

The monosynaptic component of the corticospinal projection to limb motoneurones (MNs) in monkey and man is large and easily studied. Indeed, after stimulation of the corticospinal tract, intracellular recordings from MNs innervating distal muscles in monkey rarely show any input other than monosynaptic excitation and disynaptic inhibition (Maier, Illert, Kirkwood, Nielsen & Lemon, 1996). However, recent experiments in conscious man suggest that there may also be a substantial disynaptic excitation to MNs of some, but not all, arm muscles which acts in parallel with the monosynaptic component (see review by Pierrot-Deseilligny, 1996). The interneurones that relay this effect receive excitation and inhibition from peripheral afferents. Originally this population of interneurones was thought to be analogous to the C3-C4 propriospinal system described in the cat (Illert, Lundberg, Padel & Tanaka, 1978). However, lack of evidence for a similarly arranged system in monkeys (Maier et al. 1996) necessitates reconsideration of this question. In view of this we no longer refer to this population of interneurones as ‘propriospinal-like’, but use the term ‘premotoneurones’.

In man, the evidence supporting the existence of a disynaptic premotoneuronal pathway is necessarily indirect. Experiments have relied on using H-reflexes to test the excitability of a MN pool, and transcranial magnetic stimulation over motor cortex to activate the corticospinal tract. Two arguments were used. The first employed the classical technique of spatial facilitation (Gracies, Meunier & Pierrot-Deseilligny, 1994). It was shown that combined stimulation of peripheral afferents and motor cortex could produce far more facilitation of a test H-reflex than expected from the algebraic sum of either input given alone. The time between the stimuli at which extra facilitation became apparent suggested that the interaction took place in premotoneurones. However, the arguments rely on the assumption (which was not demonstrated) that any excitatory inputs that converge directly onto the population of MNs involved in the H-reflex would summate in a linear and homogeneous fashion. The second piece of evidence supporting the idea of a premotoneuronal relay was that peripheral afferent input, which itself had no direct effect on MNs, could inhibit cortically evoked EMG responses in wrist extensor muscles (Burke, Gracies, Mazevet, Meunier & Pierrot-Deseilligny, 1994), this inhibition sparing the initial (monosynaptic) portion of the response (Mazevet, Pierrot-Deseilligny & Rothwell, 1996). Again this was compatible with inhibition of premotoneurones in the disynaptic relay from the cortex. However, because the temporal resolution of surface EMG is not high, it was not possible to be completely certain that suppression affected the disynaptic premotoneuronal linkage, rather than another oligosynaptic route.

The aim of the present investigation was to provide further evidence in man for the existence of a substantial disynaptic linkage in the transmission of corticospinal input to MNs. The experiments were performed on single motor units (MUs) of forearm muscles so that time resolution would be sufficiently high and assumptions concerning linearity and homogeneity of the pool could be disregarded. We show that the musculo-cutaneous afferent input, which itself has no effect on the excitability of forearm MNs, can nevertheless influence the response to cortical stimulation. Timing considerations show that the interaction of cortical and musculo-cutaneous inputs occurs at a premotoneuronal level.

METHODS

The experiments were carried out on six healthy subjects (including the 3 authors), aged 23-61 years, all of whom gave informed consent to the experimental procedure, which was approved by the local ethical committee. They were comfortably seated in an armchair. The elbow was semi-flexed (100-120 deg) and the forearm was supported by the arm of the chair.

Poststimulus time histograms (PSTHs) of a voluntarily activated MU were constructed for the period following a conditioning stimulation. This process extracts from the naturally occurring spike train only those changes in firing probability that are time-locked to the stimulus (Stephens, Usherwood & Garnett, 1976). A detailed description of the particular method with the PSTH technique used in this study is given elsewhere (Fournier, Meunier, Pierrot-Deseilligny & Shindo, 1986), so it will be only summarized here. PSTHs of MUs from various upper limb muscles (extensor carpi radialis (ECR); flexor carpi radialis (FCR); flexor digitorum superficialis (FDS); biceps) were constructed for the 10-50 ms following a conditioning stimulation (bin width: 0.2, 0.4, 0.5 or 1 ms). The EMG potentials of single MUs recorded with surface or needle electrodes were converted into standard pulses by a discriminator with variable trigger levels and were used to trigger first a computer and then the stimulator delivering the conditioning stimulations. The stimulation was set with respect to the discharge of the single MU, and a histogram of the firing probability after the stimulus was constructed. A histogram was also constructed for the background firing probability of the MU without stimulation. Measurements with and without stimulation were randomly alternated within the same sequence. To clarify the differences between the results obtained in the two situations, the control value in each bin was subtracted from that observed after stimulation to give the hatched columns in Fig. 1B and D and all the columns in Figs 3-6.

Figure 1.

Monosynaptic and non-monosynaptic excitation

PSTH from a single FCR MU (0.2 ms bins). The number of counts expressed as a percentage of the number of triggers is plotted against the latency after stimulation. A and C, time histograms of the discharge of the unit are compared in control conditions (□) and after stimulation (▪). B and D, difference between these two histograms. A and B, median nerve stimulation (0.9 × MT). C and D, musculo-cutaneous nerve stimulation (0.8 × MT). Dotted vertical arrows indicate the beginning of the facilitation.

Transcranial stimulation was applied over the motor cortex using a high-powered Magstim 200 (Magstim, Whitland, Dyfed, UK) with a 9 cm mean-diameter circular coil, held at the vertex, at the optimal position for a twitch of the explored muscle. The stimulus intensity was then reduced (to between 17 and 30 % of the maximal output) so that during the contraction of the MU cortical stimulation did not evoke any EMG potential (other than a possible change in the firing probability of the MU).

Peripheral stimuli were 1 ms duration electrical pulses delivered through bipolar electrodes. The median nerve was stimulated in the cubital fossa, the radial nerve in the spiral groove and the musculo-cutaneous nerve 10-15 cm above elbow level on the anterior and medial aspect of the arm (see Gracies, Meunier, Pierrot-Deseilligny & Simonetta, 1991).

In the first part of the experiment, the effects of the homonymous and of the musculo-cutaneous stimulation when given alone were investigated. The effects of peripheral (homonymous or musculo-cutaneous) volleys on the cortically evoked excitation were then investigated, the different stimuli (peripheral nerve, cortical and combined stimulation) being alternated in the same sequence. A time course of the effect of the peripheral shock was constructed by repeating the process with different interstimulus intervals (ISIs), using 0.5 or 1 ms steps. A positive ISI indicates that the peripheral stimulus was given before the cortical stimulus.

Each experiment on a single MU involved studying the time course of effects from more than one peripheral nerve at ISIs of 1 ms or less. In order to complete the study in reasonable time, we chose to give only fifty stimuli of each type at each interval. The resulting PSTHs are rather sparse, but nevertheless give reliable statistical results. This was confirmed when data from all units were grouped together for final analysis.

Expression of results and statistical analysis

Individual units

A χ2 test was used to test the difference between the peak obtained on combined stimulation and the sum of effects by separate stimuli. Since the timing pattern of homonymous and musculo-cutaneous facilitation is different (see below), the window of analysis started with the first bin where there was a facilitation on combined stimulation, and lasted for 1.2 ms (2 ms in experiments with 1 ms bins) (dotted downwards arrows in Figs 3, 5 and 6).

Group analyses

These were also performed using the non-parametric Wilcoxon test. They concerned: (i) the amount of facilitation on combined stimulation which was assessed both in the window defined above and in a ‘fixed’ window (filled upwards arrows in Figs 3, 5 and 6) starting 0.6 ms after the onset of the peak and lasting 0.8 ms (1 and 1 ms, respectively, when using 1 ms bins). The facilitation was assessed at the shortest effective ISI and normalized as a percentage of the number of triggers; (ii) the delay at which the facilitation on combined stimulation appeared in the cortical peak, assessed, at the shortest effective ISI, as the difference between the onset of the facilitation and that of the control peak (e.g. 30.8 - 30.4 = 0.4 ms in Fig. 6D). Negative values in columns 11 and 12 of Table 1 indicate that the onset of the facilitated peak occurred before the onset of the control peak (e.g. Fig. 3E). The Fischer exact test was also used to compare in two situations (homonymous and musculo-cutaneous stimulation) the number of MUs with and without significant facilitation or the number of MUs in which facilitation affected or did not affect the initial part of the peak.

Figure 6.

Modulation of the cortical excitation by peripheral stimuli in an FDS MU

PSTH from a single FDS MU (0.2 ms bins). Abscissa and ordinate, as in Fig. 1. A, isolated stimulation of the motor cortex. B-E, same legends as in Fig. 5. B and C, stimulation of the median nerve (0.6 × MT) at ISI = 0 ms. D and E, stimulation of the musculo-cutaneous nerve (0.7 × MT) at ISI = 7 ms. Dotted vertical line in A as in Fig. 3, dotted downward arrows (B and C, and D and E) and filled upward arrows (D and E) as in Fig. 3.

Table 1. Comparison of homonymous- and musculo-cutaneous-induced facilitation of the cortical peak in twenty-two motor units
    ISIAmount of facilitationDuration of initial sparing
MUSubjectBin (ms)MuscleHomo (ms)MC (ms)Difference (ms)HomoMCMC fixed delayHomo (ms)MC (ms)
123456789101112
  1. Columns are numbered from 1 to 12: 1, number of the unit; 2, subject; 3, bin width; 4, muscle; 5, earliest ISI with homonymous (Homo)-induced facilitation; 6, earliest ISI with musculo-cutaneous (MC)-induced facilitation; 7, difference between column 6 and column 5, i.e. difference between the earliest ISIs with musculo-cutaneous and homonymous facilitations after correction for differences in peripheral afferent conduction times; 8, amount of homonymous facilitation (Homo); 9 and 10, amount of musculo-cutaneous (MC) facilitation assessed from the onset of facilitation (9) or at a fixed delay after the onset of the peak (10); 11 and 12, duration of the initial sparing (see text) by homonymous (11) and musculo-cutaneous (12) facilitation.

1V. P.1ECR6.014.011.01.0
2D. M.1ECR6.011.03.00.0
3V. P.1ECR1.06.05.01.07.020.00.01.0
4A. V. L.1ECR2.07.05.08.08.012.01.01.0
5A. V. L.1ECR1.06.05.02.012.08.00.02.0
6E. P. D.1ECR6.013.06.01.0
7D. M.1ECR6.011.07.02.0
8J. C. R.1ECR0.54.03.54.018.06.01.01.0
9S. M.1ECR0.04.04.04.010.04.00.01.0
10V. P.0.2FCR3.07.04.00.014.012.0−0.20.2
11V. P.0.2FCR2.07.55.53.012.013.00.00.4
12E. P. D.0.2FCR2.06.04.54.016.010.00.20.4
13S. M.0.4FDS0.05.56.113.07.01.00.0−0.4
14E. P. D.0.4FDS0.07.07.55.06.06.00.00.4
15E. P. D.0.4FDS3.07.04.51.014.014.00.00.4
16E. P. D.0.2FDS3.09.06.55.08.08.00.00.2
17V. P.0.2FDS4.010.06.05.08.08.0−0.20.8
18V. P.0.2FDS2.07.55.55.07.04.00.00.6
19V. P.0.2FDS2.58.56.02.09.012.0−0.4−0.2
20D. M.0.2FDS0.07.07.03.011.012.0−0.20.6
21D. M.0.2FDS2.09.07.012.010.09.0−0.60.2
22E. P. D.0.2FDS1.08.07.53.015.08.00.01.0
Figure 3.

Time course of homonymous- and musculo-cutaneous-induced facilitation of the cortical peak

PSTH from a single FCR MU (0.2 ms bins). Abscissa and ordinate as in Fig. 1. A-C, effect of isolated stimulation of the motor cortex (25 %, A), of the median nerve (0.6 × MT, B) and of the musculo-cutaneous nerve (0.7 × MT, C). D-L, the effect on combined stimulation (▪) is compared with the algebraic sum of effects by separate stimuli (□). D-F, combined stimulation of the motor cortex and the median nerve at 2 ms (D), 3 ms (E) and 4 ms (F) ISIs. G-L, combined stimulation of the motor cortex and of the musculo-cutaneous nerve at 3 ms (G), 4 ms (H), 6 ms (I), 7 ms (J), 8 ms (K) and 9 ms (L) ISIs. Dotted vertical lines (23.4 ms) indicate the onset of the cortical peak obtained in A. Dotted downward arrows (E and J), window of analysis starting with the first facilitated bin and lasting for 1.2 ms. Filled upward arrows (J),‘fixed’ window starting 0.6 ms after the onset of the peak and lasting for 0.8 ms. Note that in D-L, the abscissa is expressed relative to cortical stimulation.

Differences in various motor nuclei between mean values observed in either the central delay of non-monosynaptic excitation or the time of musculo-cutaneous facilitation of the cortical peak, were tested using the non-parametric Kruskal-Wallis test.

Finally, it was verified, using variance analysis ANOVA, that different independent variables (subject, muscle, bin width) had no significant effect on the musculo-cutaneous-induced facilitation.

RESULTS

Peripheral conduction times in the different nerves

To compare the central delay of homonymous- (radial, median) and musculo-cutaneous-induced effects (see below) it was necessary to assess in all subjects the peripheral conduction time in the corresponding I a afferents. The latencies of the monosynaptic peak evoked by stimulation of each of these three nerves in the PSTH of the same biceps MU (see Cavallari & Katz, 1989) were therefore assessed using 0.2 or 0.5 ms bins. Despite the more proximal site of stimulation of the musculo-cutaneous nerve, the slower conduction velocity in I a afferents of this nerve (Cavallari & Katz, 1989; see also Miller, Mogyoros & Burke, 1995) resulted in similar monosynaptic latencies implying conduction times equal to those for radial nerve I a volleys, and equal or only slightly shorter (0.5-0.6 ms) than those for the median I a volleys.

Monosynaptic and non-monosynaptic excitation evoked by peripheral stimulation

Figure 1 shows the results obtained in an FCR MU (MU 10 in Table 1) using 0.2 ms bins. Stimulation of the homonymous median nerve at 0.9 × motor threshold (MT) evoked a highly significant (P < 0.001) increase in firing probability at a latency of 27.6 ms which lasted for 1.2 ms (Fig. 1A and B). This early peak reflects the monosynaptic I a EPSP (see Malmgren & Pierrot-Deseilligny, 1988a). Stimulation of the musculo-cutaneous nerve at 0.8 × MT produced a smaller but significant excitation (P < 0.05), with a latency of 32.2 ms, i.e. 4.6 ms longer than the homonymous monosynaptic latency (Fig. 1C and D). Since in this subject the afferent conduction time for median and musculo-cutaneous volleys was the same, these supplementary 4.6 ms represent the central delay of the non-monosynaptic excitation (Gracies et al. 1991).

The extra central delay between non-monosynaptic and monosynaptic I a excitation was explored using 0.2 ms bins in twelve biceps, twelve FCR and fourteen FDS MUs in which there was a significant (P < 0.05) non-monosynaptic effect. The mean values (biceps: 3.5 ± 0.3 ms; FCR: 4.2 ± 0.2 ms; FDS: 4.8 ± 0.4 ms) were significantly (P < 0.05) different and very close to those reported by Gracies et al. (1991) using 1 ms bins.

Facilitation of the cortically evoked peak by stimulation of peripheral nerves

Figure 2 (0.2 ms bins) illustrates the basic finding that musculo-cutaneous stimulation facilitates the cortical peak in forearm MNs (here an FDS MU). The background firing and the changes in firing probability induced by separate stimulation of the musculo-cutaneous nerve (0.7 × MT) and of the motor cortex (26 % of the maximal output) are shown in A, B and C, respectively. On combined stimulation (ISI = 8 ms) the cortical peak (26-27.2 ms, dotted lines) was significantly (P < 0.005) facilitated and this effect spared the first two bins. In the following, shorter duration intervals are used to juxtapose in the same figure the effects obtained at various ISIs after stimulation of different nerves.

Figure 2.

Modulation of the cortical excitation by a musculo-cutaneous volley

PSTH from a single FDS MU (0.2 ms bins). Abscissa and ordinate as in Fig. 1. A, background MU firing. B-D, changes in firing probablility induced by isolated (B and C) or combined (D) stimulation. B, musculo-cutaneous nerve stimulation (0.7 × MT). C, motor cortex stimulation (26 %). D, combined musculo-cutaneous and cortical stimulations at ISI = 8 ms (note that the abscissa is relative to cortical stimulation). Dotted vertical lines, window of statistical analysis.

Figure 3A shows that cortical stimulation (25 %) alone increased the firing probability of an FCR unit (same MU as in Fig. 1, 0.2 ms bins) at a latency of 23.4 ms. To reduce the risk of occlusion on combined stimulation, the stimulus intensity was then reduced to 20 % of the maximal output in the experiments illustrated in Fig. 3D-L.

This cortically evoked peak was conditioned by stimulation of the median or musculo-cutaneous nerve, at an intensity (0.6 × MT and 0.7 × MT, respectively) chosen so that the fifty stimuli used in these experiments hardly modified the firing probability of the unit when applied alone (Fig. 3B and C). Changes in firing probability evoked by combined cortical and peripheral stimulation (black columns) were then compared at different ISIs to the algebraic sum of the effects evoked by the two stimuli (cortical and peripheral) when applied alone (open columns).

The effects of median nerve stimulation are shown in Fig. 3D-F. There was no facilitation at an ISI of 2 ms (D) or 4 ms (F). At an ISI of 3 ms (E), the onset of the peak was facilitated, but facilitation did not reach statistical significance within the window 23.8-24.8 ms (dotted arrows). From the difference between the latencies of the monosynaptic I a peak (27.6 ms, Fig. 1A and B) and of the cortically evoked peak (23.4 ms, Fig. 3C), we can predict that the two volleys should arrive synchronously at the MN at an ISI of 4.2 ms (27.6 - 23.4 ms). The 3 ms ISI found here indicates that the two volleys had to be elicited at a shorter ISI than predicted (4.2 - 3 = 1.2 ms) to allow for mutual facilitation to take place. This might be explained by the small size of the I a EPSP induced by weak (0.6 × MT) median nerve stimulation (compared with stimulation at 0.9 × MT as used in Fig. 1A and B to evoke a clear monosynaptic I a peak). Accordingly, this delay was substantially reduced when a stronger median nerve stimulus was used (e.g. Fig. 5A and B).

Figure 5.

Modulation of the cortical excitation by peripheral stimuli in an FDS MU

PSTH from a single FDS MU (0.4 ms bins). A, C, E, G, I, K and M, abscissa and ordinate, open and black columns as in Fig. 3D-L.B, D, F, H, J, L and N, subtraction histograms showing the difference between the effect on combined stimulation and the algebraic sum of the effects by separate stimuli (hatched columns). A and B, median nerve stimulation (3 ms ISI). C-N, musculo-cutaneous nerve stimulation at 3 ms (C and D), 5 ms (E and F), 6 ms (G and H), 7 ms (I and J), 8 ms (K and L) and 10 ms (M and N) ISIs. Dotted downward arrows (in A and B, and I and L) and filled upward arrows (I and J) as in Fig. 3.

The effects of varying the ISI between musculo-cutaneous and cortical stimulation in this MU are shown in Fig. 3G-L. At the 3 ms (G), 4 ms (H), 6 ms (I) and 6.5 ms (not illustrated) ISIs there was no facilitation of the cortically evoked peak. At an ISI of 7 ms (J) the peak was significantly increased within the windows 23.8-24.8 ms (dotted open arrows) and 24-24.6 ms (filled arrows). Facilitation disappeared when the ISI was increased to 8 ms (K), and at ISI = 9 ms (L) the cortically evoked peak was so suppressed by musculo-cutaneous stimulation that it almost completely disappeared (P < 0.01). Since, in this subject, the afferent conduction time for median and musculo-cutaneous volleys was the same, the difference between the shortest ISI at which each peripheral nerve volley facilitated the cortically evoked peak (7 - 3 = 4 ms) may be attributed to the difference in their central delays. In other words it takes 4 ms longer for afferents in the musculo-cutaneous nerve to facilitate the cortical volley, and this delay occurs within the spinal cord itself.

This long central delay may reflect an interaction between musculo-cutaneous and cortical volleys at the postulated premotoneuronal level, but it is also compatible with a modification of the cortical EPSP at the MN by the non-monosynaptic musculo-cutaneous EPSP (see Discussion). An attempt was therefore made to estimate the amount of summation at the MN between a monosynaptic I a EPSP and the non-monosynaptic musculo-cutaneous-induced EPSP. To that end, stimulation of the musculo-cutaneous (0.8 × MT) and of the median nerve (0.7 × MT) were randomized together in the same block of trials. The ISI on combined stimulation was chosen so that the monosynaptic homonymous I a- and the non-monosynaptic musculo-cutaneous-induced excitation occurred simultaneously in the MN. Figure 4 shows in an FCR MU the effects of separate stimulation of the median nerve (A, a peak of monosynaptic excitation at a latency of 24.8 ms), of the musculo-cutaneous nerve (B, weak non-monosynaptic excitation at a latency of about 29.2-30.2 ms) and of combined stimulation (when the latter preceded the former by 5 ms). The excitation obtained on combined stimulation (15 counts in the window 24.8- 26.6 ms) was hardly greater than the algebraic sum of the two separate effects (13 counts). A similar absence of significant facilitation on combined stimulation was observed at 4 and 6 ms ISIs and in the two other subjects so explored.

Figure 4.

Effect of combining two peripheral volleys converging on one motoneurone

PSTH from a single FCR MU (0.2 ms bins). Abscissa and ordinate as in Fig. 1B and D (in each histogram, the background firing probability of the unit has been subtracted). A, isolated stimulation of the median nerve (0.7 × MT). B, isolated stimulation of the musculo-cutaneous nerve (0.8 × MT). C, combined stimulation, musculo-cutaneous stimulation preceding median nerve stimulation by 5 ms (note that the abscissa is relative to median nerve stimulation).

Figure 5 shows results obtained in an FDS MU (MU 15 in Table 1) from another subject, using 0.4 ms bins. In some panels it is possible to distinguish two waves of cortical excitation with an initial small peak (e.g. in panel G, but also in panel K). This very probably reflects multiple cortical volleys (see Nakamura, Kitagawa, Kawaguchi & Tsuji, 1997). Subtraction histograms (hatched columns in lower panel of each pair) show the difference between combined stimulation and the algebraic sum of effects by separate (cortical and peripheral) stimuli. The latency of the homonymous (median nerve, 1 × MT) monosynaptic I a peak was 28.6 ms in this MU and the earliest increase in firing probability evoked by cortical stimulation occurred at 25.4 ms. Here the difference between the two latencies (3.2 ms) corresponded almost exactly to the 3 ms ISI at which a lower intensity of median nerve stimulation (0.8 × MT, but still capable of evoking a small excitation on its own), facilitated the cortically evoked peak (Fig. 5A and B). This facilitation affected the earliest part of the peak (dotted arrows), but was very weak and did not reach statistical significance. There was no effect at ISI = 2.5 or 3.5 ms. The effects of musculo-cutaneous nerve stimulation on the cortically induced peak were investigated using a stimulus intensity (0.7 × MT) that hardly modified the firing probability of the MU when given alone. At the 3-6 ms ISIs the cortically evoked peak was not facilitated (Fig. 5C-H). At ISI = 7 and 8 ms (panels I and J, and K and L), a significant (P < 0.05) facilitation occurred which spared the initial part of the peak (dotted downwards arrows). Thus, here again musculo-cutaneous facilitation was larger, affected later bins in the histogram and occurred at a substantially longer ISI than homonymous (median) facilitation. At ISI = 10 ms the facilitation was replaced by a depression (Fig. 5M and N).

The differences between the effects of homonymous and musculo-cutaneous stimulation on the peak of cortical excitation are further documented in Fig. 6 which shows results obtained from an FDS MU from another subject. In this unit (MU 20, 0.2 ms bins) the cortical peak occurred at a latency of 30.6 ms (Fig. 6A). Median nerve stimulation at 0.6 × MT had little effect on the firing probability of the MU when applied alone, but facilitated the peak of cortical excitation at an ISI of 0 ms (Fig. 6B and C). This facilitation, which included the onset of the peak, did not reach statistical significance. At an ISI of 7 ms, musculo-cutaneous stimulation (0.7 × MT) evoked a significant (P < 0.05) facilitation which spared the initial 0.4 ms of the peak (Fig. 6D and E).

Comparison of results obtained in different situations

Table 1 shows the main results for twenty-two MUs: nine MUs from the ECR were explored using 1 ms bins, three from the FCR using 0.2 ms bins and ten from the FDS using 0.4 ms (3) or 0.2 ms (7) bins. The shortest ISI at which the homonymous I a and the musculo-cutaneous volley facilitated the cortical-induced peak is shown in columns 5 and 6, respectively. The difference between these two ISIs, after correction for different afferent conduction times when necessary (see above), is shown in column 7. In four ECR MUs the investigation involving the musculo-cutaneous nerve was completed but the MU was lost before the effects of combined stimulation of the motor cortex and radial nerve at short ISIs could be investigated.

Homonymous facilitation

Stimulation of I a fibres in the homonymous nerve produced monosynaptic facilitation of the cortically evoked peak in 17/18 units. This facilitation, estimated from its onset (see Methods) and shown in column 8, was moderate (on average 4.4 %) and, even though it was significant (P < 0.01) when the data were taken as a whole, in individual units it reached statistical significance only twice. In all but two units, the weak facilitation involved the early part of the peak (see below). The shortest ISI at which facilitation began was somewhat shorter than the predicted time of simultaneous arrival of the two volleys at the MN, the estimated central delay being on average 0.6 ms shorter than predicted. Finally it should be noted that the duration of facilitation was short, ISIs at which facilitation occurred being immediately preceded and followed by ISIs without any tendency to facilitation (Fig. 3D-F).

Musculo-cutaneous-induced facilitation

Strength of the facilitation

Musculo-cutaneous stimulation evoked a facilitation of the cortical peak that was significant (P < 0.05) in all twenty-two MUs when tested from its onset (column 9). Accordingly, this facilitation was larger than homonymous facilitation (compare columns 8 and 9) and on average the difference between these two effects (10.7 - 4.4 = 6.3) was highly significant (P < 0.002). Note that the average amount of facilitation was slightly lower when tested within the window starting at a fixed delay after the onset of the peak (9.3 vs. 10.7) (column 10, see Methods).

Shortest ISI

The shortest ISI (corrected for peripheral delay) at which significant musculo-cutaneous facilitation occurred was always substantially longer than that required for homonymous I a facilitation. The mean difference between the shortest ISI needed for musculo-cutaneous- and homonymous-induced facilitation varied with the motor nucleus: 4.5 ± 0.3 ms for ECR and 4.7 ± 0.4 ms for FCR, which are located in C6-C8; 6.3 ± 0.3 ms for FDS, which is more caudal (Kendall, Kendall & Wadsworth, 1971). These values, which were significantly different (P < 0.005), provide an estimate of the extra central delay of musculo-cutaneous facilitation in each motor nucleus.

Sparing of the initial part of the cortically evoked PSTH peak

Homonymous and musculo-cutaneous stimulation produced different patterns of facilitation in the cortically evoked PSTH peak. The difference between the first bin with homonymous or musculo-cutaneous facilitation and the onset of the control cortical peak is given in columns 11 and 12 of Table 1. On average, homonymous stimulation affected the first bin of the peak whereas those affected by musculo-cutaneous stimulation were 0.7 ms later (P < 0.002). To summarize the data in another way, homonymous stimulation affected bins within the initial 0.4 ms of the cortically evoked peak in 16/18 units tested, whereas this only occurred in 6/22 units after musculo-cutaneous stimulation (P < 0.0001).

Length of the facilitation within the peak

At the earliest ISI for facilitation, this effect lasted for 0.8-1.4 ms (two bins in experiments using 1 ms bins).

Duration of the facilitation

It was very brief, since facilitation was generally observed at only one to two consecutive ISIs (mean duration, 1.2 ms; range, 0.5-3 ms).

Following inhibition

Twelve MU (3 ECR, 3 FCR, 6 FDS) were kept long enough so that it was possible to investigate long ISIs after the end of the excitation. In all twelve MUs excitation was followed by inhibition. This inhibition appeared at ISIs only 1-3 ms longer than those producing facilitation. This inhibition was often quite dramatic (see Figs 3L, and 5M and N), although it only lasted a short time (when explored in 4 units, its duration was 1 ms).

Finally, in six units the effects of stimulation of the homonymous nerve (0.6 × MT-0.7 × MT) were also investigated at long (absolute values) ISIs. It was found that stimulation of both homonymous and musculo-cutaneous nerves facilitated the cortically evoked peak of excitation at the same long (6-8 ms) ISIs.

DISCUSSION

This study provides further evidence that the corticospinal projection to forearm MNs in man consists of both a direct monosynaptic connection and a parallel non-monosynaptic pathway through a population of premotoneurones in the cervical cord. We used magnetic stimulation to activate the corticospinal tract. As this can potentially give rise to more than one descending volley our evidence relies on comparisons of the ISI at which different peripheral nerve afferent volleys facilitated the cortically evoked excitation of voluntarily activated units. By making this comparison we assume that the peripheral volleys interact with the same cortical volleys.

For each MU, two peripheral nerves were tested: the homonymous nerve, which contains I a afferents with monosynaptic projections to the explored MN, and the musculo-cutaneous nerve, which has no monosynaptic group I projections to any forearm MNs (Cavallari, Katz & Pénicaud, 1992). Stimulation of either nerve facilitated the cortically evoked peak, but the effects differed in three important ways: (i) the central delay of the facilitation; (ii) the amount of facilitation; and (iii) whether or not facilitation spared the initial part of the cortical peak. These points are considered separately in more detail below.

Homonymous-induced facilitation of the cortically evoked PSTH peak

Facilitation

This is most probably due to synchronous arrival of the homonymous and corticospinal volleys at the MN. Summation of two individual EPSPs from different sources in the MN may be expected to be linear (Fig. 13J-M in Eccles, 1964). It could be argued that this linearity might be altered if the EPSPs are small and when they are generated via different conductances or at different MN sites. However, the most parsimonious explanation for homonymous facilitation is that in a firing MN, EPSPs can arrive at a point in the membrane trajectory of the MN where a single EPSP alone would be insufficient to fire the cell (because of the postspike after-hyperpolarization), but where a combination of two EPSPs would produce a discharge. Given the variability in membrane trajectory this should be a small effect, but could explain the occurrence of some extra facilitation.

Central delay

The cortical volley had to arrive at the MN about 0.6 ms before the I a volley in order for facilitation to occur. This is probably due to the weakness of the stimuli used in our experiments. As noted above, this interval was reduced in experiments where the homonymous I a volley was strong enough by itself to evoke an increase in firing probability of the MU. There was no other period of facilitation at ISIs earlier than that corresponding to a simultaneous arrival of the two volleys at the MN. This suggests that late cortical volleys either were not present using such low intensities of magnetic stimulation, or that they did not interact significantly with homonymous monosynaptic I a EPSPs.

Absence of initial sparing

Consistent with summation of two EPSPs at the MN, homonymous I a facilitation always influenced the onset of the cortical peak (Figs 3E, 5A and B and 6B and C).

Musculo-cutaneous-induced facilitation of the cortically evoked peak

Summation with an interneuronally mediated peripheral volley

Consistent with the lack of monosynaptic I a connections from biceps to MNs of forearm muscles (Cavallari et al. 1992), there was never any facilitation between musculo-cutaneous and cortical stimuli when the two volleys were timed to arrived simultaneously at the segmental MN (see Figs 3G, and 5C and D). Facilitation only occurred when the musculo-cutaneous volley was timed to arrive at the segmental level several milliseconds before the corticospinal volley. We argue below that this extra delay is compatible with the idea that the cortical and musculo-cutaneous volleys interact at a premotoneuronal interneurone rather than at the MN.

Evidence for summation at a premotoneuronal level

One possible explanation for the extra delay would be that, after the musculo-cutaneous input arrives at the segmental level, it is conducted through a long interneuronal chain to the MNs where it summates with the cortical volley. However, as shown in Fig. 4, summation of two inputs at the MN should give rise to little more than algebraic summation of their effects in the PSTH. This was not the case for musculo-cutaneous facilitation of the cortical response, since it always evoked a significant extra facilitation of the cortically evoked peak, over and above that expected from summation of each separate response. The simplest explanation for this is that the two inputs converge on a population of interneurones which then project onto the MN under test. Within that population some inactive cells will fail to fire in response to either input alone and will discharge only if both inputs arrive at the same time. The net result of this is that the response of the interneuronal population to two inputs will be more than the algebraic sum of the response to each input alone (spatial facilitation). This in turn will be reflected in MN discharge.

Sparing of the initial phase of cortical facilitation

The exact onset of the cortical peak was sometimes difficult to determine because of the background firing of the unit. Despite these uncertainties, it was regularly found that homonymous and musculo-cutaneous stimulation produced different patterns of facilitation in the same unit: at the shortest effective ISI, homonymous facilitation affected the earliest part of the peak, whereas musculo-cutaneous facilitation usually occurred about 0.7 ms later. This is what would be expected if cortical and musculo-cutaneous volleys converged onto common interneurones rather than directly onto the MN. Due to the synaptic delay at the interneurone, this input would arrive at the MN after the direct fast-conducting monosynaptic cortico-motoneuronal input. Thus interneuronal facilitation would be unable to affect the onset of the cortical response and initial sparing would occur. Two points arise.

First, an estimate of 0.7 ms for the extra delay in the interneuronal pathway may seem short, given the 1 ms delay generally admitted for one supplementary synapse in human spinal cord (see Day, Marsden, Obeso & Rothwell, 1984). The smaller delay in the present experiments could be a consequence of two factors which tend to speed up synaptic processing: (1) simultaneous arrival of musculo-cutaneous and cortical volleys at the premotoneurones will shorten the rise time of the resulting premotoneuronal EPSP; and (2) similarly, in the motoneurones, the rise time of the disynaptic EPSP (via the premotoneurones) will be short since it summates with the preceding monosynaptic cortical EPSP.

Second, as a cortical stimulus can evoke multiple descending volleys it is always possible that musculo-cutaneous stimulation facilitates only late volleys while sparing earlier volleys. However, several arguments point against this possibility: (i) in the same MUs, interaction with late cortical volleys was not observed when using homonymous monosynaptic I a volleys (see above); (ii) if this mechanism were responsible for the initial sparing, one would expect the late part of the peak to be spared at longer (absolute values) ISIs. This was not the case, as illustrated in Fig. 5: in this MU, facilitation appeared at an ISI of 7 ms sparing the initial 0.4 ms (panels I and J), and was still present at the 8 ms ISI (panels K and L) where the initial 0.8 ms was spared; and (iii) finally, if, all things being equal, homonymous I a and musculo-cutaneous facilitations were produced by interaction of the peripheral volley with the first and a late cortical volley, respectively, musculo-cutaneous facilitation should manifest itself at an earlier ISI than homonymous facilitation.

Which interneurones?

The musculo-cutaneous stimulus was small (0.7 × MT) and did not evoke any cutaneous sensation, suggesting activation of group I afferent fibres. Oligosynaptic excitatory I b pathways have been described in the cat (see, Jankowska, 1992) and in the human upper limb (Cavallari, Fournier, Katz, Malmgren, Pierrot-Deseilligny & Shindo, 1985), but with a central conduction time to MNs of the order of 1 ms. In contrast, the long central delay (4-6 ms) observed in the present experiments fits what has been observed for non-monosynaptic excitation in the human upper limb (see Pierrot-Deseilligny, 1996). The central delay of this effect has been shown to be greater the more caudal the motor nucleus (Pierrot-Deseilligny, 1996, and first section in Results). Accordingly, in the present experiments, the average extra central delay of musculo-cutaneous facilitation of the cortical peak was shorter in ECR and FCR MNs (4.6 ms, located in C6-C8) than in FDS MNs (6.3 ms, located in C7-Th1). Premotoneurones mediating non-monosynaptic excitation are therefore good candidates for the site of convergence between cortical and musculo-cutaneous volleys observed here.

Since the more caudal the MN pool the greater the central delay of non-monosynaptic excitation, it was postulated that the relevant premotoneurones were located above the cervical enlargement (Pierrot-Deseilligny, 1996). However, the lack of evidence for a similarly organized system in the monkey (Maier et al. 1996) has prompted us to reconsider the question. An alternative explanation could be to assume that these premotoneurones are segmental but located at the upper part (C5-C6) of the cervical enlargement or that the pathway for non-monosynaptic excitation of MNs in distal muscles includes more interneurones than that to proximal MNs.

Musculo-cutaneous-induced depression of the cortically induced peak

Musculo-cutaneous facilitation persisted for an average duration of only 1.2 ms. At first sight, this is odd since the duration of an average corticospinal EPSP in MNs is of the order of 10-20 ms (Landgren, Phillips & Porter, 1962). However, the very brief duration observed here could reflect truncation of an initial EPSP by the subsequent arrival of disynaptic IPSPs at the premotoneurones. The short onset latency of this inhibition is compatible with a spinal rather than cortical effect (see Delwaide & Olivier, 1990) and we therefore favour the idea that it is due to disynaptic inhibition of the premotoneuronal pool by the peripheral afferent volley. A similar organization was described in previous studies using PSTH and H-reflex techniques (Malmgren & Pierrot-Deseilligny, 1988b Mazevet & Pierrot-Deseilligny, 1994), and is supported by the observation that in all twelve MUs studied over long time intervals, facilitation was followed by strong inhibition of the cortically evoked peak. The reason that this inhibition affected even the initial part of the cortically evoked peak in the PSTH was discussed previously by Mazevet et al. (1996). Since inhibition never manifested itself as a trough in the PSTH when musculo-cutaneous stimulation was given alone, it is suggested that musculo-cutaneous and cortical volleys also both act on inhibitory interneurones projecting to the premotoneuronal pool.

The finding that musculo-cutaneous stimulation can almost completely suppress the cortically evoked peak (see Fig. 3L) could give the impression that a large proportion of the cortical excitation is mediated by the premotoneurones described here. In fact, the peak in the PSTH is produced by summation at the MN of the premotoneuronally mediated EPSP with the monosynaptic EPSP. This summation may be critical for evoking any response with the very small stimulus intensities used in these experiments and therefore removal of either would have a large effect. Nevertheless, the results suggest that a significant part of the cortical command is mediated through this system of premotoneurones and it will be of interest to investigate how the balance of excitatory and inhibitory inputs to premotoneurones is controlled during natural movements.

Acknowledgements

The authors wish to express their gratitude to Annie Rigaudie and Michèle Dodo for excellent technical assistance. This work was supported by grants from MESR (EA 2393), AP-HP (95/078), INSERM (CRI 9611), IRME, and the European Union.

Ancillary