INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Single-unit recording studies in cats (Almaric et al. 1983; Batson and Amassian 1986; Burton and Onoda 1978; Ghez and Kubota 1977; Houk et al. 1987; Martin and Ghez 1988, 1991; Padel and Steinberg 1978; Schmied et al. 1988), rats (Jarratt and Hyland 1999), and primates (Gibson et al. 1985a,b; Mewes and Cheney 1994; Miller and Houk 1995; Miller et al. 1993; Otero 1976; van Kan and McCurdy 2001) have shown that neurons in the magnocellular region of the red nucleus, including rubromotoneuronal cells (Fetz et al. 1989; Mewes and Cheney 1994), increase their discharge frequency during the execution of voluntary movements. Detailed analysis of these discharge patterns suggests that in most cases the increase of discharge frequency precedes the onset of the movement (Almaric et al. 1983; Ghez and Kubota 1977; Gibson et al. 1985b; Martin and Ghez 1988; Mewes and Cheney 1994; Otero 1976) and may be correlated with different parameters of the movement, such as the velocity (Burton and Onoda 1978; Gibson et al. 1985b; Mewes and Cheney 1994), or with the temporal characteristics of the electromyographic (EMG) activity that produces that movement (Miller and Houk 1995; Miller and Sinkjaer 1998; Miller et al. 1993). Furthermore, the studies of several authors (Gibson et al. 1985a, 1994, 1998; Mewes and Cheney 1994; Miller et al. 1993; Sinkjaer et al. 1995; van Kan and McCurdy 2001) suggest that the red nucleus may play a particular role during coordinated, multi-articular movements such as reaching, particularly when these movements involve the use of the hand e.g., reach-to-grasp.

During locomotion, the only information on the discharge patterns of rubral neurons comes from experiments in the decerebrate cat (Arshavsky et al. 1988; Orlovsky 1972a). The results from these experiments show that rubrospinal neurons are phasically active during locomotion in the decerebrate cat and discharge preferentially in the swing phase of locomotion where they could influence the activity of flexor muscles (Orlovsky 1972a). This is in agreement with the results from studies in both the decerebrate (Degtyarenko et al. 1993; Orlovsky 1972b) and intact cat (Rho et al. 1999) showing that microstimulation within the red nucleus during locomotion preferentially modifies the activity of physiological flexor muscles. Nevertheless, both the unit recording and microstimulation studies indicate that the red nucleus may also influence the activity of extensor muscles active at the end of the swing phase and during stance. Electrolytic lesion (Ingram and Ranson 1932), excitotoxic lesion (Muir and Wishaw 2000), or pharmacological inactivation (Gibson et al. 1994) of the red nucleus leads to evident but relatively mild locomotor deficits during overground locomotion, although the study of Ingram and Ranson (1932) reported stronger deficits when the cats walked in a cluttered environment.

Taken together, the data suggest that the red nucleus does contribute to the normal control of walking, although the nature of this contribution is unclear, in part because of the lack of any data on cell discharge characteristics in the intact, walking cat. Moreover, given the comments of Ingram and Ranson (1932) concerning the increased loss of control in more challenging circumstances and the wealth of evidence in primates demonstrating the importance of the red nucleus in the control of voluntary movements, it seems probable that the red nucleus, like the motor cortex (Amos et al. 1990; Armstrong 1988; Beloozerova and Sirota 1993; Drew 1988, 1993; Drew et al. 1996; Widajewicz et al. 1994), might contribute more strongly to the regulation of locomotion when the cat has to adapt its gait to the environment. To test this general hypothesis, we used single-unit recording techniques in the intact cat to characterize the nature of the discharge characteristics of rubral neurons during a task requiring a modification of the base locomotor rhythm. To allow direct comparison of the results from this study with those that we have previously obtained from our recordings of identified pyramidal tract neurons (PTNs) in the motor cortex (Drew 1993; Widajewicz et al. 1994), we used the identical task requiring the cats to modify their gait to step over an obstacle attached to a moving treadmill belt. This task requires that the cats adjust the spatiotemporal pattern of the muscle activity in their limbs to produce the changes in limb trajectory needed to step over the obstacle without touching it. As such, it provides an appropriate method for characterizing the discharge characteristics of rubral neurons in a situation in which the locomotor pattern has to be modified.

A preliminary report of this study has been published as an abstract (Lavoie and Drew 1997).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Training and surgery

Experiments were carried out on three cats (weight: 5.4-5.9 kg) that were initially trained to walk on a treadmill at speeds circa 0.4 m.s¹ and to step over obstacles attached to the moving belt (see Drew 1988, 1993). These were the same three animals that were used in our previously published paper detailing the effects of microstimulation of the red nucleus on the locomotor rhythm and pattern (Rho et al. 1999). Following training, the cats were prepared for surgery under general anesthesia and in aseptic conditions. All procedures followed the recommendations of the Canadian Council for the Protection of Animals and were approved by the local animal ethics committee at the Université de Montréal.

The surgical procedures used in these experiments are detailed in Rho et al. (1999). In brief, a recording chamber was attached over a craniotomy in the parietal bone to provide access to the red nucleus on the right hand side. A bundle of three microwires was implanted stereotaxically into the left brachium conjunctivum at P-7.5, V-3, L3.8, and, in two animals, a similar bundle was implanted stereotaxically into the left rubrospinal tract at P9.2, V-8.5, and L4.5. In two animals, three microwire electrodes were inserted manually into the left dorsolateral funiculus at L₂. In all animals, pairs of Teflon-insulated, braided stainless-steel wires were implanted into selected muscles of all four limbs to record EMG activity during locomotion. The selected muscles included physiological flexor and extensor muscles acting around all the major joints of the forelimb and hindlimb contralateral to the recording site (see Rho et al. 1999 for a list of these muscles and their major functions). Each animal was administered Buprenophrine (hydrochloride, 5 µg/kg) for the 48 h following the surgery. Antibiotics (cephalexin monohydrate 50 mg/kg) were administered throughout the period of study of these animals. Experiments were started 1 wk after the surgery.

Protocol

Glass-insulated, tungsten microelectrodes (impedance: 0.5-2.0 M), held in a custom-made micromanipulator attached to the recording chamber, were driven through the parietal cortex and the superior colliculus to a position previously calculated to be just dorsal to the red nucleus. The electrode was then advanced slowly into the red nucleus. In initial experiments, the exact location of the forelimb representation of the red nucleus was determined on the basis of the presence of a large, short-latency (<1 ms) field potential following stimulation of the electrodes in the brachium conjunctivum (see e.g., Figs. 2C and 5C), of large neurons with a cutaneous receptive field on the contralateral forelimb and by brief, twitch responses, restricted to the forelimb, following stimulation through the microelectrode (11 pulses at 330 Hz, pulse duration: 0.2 ms) at strengths of <20 µA (not illustrated). In these initial experiments, the electrode was then withdrawn ~1 mm and left in place for 10-15 min before being again slowly advanced into the red nucleus. The data from these initial experiments were used to directly position the electrode above the predicted location of the red nucleus in later experiments.

As the electrode was advanced, action potentials that were clear of the noise were tested for a receptive field by gentle manipulation of the body. In many penetrations, the action potentials that were initially isolated had receptive fields around the face or neck of the animal (see Fig. 1). These cells were not routinely recorded during locomotion, although a number of such cells, in all three cats, were sampled to determine whether they were phasically active during the task. Below these cells, or intermingled with them, we isolated cells with receptive fields that were restricted to or included the contralateral forelimb. From this point on, the discharge frequency of all neurons with stable, isolated action potentials were recorded during locomotion with either one or two obstacles attached to the treadmill belt. Unit and EMG activity were recorded on a 14-channel instrumentation tape recorder, and we also made simultaneous video recordings of the cat during this time. A digital time code simultaneously recorded on both media allowed synchronization of these recordings. Following the recordings during locomotion, the cat was removed from the treadmill, and the receptive field of the cell was more carefully mapped by gentle manipulation of the limbs and body of the animal. In sites at which well-modulated activity was recorded during locomotion, short trains of microstimulation at 25 µA were applied and the evoked movements, as well as the threshold for the effects noted (see Rho et al. 1999). The cat was then replaced in the treadmill and the electrode advanced until another single unit was isolated. Recording sessions normally lasted for 2-3 h and terminated when the cats were no longer willing to walk. In selected penetrations, small electrolytic lesions were made above, below or, occasionally, within the predicted borders of the red nucleus to aid in histological reconstruction of the electrode penetrations.

View larger version (45K): [in this window] [in a new window]   Fig. 1. Location of representative examples of electrode penetrations in cat RN4. A: digitized image of a photomicrograph of the red nucleus at a laterality (L) of 2.4 mm from the midline showing the location of penetrations 32 and 15. A 3rd penetration can be seen between the 2 that are identified on the figure. The figurines to the right of the photomicrograph illustrate the receptive fields of cells recorded in each of the identified penetrations. The numbers beside each of the figurines indicate the calculated location of each of the recorded cells as indicated on the photomicrograph. B and C: organized in the same manner as in A but show tracings of the histological sections instead of photomicrographs. Shaded regions on the figurines indicate cutaneous receptive fields, curved arrows indicate that the cell was activated by movement of the indicated part; the pair of small straight arrows in cell 6 in track 11 indicate that this cell was activated by pressure applied to the paw. FL, the receptive field was identified as being on the forelimb but could not be more precisely defined because the cell was lost before identification was complete; 3N, oculomotor nerve; PG, pontine gray; PT, pyramidal tract; RN, red nucleus; SN, substantia nigra; T, track (or penetration); TRN, tegmental reticular nucleus.

Following all of the recording and stimulation procedures, the same cats were used to examine the effects of microinjections of muscimol into the red nucleus (not illustrated in this report).

Data analysis

Following the experiments, the data were displayed on an electrostatic printer (Gould ES 2000), and sections of data in which the action potential and the locomotion were stable were selected for analysis. Action potentials were amplitude-discriminated and transformed into digital pulses; these were sampled, together with the EMG data, at a frequency of 1 kHz. A custom-written program was used to mark the onset and offset of the periods of EMG activity and to identify the step cycles during which the cat stepped over the obstacle as well as those cycles preceding this step. As in our previous studies (Drew 1993; Widijawicz et al. 1994), steps over the obstacle were further divided into those in which the leg contralateral to the recording site was the first to pass over the obstacle (the lead limb) and those when it was the second (trail limb). The identification of these steps was based on inspection of the simultaneously recorded videos; the latter were also used to ensure that the cat neither touched the obstacle nor hesitated during the gait modification. Control cycles were defined as those that occurred two steps before the one over the obstacle as our previous studies showed that neither EMG activity nor the discharge pattern of PTNs was different during this step than during locomotion when no obstacles were attached to the treadmill belt (Drew 1993).

Once identified, the selected cycles were used to construct averages that were triggered on the onset of either the contralateral cleidobrachialis (coClB) or the contralateral sartorius (coSrt), the onset and duration of which correspond, approximately, to the onset and duration of the forelimb and hindlimb swing period, respectively (Drew 1993; Widajewicz et al. 1994). To average the signals, both the instantaneous discharge frequency of the cell and the EMG activity during each cycle was divided into 256 bins (Drew 1993; Drew and Doucet 1991; Udo et al. 1982). The discharge activity during the steps over the obstacle was superimposed on the activity during the control cycle. The latter activity was displayed together with the 0.01 confidence interval of the standard error of the mean. Periods of activity that exceeded the upper or lower boundaries of this level of confidence for more than 25 consecutive bins were deemed to be significantly different from control activity (Drew 1993). When two obstacles were attached to the treadmill, the data from each were combined. Unit activity was also displayed in the form of rasters, triggered successively on the onset of each of the recorded muscles, to examine the timing relationships between the periods of unit activity and the period of activity of each of the recorded muscles (Drew 1993; Drew et al. 1986).

To determine whether cells were phasically modulated and during which parts of the step cycle they showed peak activity, we used a combination of methods. First, we used circular statistics and the Rayleigh test for directionality (P < 0.01) to determine which cells showed phasic (i.e., nonuniform discharge) modulation (Batschelet 1981; Drew 1993; Drew and Doucet 1991). However, because many of the cells in the red nucleus showed multiple periods of discharge activity (see RESULTS), we could not use circular statistics to routinely determine the mean phase and level of activity of these cells in the same way as for PTNs (Drew 1993). For those cells that were phasically modulated, we therefore determined the phase of neuronal discharge from normalized averages of the cell discharge, triggered on the onset of coClB or coSrt. For control steps, the overall average level of the discharge during each step cycle was calculated, and the points at which the histogram crossed this level were considered to represent the onset and offset of the period of burst activity (see e.g., Figs. 2B and Fig. 3). To quantify the data during the steps over the obstacle, similar methods were used except that we found that for four cells, some significant increases in discharge were too small to be included on the basis of the overall average. For these four cases, the phase of discharge was calculated on the basis of the overall discharge rate for the major burst as well as on the basis of mean discharge for the minor, but significant, burst; an example of one of these four cells can be seen in Fig. 6A. It should be noted that a comparison of the phase of activity of the cortical neurons obtained using the current method with that obtained using circular statistics showed only small differences in the values obtained using the two methods (compare Fig. 4D in this report with Fig. 5 in Drew 1993).

To identify the part of the step cycle during which cells were maximally active during the gait modifications, they were classified in a similar manner to our previous publications. Three of these groups were identical to those previously used to identify PTNs. For cells related to the forelimbs (see RESULTS), these were cells whose peak discharge activity occurred just prior to, or just after, the onset of the activity in the coClB but prior to the onset of activity in the wrist and digit dorsiflexor, coEDC (phase I cells); cells whose peak discharge occurred subsequent to the onset of activity in the EDC but before the end of the period of activity in the coClB (phase II cells); and cells in which the increase in discharge began >200 ms prior to the onset of activity in the coClB (early). Because we identified a group of cells in the red nucleus whose activity was temporally related to the swing period of the ipsilateral limb (see e.g., Fig. 3A), and thus discharged during the stance phase of the contralateral limb, we subdivided our previous stance classification (Drew 1993) into cells whose peak discharge was included within the period of activity of the iClB (i.e., ipsilateral swing cells); and those cells whose peak discharge occurred outside the period of the coClB activity, excluding group 4 (i.e., stance cells). In addition, on the basis of our initial inspection of the discharge patterns of these cells, we designated a further two categories: cells whose discharge activity was maximal throughout the period of swing (phase I + II cells) and those cells that showed a period of activity in phase I and another in phase II (phase I and phase II cells). For cells related to the hindlimbs, the same classifications were maintained with respect to the coSrt and the extensor digitorum brevis (EDB).

Histology

At the end of the experimental procedures, the animals were killed with pentobarbital sodium and perfused per aortum with a solution of saline, followed by 4% paraformaldehyde and 4% sucrose. The brain stem was removed, blocked, and stored in 20% sucrose overnight. Frozen sections (30-40 µm) were cut and mounted. These sections were stained with cresyl violet and used to trace the penetrations into the red nucleus, using the marking lesions as a guide.

Motor cortex

The data for the motor cortex that are included in this report are taken from the identical population of neurons, recorded in the forelimb representation of area 4, that we used in a previous publication (Drew 1993). These data have been reanalyzed using the same methods as for the red nucleus cells to allow a direct comparison of discharge frequencies and phases of activity. With the exception of the examples of neuronal activity that are used to illustrate the summary figure (Fig. 11), only the results of this synthesis are illustrated here.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Database

A total of 157 neurons from 45 penetrations were recorded during locomotion. All of these penetrations were histologically verified to lie within the boundaries of the red nucleus with most trajectories lying within the caudal two-thirds of the nucleus that is normally defined as the magnocellular region of the red nucleus (Massion 1967; Padel et al. 1972; Pompeiano and Brodal 1957; Robinson et al. 1987).

Figure 1 illustrates histology from one cat, RN4, and shows the location of some of the penetrations made at different lateralities in this cat, together with the receptive fields of the cells recorded in representative penetrations. In accord with previously published reports, we found that most cells with receptive fields confined to the contralateral hindlimb were located in the more lateral and ventral regions of the red nucleus (see e.g., tracks 15 and 32 in Fig. 1A), whereas neurons with receptive fields restricted to the forelimbs were more scattered throughout the red nucleus but were concentrated in the more medial and dorsal regions (tracks 11, 24, 25, and 29 in Fig. 1, B and C). Cells with receptive fields confined to the face, neck and proximal forelimb were normally found in the most dorsal parts of the trajectories, especially in more rostral regions (e.g., tracks 11 and 29), although cells with very distal receptive fields were also sometimes recorded in these most dorsal regions (e.g., cell 1 in track 11).

Receptive fields were tested for 138/157 of these neurons and could be identified for 129/138. Of these, the majority (72/129, 56%) had a receptive field that was restricted to the contralateral forelimb, and, for most of these (50/72) the receptive field was restricted to those parts of the forelimb including and distal to the elbow. A further 22/129 (17%) had a receptive field that was restricted to the hindlimbs. In part, this preponderance of cells with receptive fields on the forelimb, compared with the hindlimb, is a result of a bias in the recordings in that the cat was normally unwilling to continue walking by the time that the penetration had advanced to the more ventral regions of the nucleus where the hindlimb cells were concentrated. Relatively few cells were recorded with receptive fields restricted to the face and neck (15/129, 12%) or the trunk (4/129, 3%), largely because of our tendency to advance the electrode rapidly through the most dorsal regions of the red nucleus where these cells were most frequently located. Of the remaining cells, 14/129 (11%) had a receptive field that overlapped more than one of the above categories and 2 cells had a receptive field restricted to the ipsilateral forelimb. The vast majority (120/129) of these neurons were activated by brushing or light tapping of the skin; because of their extreme sensitivity to these cutaneous stimuli, we were unable to accurately determine if they received additional deeper or proprioceptive afferent input. The other neurons (9/129) were activated only by deep tapping or joint movement (see e.g., track 11 in Fig. 1C).

The electrodes implanted into the brachium conjunctivum were accurately positioned in two of the cats (RN3 and RN5). Stimulation through these electrodes orthodromically activated a substantial proportion of the cells (66/93, 71%) that we recorded in these two cats. Unidentified cells were often recorded interspersed among the identified cells, suggesting that the lack of identification was probably due to an inability to completely activate the brachium (see Eccles et al. 1975b). The electrodes directed at the rubrospinal tract in the brain stem were accurately positioned in only one of the two cats in which they were inserted. These electrodes antidromically activated 19/42 (45%) of the tested cells in this cat. The actual percentage was probably higher but the very-short-latency responses were masked by the stimulus artifact. Spinal cord electrodes were accurately positioned in one of the two cats implanted. In this animal, 4/5 of the rubral neurons with a receptive field on the hindlimb were antidromically identified from these electrodes; in this same cat, none of the 16 neurons with a forelimb receptive field were activated from these electrodes.

Cell discharge during locomotion

Although the major objective of this study was to examine the discharge characteristics of rubral neurons during voluntary gait modifications, we first present data on the background discharge characteristics during normal locomotion in the intact cat. This is necessary as no other detailed source of information is available for locomotion in the intact cat and it also allows comparison with the only other detailed study available which was obtained in the decerebrate cat (Orlovsky 1972a). Because of the nature of our task, in which the forelimbs are obviously the first to encounter the obstacle, we concentrate on the 72 cells in which the receptive field was located exclusively on the forelimb. Data for the hindlimb, face, and neck are presented more concisely at the end of RESULTS.

CELLS WITH A RECEPTIVE FIELD RESTRICTED TO THE FORELIMB. Discharge activity during control locomotion. Of the 72 neurons with receptive fields restricted to the forelimbs, 66 discharged phasically as determined by the Rayleigh test for directionality. Figure 2 shows the discharge characteristics of one phasically active rubral neuron in the control cycles between gait modifications. Because only one obstacle was attached to the treadmill belt during this recording, there were seven or eight steps between the steps over the obstacle. This rubral neuron had a receptive field that extended from the paw to the upper arm but that was most intense on the dorsum and ventral surface of the paw: the neuron was orthodromically activated from the electrodes in the brachium conjunctivum at a latency of 1.2 ms (Fig. 2C). During locomotion, the neuron discharged in a clearly rhythmical manner with one intense peak of activity at the end of the swing phase, coincident with the period of activity of the wrist and digit dorsiflexor, EDC (Fig. 2, B and D). However, as in most of the rubral neurons that we recorded, the cell did not discharge in a simple unimodal pattern but showed a more complex pattern of peaks and troughs with another prominent peak of discharge activity at the end of stance, just prior to foot lift.

View larger version (59K): [in this window] [in a new window]   Fig. 2. Example of rubral discharge during control locomotion. A: raw tracing of unit activity (clipped during recording) together with the electromyographic (EMG) activity from 3 flexor and 1 extensor muscle. B: normalized and averaged activity of the neuron and the EMGs over 41 trials; 2 complete cycles are shown. For each trace, the solid line illustrates the averaged activity; the thinner, dotted lines illustrate the interval of confidence (P < 0.01) of the SE of the mean. The dashed horizontal line through the unit trace in the 2nd cycle indicates the average activity of the neuron during the step cycle; the filled circles indicate the onset and offset of each burst of activity (see METHODS). C: orthodromic activation of the rubral neuron by stimulation of the brachium conjunctivum and receptive field of the neuron. The neuron was more sensitive to activation of the distal limb (darker shade of gray) than to the rest of the forearm. #, field potential evoked by the stimulation of the brachium conjunctivum; *, the superimposed orthodromically evoked spikes. D: raster displays and post event histograms (PEHs) of unit activity triggered from the onset of activity (vertical dashed line) in the contralateral cleidobrachialis (coClB) and the 2nd burst of activity of the contralateral extensor digitorum communis (coEDC). The raster traces are rank-ordered according to the duration of the respective EMG bursts. The thin vertical line following the dashed line indicates the time of cessation of activity in the respective EMG; the vertical dashes to the extreme right of each raster (when present) indicate the onset of activity in the next step cycle. co, contralateral; TrM, teres major; TriL, triceps brachii, lateral head.

View larger version (53K): [in this window] [in a new window]   Fig. 3. Examples of 2 other rubral neurons during control cycles. The data are arranged in the same manner as for Fig. 2. i, ipsilateral.

View larger version (33K): [in this window] [in a new window]   Fig. 4. Distribution of the periods of discharge activity of rubral and cortical cells during control locomotion. A: periods of activity for 66 phasically modulated rubral neurons. Each horizontal line indicates one period of activity as defined in METHODS. , the time of peak activity in each cell. The cells are rank-ordered according to the phase of onset of the earliest period of activity. The cells illustrated in Figs. 2 and 3 are identified to the left side of the figure. B: phase of maximal peak activity of the 66 rubral neurons (). , the phase of peak discharge for all periods of phasic activity (that is 1 value for each - - - in A). C: distribution of the maximal discharge frequency of each rubral neuron. D-F: similar displays for 73 phasically active pyramidal tract neurons (PTNs) recorded from the motor cortex.

View larger version (49K): [in this window] [in a new window]   Fig. 5. Modulation of rubral discharge activity during a gait modification. A: untreated traces of unit and EMG activity during a lead step over the obstacle. B: averaged activity of the neuron during the gait modification. The thicker line indicates the averaged activity during the step over the obstacle (n = 27). The thin solid line, together with the dotted lines, illustrates the average activity together with the interval of confidence of the SE during control locomotion (n = 54). The circles indicate periods of phasic activity as defined in METHODS. C: orthodromic activation of the cell from the brachium conjunctivum and antidromic activation from the rubrospinal tract (RST). The asterisk indicates spontaneous action potentials that collided with antidromically evoked spikes. D: raster displays and PEHs triggered on the contralateral brachialis (coBr) during control walking and during the steps over the obstacles.

View larger version (58K): [in this window] [in a new window]   Fig. 6. Examples of the different types of rubral discharge observed during the gait modifications in the lead condition. A-D: 4 different neurons. The traces are organized as in Fig. 5.

View larger version (35K): [in this window] [in a new window]   Fig. 7. Periods of activity of the neuronal discharge during the steps over the obstacle (lead condition). The data are organized in the same manner as in Fig. 4 with the following exceptions. A and D: solid black lines indicate periods of phasic activity that were significantly different from control values according to the criteria given in METHODS while dotted lines indicate periods of phasic activity that were not significantly different from control (see text). B and E: shaded bars indicate the phase of peak activity of the cell while open bars indicate the peak periods of all periods of activity, regardless of whether they were significantly different from control or not.

View larger version (34K): [in this window] [in a new window]   Fig. 8. Comparison of discharge activity in the lead and trail conditions. The data are organized as in Fig. 5. The EMG data in D are taken from the cell illustrated in C. Note that in the lead condition the left limb (coClB) precedes the right limb (iClB), whereas in the trail condition the reverse is the case.

View larger version (17K): [in this window] [in a new window]   Fig. 9. Comparison of the phase of peak activity (A) and the magnitude of the peak discharge (B) in the lead and trail conditions. Data are plotted only for neurons that showed significantly increased activity during the swing phase in both conditions. In cases in which neurons showed multiple increases in activity during the swing period, only the phase of the earliest period of activity is plotted.

CELLS WITH A RECEPTIVE FIELD RESTRICTED TO THE HINDLIMBS OR TO THE FACE AND NECK. Twenty (20/22) of the cells with a receptive field limited to the hindlimb were phasically active during control locomotion. As for those cells with a receptive field restricted to the forelimbs, most of these (18/20) showed a period of increased discharge during the swing phase of the hindlimb and many of these (15/18) also showed a second period of increased activity during the stance phase. In some cases (5/15), this latter period of activity was temporally related to the swing period of the ipsilateral hindlimb. As for the forelimb cells, a proportion (6/18) of neurons exhibited multiple periods of increased activity during swing.

View larger version (25K): [in this window] [in a new window]   Fig. 10. A: example of a neuron which had a receptive field restricted to the hindlimb. B: example of a neuron in which the receptive field was located on the face and vibrissae. Traces in A are organized as in Fig. 5, traces in B show only the activity during the gait modification as this cell was silent in the control periods. EDB, extensor digitorum brevis; Srt, sartorius; VL, vastus lateralis.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This paper provides the first detailed report of the activity of rubral neurons during both unobstructed locomotion and in circumstances that required voluntary gait modifications of the fore- and hindlimbs. During unobstructed treadmill locomotion, the majority of rubral neurons were phasically modulated, with the maximum discharge frequency occurring during the swing phase of locomotion of either the contralateral forelimb or of the contralateral hindlimb. During the voluntary gait modifications, a majority of phasically modulated cells showed significant increases in their discharge frequency during the passage of either the fore- or hindlimbs over the obstacle.

Database and methodological considerations

Although, for technical reasons, only a few cells could be specifically identified as spinal projection neurons, the majority of the cells that we recorded had large action potentials that could be recorded over distances of several hundreds of micrometers and were, therefore, probably generated by neurons with a large soma. As most, if not all, large neurons in the feline red nucleus project to the spinal cord (Hayes and Rustioni 1981), it is likely that the majority of the cells that we recorded were rubrospinal and would, therefore, exert direct influences on spinal interneurons.

As in our motor cortical studies (Drew 1993; Widajewicz et al. 1994), we initially classified our population of cells into fore- and hindlimb related cells based on their receptive fields. That this is an appropriate method of classification for rubral neurons is suggested by our results showing that cells with receptive fields restricted to the forelimbs discharged only during the motor adjustments of the forelimbs, while those with receptive fields restricted to the hindlimbs discharged only during the motor adjustments of the hindlimbs. Moreover, our results showing that none of the cells with receptive fields restricted to the forelimbs could be activated from the lumbar spinal cord suggest that the majority of these rubral neurons exerted an influence primarily on either forelimb or hindlimb musculature. This segregation of motor function is in agreement with previous anatomical and physiological studies on the topography and input/output relationships of rubral neurons (Eccles et al. 1975a,b; Ghez 1975; Nishioka and Nakahama 1973; Padel et al. 1972; Pompeiano and Brodal 1957; Rho et al. 1999; Robinson et al. 1987) and particularly with the finding of Shinoda et al. (1977) that only a few of the rubral neurons (2/40) that projected to the cervical cord also sent an axon to lumbar (L₁) levels.

Rubral activity during locomotion

UNOBSTRUCTED TREADMILL LOCOMOTION. Almost all (92%) rubral neurons with a receptive field restricted to the fore- or hindlimb showed a phasic modulation of their discharge activity during unobstructed treadmill locomotion with the peak discharge of the majority of the modulated neurons (89% forelimb and 90% hindlimb) centered around the swing phase (phases: 0.9-0.4). These results, in general, conform with the data of Orlovsky (1972a), who recorded the activity of rubrospinal cells projecting to the lumbar cord in thalamic cats walking on a treadmill. In those experiments, he reported that the majority of neurons (83%) were also phasically modulated during locomotion and that most (62% of the total) discharged during the swing phase of locomotion. However, in comparison to the results reported by Orlovsky (1972a), we found a larger number of our neurons (55% of modulated neurons compared with 13% in his study) discharged both during swing and stance. It is possible that this difference in the two studies might be related to differences in functional control over the hindlimb as compared with the forelimb. However, we also found a high proportion of neurons discharging during both swing and stance in our smaller population of hindlimb related neurons. As such, it is more likely that this difference reflects either differences in the preparations used or differences in the resolution of the analytical techniques. The latter reason probably also explains the lack of any mention of cells discharging in multiple bursts during the swing phase in Orlovsky's report.

VOLUNTARY GAIT MODIFICATIONS. General comments. Most rubral neurons increased their discharge frequency during the gait modifications required to step over the obstacles. In general, the basic characteristics of the discharge activity in individual cells, including the preponderance of phasic activity centered around the swing phase of locomotion, the presence of neurons that discharged during both swing and stance and particularly the presence of multiple peaks of activity in individual neurons during swing were similar during unobstructed locomotion and during voluntary gait modifications. This suggests that the red nucleus, like the motor cortex, contributes both to the production of the base locomotor rhythm and to the increase in the level of the EMG activity when the cats need to modify their gait. However, many neurons did not simply show a change in the magnitude of their activity but also exhibited changes in their relative phase of activation (see Figs. 5B) or showed phasic activity during the gait modification where none was present during the unobstructed locomotion (see Fig. 6). This latter characteristic resulted in neurons exhibiting more phasic peaks of activity during the gait modifications than during the unobstructed locomotion. These changes in the magnitude, duration and timing of the cell discharge patterns, together with the appearance of new periods of neuronal activity during the gait modification, suggests that the red nucleus, like the motor cortex, contributes to the specification of the appropriate spatiotemporal EMG activation patterns that are required to produce the change in limb trajectory required to step over the obstacle (see also Fig. 11).

View larger version (34K): [in this window] [in a new window]   Fig. 11. Summary of the major modifications of EMG activity and of the major cell types observed during the swing phase of the modified cycle in the red nucleus (RN) and the motor cortex (MC). Data are shown only for cells with a forelimb receptive field for the RN and for PTNs recorded in the forelimb representation of the MC. Left: the sequential nature of the activation of selected EMGs during the swing phase starting with the shoulder retractor, TrM, which showed peak activation early in phase I at an average phase of 0.1 (i.e., 1/10 of the step cycle) and progressing through to the wrist dorsiflexor, EDC, which was activated during phase II at a phase of 0.36. The ClB discharges throughout swing. Middle and right: selected unit discharge patterns recorded from the RN and from the MC, respectively, whose activity temporally covaried with the illustrated muscles during both the lead and trail condition. Thinner lines indicate activity during control cycles and thicker lines the activity during the gait modifications; all activity is synchronized to the onset of the ClB. The traces show the activity only for the step preceding passage of the limb over the obstacle and for the modified step itself. Percentage values for the RN and MC cells indicate the proportion of neurons showing the indicated pattern relative to the total number of neurons showing increased activity during swing. The filled circles, and adjacent values, indicate the average phase of peak activity calculated from all units that showed a similar pattern of temporal covariation to those illustrated; the accompanying horizontal lines, where visible, indicate 1SD. The averaged EMG records were obtained simultaneously with the PTNs illustrated in the same row (filled circles above the EMGs similarly indicate the average phase of peak activity).

Integration with the locomotor rhythm

Although the multiple periods of activity observed in some of these rubral neurons may provide a neural mechanism for ensuring that periods of increased EMG activation at different joints during different phases of the movement are produced at the appropriate time, it does raise the problem as to how the nervous system ensures that each period of increased activity acts only on the appropriate muscles. For example, why does the period of increased activity during stance not produce inappropriate activation of muscles that are activated only during swing and vice versa? There are several possible explanations.

First, during locomotion, this pattern of activity is being superimposed onto the basic locomotor rhythm that itself is capable of determining much of the underlying structure of the step cycle. For example, during swing, microstimulation preferentially activates physiological flexor muscles, whereas during stance, it modulates the activity in physiological extensors (Rho et al. 1999). The rhythmical changes in polarization in these different groups of muscles therefore provides a simple method to allow single neurons to regulate activity in different groups of muscles. A similar mechanism may also function at finer resolution and provide a means for single neurons to differentially regulate the activity of different muscles during the swing phase. For example, our microstimulation study (Rho et al. 1999, Fig. 5) shows that stimulation at a single site in the red nucleus evoked maximal responses in the TrM, Br, and EDC at different times during the swing phase. Maximal responses in the TrM were evoked at the onset of the swing phase while maximal responses in the EDC were evoked at the end of the swing phase. It is therefore possible that the phasic modulations of the level of excitability of spinal interneurons is sufficient to induce some functional organization of the descending signals originating in the red nucleus. We have suggested (Drew 1991; Drew et al. 1996; Prentice and Drew 1997) that such temporal sculpting may also provide a means of focusing the effects of the cortical descending command during locomotion, thus partially compensating for the large spatial distribution of the terminal arbors (Futami et al. 1979). Similar spinal mechanisms would also contribute to the appropriate integration of the signals modifying activity in both contralateral and ipsilateral limbs.

In addition to these passive mechanisms, the final expression of the descending signal will also depend on the integrated activity of the overall population of neurons providing input to the interneuronal circuits controlling the limb, including the large population of cortical and rubral neurons that increase their discharge in a single, discrete part of the modified cycle. This concentrated signal, at specific phases, will likely have the effect of facilitating transmission in spinal circuits to certain muscle groups while decreasing transmission in others. In this respect, the modification of muscle activity produced by the rubral neurons discharging in multiple bursts may be contingent on the activity of those neurons with more discrete activity. This is similar to the suggestion that we have previously made (Prentice and Drew 2001) that the efficacy of descending reticulospinal volleys may be contingent on the activity in other descending pathways, including the corticospinal tract.

Source of modulation

The major afferent input to the red nucleus is from the interpositus nucleus (Massion 1967; Toyama et al. 1968) with weaker inputs from the motor cortex (Padel et al. 1973; Tsukahara et al. 1967, 1968) and from the spinorubral pathway (Vinay and Padel 1990). All of these may serve to modulate the activity of the rubral neurons although the greater strength and faster time course of the excitatory postsynaptic potentials (EPSPs) produced by stimulation of the interpositus nucleus with respect to those observed following stimulation of the motor cortex (Tsukahara et al. 1967) suggests that the cerebellar input is likely to be the major determinant of the pattern of activity observed in the rubral cells. In this respect, it is important to note that recordings of interpositus neuronal activity in both the decerebrate (Orlovsky 1972c) and in the intact cat (Armstrong and Edgley 1984) show clear phasic activity in these cells, particularly during the swing phase of unobstructed locomotion and that cooling of the interpositus nucleus during locomotion leads to a reduction in the amplitude of the flexion phase of locomotion (Udo et al. 1979, 1980), presumably due to disinhibition of the interpositorubral pathway. In a complementary manner, injection of muscimol into the interpositus nucleus also leads to hypoflexion and to some changes in EMG timing patterns during the swing phase of locomotion (unpublished observations) (see also Bracha et al. 1999; Rathelot et al. 1996).

It seems probable that the interpositus nucleus is also the primary source of input to the red nucleus during the voluntary gait modifications, although in this case, there is no direct evidence to support this supposition. Indeed, the only study that has examined interpositus discharge during a voluntary locomotor task reported little change in activity compared with unobstructed treadmill locomotion (Armstrong and Marple-Horvat 1996). However, as the authors emphasized, only a small number of neurons were examined, and direct comparison of the discharge activity during ladder and treadmill locomotion was not possible. Moreover, experiments in which injections of muscimol have been made into the interpositus nucleus showed a clear disruption of the ability of cats to successfully step over obstacles (Rathelot et al. 1996), or to walk from rung to rung of a horizontally positioned ladder (Bracha et al. 1999), suggesting a contribution from the interpositus nucleus in modulating this activity. Last, there is abundant evidence from experiments in cats and primates trained to make forelimb reaching movements to suggest that the interpositus nucleus is critically involved in the execution of such voluntary movements (Burton and Onoda 1978; Fortier et al. 1989; Gibson et al. 1998; Harvey et al. 1979; Martin et al. 2000; Milak et al. 1995; Thach 1970; van Kan et al. 1994).

The role of the motor cortex in determining the level of activity in rubral cells is less clear. Although we have already discussed the strong resemblance in the pattern of activity observed in these two structures during the gait modification, most of the synaptic contacts between cortical cells and rubral neurons are found on the distal dendrites and produce only small and slow EPSPs (Tsukahara et al. 1967). It seems more likely then that cortical input would modulate rubral activity rather than being a prime determinant of the pattern of discharge. This would be in agreement with the fact that the pattern of activity of the rubral neurons, although not the level, is quite similar in the intact and the decerebrate cat during unobstructed treadmill locomotion. Nevertheless, we cannot rule out the possibility that the multiple periods of activity during the swing phase might be the result of the combined influence of the cortical and cerebellar inputs.

The role of the direct spinorubral pathway in determining the modulatory activity of rubral neurons is also unclear. Although peripheral afferent input is transmitted directly, and at short latency, to rubrospinal neurons by this path, Orlovsky (1972a) demonstrated that the modulatory pattern of rubral neurons is mostly lost following cerebellectomy (see, however, Vinay et al. 1993), suggesting that the direct spinorubral pathway is not a major determinant of the pattern of rubral activity during locomotion.

Comparative aspects

The fundamental results obtained during this locomotor study are very similar to those obtained by other research groups examining the role of the red nucleus in the control of reaching movements (see INTRODUCTION for references), suggesting a generally similar mode of control of reaching and voluntary gait modifications. At a general level, the discharge rates observed during the gait modification were generally as high as those observed during reaching movements in both cats and primates, suggesting a similar level of influence over the muscle activity in the two different types of activity. Moreover, in agreement with the results obtained in primates (Gibson et al. 1985a,b; Mewes and Cheney 1991; Miller et al. 1993; Sinkjaer et al. 1995; van Kan and McCurdy 2001) the temporal aspects of the discharge activity in a majority of rubral neurons (56%) was best related to the activity of physiological flexor muscles of the wrist and digits, such as the ECR and EDC. This is also in agreement with a recent study in rats (Jarratt and Hyland 1999) showing that most rubral neurons discharged toward the end of a reaching task corresponding to extension of the wrist. This suggests a parallel in the importance of the red nucleus in controlling the wrist during coordinated reaching and grasping movements and during locomotion. It is probable that a locomotor task in which paw placement needed to be more precisely controlled might show an even greater modification of the neuronal discharge at this phase of the step cycle. However, while these results suggest a powerful contribution to the control of distal musculature, it should nevertheless be emphasized that a similar proportion of neurons might influence proximal musculature, some of them simultaneously. As emphasized by several authors (Belhaj-Saif et al. 1998; Cheney et al. 1991; Gibson et al. 1998; Mewes and Cheney 1991), the rubral projection likely contributes to the control of both proximal and distal musculature and may fire preferentially during coordinated movements of the whole limb that are associated with control of the hand (Gibson et al. 1998; Mewes and Cheney 1994; Miller et al. 1993; van Kan and McCurdy 2001).

One major difference between the results of this study and those of most other studies examining reaching movements is the apparent absence of neurons showing multiple periods of increased activity during reaching. In most studies on reaching, neurons are described as discharging in a single burst of activity during the transport period of the reach, although a recent paper on reaching in the rat (Jarratt and Hyland 1999) shows multiple periods of activity similar to those described in the present paper. Similarly, we are aware of no reports of rubral neurons discharging during ipsilateral limb movements in primates. Whether these are true differences between rubral activity in cats and rats compared with primates or simply due to differences in task or to different analytical techniques is not clear.

Concluding remarks

During locomotion, the rubrospinal system is normally considered to play the relatively simple role of adjusting the overall level of flexor muscle activity (Orlovsky 1972b). As stated in the INTRODUCTION, this view is based primarily on the lack of any major deficits following lesion of the nucleus, partly on the fact that neuronal activity in this structure during locomotion had been previously recorded only during locomotion in the decerebrate animal and partly on the finding that stimulation of this structure has no effect on cycle timing. In contrast, the results presented in this study suggest that the red nucleus plays an important role in the regulation of the locomotor cycle, especially during conditions that require adjustment of the spatiotemporal muscle activity patterns in the limb. In this respect, as detailed in the preceding text, the results agree with the view obtained from neuronal recording studies in animals trained to make reaching movements that the rubrospinal system, together with the corticospinal pathway, provides a major part of the descending signal that is used to modify voluntary movement. Nevertheless, as discussed in the preceding paragraphs, the functions of the two systems are not identical, and, as argued by others (e.g., Martin and Ghez 1988), they likely play complementary roles in the control of movement determined by their afferent inputs and their spinal projection patterns.

In particular, the multiple periods of increased activity during the step over the obstacle suggest that individual rubral neurons, in contrast to motor cortical neurons, contribute to the regulation of muscle activity in both phase I (the transport phase) and phase II (the placement phase). This suggests a contribution to the regulation of intralimb coordination. In addition, the multiple activation of some neurons during swing and stance and time-locked to contralateral and ipsilateral swing suggest that some neurons may also play a role in interlimb coordination at least during locomotion. As such, the general characteristics of the rubrospinal tract fall between the high degree of specificity shown by the corticospinal tract and the more diffuse action of the reticulospinal tract. We suggest that these general characteristics may be related, in part, to the evolutionary lineage of the red nucleus. The red nucleus evolved with the first land animals and has been suggested to be tightly linked to the control of the limbs during locomotion (Keifer and Houk 1994; ten Donkelaar 1988). In these more primitive animals, limb movements were relatively stereotyped, and there was probably little requirement for a descending control system that would allow specific and fractionated control of individual joints. On these general characteristics there would then be superimposed further evolutionary changes related to the different behavioral requirements of different species and the level of development of the corticospinal tract. In the cat, these evolutionary pressures have resulted in two complementary systems involved in the voluntary control of limb movement with both pathways providing a descending signal that can be used to regulate activity in small groups of synergistic muscles, particularly in the distal limb but with the rubrospinal system providing an additional signal that might aid in the coordination of intra- and interlimb coordination.