A fundamental question in vertebrate locomotion is whether distinct spinal networks exist that are capable of generating rhythmic output for each group of muscle synergists. In many vertebrates including the lamprey, it has been claimed that burst activity depends on reciprocal inhibition between antagonists. This question was addressed in the isolated lamprey spinal cord in which the left and right sides of each myotome display rhythmic alternating activity. We sectioned the spinal cord along the midline and tested whether rhythmic motor activity could be induced in the hemicord with bath-applied d-glutamate or N-methyl-d-aspartate (NMDA) as in the intact spinal cord or by brief trains of electrical stimuli. Fast rhythmic bursting (2–12 Hz), coordinated across ventral roots, was observed with all three methods. Furthermore, to diminish gradually the crossed glycinergic inhibition, a progressive surgical lesioning of axons crossing the midline was implemented. This resulted in a gradual increase in burst frequency, linking firmly the fast hemicord rhythm [6.6 ± 1.7 (SD) Hz] to fictive swimming in the intact cord (2.4 ± 0.7 Hz). Ipsilateral glycinergic inhibition was not required for the hemicord burst pattern generation, suggesting that an interaction between excitatory glutamatergic neurons suffices to produce the unilateral burst pattern. In NMDA, burst activity at a much lower rate (0.1–0.4 Hz) was also encountered, which required the voltage-dependent properties of NMDA receptors in contrast to the fast rhythm. Swimming is thus produced by pairs of unilateral burst generating networks with reciprocal inhibitory connections that not only ensure left/right alternation but also downregulate frequency.
The rhythmic movements of the body and limbs that propel vertebrates through space, be it swimming, walking, or flying, are mainly generated by specialized circuits confined to the spinal cord. These networks, called central pattern generators (CPGs) for locomotion, normally integrate sensory feedback about body position and the environment together with supra-spinal motor commands (see Grillner 1985). This allows, for example, a squirrel to run along the branches of a tree with unfailing precision. Nonetheless, spinal CPGs are capable of producing stereotyped locomotor output also in complete isolation. This general control structure has emerged from experiments on a wide variety of animals including lampreys, tadpoles, turtles, chicks, rats, cats (see Grillner 1981; Kiehn et al. 1997; Stein and Smith 1997), and recently, also humans (Dimitrijevic et al. 1998).
Locomotion is produced by a complex motor pattern with rhythmic and alternating contractions of antagonistic muscles: in the lamprey and many aquatic vertebrates between left and right sides of the body and in terrestrial and airborne species also between flexors and extensors at each joint of the leg or wing. The coordination between different muscle groups is to some degree flexible (see Stein and Smith 1997) and can even be recombined (compare forward and backward locomotion). This ability of the spinal networks to flexibly reshape the motor pattern has led to the proposal of a modular organization of the CPGs for locomotion (Grillner 1981). Each group of muscle synergists at a joint would be controlled by a dedicated module, the unit burst generator (UBG), autonomously capable of rhythmic output. UBGs would then be dynamically interconnected by reciprocal inhibition or excitation depending on whether alternation, co-activation, or more complex patterns are required. In the lamprey (Buchanan 1999) and other vertebrates, it has instead been suggested that locomotor burst generation depends crucially on reciprocal inhibition between antagonistic centers.
The different flexor and extensor motor nuclei in the tetrapod spinal cord are not sufficiently separated to allow an experimental functional isolation of potential unit burst generators (see Kiehn and Kjaerulff 1998). The situation is somewhat more favorable in the mudpuppy in which the main groups of flexors and extensors in the forelimb area can be separated into two centers (Cheng et al. 1998). Other evidence supporting UBGs mainly comes from experiments in which rhythmic ventral root activity was observed in the presence of antagonists of inhibitory neurotransmission (neonatal rat: Cowley and Schmidt 1995; Kremer and Lev-Tov 1997) or in which unilateral bursting occurred in one muscle group without concomitant activity in the antagonists (cat: Grillner and Zangger 1979; turtle: Stein et al. 1995). In the frog embryo, rhythmic ventral root discharge was observed after longitudinal hemisection of the spinal cord in the fast pattern of “single-spike alternation” of tadpole swimming (Soffe 1989).
In the lamprey, it was shown that the spinal cord maintains its rhythm-generating capacity during a blockade of glycinergic transmission with strychnine with either a fast or a very slow bursting being reported (Alford and Williams 1989; Aoki et al. 2001; Cohen and Harris-Warrick 1984; Hagevik and McClellan 1994). Although this supports the existence of unilateral rhythmic networks not requiring reciprocal inhibition for their operation, it leaves open the question of whether these two rhythms are related to the “normal” operation of the CPG for swimming. Moreover, direct tests of the UBG hypothesis, with a longitudinal midline section separating the two sides, led Buchanan (1999) to the opposite conclusion that hemicords are unable to generate a rhythmic burst pattern.
We have further investigated this and are now able to show that the lamprey hemicord can express two clear patterns of locomotor rhythmicity, a fast and a slow rhythm. We demonstrate that the fast rhythm, evoked under a broad spectrum of activating protocols, is directly linked to the operation of the network in the intact cord during swimming. Crossed connections are thus not required for the generation of the swimming rhythm. Furthermore the unilateral motor pattern is independent of ipsilateral glycinergic inhibition and presumably due to a burst-generating network consisting of interacting glutamatergic interneurons.
A total of 55 adult lampreys (Lampetra fluviatilis) were anesthetized by immersion in tricaine methane sulfonate (MS-222; Sigma, St. Louis, MO; 200 mg/l) and decapitated caudal to the gills. The spinal cord preparations (normally between 8 and 13 segments) were dissected from the region between the gills and the first dorsal fin together with the dorsal half of the notochord as a mechanical support. After pinning down these preparations in a silicone elastomer (Sylgard)-lined chamber, the meninges were peeled off from the dorsal surface of the cord. During the dissection and thereafter, preparations were perfused with cooled physiological solution (5–10°C). The Ringer solution [containing (in mM) 138 NaCl, 2.1 KCl, 1.8 CaCl2, 1.2 MgCl2, 4 glucose, 2 HEPES, and 0.5 l-glutamine] was bubbled with O2 and adjusted to pH 7.4 with NaOH.
Glass pipette suction electrodes were gently placed on the ventral roots to record efferent locomotor activity. Signals were sent to a differential AC amplifier (Model 1700, A-M Systems, Everett, WA) and band-pass filtered between 100 and 500 Hz. The output from the amplifier was acquired at a sampling rate of 2 kHz per channel on a PC equipped with an A/D converter (Digidata 1320A) and running Clampex 8 software (both from Axon Instruments, Foster City, CA). In some experiments, the hemispinal cord was stimulated by a single train of 1–30 pulses (2-ms pulse duration, 30-Hz frequency) delivered to either end of the spinal cord by a large suction electrode applied on the dorsal surface. This was sufficient to elicit a long-lasting bout of motor activity.
With the aid of a stereoscopic microscope and trans-illumination oriented at an optimal angle of incidence, relevant features of the spinal cord could be viewed and used as a guide during the lesions. The midline appeared as a distinct dark line on a lighter background or as a bright line on a gray background depending on the lighting. The ventromedial column and cell layer were also readily identified. Hemispinal cords were obtained either by sectioning sagittally along the midline (n = 53) or by following a slightly lateral path, lying halfway between the midline and the contralateral cell layer (n = 34). The two types of preparation were equal with regard to the fast locomotor burst activity (2–10 Hz), whereas there was a tendency for those with the lateral section to display an earlier recovery of the N-methyl-d-aspartate (NMDA)-induced slow burst rate (0.1–0.4 Hz) after the lesion. Mid- and para-sagittal preparations are dealt with together in the following text. Sectioning was performed either with the tip of a fine hypodermic needle (0.4 mm OD; 20 of 55 animals) or using a Micro Feather ophthalmic scalpel (Feather Safety Razor, Osaka, Japan) with a tip angle of 15°, mounted on a micromanipulator. Both instruments were lowered into the spinal cord along the line of sectioning (midline or lateral cut). In later experiments, the scalpel was employed due to its bilateral symmetry, which permitted both left and right hemicords to be used.
Efferent locomotor activity was induced both in intact and hemispinal cords by perfusing either 0.5–1 mM d-glutamate (Sigma) or 75–150 μM NMDA (Tocris, Bristol, UK). Both agonists have routinely been used to elicit fictive swimming in the lamprey spinal cord (Cohen and Wallén 1980; Grillner et al. 1981). To study the contribution of the voltage dependence of NMDA receptors to rhythm generation, magnesium-free Ringer was used. This was obtained by replacing MgCl2 with NaCl. In the absence of Mg2+, lower concentrations of d-glutamate and NMDA were used to elicit locomotor activity (Brodin and Grillner 1986). Glycinergic inhibition from ipsilaterally projecting interneurons was blocked by perfusing 1 μM strychnine hemisulfate salt (Sigma) for 1 h (Buchanan and Grillner 1988; McPherson et al. 1994).
Ventral root recordings, as a rule 3 min in length (except when activating the hemicords electrically), were analyzed using custom scripts run within Axograph 4.6 software (Axon Instruments) on a Macintosh G4 computer (Apple Computer, Cupertino, CA). The main parameters characterizing rhythmic activity were extracted from the autocorrelogram (burst frequency and rhythmic quality) and the cross-correlogram (coordination between ipsi- or contralateral hemisegments) of ventral root recordings. Before calculating these functions, the raw recordings were digitally high-pass filtered (smooth cutoff at 40 Hz) to remove any DC offset introduced after the hardware filter. Then for each recording, the amplitude of baseline noise was automatically estimated, and a multiple of this value was selected as a threshold below which all samples were set to zero. This procedure sets baseline noise to zero and is useful in preventing that fluctuations in noise amplitude, which may occur during a long experiment, influence the estimation of rhythmic quality. Finally, recordings were rectified and correlograms calculated (normalized between 0 and 1). Figure 1 shows three examples of ventral root recording of decreasing rhythmic quality and their corresponding autocorrelograms. Cycle period (the inverse of burst frequency) was taken as the delay of the second peak in the autocorrelogram (Fig. 1). Rhythmic quality was described using a numerical value, referred to as the coefficient of rhythmicity (Cr), which ranges between 0 and 1. This was defined as Cr = (α-β)/(α+β) where α and β are the height of the second peak and the first trough in the autocorrelogram, respectively (Fig. 1). The higher the coefficient, the more rhythmic the activity at the ventral root, as can be seen by comparing the three samples shown in Fig. 1, with their corresponding Cr. For a discussion of similar techniques, see Buchanan (1999). Only records having a Cr ≥ 0.01 were considered rhythmic. In all these cases, the raw recordings were inspected by eye, and rhythmicity at the frequency predicted by the autocorrelogram was confirmed.
Data are expressed as means ± SD and Student's paired t-test was used for statistical comparisons. Unless stated otherwise, the value of rhythmic quality associated to each preparation is the highest observed in the course of the experiment, calculated on 3 min of continuous recording (see preceding text). The reported duration of the locomotor bouts following electrical stimulation is also the longest obtained from each individual hemicord.
Much of our current knowledge of the factors that contribute to the generation of swimming in the lamprey has come from experiments in which fictive locomotion was induced in the isolated spinal cord (see Grillner et al. 2000). Most studies have used either bath-applied d-glutamate or NMDA to evoke fictive locomotion (Cohen and Wallén 1980; Grillner et al. 1981; Wallén and Williams 1984), the difference being that d-glutamate activates both AMPA and NMDA receptors to a similar degree (Zhang et al. 1996). Both agonists were therefore tested on the hemispinal cord preparations.
Hemispinal cords express a fast rhythm in d-glutamate
Fictive swimming was induced in nine intact spinal cord preparations by perfusion of 0.5 to 1 mM d-glutamate. The quality of rhythmic activity in the intact cords, as expressed by the coefficient of rhythmicity, was always high (average Cr = 0.85 ± 0.06; range = 0.75–0.94; see METHODS). Hemisectioning was performed when the rhythm had reached a steady state to have a stable control value. In a further eight spinal cord pieces, d-glutamate (0.5–1 mM) was perfused only after hemisection. Altogether 28 hemicords were studied.
Ventral root activity could be evoked by d-glutamate immediately after the lesion, except in a few cases in which the hemicord remained inactive for up to 1 h. Rostral and caudal ventral roots were rhythmically bursting (Fig. 2A1) and coordinated with each other (Fig. 2A1, bottom). All hemicord preparations (n = 28, from 17 animals) expressed fast rhythmic bursting in d-glutamate, but the Cr was lower than in the intact spinal cord (average Cr = 0.27 ± 0.13; range = 0.08–0.56). This burst activity was readily visible in the raw recordings and was confirmed by the autocorrelation analysis (Fig. 2A2). In all five preparations where it was tested, bursting was coordinated along the hemispinal cord with a small lag (-3 to +2% of the cycle period, per segment) observed in the cross-correlogram (Fig. 2A3, inset).
Figure 2B shows the burst rates observed before and after the hemisection in different preparations. We subdivided the frequency axis in consecutive intervals. For each interval, the number of intact and lesioned preparations is represented. The histograms (Fig. 2B) show that in the intact spinal cords the burst frequency ranged between 1 and 3 Hz (▦), whereas in the hemicords it was faster and ranged between 3 and 10 Hz (□). In the preparations in which the frequency was measured before and after the lesion, a significant increase was observed (n = 15; P < 0.0001); the frequency after hemisection being on average 270 ± 81% of control.
To explore if the frequency could still be modulated in the hemicord preparations, we doubled the concentration of d-glutamate from 0.125 up to 2 mM in steps (Fig. 2C). Within the range 0.25–2 mM, a doubling of d-glutamate concentration always induced an increase in burst frequency (n = 6; average increase 34 ± 18%). Little or no activity was present at 0.125 mM and, in some cases, even at somewhat higher concentrations, whereas at ≥2 mM, the hemicords tended to become quiet after a short time.
To determine the minimum substrate necessary to generate this unilateral rhythm, two hemicords (10 segments long) were transected into pieces of different length. Rhythmic activity was still expressed by a 2.5-hemisegment-long piece (Cr = 0.26), which was the shortest tested and a 3-hemisegment one displayed synchronized bursting across two ventral roots (Fig. 2D). The burst frequency of these short hemicords did not differ from that expressed by the original hemicord preparations before the transections (Fig. 2D).
Hemispinal cords express a fast and a slow rhythm in NMDA
A similar approach to the experiments described above was used with NMDA. In part of the intact spinal cord pieces (32 of 43), 75–150 μM NMDA was perfused prior to hemisection, thus inducing fictive swimming (average Cr = 0.75 ± 0.19; range = 0.37–0.99). In the hemicords, two types of rhythm were observed, a fast and a slow rhythm. The fast rhythm (2–10 Hz) was observed in one-third of the preparations in which it was investigated (14 of 42 hemicords from 11 animals). Figure 3A1 shows the fast burst pattern and the coordination between two adjacent hemisegments. This can be seen more clearly by comparing rectified/integrated recordings, which display the intensity of ventral root activity over time (Fig. 3A1, bottom). The quality of the fast hemicord rhythm with NMDA was reduced as compared with d-glutamate (average Cr = 0.03 ± 0.015; range = 0.01–0.05), and distinct bursts could be present only in sections of the recordings (Fig. 3A1). The fast rhythm, however, was always detected or confirmed by the autocorrelogram (n = 14; Fig. 3A2). Cross-correlation analysis shows that this fast rhythm is expressed approximately synchronously in two adjacent ventral roots (Fig. 3A3). The fast rhythm was generally detected within the first hours after hemisection and often diminished or vanished as the slow rhythm (see following text) appeared and gained in strength.
A slow rhythm (0.1–0.4 Hz) was expressed in 40 (from 25 animals) of 47 hemicords investigated (Fig. 3B). We observed a high variability from animal to animal. In some cases, it appeared within 1 h (average 1st hour Cr for all preparations = 0.09 ± 0.13; range = 0.00–0.56), whereas in others only several hours after hemisection, gaining progressively in quality during the course of 1–2 days. Irrespective of the time of onset after hemisection, the slow rhythm generally attained a good quality (average Cr = 0.28 ± 0.19; range = 0.01–0.93). Bursting was almost synchronous across nearby ventral roots (Fig. 3B). The slow rhythm was observed in both the left and right hemicord obtained from the same piece of cord (n = 4).
Among the 14 hemicords that expressed the fast rhythm, 12 also expressed the slow rhythm during the experiment. Of these, 10 expressed both rhythms simultaneously.
The histogram in Fig. 3C summarizes the burst frequencies observed before and after the hemisection, during perfusion with NMDA. While fictive swimming in the intact spinal cord was in the range from 0.7 to 3 Hz (▦), the fast rhythm in the hemicords occurred between 2 and 10 Hz (□ to right) and the slow rhythm between 0.1 and 0.4 Hz (□ to left). It is clear from Fig. 3C, that the fast and the slow rhythm are separated by a gap in which rhythmicity was never observed. In preparations where burst frequency was measured both before and after hemisection, the fast rhythm was significantly faster than in control fictive swimming (n = 14; P < 0.001; 298 ± 151% of control), whereas the slow rhythm was significantly slower (n = 30; P < 0.0001; 15 ± 7% of control).
In three hemicords, we varied the concentration of NMDA in the range 38–300 μM (Fig. 3D). We could follow the slow rhythm in all three preparations, but the fast rhythm was present only in one of them (Fig. 3D, top). In both rhythms, a doubling of the NMDA concentration evoked an increase in burst frequency (average increase: 27 ± 9 and 42 ± 40%, fast and slow rhythm, respectively).
The slow rhythm often observed in NMDA was not detected in any of the experiments with d-glutamate (see preceding text). To consolidate this negative finding, we evoked the slow rhythm in four hemicords using NMDA (average Cr = 0.27 ± 0.20; range = 0.10–0.50), and then replaced NMDA with 0.5–2 mM d-glutamate. None of these preparations re-expressed the slow rhythm in d-glutamate (not shown).
Hemispinal cords express a fast rhythm after electrical stimulation
To observe the output of the unilateral locomotor networks under a broader range of conditions, we attempted to activate them by brief electrical stimulation of the hemicord, in the absence of exogenous d-glutamate or NMDA. In the intact spinal cord, electrical stimulation of the cut end produces, as a rule, a short episode of ventral root activity outlasting the stimulus by only a few seconds. The duration of such episodes appears to be limited by the intervening action of crossed inhibition (Fagerstedt et al. 2000). Because one main consequence of splitting the cord along the midline is the removal of inhibition from the contralateral side, a prolonged response to electrical stimulation appeared likely in the hemicord.
After dissection, preparations were perfused with standard Ringer solution and hemisected along the midline. Stimulation electrodes delivered a single pulse or a brief train of pulses (1–30) at one end of each hemicord. Figure 4A shows the activity in two hemicords obtained from the same spinal cord piece and stimulated with the same parameters. Both display a similar bout of activity composed of clear bursts as can be seen in the lower records with higher time resolution. In all hemicords tested (n = 8, from 5 animals), a prolonged bout of rhythmic burst activity in the ventral roots was evoked (average bout duration: 135 ± 54 s). Activity within the bout was organized in distinct bursts in a frequency range from 12 to 3 Hz (Fig. 4, A and B), with the lower frequencies occurring toward the end of the bout. The fast rhythm was of a good quality (average Cr = 0.31 ± 0.13; range = 0.18–0.58) and could be evoked immediately after the lesion and up to 3 days later. The ability of the lamprey hemicord to respond to electrical stimulation with a long bout of rhythmic burst activity is similar to that of the intact and hemispinal cord of the embryo of Xenopus laevis (Soffe 1989).
We also tested the rhythm-generating capability of single spinal hemisegments, which were obtained by transecting the hemicord midway between consecutive ventral roots. A stimulation electrode was placed at one end of a hemisegment and a recording electrode placed on the ventral root. Single hemisegments responded to electrical stimulation with shorter bouts of activity (24 ± 15 s) compared with the longer hemicords. In 14 hemisegments of 29 investigated this activity was rhythmic (average Cr = 0.10 ± 0.05; range = 0.05–0.20) with frequencies in the range 13–6 Hz. Figure 4C shows sample records from two different preparations.
Progressive midline lesions in d-glutamate: the fast hemicord rhythm is linked to fictive swimming
While the intact cords express rhythmic and alternating locomotor activity in the range 0.7–3 Hz (d-glutamate or NMDA), the hemicords may display two very distinct outputs: a fast rhythm with a frequency of 2–12 Hz (d-glutamate, NMDA, electrical stimulation) and a slow rhythm between 0.1 and 0.4 Hz (NMDA). Because rhythm generation in these two nonoverlapping time frames is likely to involve different neuronal mechanisms, the question arises as to which of the two rhythms is directly related to the operation of the network during fictive swimming. We approached this question by studying the transition between intact and hemisected spinal cord by progressively reducing the number of axons crossing the midline while monitoring the ventral root output. This was done by producing intermittent micro-lesions along the midline, such that the cuts were intercalated with an unlesioned midline (Fig. 5A1). Using a tiny scalpel mounted on a micro-manipulator, it was possible to perform cuts of the same length and at the same interval (∼3 per segment) along the entire piece of cord. This density of micro-lesions should be sufficiently high to avoid creating a chain of autonomously functional intact and hemisected spinal cord sections because most crossed axons extend more than one segment rostrally or caudally (see Fig. 3 in Ohta et al. 1991). Once a stable motor output was observed at a certain percentage midline section (100 × cut length/total length), the lesions could be further extended to a new percentage value, progressing for example from 15 to 55, 75, and 100% midline section (Fig. 5A1). The ratio of lesioned length over total length was estimated by eye under the dissection microscope.
Figure 5A1 shows that the d-glutamate-induced burst frequency at 55% section is intermediate between those in control (0%) and after complete hemisection (100%). With every increase in midline section, the burst frequency increased in both ipsi- and contralateral ventral roots (Fig. 5A1, A2). This was true for all preparations (n = 4), which are represented in graphs of burst frequency versus percentage midline section (Fig. 5, A2, B, C, and D). The quality of burst activity (Cr) is given by the size of the filled circles, as indicated in Fig. 5D. The left-right alternation of fictive swimming remained at 75% of sectioned midline but was absent at 90% when the two sides became uncoupled. The motor output expressed in the intermediate preparations forms a continuum between fictive swimming in the controls (0% section) and the fast rhythm obtained after full hemisection (100%) both in terms of frequency and of rhythmic quality (Cr). This continuity of motor behavior during progressive hemisection suggests that the fast rhythm represents the output of the CPG for swimming, in the absence of crossed inhibitory connections.
Progressive midline lesions in NMDA: the fast and the slow hemicord rhythm and their relation to fictive swimming
In NMDA, the hemicords may express both fast and slow rhythms, and it was therefore important to perform the intermittent cuts experiments also during these conditions. Fictive swimming in the intact cord displays as a rule only one burst pattern (Fig. 6A1, 0% section), while most intermediate preparations displayed two rhythms simultaneously, such that bursts at a low frequency also contained high-frequency bursts (Fig. 6A1, 40% section). Both the fast and slow rhythms alternated bilaterally as can be seen in the raw recordings at low and high time resolution as well as in the cross-correlogram (Fig. 6A1, 40% section). On completion of the progressive lesions, the hemicords displayed the slow rhythm alone (Fig. 6A1, 100% section) as they typically do when the slow rhythm is strong (see preceding text). It is evident that the fast rhythm in the intermediate preparations (n = 4) derives from fictive swimming in control, as its frequency increases progressively with the increase in percentage midline lesion (Fig. 6, A2, B, and C). The pattern of continuity between fictive swimming and the fast rhythm that was established in d-glutamate is thus confirmed in NMDA. The slow rhythm, on the other hand, appears in the intermediate preparations already at a low frequency and remains at this rate until the cut is complete (Fig. 6, A2 and B—D). In one case, the slow rhythm was already detectable in the intact cord superimposed on fictive swimming, and it could be followed through several consecutive cuts up to the slow hemicord rhythm (Fig. 6D), while maintaining approximately the same frequency. Taken together these data confirm that the fast rhythm, generated by the lamprey hemispinal cord under an array of different activating conditions, is directly linked to fictive swimming.
Role of the voltage-dependent properties of NMDA receptors
The results presented thus far suggest that the two types of rhythm expressed in the hemicord are not just separated by an order of magnitude in burst frequency but are likely to be mediated by different network configurations (see DISCUSSION). The slow rhythm was only observed in NMDA. NMDA receptors (NMDARs) display a voltage dependence due to the pore-blocking action of magnesium ions (Dale and Grillner 1986; Nowak et al. 1984). To test if this voltage dependence is important for the generation of rhythmicity in the hemicord, Mg2+ was removed from the physiological solution as previously studied in the intact spinal cord (Brodin and Grillner 1986).
In four hemicords, the slow NMDA-induced rhythm (75 or 150 μM; 0.2–0.3 Hz; average Cr = 0.09 ± 0.05; range = 0.05–0.17) was evoked, and then NMDA, as well as Mg2+ ions, was washed out. The NMDA concentration was subsequently increased in steps starting from 10 μM, which was the lower threshold for activation (NMDA is more effective in Mg2+-free solution) (Brodin and Grillner 1986). At ≥40 μM, ventral root activity progressively vanished as observed previously in the intact spinal cord when using Mg2+-free Ringer (Brodin and Grillner 1986). Within this effective range of NMDA concentrations, a slow rhythm was never detected (Fig. 7A), not even with the aid of autocorrelation analysis (not shown). On the other hand, a fast rhythm was now present in all preparations (n = 4; 4–5 Hz; average Cr = 0.09 ± 0.04; range = 0.04–0.14; Fig. 7A, bottom).
In another two hemicords, d-glutamate was perfused to induce the fast rhythm (1 mM; 6–7 Hz; average Cr = 0.27). Both d-glutamate and Mg2+ ions were then washed out for ≥1 h, at which point no activity was present at the ventral roots. d-glutamate was then reperfused in Mg2+-free solution, first at 125 μM and subsequently at 250 μM. The fast rhythm was expressed in the absence of Mg2+ (Fig. 7B) in both preparations tested (5–6 Hz; average Cr = 0.27). A consistent change in quality with Mg2+-free compared with control was thus not manifest.
In summary, the slow rhythm, but not the fast, is crucially dependent on the voltage sensitivity of NMDAR gating. Moreover, when the slow rhythm was abolished in 0 Mg2+, the fast motor pattern emerged.
Fast and the slow hemicord rhythms do not require ipsilateral glycinergic inhibition
Another important question is whether inhibitory neurons participate in shaping patterned activity in the ipsilateral networks of the hemispinal cord, for example by promoting burst termination on the ipsilateral side. In the lamprey, two types of glycinergic neurons that project ipsilaterally have been identified: the lateral interneuron (LIN) (McPherson et al. 1994; Rovainen 1974) and the small ipsilateral inhibitory interneuron (SiIN) (Buchanan and Grillner 1988).
To test this possibility, we blocked glycinergic synaptic transmission by applying strychnine (1 μM) during both the fast rhythm in d-glutamate (n = 6, from 3 animals) and the slow rhythm in NMDA (n = 6, from 5 animals). The effects were investigated after ≥1 h of drug perfusion. Both rhythms remained in strychnine (Fig. 8), indicating that glycinergic inhibition is not essential for rhythm generation in the hemicord. In the case of the fast rhythm (Fig. 8A), the frequency was not significantly affected (average frequency change = - 6 ± 14%), whereas there was a consistent decrease in rhythmic quality in all preparations (average Cr change = - 44 ± 18%). The frequency of the slow rhythm (Fig. 8B) was unaffected in three of six preparations tested, whereas it was somewhat decreased in the remaining three (average frequency change -27 ± 28%). Remarkably, all three cases in which the frequency decreased and one where it did not change were associated with a substantial improvement in quality (Fig. 8B; average Cr change = 153 ± 220%).
From RESULTS follows that rhythmic bursting can be generated by a completely hemisected spinal cord and thus without contralateral inhibition. The hemicord can express two distinct motor patterns, fast and slow bursting, so defined in relation to the frequency of fictive swimming before hemisection. The fast rhythm (2–12 Hz) could be induced both pharmacologically and by electrical stimulation in preparations as small as one single hemisegment. The slow rhythm (0.1–0.4 Hz) appeared only with NMDA and required the NMDAR voltage dependence. Ipsilateral glycinergic inhibition is not essential for the generation of either rhythm. Of the two motor patterns expressed by the hemicord, only the fast is firmly linked to fictive swimming. This was demonstrated by progressively reducing the amount of crossed connections and showing that ventral root bursting accelerates, forming a continuum between fictive swimming before hemisection and the fast rhythm in the hemicord. The slow rhythm may thus either represent a nonlocomotor-related motor behavior or slow swimming dependent presumably on different network mechanisms.
Past attempts to demonstrate rhythmic output in the lamprey hemicord have been negative or unclear. Ventral root bursting was reported to be replaced by tonic/irregular activity after hemisection of the isolated spinal cord (Buchanan 1999; Grillner et al. 1983) or after photoablation of commissural neurons (Buchanan and McPherson 1995). These different findings made Buchanan (1999, 2001) conclude that crossed inhibitory connections are required for locomotor rhythm generation. Signs of a fast rhythm were, however, reported in a brain-stem-hemicord preparation during reticular stimulation and after electrical stimulation of the hemicord in strychnine (Grillner et al. 1986).
The reasons for the failure to detect unilateral bursting in previous studies may be due to several factors. A fast rhythm was occasionally observed in NMDA by Buchanan (1999) but considered unrelated to swimming due to its higher frequency. Moreover, in d-glutamate, the conditions are more favorable than in NMDA used by Buchanan. The slow NMDA rhythm may have escaped attention because it often requires a long time to develop. The present demonstration of fast bursting initiated by electrical stimulation is complementary because it does not depend on a pharmacological activation.
Hemicord locomotor networks do not require glycinergic inhibition
An important consideration is the role of inhibition in the generation of hemicord rhythmicity. The question, if small ipsilateral glycinergic interneurons (Buchanan and Grillner 1988) are involved in the burst generation, had been unresolved, although simulations indicate that they are not needed (Hellgren et al. 1992). Our results with strychnine provide direct evidence that glycinergic inhibition is not required for burst generation to occur and neither for burst frequency regulation of either the fast nor the slow rhythm.
Mechanisms of hemicord burst generation
Why did the fast rhythm not attain as high quality (Cr) in NMDA as with d-glutamate and electrical stimulation? One possibility is that the plateau properties endowed onto spinal neurons by specific NMDAR activation (Wallén and Grillner 1987), although important at low burst frequencies (Brodin and Grillner 1986), might instead be an obstacle at high frequencies. This interpretation is supported by the fact that in NMDA, the highest quality fast rhythm (Cr) was observed in Mg2+-free solution. In d-glutamate, the activation of both AMPA and NMDA receptors will result in a lesser contribution of the voltage-dependent NMDA properties (Zhang et al. 1996).
The excitatory glutamatergic interneurons (EINs) excite each other (Parker and Grillner 2000) and activate motoneurons (Buchanan and Grillner 1987), in both cases via mono-synaptic connections. Rhythm generation in the hemicord can thus be accounted for by an interaction between ipsilateral EINs and does not require glycinergic inhibition. A fast rhythmic motor pattern, as observed in the hemicord, could be produced by a population of EINs that, when active, become synchronized through their mutual excitation. Accordingly, the interval between consecutive spikes or bursts of spikes, mainly dictated by the slow afterhyperpolarization (sAHP), would synchronize across EINs. During these periods of quiescence in the interneurons, there will be no action potentials in the motoneurons, thus separating consecutive bursts at the ventral roots. The sAHP is mainly due to activation of Ca2+-dependent K+ channels (KCa) but also to a smaller component probably representing Na+-activated K+ channels (Cangiano et al. 2002; Wallén et al. 2002). KCa activation has previously been shown to contribute to burst termination in the intact spinal cord (El Manira et al. 1994). This comparatively simple neuronal organization can account for the “fast” burst generation in the hemicord both during d-glutamate and after electrical stimulation and in the intact cord as demonstrated in simulations (Hellgren et al. 1992; Kotaleski et al. 1999).
In NMDA, the fast rhythm is less frequent and often replaced with a slower rhythm (0.1–0.4 Hz) with a reciprocal relationship between the two. Clear analogies exist between the slow rhythm and the NMDA-induced membrane potential oscillations (with a depolarizing plateau) evoked in tetrodotoxin in spinal neurons (Sigvardt et al. 1985; Wallén and Grillner 1985, 1987). The latter have similar frequencies, disappear if NMDA is substituted with d-glutamate (Zhang et al. 1996), and require the voltage-dependent block of NMDARs by Mg2+ ions. If a proportion of the EINs would display plateau potentials, they would also mutually excite each other and therefore be active together. The termination of the plateau potentials is due to activation of KCa channels caused by the Ca2+ entry during the plateau. As the EINs terminate firing one after the other, the remaining cells will lose the input from surrounding EINs. This mechanism could thus account for the slow burst generation.
Finally, in the transition from a fast to a slow NMDA-induced rhythm, the two can be expressed simultaneously. This can be explained if part of the EINs display plateau potentials and generate the slow rhythm, whereas other EINs become synchronized and tick-on in the fast rhythm. The action potentials generated by plateauing EINs could still fire in synchrony with the remaining “fast” EINs. A simple arrangement like this can account for the two superimposed rhythms. Slow rhythms superimposed on fictive swimming have also been observed in the intact lamprey spinal cord (Aoki et al. 2001) and in the tadpole (Reith and Sillar 1998).
Glycinergic blockade during fictive swimming in the intact spinal cord mimics hemisection
A partial glycinergic blockade by strychnine leads to an increased locomotor frequency (Cohen and Harris-Warrick 1984; Grillner and Wallén 1980; McPherson et al. 1994). This is in agreement with the frequency changes observed here with progressive midline lesions. With a full blockade of glycinergic inhibition, a tonic activity has been reported (Grillner and Wallén 1980; McPherson et al. 1994) or a fast (1–5 Hz) or very slow synchronous bilateral bursting (Alford and Williams 1989; Cohen and Harris-Warrick 1984; Hagevik and McClellan 1994). The fast bilateral bursting can be explained by a crossed excitatory interaction and is in the hemicord fast rhythm range.
The contralateral phasic inhibition in the intact spinal cord is thus responsible for generating not only the left-right alternation but also for slowing down the burst frequency. Between each burst there will be a period of inhibition that delays the occurrence of the subsequent burst and thereby prolongs cycle duration.
Organization of the spinal locomotor networks in other vertebrates
The spinal network for swimming in the frog embryo (Xenopus laevis) shares many features with that of the lamprey. The frog embryo swims with alternating left-right contractions of the body that, however, are generated by single motoneuronal action potentials in each swim cycle. Unilateral activity can be evoked in the Xenopus hemicord having a frequency higher than that of fictive swimming before hemisection (Soffe 1989). In analogy, low doses of strychnine increase the locomotor rate in the intact cord (Dale 1995).
In mammals and other tetrapods, the evidence for a UBG organization is more indirect. Unilateral bursting may occur in flexors without concomitant extensor activity, and activity may occur in proximal muscles without rhythmic bursting in distal muscles (Grillner 1985; Grillner and Zangger 1979; Kiehn and Kjaerulff 1998; Stein et al. 1995). In the rat embryo, synchronous bursting occurs initially in all muscles of the two hindlimbs, somewhat later left and right limbs start to alternate, and finally also flexors and extensors within a limb (Nishimaru and Kudo 2000). In the newborn rat, a blockade of reciprocal inhibition during fictive stepping, increases the locomotor rate, as in the lamprey and tadpole, and a full blockade can induce slow bursts synchronized across all muscle groups (Cowley and Schmidt 1995; Kremer and Lev-Tov 1997). In the mudpuppy, forelimb flexor and extensor activity could be surgically separated (Cheng et al. 1998). These data taken together suggest that there are different rhythm-generating centers for each group of muscles and that reciprocal inhibition is not required for rhythmogenesis.
In conclusion, the lamprey locomotor CPG contains left and right rhythm-generating networks distributed along the spinal cord. These unilateral networks internally rely solely on excitation but are strongly downregulated in frequency by reciprocal inhibition. These results, taken together with those obtained in other vertebrates, provide support for the notion of a neural organization of the vertebrate motor system with a series of UBGs that can be recombined in a flexible way to produce different motor patterns.
We are grateful to Drs. A. el Manira, O. Kiehn, and P. Wallén for valuable comments on the manuscript. We also appreciate M. Bredmyr for skillful laboratory assistance and J. D. Woolley for participating in some initial experiments.
This project was funded by the Swedish Research Council VR-M-3026, Marianne and Marcus Wallenberg Foundation, Karolinska Institutet funds, European Union Grant QLG3-CT-2001-01241. L. Cangiano has been a recipient of a Marie Curie doctoral fellowship from the European Commission.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2003 by the American Physiological Society