Corresponding author B. J. Schmidt: Department of Physiology, Room 406, Basic Medical Sciences Bldg, University of Manitoba, 730 William Avenue, Winnipeg, Manitoba, Canada R3E 3J7. Email: email@example.com
This study examines whether propriospinal transmission contributes to descending propagation of the brainstem locomotor command signal in the in vitro neonatal rat spinal cord. Using double bath partitions, synaptic transmission was suppressed in the cervicothoracic region while monitoring locomotor-like activity on lumbar ventral roots evoked by either chemical or electrical stimulation of the brainstem. Locomotor-like activity induced by electrical stimulation was more stable (cycle period coefficient of variation (CV) 11.7 ± 6.1%) than the rhythm induced by chemical stimulation (CV 31.3 ± 6.4%). Ca2+-free bath solution, elevated Mg2+ ion concentration, excitatory amino acid receptor antagonists (AP5 and/or CNQX), and the muscarinic receptor antagonist, atropine, were used in attempts to block synaptic transmission. Each of these manipulations, except muscarinic receptor blockade, was capable of blocking locomotor-like activity induced by brainstem stimulation. However, locomotor-like activity induced by higher intensity electrical stimulation of the brainstem (1.2–5 times threshold) was relatively refractory to synaptic suppression using AP5 and CNQX, and Ca2+-free solution was more effective if combined with high Mg2+ (15 mm) or EGTA. Enhancement of neuronal excitation in the cervicothoracic region, using Mg2+-free bath solution, facilitated brainstem activation of locomotor-like activity in the lumbar cord, consistent with a propriospinal mechanism of locomotor signal propagation. Blockade of brainstem-induced locomotor-like activity was related to the number of cervicothoracic segments exposed to synaptic suppression, being most effective if five or more segments were included. These results provide direct evidence that propriospinal pathways contribute to bulbospinal activation of the locomotor network in the in vitro neonatal rat brainstem–spinal cord preparation, and suggest that a propriospinal system is recruited in parallel with long direct projections that activate the locomotor network.
Development of successful spinal cord repair strategies demands an understanding of the anatomical and physiological substrates of normal functions such as locomotion. Two well-established features of the mammalian locomotor pattern-generating network are that components of this circuitry can remain relatively intact in spinal cord segments caudal to the site of injury and the network can be activated by relatively simple unmodulated excitatory input. Thus, an important and realistic challenge for spinal cord research is to devise methods of re-establishing supraspinal input to locomotor rhythm-generating elements located caudal to spinal cord injury. A major obstacle in achieving this goal is the lack of identification and characterization of descending locomotor command pathways. Another impediment is the limited capacity of axons to regenerate over long distances in the central nervous system. On the other hand, if it could be shown that locomotor command signals are relayed through a polysynaptic network of spinal neurons, restoration of neural transmission for short distances across the site of a spinal cord injury may successfully restore supraspinal control over spinal locomotor circuits.
Descending projections located in the ventrolateral white matter of the spinal cord are essential for the activation of locomotion in acute cat (Steeves & Jordan, 1980, 1984; Noga et al. 1991, 2003) and rat preparations (Stelzner & Cullen, 1991; Iwahara et al. 1991; Magnuson & Trinder, 1997). However, locomotor capacity changes over time and with training such that no specific white matter region appears essential for activation of locomotion in animals with chronic partial spinal cord lesions (for review see Rossignol et al. 1996). Although results of acute cord lesion experiments implicate specific descending brainstem projections, known to travel in specific fasciculi, the contribution of direct versus propriospinal fibres is difficult to deduce from simple lesion experiments alone.
Anatomical and physiological studies have demonstrated a variety of rostrocaudal and caudorostral propriospinal connections in the mammalian spinal cord. Some of these projections extend over many segments, for instance between the cervical and lumbar enlargements (e.g. Barilari & Kuypers, 1969; Miller et al. 1973, 1998; Matsushita et al. 1979; Skinner et al. 1979; Alstermark et al. 1987; Conta & Stelzner, 2004). Others project over just a few segments (e.g. Yezierski et al. 1980; Sherriff & Henderson, 1994; Conta & Stelzner, 2004). Despite abundant anatomical evidence of propriospinal neurons, and studies examining coupling between lumbar and cervical segments during locomotion (e.g. Miller et al. 1975; English, 1989; Ballion et al. 2001; Juvin et al. 2005) only a few studies have addressed whether the brainstem makes use of propriospinal relays to transmit the locomotor command signal. Several reports suggested that a polysynaptic neuronal relay in the dorsolateral aspect of the cat spinal cord, continuous with the pontomedullary locomotor region (PLR), may activate the spinal locomotor network (Kazennikov et al. 1983; Shik, 1983; Yamaguchi, 1986; Kazennikov & Shik, 1988). However, if this dorsal propriospinal system is involved in the descending activation of locomotion, it appears unessential. Acute lesions of the dorsolateral fasciculus (at the C2–3 level) do not interfere with hindlimb stepping induced by electrical stimulation of the PLR or mesencephalic locomotor region (MLR; Noga et al. 1991, 2003). In addition to projections to the dorsal propriospinal system, the PLR, like the MLR, sends inputs to the medial reticular formation which in turn project to the spinal cord (Noga et al. 1991). Thus, long direct reticulospinal projections in the ventrolateral fasciculus have been considered critically important for descending transmission of the brainstem locomotor command.
The present work examines the effect of blocking synaptic transmission in the cervicothoracic region on lumbar locomotor-like output induced by brainstem stimulation. The results provide direct physiological evidence that propriospinal pathways contribute to bulbospinal activation of the locomotor network in the in vitro neonatal rat brainstem–spinal cord preparation. Preliminary results were presented previously in abstract form (Zaporozhets & Schmidt, 2001; Schmidt et al. 2003).
The experimental protocols used in this study were in compliance with the guidelines set by the Canadian Council on Animal Care and the University of Manitoba. Experiments were performed on Sprague-Dawley rats (1–5 days old). The technique for isolation of the spinal cord, extracellular recording, and neurochemical induction of rhythmic activity has been previously described (e.g. Cowley & Schmidt, 1995). In brief, animals were anaesthetized with ether, decerebrated at the mid-collicular level, eviscerated, and placed in a bath chamber containing artificial cerebrospinal fluid (ACSF) composed as follows (mm): NaCl 128, KCl 3.0, NaH2PO4 0.5, CaCl2 1.5, NaHCO3 21, and glucose 30, equilibrated to pH 7.4 with 95% O2–5% CO2. Unless otherwise indicated, the MgSO4 concentration was 1.0 mm. Experiments were conducted at room temperature (ACSF approximately 22°C). The brainstem bath was separated from the spinal cord bath using a partition at C1 in all experiments (see Cowley & Schmidt, 1997). A second partition was located in the cervicothoracic region as indicated in Results.
Ventral root activity was recorded using glass suction electrodes at L2 and L5, which monitor flexor and extensor phases of the step cycle, respectively (Cowley & Schmidt, 1994a; Kiehn & Kjaerulff, 1996). In some experiments, C8 cervical root activity was also monitored. The recordings were band-pass filtered (30–3000 Hz), digitized and captured using AxoScope (v9.0 Axon Instruments) software. AxoScope files were converted to an appropriate binary format for further analysis using special purpose software (developed by the Spinal Cord Research Centre, University of Manitoba).
Neurochemicals were applied from concentrated stock solutions (10 mm) to a static bath that was continuously oxygenated and agitated. All concentrations refer to final bath concentrations which ranged as follows: N-methyl-d-aspartate (NMDA, 2–10 μm), 5-hydroxytryptamine (5-HT, 15–50 μm), dihydrokainate (DHK 200 μm), atropine sulphate (20–80 μm), d-2-amino-5-phosphonovaleric acid (AP5, 20–80 μm, and 6-cyano-7-nitro-quinoxaline-2,3-dione (CNQX, 8–10 μm), bicuculline (BIC, 10–20 μm), and ethylene glycol-bis(2-aminoethyl-ether)-N,N,N′,N′-tetraacetic acid (EGTA, 3–5 mm). A period of 25–60 min was allowed to elapse after neurochemical manipulation before testing the effect of brainstem stimulation. The same preparation was used in some instances for more than one test protocol. In these cases all bath compartments were subjected to a prolonged washout (at least 1 h), using normal ACSF, between test protocols.
Locomotion was induced by either neurochemical or electrical stimulation of the brainstem. Neurochemical stimulation involved applying, in varying combinations, NMDA, 5-HT, the inhibitory receptor antagonist BIC, and/or Mg2+ ion removal to further enhance neuronal excitation (Smith et al. 1988; Atsuta et al. 1991). The brainstem was separated from the spinal cord bath by a partition at the cervicomedullary level (C1). In other experiments, electrical stimulation of the brainstem was used to induce locomotor-like rhythms according to methods previously described (Zaporozhets et al. 2004). In brief, an ACSF-filled glass electrode, with a tip diameter of 200–300 μm, was placed in contact with the ventral surface of the brainstem. Monophasic rectangular current pulses (4–20 ms, 0.5–10 mA, 0.8–2.0 Hz) were delivered using bipolar stimulation.
In order to assess the stability of rhythmic activity induced by neurochemical and electrical stimulation of the brainstem the coefficient of variation (CV), determined as the standard deviation of successive cycle periods in a given episode of rhythmic activity divided by the mean cycle period, allowed analysis of the regularity of the interburst intervals during the episode. Mean CVs for grouped data are presented with the standard deviation.
Circular statistics were used to examine left–right, flexion–extension (L2–L5), and cervical–lumbar phase relationships. Phase values were displayed graphically as data points on a polar plot (Zar, 1974; Batschelet, 1981). The cycle period was measured for each cycle in a given episode of rhythmic discharge on one root (A) of the pair under analysis. The latency to the onset of discharge on the other root (B) was then determined for each cycle. A phase value for each step cycle of the episode was calculated using the formula:
The mean phase value for each episode was displayed as a vector with length r (Zar, 1974). The length of the vector ranged from 0 to 1 and is a measure of the concentration of phase lags around the mean phase value for the episode. Rayleigh's circular statistical test (Zar, 1974; Batschelet, 1981) was applied to determine whether the phase lags in a given episode were significantly dispersed, indicating no phase relationship, or concentrated, suggesting coupling of discharge in the two roots under comparison. If r was greater than the critical Rayleigh's value (cR) for a given P value then the relationship was considered phase-related.
The regularity of locomotor-like rhythmic discharge evoked by electrical stimulation of the brainstem was compared to the regularity of rhythms induced by chemical stimulation of the brainstem and by whole cord chemical stimulation. Cycle periods were shorter (2.3 ± 1.2 s) and more regular (mean coefficient of variation (CV) = 11.7 ± 6.1%, n= 6 preparations) during locomotor-like activity induced by electrical stimulation of the brainstem compared to measurements obtained during chemical activation of the brainstem (mean period 7.4 ± 4.4 s; mean CV = 31.3 ± 6.4%, n= 6 preparations). Whole cord (C1 to conus) application of neurochemicals (15–40 μm 5-HT combined with 2 μm NMDA and 200 μm DHK) resulted in a mean cycle duration of 5.4 ± 1.2 s; the CV was 10.6 ± 7.1% (n= 6) which is similar to the value obtained during electrical stimulation of the brainstem. Thus, CV analysis supported our impression that locomotor-like patterns elicited by neurochemical stimulation of the brainstem were less regular compared to whole-cord application of neurochemicals or electrical stimulation of the brainstem. In addition, as will be shown, locomotor-like activity evoked by neurochemical stimulation of the brainstem was more sensitive to synaptic blockade in the cervicothoracic segments compared to rhythmic activity induced by electrical stimulation of the brainstem. Because chemical and electrical stimulation methods have different advantages and disadvantages (see Discussion), both types of stimulation were used, in a complementary fashion.
Neurochemical stimulation of the brainstem
Locomotor-like rhythm, evoked by application of neurochemicals to the brainstem bath, was characterized by phasic alternation of ipsilateral L2 (predominantly flexor) and L5 (predominantly extensor) ventral root discharge and left–right alternation of discharge at the same segmental level (Fig. 1B). In addition to the bath partition at the cervicomedullary junction, a second partition was established at either a cervical or thoracic level. Ca2+ ions were removed from the cervical (or cervicothoracic) bath in an attempt to abolish synaptic transmission in these segments. Suppression of synaptic transmission in the cervicothoracic region completely terminated lumbar rhythmic activity in 13/27 preparations (Fig. 1C), and partially blocked rhythmic activity in 6/27 preparations. ‘Partially blocked’ refers to preparations still showing some phasic ventral root activity, but the activity in the presence of synaptic suppression was clearly perturbed compared with the control activity. Perturbations included the following, either alone or (more usually) in combination: phasic discharge was no longer evident on some ventral roots or only briefly observed, burst amplitude was reduced, or the rhythm became irregular or characterized by non-locomotor-like prolonged bursts. Removal of Ca2+ ions failed to block rhythmic activity in 8/27 preparations. This was most probably due to exposing an insufficient number of segments to Ca2+-free ACSF as discussed below, although the possibility of incomplete Ca2+ ion removal is not excluded (see Discussion). Even in those preparations where Ca2+-free ACSF abolished all phasic discharge, tonic lumbar ventral root activity persisted (compare Fig. 1C and A, also see Fig. 3C). The persistent discharge may have been due to bulbospinal transmission mediated by direct axonal projections (unaffected by the Ca2+ removal) or incomplete suppression of propriospinal synaptic transmission. The suppressive effect of Ca2+-free ACSF on locomotor rhythm production was reversible upon restoration of normal ACSF in the cervicothoracic bath compartment (Fig. 1D).
In 12 preparations, chemical stimulation of the brainstem elicited either erratic phasic lumbar root discharge or tonic discharge only (Fig. 2A). When neuronal excitation in the cervicothoracic region of the spinal cord was enhanced by removing Mg2+ ions, thereby abolishing the voltage-dependent blockade of NMDA receptors, rhythmic locomotor-like patterns were observed in five of these preparations (Fig. 2B). Because neurochemical excitation of the cervicothoracic region alone, for instance using NMDA and 5-HT (Cowley & Schmidt, 1997), can induce phasic discharge in the lumbar region, it is possible that removal of Mg2+ ions in the cervicothoracic bath may have promoted lumbar rhythmic activity independent of any requirement for activation of brainstem sites. However, the results shown in Fig. 3 indicate that the rhythmic discharge recorded on lumbar ventral roots, in the presence of Mg2+-free ACSF in the cervicothoracic region, did indeed require brainstem activation. Mg2+-free ACSF alone provoked chaotic non-rhythmic bursts in the lumbar region (n= 4, Fig. 3A). Neurochemical stimulation of the brainstem was required to produce locomotor-like activity (n= 3, Fig. 3B). In one of these four preparations neurochemical stimulation of the brainstem failed to produce locomotor-like activity. Suppression of synaptic transmission, using Ca2+-free ACSF and normal Mg2+ concentration in the cervicothoracic region, abolished the rhythmic component elicited by brainstem stimulation (Fig. 3C).
The NMDA receptor antagonist AP5 (80 μm) applied in conjunction with the kainate/quisqualate receptor antagonist CNQX (8 μm) to the cervicothoracic region blocked lumbar rhythmic activity completely in 6/10 preparations (Fig. 4A) and partially in 2/10 preparations, when locomotor activity was elicited using chemical stimulation of the brainstem. AP5 alone blocked rhythmic activity completely in 6/8 preparations and partially in 2/8 preparations. However, application of CNQX (8–10 μm) alone to the cervicothoracic regions failed to block lumbar locomotor-like activity in 9/9 preparations. In contrast to the effect of removing Mg2+ ions noted above, increased Mg2+ ion concentration (10 mm) suppressed transmission of the brainstem locomotor command signal (Fig. 4B). Although cholinergic propriospinal neurons have been implicated in the control of locomotion (Jordan & Schmidt, 2002), and 20–50 μm atropine blocks cholinergic receptor activation in this preparation (Smith et al. 1988), application of atropine to the cervicothoracic regions in concentrations up to 80 μm failed to abolish lumbar rhythmic activity evoked by chemical stimulation of the brainstem (n= 7/7).
The ability of synaptic suppression (using Ca2+ removal, AP5, and/or elevation of [Mg2+] in the cervicothoracic region) to block locomotion induced by chemical stimulation of the brainstem was related to the number of cervicothoracic segments manipulated. When four segments or less were included, locomotor-related transmission was suppressed in only 3/12 preparations, and the observed suppression was only partial (Fig. 5A). On the other hand, synaptic inhibition of five or more segments was effective in 31/33 preparations (complete blockade in 23, partial blockade in 8, no blockade in 2). Suppression of synaptic activity in rostrocaudal lengths of spinal cord that were 5 mm or greater was usually effective in blocking bulbospinal transmission (28/33 preparations) whereas lengths shorter than 5 mm were usually ineffective (Fig. 5B). In the neonatal rat preparation a single spinal cord segment is approximately 1 mm in length, although there is variability depending on the rostrocaudal region of the cord and age of the neonatal animal.
Electrical stimulation of the brainstem
In 11/11 preparations the locomotor-like activity evoked using threshold levels of electrical stimulation of the brainstem was abolished after replacing cervicothoracic ACSF with Ca2+-free ACSF for 30–45 min (Fig. 6A). However, in 10/11 preparations Ca2+-free ACSF failed to abolish locomotor-like activity when the intensity of brainstem electrical stimulation was increased slightly above the threshold (T) level used to induce rhythmic activity in normal ACSF (Fig. 6A). The exception to this observation is shown in Fig. 8 where the lumbar rhythmic activity remained abolished in Ca2+-free ACSF during stimulation at 4T. We previously demonstrated that activation of spinal locomotor networks is not the result of electrotonic spread of the brainstem stimulus, even if high stimulus intensity (4T) is used (Zaporozhets et al. 2004). Therefore, in the present experiments the effectiveness of higher brainstem stimulus intensity at inducing locomotor-like activity was not related to electrotonic spread of the brainstem stimulus through the Ca2+-free compartment.
The failure of Ca2+-free ACSF to abolish lumbar ventral root locomotor-like activity evoked by slightly increased levels of electrical stimulation raised the question of inadequate removal of Ca2+ ions from the extracellular environment of the spinal cord. Therefore, to enhance Ca2+ ion removal, we added the Ca2+ chelator EGTA (3–5 mm, n= 5). In all five preparations, the combined use of Ca2+-free ACSF and EGTA blocked locomotor activity evoked by electrical stimulation, even when stimulated at 4T (Fig. 7A).
In four other preparations Ca2+-free ACSF was combined with high concentrations of Mg2+ (15 mm) to further suppress synaptic transmission (Czeh & Somjen, 1989). Lumbar ventral root locomotor-like activity in response to electrical stimulation of the brainstem remained abolished in each preparation, even when the brainstem stimulus was increased to 4T (Fig. 7B).
In 8/8 preparations AP5 (80 μm) abolished locomotor-like activity induced by threshold levels of brainstem electrical stimulation (Fig. 6B). Similar to the effect of Ca2+-free ACSF alone, AP5 failed to abolish rhythmic activity if the brainstem electrical stimulus was increased slightly (1.2–2T). CNQX abolished locomotor-like activity evoked by threshold electrical stimulation of the brainstem in 8/16 preparations; however, at higher levels of brainstem stimulation (1.2–5T) locomotor-like activity was elicited in all 16 preparations. Three preparations were exposed to combined AP5 and CNQX. Locomotor-like activity could still be elicited but supra-threshold levels of brainstem electrical stimulation (1.2–1.5T) were required.
In three preparations cervical ventral root activity was also monitored (Figs 7A and 8). In all three preparations, both the rhythmic and tonic discharge of cervical roots was completely abolished by Ca2+-free ACSF in the cervicothoracic compartment (C1–T8). In addition to suppressing the excitation of interneurons, Ca2+ removal in the cervical region reduces last-order synaptic drive to cervical motoneurons. This may account for the greater suppression of cervical than lumbar root discharge, in response to Ca2+ removal, under these experimental conditions. In the lumbar region the excitation of motoneurons, either directly from long bulbospinal projections or indirectly via long bulbospinal axons reaching lumbar neurons that project to motoneurons, is not exposed to synaptic suppression and therefore some tonic discharge persists during brainstem stimulation (Figs 1C, 3C, 7A and B, and 8). Rhythmic activity persisted in the lumbar region in 2/3 preparations despite complete suppression of cervical root activity (Fig. 7A). Thus, the absence of rhythmic activity in the cervicothoracic cord is not in itself sufficient to abolish lumbar locomotor-like activity induced by brainstem stimulation.
The main observation of this study is that, during brainstem stimulation, suppression of synaptic transmission in the cervicothoracic cord abolishes locomotor-like activity in the lumbar region. This finding provides direct evidence that propriospinal relays contribute to the descending activation of the locomotor network. Moreover, the results suggest that the propriospinal system is essential in this role, at least in the neonatal rat preparation.
The possibility that propriospinal neurons propagate the mammalian locomotor command signal has received relatively little experimental attention in the past. Technically it is difficult to block synaptic activity in classic in vivo preparations, such as the decerebrate cat or rabbit, without also interfering with direct bulbospinal axonal conduction. In vitro rodent spinal cord preparations offer the opportunity to selectively inhibit synaptic activity in the spinal cord. In the present series of experiments, we used chemical (Smith et al. 1988; Atsuta et al. 1991) and electrical (Atsuta et al. 1988; Zaporozhets et al. 2004) stimulation of the brainstem to activate locomotion. Chemical stimulation, which activates the neuronal soma but not axons of passage in the brainstem, did produce locomotor-like activity. However, a stable rhythm was more difficult to establish compared to rhythms elicited by direct application of neuro-active substances to the spinal cord. Therefore, we also employed electrical stimulation of the brainstem to produce more robust locomotor-like activity, comparable to rhythms induced by whole cord application of neurochemicals. However the disadvantage of electrical stimulation is that the higher levels of descending excitation associated with this method may be more likely to recruit redundant systems and obscure attempts to determine key propriospinal pathways and their transmitter content. Therefore, both chemical and electrical stimulation methods were used in the present study.
The specific locomotor-related bulbospinal projections activated by chemical and electrical stimulation in this study are unknown. Moreover, the various combinations and concentrations of neurochemicals and different electrical stimulus intensities employed probably activated a range of neural elements in the brainstem. Despite these variables, blockade of cervicothoracic synaptic activity consistently suppressed lumbar locomotor output, supporting a role for propriospinal pathways. Candidate locomotor-related cell populations or axons in the brainstem, activated during these experiments, include reticulospinal cells originating in the pontomedullary medial reticular formation (MedRF) that project via the ventrolateral fasciculus of the spinal cord, as well as MedRF inputs from the mesencephalic and pontomedullary locomotor regions (Noga et al. 1991, 2003). Our stimulation methods are likely to include medullary para-raphe (parapyramidal) region serotonergic neurons, the activation of which was recently shown to induce locomotor-like activity in the neonatal rat spinal cord (Liu & Jordan, 2005).
Blockade of descending locomotor activation was more difficult, although not impossible, when electrical rather than chemical stimulation of the brainstem was used to induce locomotion. Several factors may account for this observation. Electrical stimulation, unlike chemical stimulation, recruits axons of passage in addition to exciting neuronal cell bodies in the brainstem, presumably resulting in more intense descending neural excitation. Some descending projections may consist of long direct fibres that are unaffected by synaptic blockade in the cervicothoracic region. If electrical stimulation is more effective than chemical stimulation of the brainstem at recruiting long direct projections, then suppression of synaptic relays in the cervicothoracic region would have less influence on the propagation of bulbospinal activity induced by electrical compared to chemical stimulation. In addition, Antonino-Green et al. (2002) suggested that electrical stimulation may antidromically excite an ascending spinoreticular projection thereby contributing to locomotor network activation. Chemical stimulation of the brainstem would not directly activate such a system.
Extracellular Ca2+ ions are required for transmitter release evoked by depolarization of the presynaptic terminal (Katz & Miledi, 1970). Reducing extracellular Ca2+ ion concentration decreases synaptic transmission in the in vitro rodent spinal cord (Kuno & Takahashi, 1986; Czeh & Somjen, 1989; Lev-Tov & Pinco, 1992; Li & Burke, 2001); in fact, Ca2+-free ACSF completely abolishes mono- and polysynaptic segmental reflex transmission (Tresch & Kiehn, 2000) and in the present study no evidence of cervical ventral root discharge was seen after exposing cervical segments to Ca2+-free ACSF (Figs 7A and 8). However, perfusion with nominally Ca2+-free ACSF does not guarantee that the concentration of extracellular Ca2+ in the spinal cord is sufficiently low to block synaptic transmission (Kuwana et al. 1998; Piccolino et al. 1998). Moreover, low Ca2+ ion concentration alone may actually increase neuronal excitability by reducing the surface-charge screening effect that Ca2+ ions exert on the negatively charged external surface of cell membranes (for review see Piccolino et al. 1998). Lower concentrations of Ca2+, sufficient to reverse the normal Ca2+ ion concentration gradient between extracellular and intracellular (nerve terminal) compartments, can be achieved using a chelator (Piccolino et al. 1998). Thus, in the present study, addition of EGTA enabled blockade of locomotor-like activity elicited by supra-threshold levels of electrical stimulation, whereas Ca2+-free ACSF alone did not. Alternatively, elevation of extracellular Mg2+ ion concentration was used to suppress neurotransmitter release (Katz & Miledi, 1963; Kuno & Takahashi, 1986; Czeh & Somjen, 1989; Lev-Tov & Pinco, 1992; Kuwana et al. 1998). Mg2+ ions also restore the membrane surface-charge screening effect that is lost when Ca2+ ions are removed from ACSF, thereby preventing increased membrane excitability in low Ca2+ solutions (for review see Bernath, 1992). In the present series of experiments, increased Mg2+ ion concentration, in combination with Ca2+-free ACSF, effectively blocked the descending locomotor signal in response to electrical stimulation of the brainstem.
In some preparations, AP5 application to the cervicothoracic region blocked locomotor-like activity evoked by chemical or threshold electrical stimulation of the brainstem, and CNQX also blocked locomotor-like activity induced by threshold levels of electrical stimulation. These results are consistent with a contribution from glutamatergic receptors in propriospinal transmission of the locomotor signal. However, in some preparations of either type (i.e. chemically induced or electrically stimulated) neither AP5 nor CNQX, alone or in combination, blocked brainstem-activated locomotion. This was uniformly the case when supra-threshold levels of electrical stimulation were used. The variable degree to which long direct brainstem projections, capable of reaching the lumbar cord, are recruited may account for the apparent refractoriness of the propriospinal system to excitatory amino acid neurotransmitter blockade under certain conditions. Non-glutamatergic-dependent excitatory mechanisms may also account for persistent propriospinal transmission in the presence of AP5 and/or CNQX, given that neither NMDA nor kainate/quisqualate receptors are essential for locomotor network operation in the neonatal rat (Beato et al. 1997; Cowley et al. 2005).
Propriospinal transmission of the locomotor signal does not exclude participation of long direct projections from the brainstem. Brainstem monoaminergic projections, even single fibres, give off multiple axon collaterals throughout the rostrocaudal extent of the spinal cord (Huisman et al. 1984; Skagerberg & Bjorklund, 1985; Bowker & Abbott, 1990). Thus, brainstem neurons, such as the reticulospinal system, may transmit the descending locomotor command signal in parallel via (a) long direct projections to caudal regions of the cord and (b) more rostral inputs that target the descending propriospinal system. The rostral inputs may derive from collaterals of long projections or from separate shorter bulbospinal projections.
Although the motor output of the cervical cord is not critical for biped locomotion, rostrocaudal propagation of the brainstem locomotor command signal in both quadrupeds and bipeds alike may nevertheless depend on propriospinal relays in the cervicothoracic region. In the present series of experiments, locomotor-like rhythms could still be elicited in the lumbar region, after synaptic suppression in the cervicothoracic cord, despite the absence of rhythmic cervical ventral root output (Fig. 7A). This result suggests that locomotor command signal propagation through the cervical region may be independent of cervical locomotor circuitry. Alternatively, the absence of cervical motor output in these experiments could also be due to relatively greater sensitivity of cervical motoneuron activation to Ca2+-free ACSF compared to interneurons and/or propriospinal neurons. In this case, cervical locomotor network interneurons may still be rhythmically active despite the absence of cervical ventral root discharge. Intraspinal recordings of interneuron activity under these conditions would clarify this issue. Indeed, the fact that subsequent measures to enhance Ca2+ ion removal (addition of EGTA) in the cervical region successfully abolished lumbar rhythmic activity is compatible with the idea that synaptic transmission was incompletely suppressed using Ca2+-free ACSF alone, despite the absence of cervical ventral root output.
The present results suggest propriospinal relays may be essential, at least under some experimental conditions, for descending activation of locomotion in the in vitro neonatal rat preparation. Propriospinal compensatory mechanisms have been shown to contribute to the recovery of locomotor function after spinal transection in the larval lamprey (McClellan, 1994; Rouse & McClellan, 1997) and chick embryo (Sholomenko & Delaney, 1998). However, it remains to be determined whether restoration of propriospinal connections in mammalian preparations is sufficient to re-establish locomotion in the absence of long direct brainstem projections. Bareyre et al. (2004) showed that corticospinal tract axons lesioned in the cervical cord of adult rats spontaneously sprout and make new contacts with intact cervical propriospinal neurons that in turn project to the lumbar cord. The present study showed that enhancement of neuron excitation in the cervicothoracic region facilitated brainstem activation of lumbar locomotor-like activity. In combination, these observations suggest that propriospinal systems are an attractive target for regeneration and functional recovery strategies aimed at restoration of locomotor function after spinal cord injury.
This work was supported by the Canadian Institutes of Health Research and NIH: NS40903-02. K.C.C. was supported by the Will-to-Win Scholar Fund.