Modulation of recurrent inhibition from knee extensors to ankle motoneurones during human walking

Abstract

The neural control for muscle coordination during human locomotion involves spinal and supraspinal networks, but little is known about the exact mechanisms implicated. The present study focused on modulation of heteronymous recurrent inhibition from knee extensors to ankle motoneurones at different times in the gait cycle, when quadriceps (Quad) muscle activity overlaps that in tibialis anterior (TA) and soleus (Sol). The effects of femoral nerve stimulation on ankle motoneurones were investigated during treadmill walking and during tonic co-contraction of Quad and TA/Sol while standing. Recurrent inhibition of TA motoneurones depended on the level of background EMG, and was similar during walking and standing for matched background EMG levels. On the other hand, recurrent inhibition in Sol was reduced in early stance, with respect to standing, and enhanced in late stance. Reduced inhibition in Sol was also observed when Quad was coactivated with TA around the time of heel contact, compared to standing at matched background EMG levels in the two muscles. The modulation of recurrent inhibition of Sol during walking might reflect central and/or peripheral control of the Renshaw cells. These modulations could be implicated in the transition phases, from swing to stance to assist Sol activation during the stance phase, and from stance to swing, for its deactivation.

During human walking, the activity of muscles acting at different joints must be well synchronized to ensure upright posture and the ongoing locomotor rhythm. Given their organization and their control by peripheral and descending inputs, this may be achieved by modulation of the activity of spinal neural networks (see Nielsen, 2003). Two of the neural pathways which are likely to make an important contribution to muscle coordination during walking are monosynaptic excitation and recurrent inhibition. They are produced in spinal motoneurones by group Ia afferents and motor axon discharge, respectively, and are more widely distributed in the human lower limb (Meunier et al. 1993, 1994) than in the cat and baboon hindlimb (Eccles et al. 1957; Eccles & Lundberg, 1958; Hultborn et al. 1971; Hongo et al. 1984). It has been suggested that these trans-joint connections have evolved to assist bipedal stance and gait (see Pierrot-Deseilligny & Burke, 2005).

Quadriceps (Quad) group Ia afferents and recurrent collaterals from its motoneurones have been shown to influence the activity of both tibialis anterior (TA) and soleus (Sol) motoneurones (Fig. 1A; Meunier et al. 1994). This antagonistic muscle pair thus receives common inputs from Quad and the question then arises as to how the motor command is focused on the relevant motoneurone pool when activity in Quad overlaps successively TA (around the time of heel contact) and Sol (stance phase) activity during walking; Ia monosynaptic excitation and recurrent inhibition from Renshaw cells are of special interest. During walking, modulation of the activity of interneurones mediating presynaptic inhibition of group Ia terminals (Hultborn et al. 1987a,b; see Rudomin, 2002) could help to gate monosynaptic Ia excitation to the different ankle motoneurones, and thus facilitate their respective activation with Quad. Cat and human experiments suggest that presynaptic inhibition is generally enhanced during locomotion keeping homonymous monosynaptic Ia excitation low (Morin et al. 1982; Capaday & Stein, 1986; Gossard, 1996); less is known on heteronymous pathways (Faist et al. 1996a), in particular whether there is a specific modulation of the monosynaptic Ia excitation of ankle dorsiflexors from knee extensors at the time of coactivation. In cats, Renshaw cell activity changed during fictive locomotion, but it is unclear whether this simply reflects the motoneuronal discharge or whether there is a possibility of selectively modulating the amount of recurrent inhibition in different parts of the step cycle (Pratt & Jordan, 1980). In humans, it has been shown that recurrent inhibition may be both facilitated and inhibited during various voluntary movements (see Katz and Pierrot-Deseilligny, 1999), and there is some evidence to suggest a reduction of heteronymous recurrent inhibition during postural tasks (from Quad to ankle muscles; Barbeau et al. 2000) and locomotion (from Sol to Quad; Iles et al. 2000).

Figure 1. Experimental design.

A, diagram of the spinal pathways controlling TA and Sol motoneurones, fed by group Ia afferents (dashed line) and motor axon recurrent collaterals (continuous line) from quadriceps (Q). The open circle and Y-shaped ending represent the motoneurone soma and excitatory synapses, and the filled circles, the inhibitory neurones (Renshaw cells). Arrows indicate presynaptic inhibition of group Ia terminals mediated by PAD (primary afferent depolarization) interneurones. B and C, raw (B) and rectified (averaging of 20 step cycles; C) EMG activity in VL, TA and Sol (traces from top to bottom) in one subject walking at 4 km h⁻¹. Dots in B indicate the heel contact, which was used as the trigger for EMG averaging in C (transition between swing 0 and stance 1). D, difference (in μV) between TA EMG activity conditioned by FN stimulation (30%M_max) and its mean control value (plotted against the latency after FN stimulation; ms) in one subject during tonic co-contraction of TA and Quad while standing; vertical bars are ± 1 s.e.m. In this subject, the distance between L2 vertebra and FN and CPN was, respectively, 23 and 68.5 cm, and given their conduction velocity (60 and 70 m s⁻¹ in FN and CPN, respectively), the afferent delay for group Ia fibres after FN and CPN stimulation to reach spinal motoneurones was, respectively, 3.8 (23/60) and 9.7 ms (68.5/70). The TA H-reflex latency (after CPN stimulation) was 30.3 ms and FN increased EMG activity at 24.5 ms, so there was a 5.8 ms difference, which almost corresponds to the theoretical difference (5.9 = 9.7 − 3.8) suggesting that the FN-induced increase in TA EMG activity was mediated by the monosynaptic group Ia pathway. Similar calculations were done in each subject, and for EMG modulations in Sol (see Meunier et al. 1990).

In the present study, we investigated the modulation of heteronymous Ia excitation and recurrent inhibition from Quad to ankle motoneurones. The effect of femoral nerve (FN) stimulation on TA and Sol motoneurones was assessed by studying the modulation of rectified EMG averages and the size of motor evoked potentials (MEPs), at the end of the swing phase (effect of FN stimulation on TA motoneurones) and during the stance phase (effect on Sol motoneurones) of treadmill walking, when Quad activity overlaps that in TA and Sol, respectively. The FN-induced inhibition of Sol H-reflex was investigated around the time of heel strike, when Quad and TA are coactivated, to test the modulation of recurrent inhibition of Sol motoneurones during activation of its antagonist. The results were compared to those obtained during tonic co-contraction of Quad to either TA or Sol while standing.

Methods

The experiments were carried out in 14 healthy volunteers (22–45 year) who all gave written informed consent to the experimental procedures. The study was performed according to The Code of Ethics of the World Medical Association (Declaration of Helsinki), and was approved by the local ethics committees of the Pitié-Salpêtrière Hospital.

Recordings

The EMG activity was recorded with bipolar surface electrodes (EMG sensors DE-2.1; Delsys Inc., Boston, MA, USA) placed over the muscle belly of Vastus Lateralis (VL; lateral head of Quad), Tibialis Anterior (TA) and Soleus (Sol). EMG activity was amplified (×1,000–10,000; Delsys Bagnoli System 4 Ch), and filtered (EMG bandwidth 20–450 Hz) before being digitally stored (2 kHz sampling rate) on a personal computer for later off-line analysis (Notocord-hem 3.4; Notocord SA, Croissy Sur Seine, France). Recordings were done during treadmill locomotion (Biodex Medical Systems Inc., Shirley, NY, USA), and a pressure transducer was placed on the heel of the shoe in order to detect the time of heel strike. At the beginning of the experiment, the subjects walked on the treadmill for 5–10 min before recordings, to accustom themselves to treadmill walking, and to determine their preferred speed (3–4 km h⁻¹). Recordings were also done during tonic co-contraction of Quad and TA and of Quad and Sol while standing with EMG levels within the same range to those recorded during walking.

Peripheral stimulations

Peripheral stimulations consisted of rectangular electrical pulses of 1 ms duration delivered at fixed intervals after heel strike (triggered by the pressure transducer under the heel) or during tonic co-contraction while standing. The maximal amplitude of the M response (M_max) in each corresponding EMG activity was first measured during both tasks, and at each investigated interval during walking. Femoral nerve (FN) was stimulated between cathode (2.5 cm diameter brass hemisphere) placed in the femoral triangle and anode on the back of the thigh (40 cm² plate), and stimulus intensity was adjusted (i) so as to produce a constant M response of 30%M_max in VL EMG, which was monitored throughout the experiment, and (ii) at ∼0.8 × the threshold for M response (MT) determined during walking and tonic contraction while standing (and below H-reflex threshold). The comparison between the two intensities was done to check if EMG suppression was evoked only if Quad motor axon discharged. Common peroneal nerve (CPN) was stimulated through bipolar surface electrodes (Maersk Medial Ltd, Redditch, UK; proximal cathode) at the level of the neck of the fibula to measure TA H-reflex latency. Posterior tibial nerve (PTN) stimulation was applied between a cathode (2.5 cm diameter brass hemisphere) placed in the popliteal fossa and an anode above the patella (40 cm² plate), and stimulus intensity was adjusted so as to produce H-reflexes in Sol EMG of similar size during the two motor tasks (5–10%M_max; Crone et al. 1990).

Cortical stimulation

Trans-cranial magnetic stimulation (TMS) over the primary motor cortex was used to produce MEPs in TA and Sol EMG. The magnetic field was generated through a double cone-coil (Magstim Rapid, Whitland, UK) held at the optimal position for evoking an MEP in one of the ankle muscles, which was determined during tonic ankle plantarflexion (for Sol) and dorsiflexion (for TA) while standing on the treadmill; activities in TA and Sol were simultaneously recorded to ensure that the response was evoked in the target muscle and was not caused by cross-talk of an MEP produced in its antagonist. A custom-made prosthesis, with the same shape as the coil, was used to fix the coil over the head; a band was used to tighten the coil and prosthesis over the head. The coil cable was held by an elastic restraint, which was fixed to the treadmill body-weight support system. This setup reduced the weight of the coil and the cable. The coil position was thus stable, despite the up-and-down oscillations during walking, which were softened by the elastic restraint. This was checked by asking subjects if they felt the coil moving while walking, and by monitoring the MEP threshold, its shape and test size throughout the duration of the experiment. Recordings during walking and tonic co-contraction while standing were done during the same experiment to ensure that the coil position was the same. TMS intensity was adjusted so as to produce test MEP within the same proportion of M_max (15–25% in TA and 5–15% in Sol) during both motor tasks.

Experimental protocols

Stimulus trigger delay after heel strike

At the beginning of the experiment, the subject was asked to walk for 1–2 min at his preferred speed (3–4 km h⁻¹) to determine the stimulus trigger delays after heel strike according to his walking EMG pattern (Fig. 1B and C). At this speed, the step cycle was reasonably stable, and on average the variability was 5 ± 1% of the mean cycle duration. FN stimulation was delivered when ankle muscle activity overlapped that in VL (dotted area in Fig. 1C) and that in rectus femoris (another head of Quad, grey area in Fig. 1C; Nene et al. 2004). TA EMG modulations could not be explored at the beginning of the swing phase since FN stimulation, evoking an M response of 30%M_max, perturbed the gait cycle, which made a comparison between control and conditioned TA EMG impossible. Indeed, the tip of the toe scraped the treadmill belt, due to the FN-induced knee extension, which delayed the ankle dorsiflexion. Therefore, the effect of FN stimulation was not investigated during the first TA EMG burst (∼600 ms after heel strike; Fig. 1C) corresponding to the TA shortening contraction but only during the second EMG burst (∼1000 ms after heel strike) corresponding to the TA lengthening contraction (den Otter et al. 2004).

In each subject, FN stimulation was thus delivered (see vertical arrows in Fig. 1C) (i) for TA investigation: in the ascending phase of the second EMG burst (onset TA; 850–1000 ms after heel strike, depending on the subject), at the peak of activity (middle TA; 900–1150 or 0 ms after heel strike), and within the descending phase of the EMG burst (late TA; 1050 or 0–50 ms after heel strike); and (ii) for Sol: in the ascending phase of the EMG burst (onset Sol; 60–150 ms after heel strike), at the beginning of the plateau (early Sol; 100–180 ms after heel strike), and before the EMG burst descending phase (late Sol; 300–480 ms after heel strike).

Tonic co-contraction while standing

Subjects stood on the treadmill and performed tonic co-contraction of TA and Quad and of Sol and Quad. EMG activities were displayed on the screen of the computer for visual feedback to help subjects to produce EMG activity similar to those recorded during walking. Sol EMG activity was also recorded when subjects sustained their maximal voluntary contraction (MVC) for a few seconds, while standing on the tip of their toes and one of the investigators exerted a resistance on their shoulders. The EMG activity produced during walking and tonic co-contraction while standing was expressed as a percentage of the EMG activity produced during MVC.

Assessment of the FN effect on TA and Sol motoneurones

The effect of the FN stimulation on TA (10 subjects) and Sol (11 subjects) motoneurones was first investigated in stimulus-triggered averaging of rectified EMG activity. Recordings with stimulation (conditioned EMG; n = 50) were randomly alternated (∼0.8 Hz) with recordings without stimulation (control EMG; n = 50; Figs 1D, 2 and 5).

Figure 2. FN-induced modulation of rectified and averaged TA EMG.

A, C, E, G and I, control (grey line) and conditioned (black line) TA EMG (μV), in one subject, during tonic co-contraction of Quad and TA while standing (A) and walking (3.6 km h⁻¹), at the end of the swing phase (1100 ms after heel strike, Onset TA, C and 1150 ms after heel strike, Middle TA, E and I) and at the beginning of stance (0 ms after heel strike, Late TA, G), are plotted against the latency (ms) after FN stimulation adjusted so as to evoke an M response of 30%M_max (A, C, E and G) and at 0.8 × MT (I). Vertical bars are ± 1 s.e.m.B, D, F, H and J, mean VL EMG activity after FN stimulation, recorded simultaneously with TA EMG, during tonic (B) and walking (D–J), is expressed as a percentage of M_max in VL, which was estimated for each situation.

Figure 5. FN-induced modulation of rectified and averaged Sol EMG.

Same legend as in Fig. 2. Walking speed was 3.6 km h⁻¹ and FN stimulation was delivered during tonic co-contraction of Quad and Sol while standing (Tonic, A and B), and during walking 80 (Onset Sol, C and D), 100 (Early Sol, EF) and 320 ms (Late Sol, G-J) after heel strike. FN stimulation was adjusted so as to produced M response of 30%M_max (A–H) and at 0.8 × MT (I and J).

Since presynaptic inhibition can contribute to FN-induced EMG suppression and since the MEP is not sensitive to presynaptic inhibition (Nielsen & Petersen, 1994), the FN stimulation was used to condition MEPs in TA and Sol (5 subjects; Fig. 4) at interstimulus intervals between –12 and 50 ms (depending on the subject). Test MEPs (without FN stimulation; n = 12) were randomly alternated (∼0.8 Hz) with conditioned MEP (with FN stimulation; n = 12). During walking, TA MEPs were investigated at delays corresponding to the onset and middle TA activity, and Sol MEPs were studied at the onset of Sol activity and later in the EMG burst, before its deactivation (see above).

Figure 4. FN-induced changes of MEP size.

A and C, mean test (grey line) and conditioned (black line) MEPs after FN stimulation adjusted so as to produce M response in VL of 30%M_max. Test and conditioned MEPs are expressed as a percentage of M_max (estimated in each situation), and the interval between FN stimulation and TMS was 20 ms. A, TA MEPs evoked in one subject during tonic co-contraction of Quad and TA while standing (Tonic), and during walking (4 km h⁻¹), when TMS was delivered 950 ms (Onset TA) and 1025 ms after heel strike (Middle TA). C, Sol MEPs evoked in one subject during tonic co-contraction of Quad and Sol while standing (Tonic), and during walking (4 km h⁻¹), when TMS was delivered 100 ms (Onset Sol) and 300 ms after heel strike (Late Sol). B and D, full time course of the effects evoked in the same subjects as in A and C, respectively. Conditioned MEPs (expressed as a percentage of its test size) are plotted against the interval between FN stimulation and TMS (ms) combined during tonic contraction (open circles and dashed line) and during walking: grey circles and continuous grey line for Onset TA (B) or Sol (D), and black circles and continuous black line for Middle TA (B) or Late Sol (D). In B, at −5 ms and 50 ms, the filled circles overlap the open circles. E and F, grouped data showing the mean conditioned MEPs (% the test size) obtained at long intervals (between 10 and 40 ms for TA in E, and between 10 and 30 ms for Sol in F) during tonic (white column) and walking (light grey for Onset TA and Sol and dark grey for Middle TA and Late Sol). Vertical bars are ± 1 s.e.m.*P < 0.05, Wilcoxon test.

Lastly, we wondered whether recurrent inhibition from Quad to one of the ankle muscles could be modified during the activation of its antagonist. This was only possible by testing the effect of FN stimulation on Sol H-reflexes (5 subjects; Fig. 6) during TA activation (1000 or 0 ms after heel strike and during Quad and TA co-contraction while standing). Test H-reflexes (without FN stimulation; n = 20) were randomly alternated (∼0.3 Hz) with conditioned H-reflexes (with FN stimulation; n = 20). Intervals between 5 and 35 ms were investigated in each subject.

Figure 6. FN-induced inhibition of Sol.

H-reflex during TA activation A, mean Sol H-reflexes (% its test size), conditioned by FN stimulation (evoking an M response in VL of 30%M_max), are plotted against the interval between FN and PTN stimulation (ms), in one subject during tonic co-contraction of Quad and TA while standing (Tonic, open circles and dashed line) and during walking (4 km h⁻¹), when PTN stimulation was delivered 1075 ms after heel strike, i.e. during maximum TA activity EMG in end swing (Middle TA, black circles and continuous line). B, grouped data (6 subjects) obtained at intervals between 10 and 25 ms during tonic contraction (white column) and in middle TA activity during walking (grey column). Vertical bars are ± 1 s.e.m.*P < 0.05 Wilcoxon test.

To compare measurements in both tasks, it was ensured that test MEPs and H-reflexes had a similar size (see above).

Quantitative and statistical analysis

Figure 1D shows that FN stimulation increased TA EMG activity at 24.5 ms before long lasting EMG suppression, and similar modulations have been described in Sol EMG activity (Meunier et al. 1996; Barbeau et al. 2000). The areas of rectified control and conditioned EMG were analysed within two windows to assess the amount of FN-induced facilitation and inhibition. The beginning of the windows was determined according to the H-reflex latency in VL, TA and Sol, and conduction velocity in group Ia afferents in FN, CPN and PTN. The method used is given in more detail in previous studies (Meunier et al. 1990; Meunier et al. 1996; Barbeau et al. 2000), and only appears in the figure legends in the present study (see Fig. 1D). Window duration for facilitation was limited to the duration of the peak in EMG (e.g. dotted area between 24.5 and 30 ms in Fig. 1D). This duration was mainly determined during tonic co-contraction since FN stimulation produced less facilitation during walking (see left part of Figs 2 and 5). The analysis window for inhibition started 10 ms after the facilitation, and its duration was fixed to 12 ms in order to limit the analysis to spinal transmission (e.g. dotted area between 34.5 and 46.5 ms in Fig. 1D; see Nielsen et al. 1997; Barbeau et al. 2000). The same analysis window was used for each motor task, and for each delay during walking. The resulting mean control EMG and the difference between conditioned EMG and mean control EMG, reflecting the amount of facilitation and inhibition, were expressed as a percentage of M_max in TA and Sol for interindividual comparison. MEPs in TA and Sol (with and without FN stimulation) were rectified for surface analysis within a window corresponding to their latency and duration. The Sol H-reflex amplitude was used to assess the effect of the FN stimulation. The conditioned responses were expressed as a percentage of the corresponding mean test response whose size was similar between the two motor tasks (see above).

Conditioned and test responses (EMG areas, MEP and reflex sizes) were compared in each individual by using Student's paired t test. EMG grouped data were analysed using Pearson's correlation (pooling multiple subject data for linear analysis; Poon, 1988) to test the relation between control EMG and the amount of facilitation and inhibition; the difference between motor tasks (tonic versus different times in the gait cycle) was tested using ANCOVA. Conditional on a significant F-value, post hoc Newman–Keuls tests were performed for comparisons of two means. Given the data distribution and heteroscadicity, mean conditioned MEPs and H-reflexes were compared between the various motor tasks using Kruskal–Wallis tests. Conditional on a significant H-value, Wilcoxon tests were performed for comparisons of two means. For all tests, the significance level was 0.05. Mean data are indicated ± 1 standard error of the mean (s.e.m.).

Results

Effects on TA motoneurones

Rectified TA EMG averaging

Figure 2 shows that FN stimulation significantly increased (paired t test; P < 0.01) TA EMG activity at 24 ms during tonic co-contraction of Quad and TA while standing (Fig. 2A) but not (onset TA Fig. 2C and late TA Fig. 2G) or hardly during walking (middle TA Fig. 2E and I) in one subject. On the other hand, significant FN-induced long lasting EMG suppression was observed during both tasks, and whatever time in the gait cycle, when FN stimulation was adjusted so as to evoke a constant M response in VL EMG of 30%M_max (Fig. 2B, D, F and H). No EMG suppression was evoked when stimulus intensity was adjusted below the M response and H-reflex threshold (0.8 × MT, Fig. 2I and J). In all the 10 subjects, FN depressed TA EMG activity only when FN stimulation produced M and/or H-reflex in VL EMG (checked during tonic and walking contraction).

On average, FN stimulation produced EMG facilitation at 23.7 ± 0.2 ms, which was very weak (0.38 ± 0.17%M_max) and rarely significant during standing (5/10 subjects). During walking, FN-induced facilitation was almost absent (−0.05 ± 0.13% onset TA and –0.25 ± 0.20% middle TA), and reached statistical significance in only one subject in late TA (mean 0.29 ± 0.28%).

FN stimulation suppressed TA EMG activity while standing in all the 10 subjects (significantly in 9/10 subjects) and grouped data were analysed to compare the amount of inhibition between tasks. The mean control EMG activity in TA during tonic co-contraction was between those recorded at the delays corresponding to onset and middle TA during walking. Group data illustrated in Fig. 3A show a significant correlation between the control EMG and the amount of inhibition (Pearson's correlation, r = 0.68, P < 0.001), whatever the motor task. ANCOVA was then used to test if the correlation between the level of control EMG and the amount of inhibition changed depending on the motor task. The amount of inhibition was thus compared between tasks using the control EMG as covariate, and the result was not significant (P = 0.14; Fig. 3C). This suggests that the FN-induced inhibition in TA EMG depended on its background activity while the subjects were standing or walking.

Figure 3. Task related changes in the FN-induced EMG suppression.

A and B, the difference between conditioned and control EMG (%M_max), used to estimate the amount observed in TA (A) and Sol (B), is plotted against the control EMG (surface estimated within the same analysis window as the conditioned EMG, and expressed as a percentage of M_max). Results obtained in each subjects during tonic co-contraction and during walking (3 delays) are represented in the same scatter diagram, so there are 4 dots per subject. C and D, mean amount of inhibition observed in TA (10 subjects; C) and Sol (11 subjects; D) during tonic (white column) and walking (light grey for Onset TA and Sol, middle grey for Middle TA and Early Sol, and dark grey for Late TA and Sol). Vertical bars are ± 1 s.e.m.*P < 0.05 post hoc Newman–Keuls test.

TA MEPs

Figure 4A shows that conditioned TA MEPs in one subject were significantly smaller (paired t test, P < 0.01) than test MEPs when FN stimulation preceded TMS by 20 ms during tonic and walking. Full time course of the effect in the same subject is illustrated in Fig. 4B showing similar FN-induced changes during tonic and walking (at two delays) with facilitation at short interval (–5 ms) and inhibition at longer intervals (20–30 ms).

On average, test MEPs had similar size (19.1 ± 2.6%M_max during tonic versus 19.1 ± 3.3 onset TA versus 21.3 ± 2.9 middle TA; Kruskal–Wallis, P = 0.70) and were evoked with TMS outputs within the same range (40.8 ± 3.5% maximum output during tonic versus 43.2 ± 3.9 onset TA versus 41.2 ± 3.5 middle TA; P = 0.85). FN stimulation significantly increased the MEP size in 4/5 subjects (between –12 and 0 ms, depending on the subject) during tonic co-contraction, and in only one subject during walking. The mean amount of facilitation tended to be larger during standing (120.4 ± 3.8% test size) as compared to walking (105.9 ± 8.2 and 100.9 ± 8.8%, at onset and middle TA, respectively) but the difference did not reach statistical significance (Fig. 4E, Kruskal–Wallis, P = 0.08). At longer intervals (between 10 and 40 ms), FN significantly reduced the MEP size in 3/5 subjects during tonic and middle TA during walking, and in 4/5 at onset TA during walking. On average, this long-interval inhibition did not significantly change between tasks (Fig. 4E; P = 0.81).

Effects on Sol motoneurones

Rectified Sol EMG averaging

Figure 5 shows that FN stimulation significantly increased (paired t test, P < 0.01) Sol EMG activity at 28.5 ms (peaking at 32.5 ms) during tonic co-contraction of Quad and Sol while standing (Fig. 5A) but not or hardly at all during walking (onset Fig. 5C, early Fig. 5E, and late Sol Fig. 5G) in one subject. This facilitation was followed by significant long lasting EMG suppression during tonic co-contraction, which was also observed when FN stimuli were delivered at the end of the walking stance phase (Fig. 5G) but hardly at all when triggered earlier (Fig. 5C and E). Such EMG suppression was only produced when FN stimulation was adjusted so as to evoke a constant M response in VL EMG of 30%M_max (Fig. 5B, D, F and H). No EMG suppression was evoked when stimulus intensity was adjusted below M response and H-reflex threshold (0.8 × MT, Fig. 5I and J). In all the 11 subjects investigated the same way, FN depressed Sol EMG activity only when FN stimulation produced M and/or H-reflex in VL EMG (checked during tonic and walking contraction).

On average, FN-induced Sol EMG facilitation was observed at 26.6 ± 0.7 ms. It was very weak (0.62 ± 0.24%M_max) and rarely significant during standing (4/11 subjects). During walking, FN-induced facilitation was even more difficult to evoke (0.18 ± 0.09% onset Sol), and reached statistical significance only in 2/11 subjects in early Sol (mean 0.27 ± 0.11%), and in 3/11 in late Sol (mean 1.03 ± 0.52%).

FN stimulation suppressed Sol EMG activity while standing in all the 11 subjects (significantly in 9/11 subjects) and grouped data were analysed to compare the amount of inhibition between tasks. Figure 3B shows the significant correlation between the control EMG and the amount of inhibition (Pearson's correlation, r = 0.55, P < 0.001), whatever the motor task. ANCOVA, by using the control EMG as covariate, revealed highly significant change in FN-induced inhibition between tasks (P < 0.01; Fig. 3D), which suggests that its level was task and walking phase dependent. Post hoc analyses revealed significant (Newman–Keuls, P < 0.05) decrease of inhibition at the onset of Sol activity during walking (as compared to tonic and late Sol) and significant (P < 0.05) increase at the end of stance (late Sol as compared to onset Sol).

The mean control EMG activity in Sol during tonic co-contraction corresponded to that recorded in early stance during walking (onset and early Sol), i.e. 15–20% of the mean EMG level recorded during MVC. At the end of the stance phase (late Sol), the mean control EMG reached 35–40% that of MVC. FN stimulation produced an H-reflex in VL EMG in 9/11 subjects. The size of the reflex was within the same range whatever the motor task: 19.6 ± 3.5%M_max in tonic versus 22.7 ± 4.0% for onset Sol versus 19.5 ± 4.7% for early Sol versus 12.5 ± 6.3% for late Sol (ANOVA, P = 0.62). At the end of stance (late Sol), the VL H-reflex was larger in three subjects and smaller in three other subjects, as compared to early stance or during tonic co-contraction.

Sol MEPs

Figure 4C shows that conditioned Sol MEPs in one subject were significantly smaller (P < 0.01, paired t test) than test MEPs when FN stimulation preceded TMS by 20 ms during tonic contraction and at the end of the stance phase (Late Sol) but larger at the onset of Sol activity during walking. The full time course of the effect in the same subject is illustrated in Fig. 4D showing that FN stimulation similarly increased MEP size during tonic and walking at short intervals (–10 ms). At longer intervals (10–20 ms), FN stimulation similarly reduced the MEP size during tonic contraction and at the end of the stance phase (Late Sol), but tended to enhance the MEP size at the onset of Sol activity during walking.

On average, test MEPs had similar size (6.9 ± 2.3%M_max during tonic versus 7.7 ± 3.3% onset Sol versus 7.9 ± 1.8% late Sol; Kruskal–Wallis, P = 0.77), and were evoked with TMS outputs within the same range (51.2 ± 2.9% maximum output during tonic versus 50.4 ± 3.1% onset Sol versus 48.6 ± 2.2% late Sol; P = 0.45).

FN stimulation significantly increased the MEP size in 3/5 subjects (between –10 and –5 ms, depending on the subject) during tonic co-contraction, in only one subject in early stance, and in 2/5 in late stance. The mean amount of facilitation tended to be larger during standing (119.8 ± 6.7% test size) compared to walking (93.2 ± 12.3 and 107.2 ± 5.1% at onset and late Sol, respectively) but the difference did not reach statistical significance (Fig. 4E; Kruskal–Wallis, P = 0.18). At longer intervals (between 10 and 30 ms), FN significantly reduced the MEP size in all the 5 subjects during tonic co-contraction and in late stance but in only 2 in early stance. On average, this long-interval inhibition significantly changed depending on the motor task (Fig. 4F; P < 0.05) with significant decrease of inhibition at the onset of Sol activity during walking, as compared to tonic (Wilcoxon, P < 0.05) or late stance (Late Sol; P < 0.05).

FN-induced inhibition of Sol H-reflexes during TA activation

Figure 6 shows that FN stimulation significantly (paired t test, P < 0.01) reduced Sol H-reflex size at intervals between 5 and 35 ms in one subject during tonic co-contraction of Quad and TA while standing and to a lesser extent at the end of the walking swing phase, when TA was strongly activated (Middle TA, Fig. 6A).

The mean Sol H-reflex test size was similar during the two motor tasks (4.0 ± 1.0% during walking versus 5.1 ± 1.0%M_max during tonic contraction; Wilcoxon, P = 0.08). Grouped data obtained by averaging results at intervals between 10 and 25 ms (Fig. 6B) revealed significant decrease of FN-induced Sol H-reflex inhibition in end swing and beginning of stance (1000–1075 ms and 0 ms after heel strike), as compared to tonic contraction while standing (Wilcoxon, P < 0.05).

Discussion

The main result of the present study is that FN-induced inhibition in Sol motoneurones was modulated during walking. When Quad and Sol activity overlapped during the gait cycle (early and late stance phase), the inhibition was smaller at the onset of Sol activity in early stance, as compared to standing, and was stronger at the end of stance, as compared to early stance. Moreover, around the time of heel strike, when Quad was activated with TA during walking, the FN-induced inhibition in Sol motoneurones decreased as compared to tonic co-contraction of Quad and TA while standing.

Pathways mediating FN-induced changes in TA and Sol motoneurones

FN-induced facilitation

FN-induced monosynaptic group Ia excitation in TA and Sol motoneurones was first shown by studying single motor unit poststimulus histograms (Meunier et al. 1990, 1993). Since FN stimulation can modify rectified EMG averaging, H-reflex and MEP size with similar characteristics, it has been suggested that the resulting facilitation could be mediated through the same pathway (Meunier et al. 1996; Barbeau et al. 2000; Fig. 1A). In the present study, FN stimulation hardly evoked excitation (with similar characteristics) in TA and Sol motoneurones during walking, and our results are thus consistent with previous studies showing a decrease in group Ia excitation during walking, as compared to tonic contraction while standing, which has been attributed to increased presynaptic inhibition of group Ia terminals mediated by PAD interneurones (Fig. 1A; Morin et al. 1982; Capaday & Stein, 1986; Faist et al. 1996a).

FN-induced inhibition

A long lasting inhibition in TA and Sol motoneurones has been observed after FN-induced monosynaptic group Ia excitation in several previous studies (Meunier et al. 1990, 1994, 1996). In these studies, it was suggested that Renshaw cells, activated by Quad motor axon recurrent collaterals, are likely to mediate the inhibition; ischaemia experiments have further supported this hypothesis (Barbeau et al. 2000). It has also been proposed that FN-induced EMG suppression should be analysed within an analysis window starting 10 ms after group Ia excitation to exclude possible contribution of group Ib inhibition, and during 12 ms to limit the analysis to spinal transmission (Barbeau et al. 2000). Similar to the previous studies, FN-induced EMG suppression was evoked in the present study only when producing Quad motor axon discharge during both tasks, and given the analysis window and interstimulus intervals for changes in MEP and H-reflex size, we assume that the amount of inhibition compared during the two motor tasks mainly reflected heteronymous recurrent inhibition from Quad to TA and Sol motoneurones.

Modulation of recurrent inhibition during walking

To our knowledge, modulation of recurrent inhibition during human walking has been investigated to a lesser extent than monosynaptic Ia excitation (Iles et al. 2000). However, a preliminary study reported in abstract form revealed modulations of the FN-induced inhibition in TA and Sol motoneurones during the gait cycle; inhibition was reduced in the transition phases, as compared to tonic contraction (Faist et al. 1996b). In the present study, changes in the amount of inhibition in TA motoneurones paralleled those in the TA background EMG activity, and only inhibition in Sol motoneurones was task and phase dependent (independent of its background EMG activity). In the former study, supra-maximal FN stimulation was used to evoke EMG suppression, which could lead to occlusion. Such saturation could then have masked distinct modulation between TA and Sol motoneurones. In addition, it seems that EMG suppression was analysed in all its duration, which might reflect changes at both spinal and supra-spinal levels (cf. Barbeau et al. 2000).

The present study revealed a distinct modulation of recurrent inhibition from Quad to TA and Sol motoneurones during walking. Only inhibition to Sol motoneurones changed during walking, with depressed inhibition at the onset of Sol activity in early stance and enhanced inhibition in late stance. These modulations are different from those reported during isolated Sol contractions. Indeed, recurrent inhibition (homo- and heteronymous pathways) has been shown to decrease with the level of muscle activity during tonic contraction (Hultborn & Pierrot-Deseilligny, 1979; Iles & Pardoe, 1999), whereas recurrent inhibition increased with the level of muscle activity during walking (15–20 and 35–40% MVC in early and late stance, respectively). Moreover, recurrent inhibition is reduced towards the end of ramp-and-hold ankle plantarflexion (Nielsen & Pierrot-Deseilligny, 1996) but during walking, inhibition increased with Sol activity. Lastly, homonymous recurrent inhibition has been reported to be enhanced during antagonist contraction (Katz & Pierrot-Deseilligny, 1984), while heteronymous recurrent inhibition is not modified (Iles & Pardoe, 1999). During walking, the heteronymous recurrent inhibition from Quad to Sol motoneurones was reduced when Quad was coactivated with TA around heel strike, as compared to tonic co-contraction of TA and Quad while standing.

Possible mechanisms underlying changes in recurrent inhibition to Sol motoneurones

Activation of Renshaw cells

Recurrent inhibition from Quad results from the activation of its motor axons by the direct M-wave, the reflex discharge, and its background activity during movement. Any change at one of these levels could influence the amount of heteronymous recurrent inhibition produced in Sol motoneurones but: (i) M-wave: attention was paid to keeping the M response in VL EMG within the same proportion of M_max measured for each condition, which suggests that the electrically evoked motor axon discharge was the same during each motor task; (ii) reflex discharge: the VL H-reflex size (see Fig. 5) was on average within the same range in the various conditions, and (iii) VL background activity: it has been shown that recurrent inhibition to Sol motoneurones does not depend on the level of Quad contraction (Pierrot-Deseilligny et al. 1977), and in heteronymous pathways, the contraction of the test muscle (Sol in the present study) influences more the level of recurrent inhibition than that of the source muscle (Quad; Iles et al. 2000). Moreover, although variations in Quad activity occurred during both TA and Sol activation, only recurrent inhibition in Sol motoneurones was modified during walking.

From the present experiments, it is difficult to draw any conclusion as regards the mechanisms responsible for the changes in recurrent inhibition of Sol motoneurones during walking. Nevertheless, some of the possibilities should be mentioned:

Descending control on Renshaw cells

Recurrent inhibition is depressed during weak voluntary contraction (Hultborn & Pierrot-Deseilligny, 1979; Rossi & Mazzocchio, 1991) and, accordingly, TMS over the primary motor cortex decreased recurrent inhibition (Mazzocchio et al. 1994). Motor cortex contributes to both Sol and TA activation during walking (Petersen et al. 1998, 2001; Christensen et al. 2001), and heteronymous recurrent inhibition was modified only in Sol motoneurones. Our results might thus suggest a differential corticospinal control on Renshaw cells, with specific inhibition of those projecting on Sol motoneurones. The vestibular influence is most important around the time of heel strike and during the double support phase (Bent et al. 2004), i.e. when recurrent inhibition of Sol motoneurones decreased during walking. The vestibulospinal inputs to Renshaw cells (Rossi et al. 1987) may thus account for the depression of recurrent inhibition during walking at that time in the gait cycle. The vestibular system is involved in foot placement during walking (Bent et al. 2004), which requires a precise control of both ankle flexors and extensors (Iles et al. 2007), and recurrent inhibition was only depressed in extensor motoneurones. This might thus reflect a specific vestibulospinal control on Renshaw cells controlling Sol motoneurones during walking.

Changes in peripheral inputs

Peripheral afferents of various origins are involved in the control of human walking (see Zehr & Stein, 1999). A well-known example is the reversal control of cutaneous afferents on ankle muscles during walking (Duysens et al. 1990; van Wezel et al. 2000). On the other hand, unloading Sol during walking causes a drop in its EMG activity, which is thought to reflect the contribution of peripheral afferents to motoneurone activation; this contribution is maximal at the onset of the stance phase and involves muscle spindle group Ib and group II afferents (Sinkjaer et al. 2000). Accordingly, partial removal of the body weight during walking reduces EMG activity in ankle extensors, but activity in flexors is enhanced (Finch et al. 1991; Bastiaanse et al. 2000). This suggests that muscle afferents (group Ib/group II) might contribute to TA activation to a lesser extent than Sol during walking (see Zehr & Stein, 1999), which could account for the different modulation of recurrent inhibition between ankle muscles. Moreover, since afferents from single- and multijoint muscles may be differentially interpreted, peripheral inputs from the gastronecmius muscles (acting at both ankle and knee level) may also account for the specific control of Renshaw cells controlling extensor motoneurones during locomotion (Sturnieks et al. 2007).

Cutaneous and muscle spindle group II hypotheses

In cats, cutaneous and muscle spindle group II inputs inhibit Renshaw cell activity (Wilson et al. 1964; Fromm et al. 1977), with specific depression of recurrent inhibition of extensor motoneurones for the latter. This may account for the depression of recurrent inhibition during walking, and if an analogous mechanism exists in humans, the control by group II afferents would account for the specific depression of recurrent inhibition from Quad to Sol at the onset of the walking stance phase. Accordingly, we have shown that group II excitation is particularly enhanced at the beginning of stance (Marchand-Pauvert & Nielsen, 2002a), and might contribute to the gait posture (Marchand-Pauvert et al. 2005; Iglesias et al. 2008). However, Renshaw cells are not inhibited during cat fictive locomotion (Pratt & Jordan, 1980).

Group Ib hypothesis

Group Ib inputs produce excitation in extensor motoneurones during cat fictive locomotion (Pearson and Collins, 1993; Gossard et al. 1994), and there is some evidence for similar mechanisms in humans (Stephens & Yang, 1996; Marchand-Pauvert & Nielsen, 2002b; Faist et al. 2006). Since short (oligosynaptic pathway) and long latency (polysynaptic pathway; see McCrea, 1998; McCrea & Rybak, 2008) group Ib EPSPs occur at the motoneurone level, it is thus conceivable that the summation (whether linear or not) of excitatory and inhibitory inputs at the motoneurone level makes them more or less susceptible to Renshaw inhibition in early stance, which could thus account for the modulation of recurrent inhibition of Sol motoneurones during locomotion (Burke et al. 1971; Segev & Parnas, 1983; Powers & Binder, 2000). Similar mechanisms of synaptic integration might also account for the decrease in recurrent inhibition of Sol motoneurones around the time of heel contact, when Quad is coactivated with TA, as compared to standing, since potent group Ib excitation is evoked during the flexor phase to initiate the transition between swing and stance (Conway et al. 1987; Gossard et al. 1994). Lastly, changes in load on extensors during locomotion might contribute to the increase in recurrent inhibition at the end of the stance phase by influencing group Ib transmission (Gossard et al. 1994; Faist et al. 2006).

Alternatively, changes in recruitment gain and/or activation of fast motor units, which increases the discharge in recurrent collaterals (Wand & Pompeiano, 1979), may also account for the enhanced recurrent inhibition in late stance.

In decerebrated cats, Renshaw cell activation is not modified during fictive locomotion, and there is a linear relationship between motoneurone firing rate and the amount of recurrent inhibition (Pratt & Jordan, 1980; McCrea et al. 1980; Noga et al. 1987). This might account for the parallel changes of FN-induced recurrent inhibition in TA and of its background EMG activity at the end of the swing phase.

Functional significance

The decrease in recurrent inhibition from Quad to Sol around the time of heel strike, and in early stance, may help initiate the transition from swing to stance and thus favour ankle extensor activation. In addition, since Renshaw cells also inhibits interneurones mediating reciprocal group Ia inhibition (Hultborn et al. 1971; Baret et al. 2003), the decrease in recurrent inhibition of Sol motoneurones may favour reciprocal inhibition of TA motoneurones at the same time. In contrast, at the end of stance, the enhanced recurrent inhibition of Sol motoneurones could help release TA motoneurones from reciprocal inhibition and thus could initiate the transition from stance to swing. This has been suggested as one of the possible roles of Renshaw cells during cat fictive locomotion (McCrea et al. 1980; Noga et al. 1987), and such a control is thought to assist muscle synergies also in humans (Hultborn & Pierrot-Deseilligny, 1979; Nielsen & Pierrot-Deseilligny, 1996). Moreover, given the distribution of heteronymous recurrent inhibition in the human lower limb (Meunier et al. 1994), the enhanced inhibition from Quad to Sol at the end of the stance phase must result in decreased activity of the Renshaw cells activated by Sol motoneurones, and thus reduced recurrent inhibition from Sol to Quad at the end of the walking stance phase (Iles et al. 2000). This mechanism may assist hip flexion by favouring rectus femoris (another head of Quad) activation at this stage of the gait cycle (Nene et al. 2004). Lastly, since the ability to modulate recurrent inhibition appropriately during movement is lost in patients with movement disorders (see Pierrot-Deseilligny & Burke, 2005), abnormal recurrent inhibition during locomotion might be involved in aberrant muscle synergies during pathological gait.