INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The properties enabling a neuronal network to generate rhythmic activity have been extensively studied in the isolated preparation of the neonatal rat spinal cord (Cazalets and Bertrand 2000; Kiehn and Kjaerulff 1998). The patterned discharges generated by this intrinsic network [called the central pattern generator (CPG)] (Marder and Calabrese 1996) can be monitored by recording from motoneurons or their axons. One particular pattern (termed fictive locomotion), consisting of fast motor discharges alternating between flexor and extensor motor pools on the two sides of the spinal cord, is widely thought to be the expression of the locomotor CPG (Kiehn and Kjaerulff 1998; Nishimaru and Kudo 2000; Rossignol and Dubuc 1994). Such a rhythm is elicited by applying neurotransmitters (reviewed by Kiehn and Kjaerulff 1998) such as excitatory amino acids, biogenic amines [especially 5-hydroxytryptamine (5-HT)], and acetylcholine, or by electrical stimulation of dorsal root fibers (Marchetti et al. 2001). Conversely, spontaneous rhythmic activity lacking alternation and thus unable to support locomotion is induced after pharmacological block of glycine and GABA receptors (Bracci et al. 1996).

During the course of our work on the action of neuropeptides on rat spinal neurons, we recently observed a novel type of intense, long-lasting bursting induced by senktide (Barbieri and Nistri 2001), a selective agonist for NK₃ receptors (a subclass of the tachykinin receptors mediating excitation by substance P and endogenous neurokinins). This phenomenon is normally not induced by activation of other tachykinin receptors (Fisher et al. 1994) and is interesting because tachykinins are regarded as major spinal excitatory transmitters (reviewed by Rekling et al. 2000). Furthermore, substance P can accelerate fictive locomotion in the rat (Barthe and Clarac 1997) or fictive swimming in the lamprey (Parker et al. 1998), although it cannot per se elicit these activities. As substance P is a broadly acting agonist effective on all major classes of receptors (termed NK₁, NK₂, and NK₃) (Maggi et al. 1993; Regoli et al. 1994), one possibility is that simultaneous activation of receptors mediating contrasting actions could prevent the onset of the CPG activity. Alternatively, the pathways expressing tachykinin receptors may have only a minor involvement in the locomotor network.

While NK₂ receptors are minimally present in the immature rat spinal cord (as shown by ligand binding and electrophysiological studies) (Baranauskas et al. 1995), this preparation does contain NK₁ and NK₃ receptors (Otsuka and Yoshioka 1993), which, however, differ in their distribution. In fact, NK₃ receptors are diffusely present throughout the spinal gray matter (Beresford et al. 1992; Linden et al. 2000; Mileusnic et al. 1999; Seybold et al. 1997), while NK₁ receptors are most abundant in the deep layers of the dorsal horn and absent in lamina II (Liu et al. 1994; Ogawa et al. 1985).

The present study aimed at clarifying the basic properties of bursting evoked by NK₃ receptor activity, its segmental distribution, and its potential interaction with the activity generated by the locomotor CPG of the neonatal rat spinal cord in vitro.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Experiments were carried out on lumbar spinal cord preparations (comprising a region from mid-thoracic level to conus medullaris) isolated from neonatal Wistar rats (0-4 days old) under urethan anesthesia (0.2 ml ip of a 10% wt/vol solution) as previously described (Bracci et al. 1998). This procedure is in accordance with the regulations of the Italian Animal Welfare Act and is approved by the local authority veterinary service.

The spinal cord was superfused (7.5 ml/min) with Krebs solution of the following composition (in mM): 113 NaCl, 4.5 KCl, 1 MgCl₂7H₂O, 2 CaCl₂, 1 NaH₂PO₄, 25 NaHCO₃, and 11 glucose, gassed with 95% O₂-5% CO₂; pH 7.4 at room temperature. All agents were bath-applied via the superfusing solution at the concentrations mentioned in the text.

DC-coupled ventral root (VR) recordings (usually from pairs of L₂ and L₅ VRs bilaterally) were obtained with glass suction microelectrodes containing an Ag-AgCl pellet and filled with Krebs solution. Dorsal root (DR) electrical stimuli, delivered via miniature bipolar suction electrodes, were employed to elicit VR reflexes (recorded from the ipsilateral VR of the same segment). Dorsal root-evoked dorsal root potentials (DR-DRPs) were induced by repeated electrical stimuli applied to the ipsilateral DR (usually L₅) and were antidromically recorded from the severed end of a DR (usually L₄). In all instances stimulus intensity (1-20 V range; 0.1 ms duration) was calculated in terms of threshold (Th), defined as the minimum intensity to elicit a detectable response in the homolateral VR (on average, Th = 1.8 ± 0.9 V, mean ± SD, n = 45). For evoking DR-DRPs stimulus intensity was set at 2-5 times Th.

Intracellular recordings were obtained under current-clamp conditions with sharp electrodes filled with either 3 M KCl (30-60 M resistance), or 2 M KMeSO₄ (60-120 M resistance). L₃-L₅ motoneurons were identified functionally by antidromic stimulation (Fulton and Walton 1986) delivered to the corresponding VR. During intracellular experiments the activity of one or more VRs was also recorded. DC-coupled VR and motoneuron recordings were amplified, displayed on-line on a chart recorder, and digitally stored on DAT tape (acquisition rate, 11 kHz) or on computer hard disk. In some experiments, QX-314 · Cl (300 µM) was added to the intracellular solution to block Na⁺-dependent spikes and slow inward rectifiers (Perkins and Wong 1995) that would otherwise mask recurrent synaptic potentials. The input resistance of motoneurons was measured by delivering hyperpolarizing current steps (0.04-0.1 nA, 15-30 ms) through the intracellular electrode: motoneurons had input resistance of 50 ± 10 M (n = 8) when recorded with KCl electrodes and 80 ± 50 M (n = 9) when recorded with KMeSO₄ electrodes.

Data were quantified as means ± SD; statistical significance was assessed with the Student's t-test, or ANOVA plus Tukey test. The accepted level of significance was P = 0.05. Period (T) was defined as the time between the onset of two cycles of oscillatory activity. When period values were averaged for a pool of preparations, data from each spinal cord were calculated as the mean of at least five cycles. Phase between two roots was calculated as reported earlier (Marchetti et al. 2001) and expressed in angular degrees whereby the value of 180° represents complete phase alternation and 0 or 360° full phase coincidence. The strength of coupling between left/right and ipsilateral L₂/L₅ VRs was analyzed with circular statistics (Kjaerulff and Kiehn 1996) in which R (which ranges from 0 to 1) is the concentration of phase values around the mean phase lag (; expressed as angular degrees). The Rayleigh test with the small-sample modification (Drew and Doucet 1991) was used to establish the statistical significance of these values.

NK₃ agonists were applied for a maximum of 4 min at intervals of at least 30 min to minimize tachyphylaxis (Barbieri and Nistri 2001). Senktide and [MePhe7]neurokinin B were purchased from NeoSystem Laboratoire. The NK₃ antagonist SR 142801 (osanetant) was kindly donated by Dr. Edmonds-Alt, Sanofi Recherche, Montpellier, France. 5-HT, GABA, and glycine were purchased from Sigma; N-methyl-D-aspartate (NMDA), 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX), and D-amino-phosphonovalerate (APV) were purchased from Tocris.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Depolarization, fast oscillations, and delayed bursting evoked by senktide

Figure 1 shows an example of the extracellularly recorded effect of the NK₃ agonist senktide (100 nM) on the neonatal rat spinal cord. With a 10 s latency from the start of application (which lasted 4 min; Fig. 1A), all four VRs (right and left L₂ and L₅) gradually developed a depolarization that, after about 60 s, abruptly increased in amplitude and presented intense, fast oscillatory activity (T = 1.70 ± 0.04 s). Despite the continuous presence of senktide, oscillations lasted 48 s only, as VRs spontaneously begun repolarizing. Figure 1B (continuous records from Fig. 1A; note doubling of amplifier gain) shows that, during agonist wash out, rhythmic depolarizing bursts appeared (T = 43 ± 10 s for the example shown in Fig. 1). Some bursts (although not all of them; see for instance last episodes of lL₅ and rL₅) had an intraburst structure consisting of oscillations with average T = 1.8 ± 0.3 s. These events will be further characterized below. Return to control resting activity appeared 16 min later (not shown).

View larger version (35K): [in this window] [in a new window]   Fig. 1. Effect of the NK3 agonist senktide on the neonatal rat spinal cord. A: effect of the NK3 agonist senktide (100 nM) extracellularly recorded from 2 pairs of ventral roots (VRs; left and right L2 and L5: lL2, lL5, rL2, rL5). Vertical arrow indicates the start of the 4 min application of senktide. B: (continuous records from A; note doubling of amplifier gain) during agonist wash out, rhythmic, depolarizing, bursts appear (T = 43 ± 10 s), which can include an intraburst structure consisting of oscillations with average T = 1.8 ± 0.3 s. Horizontal arrow shows control baseline of lL2 VR prior to senktide application for this level of DC amplification. Return to control resting activity requires 16 min (not shown).

All 68 preparations showed depolarization on senktide (50-200 nM) application, although only 53% showed oscillations during agonist application. The concentrations used were large enough to saturate NK₃ receptors (Barbieri and Nistri 2001; Fox et al. 1996). The oscillations during the depolarization evoked by senktide lasted on average 52 ± 26 s (range 16-104 s), with T = 2.8 ± 0.8 s (range 1.6-4 s). The extent and statistical significance of phase coupling between oscillations in the four VRs is presented in Table 1. The majority (44/68) of spinal cords showed only left/right homosegmental alternation at L₂ as well as L₅ level. A substantial minority (20/68) presented both homosegmental and homolateral (involving flexor and extensor motor pools) alternation. The remaining eight preparations had unstable phase shifts during the oscillatory patterns.

Characteristics of delayed bursting activity developing after senktide wash out

For 5-20 min after senktide wash out, all preparations showed sustained firing as indicated by the high-frequency noise recorded from VRs before reattaining baseline polarization. In most cases (70%) this stage was accompanied by bursts present on at least one VR. Individual burst duration was on average 8 ± 2 s (n = 40 spinal cords), which clearly separated these late-onset bursts from the shorter lasting oscillations appearing during the application of senktide. Bursts could display either left/right alternation (with synchronous homolateral activity, 52% of preparations), or variable phase coupling (42%) or, more rarely, full synchronicity (5%). An example of the prevailing response pattern, namely synchronous homolateral bursting (L₂ and L₅), is shown in Fig. 2A. Interburst interval (IBI; time between the start of a burst and the onset of the next one) variability was assessed by calculating the coefficient of variation (CV) of the IBI for each preparation. The mean CV was 0.4 ± 0.2, while the mean IBI was 55 ± 21 s (n = 40 preparations).

View larger version (37K): [in this window] [in a new window]   Fig. 2. Characteristics of late bursting. A: example of synchronous homolateral bursting during senktide wash out recorded from lL2 and lL5 VRs. B: polar plots of phase lags for the oscillations shown in A reveal full homolateral synchronicity both for "slow" (T = 1.35 ± 0.08 s) oscillations and "fast" (T = 0.22 ± 0.09 s) ones. C: expansion of the time scale of 1 of the bursts recorded on wash out (shown in A) reveals oscillatory (T = 1.35 ± 0.08 s, "slow") intraburst structure. D: further expansion of the time scale from C shows, within the same burst, an additional class of faster oscillations with T = 0.22 ± 0.09 s.

We examined whether the late bursting activity appearing during senktide wash out was a stereotypic response by a network locked into a patterned operation or whether it was a rhythmic program modulated by afferent synaptic inputs. For this purpose on four spinal cord preparations, after wash out of senktide (100 nM), we attempted to elicit bursts by repeated electrical stimuli (0.033 Hz; 2-3 times Th) one L₅ DR. Bursts could be observed on all VRs and had an average duration of 10 ± 4 s with 1:1 entrainment. In no instance did bursts appear on DRs.

Complex intraburst oscillations during senktide wash out

A considerable degree of complexity emerged when the intraburst oscillatory structure was analyzed. This is exemplified by Fig. 2B that shows the oscillatory (T = 1.35 ± 0.08 s) structure of one spontaneous burst generated during continuous recording from two homolateral VRs (starting from 8 min senktide wash out). Further expansion of the time scale (Fig. 2D) reveals, within this burst, an additional class of faster oscillations with T = 0.22 ± 0.09 s. The slower intraburst oscillations were recorded in 25% of preparations (average T = 2.0 ± 0.5 s), while the faster oscillations (average T = 0.22 ± 0.04 s) were seen in 62% of cases. The remaining spinal cords showed only tonic firing during bursts.

It was difficult to establish whether intraburst oscillations had phase coupling because this type of analysis required synchronicity of bursts as well as presence of slow oscillations, two conditions that limited the number of samples to 10 preparations. The polar plots shown in Fig. 2C indicate that slow as well as fast oscillations were synchronous between lL₂ and lL₅ VRs in the preparation illustrated in Fig. 2A. This condition was observed in all 10 spinal cords that presented homolateral synchronicity with  = 5° and  = 12° for fast and slow oscillations, respectively (R = 0.95 and 0.92).

Action of senktide was mimicked by [MePhe7]neurokinin B

The responses induced by senktide raised the question of receptor selectivity. Thus, we tested another NK₃ agonist, namely [MePhe7]neurokinin B (100 nM) that, as shown in Fig. 3, closely reproduced the pattern of effects evoked by senktide. In fact, a gradually developing depolarization was suddenly followed by a larger depolarization plateau containing oscillatory activity slowly waning despite sustained agonist application. During wash out, bursting activity emerged with intraburst oscillations. On eight preparations, [MePhe7]neurokinin B induced VR depolarization characterized by oscillations with T = 3 ± 1 s. During wash out late bursts occurred with 44 ± 12 s IBI and 9 ± 3 s duration. Bursts comprised fast oscillations with T = 0.21 ± 0.05 s.

View larger version (33K): [in this window] [in a new window]   Fig. 3. Effect of the NK3 agonist [MePhe7]neurokinin B. A: activity induced by the NK3 agonist [MePhe7]neurokinin B (100 nM) closely reproduces the pattern evoked by senktide: depolarization first develops gradually, and more intensely after about 20 s, together with oscillatory activity. Depolarization and oscillations slowly fade away despite sustained agonist application. B: during wash out, bursting activity emerges with intraburst oscillations (note doubling of amplifier gain). Bursts occur synchronously (with a period of about 15 s) on lL2 and lL5 VRs, and with variable period on the contralateral rL5 VR. C: intraburst structure comprises fast oscillations with T = 0.23 s (3 × amplifier gain with respect to A).

Pharmacological characterization of delayed bursting induced by senktide

Delayed bursting could have been due to rhythmic network activity either caused by slow wash out of senktide persistently activating NK₃ receptors, or to a distinct pattern consequent to former NK₃ activation. Another possibility might have been that delayed bursting originated from motoneurons made hyperexcitable by previous application of senktide.

We designed two protocols to test these possibilities. First, we applied senktide for 4 min to generate spinal cord depolarization (with oscillations) and then washed out senktide by applying a Krebs solution containing 10 µM SR 142801, a selective NK₃ receptor antagonist (Beaujouan et al. 1997; Emonds-Alt et al. 1995). This approach is exemplified in Fig. 4A, where delayed bursting continued on all four VRs despite the presence of the antagonist. Similar results were obtained on five preparations in which the IBI value for delayed bursting was 59 ± 8 s in control solution versus 68 ± 9 s in SR 142801 solution. The corresponding data for mean duration of individual bursts were 10 ± 3 s and 9 ± 5 s, respectively. None of these differences was statistically significant.

View larger version (30K): [in this window] [in a new window]   Fig. 4. Delayed bursting does not require persistent NK3 receptor activation. A: VR records from right and left L2 and L5 VRs during application of the NK3 antagonist SR142801 (10 µM) that had started immediately after senktide wash out (not shown). Traces are shown at 13 min wash out (see arrow) during continuous exposure to SR142801 and present delayed bursting activity with no changes in interburst interval or burst duration. B: effect of 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX; 10 µM) and D-amino-phosphonovalerate (APV; 50 µM) during senktide wash out (different preparation from A) shown for lL5 VR. The record starts at 5 min senktide wash out (see arrow) in the presence of CNQX and APV. Bursting is completely abolished on all 4 VRs (for simplicity only lL5 trace is shown here). Note that DR stimuli (see large deflection artifacts) fail to elicit polysynaptic reflexes in CNQX plus APV solution.

Second, we tested whether delayed bursting required excitatory glutamatergic synaptic transmission. For this purpose, on five preparations (different from those tested with SR 142801), after observing senktide-evoked depolarization and oscillations for 4 min, we washed out the tachykinin agonist with a solution containing 10 µM CNQX and 50 µM APV as exemplified in Fig. 4B. In this case, all bursting activity was eliminated and electrical stimulation of one L₅ DR (2-3 times Th intensity) failed to induce any response detectable from the corresponding L₅ VR (see artifacts on last part of record in Fig. 4B). Similar results were observed in the other four preparations.

These findings suggest that delayed bursting did not require persistent NK₃ receptor activity but it required network-based excitatory glutamatergic transmission.

Disinhibition as a cause for rhythmic activity evoked by senktide?

As reported earlier, on the majority of preparations the senktide-induced oscillations and delayed bursting were alternating at segmental level but lacked flexor-extensor motor pool alternation on the same side. This curious phenomenon raised the possibility that, somehow, activation of NK₃ receptors functionally affected coupling between lumbar segments perhaps because their GABAergic and glycinergic connections were weakened. As a first approximation we investigated whether, during the bursting stage, receptors for GABA or glycine remained responsive to application of exogenous GABA or glycine. On seven preparations in which senktide induced late bursting developing synchronously on the same side, but alternating at segmental level, application of GABA (0.5 mM; a concentration that strongly blocks reflex activity) (Rozzo et al. 1999), fully eliminated late bursts as shown by the example in Fig. 5A. This action of GABA was reversible on wash as bursts returned at a time corresponding to 10 min of senktide wash out. However, in four other preparations exposed to 0.5 mM GABA, late bursting persisted with no significant changes in IBI (48 ± 13 s in control vs. 55 ± 20 s in GABA solution; P > 0.05, paired t-test) or in single burst duration (7.9 ± 2.1 s in control vs. 8.5 ± 2.4 s in GABA solution; P > 0.05; Fig. 5B). In this subgroup of preparations, three had segmental alternation and ipsilateral synchronicity before GABA application, while one had synchronous activity. In the presence of GABA, such patterns remained unchanged. Resistance of bursting to GABA by these preparations did not involve general loss of GABA effectiveness as the DR-evoked (3 times Th intensity) VR reflexes were significantly (P < 0.001) and reversibly depressed (by 68 ± 18%) in all preparations tested (n = 11).

View larger version (29K): [in this window] [in a new window]   Fig. 5. Complex effects of exogenous GABA on late bursting appearing during senktide wash out. A, left: bursts (duration 12 s, period 45 s) recorded from rL5 VR after 10 s senktide wash out. Between two bursts electrical stimulation of rL5 DR evoked burst lasting 10 s. A, right: after 4 min senktide wash out and in the presence of 0.5 mM GABA bursts are suppressed and DR stimulation becomes ineffective (see artifact indicated by arrow). B, left: bursts (duration 8 s, period 40 s) recorded in control solution from a rL5 VR at 4 min senktide wash out. Different preparation from the one shown in A. B, right: in the presence of GABA, bursts continue with the same duration and interburst interval. The time scale expansion below the trace demonstrates that the intraburst structure also comprises fast oscillations, with 0.26 s period.

In addition to checking for the ability of exogenous GABA to suppress bursting, we sought to clarify whether endogenously released GABA retained its effects at the time of senktide-induced bursting. For this purpose we monitored changes in the GABAergic DR-DRP (Curtis et al. 1971; Levy 1977; Nicoll and Alger 1979; Nistri 1983). DR-DRP amplitude did not significantly decline (80 ± 10% of control; n = 5; P > 0.05) during application of senktide, but it was significantly depressed (70 ± 20%; n = 5; P < 0.05) during wash out. These data indicate that a deficit in GABAergic transmission occurred during late bursting.

Application of glycine (0.5 mM) after the first 2-5 min of senktide wash blocked bursts in seven preparations, out of which six demonstrated ipsilateral synchronous bursting and one fully synchronous activity. In two further preparations with ipsilateral synchronous bursting, this phenomenon continued in the presence of glycine: in fact, neither IBI (44 ± 6 s in control vs. 54 ± 10 s in glycine) nor burst duration (10 ± 2 s in control vs. 11 ± 3 in glycine) was significantly altered. On all nine spinal cord preparations, the DR-induced VR reflex was significantly (P < 0.05) inhibited by glycine (by 65 ± 20%).

Senktide-induced oscillations and bursting recorded intracellularly from motoneurons

The widespread distribution of NK₃ receptors in the rat spinal cord makes it difficult to identify the cells responsible for oscillations and bursting induced by senktide. Since in each case this rhythmic activity converged on lumbar motoneurons, we recorded intracellularly from these cells to address the following issues. 1) Did motoneurons actively participate in bursting? 2) What electrophysiological properties characterized the oscillatory activity? 3) Were bursting and oscillations associated with changes in motoneuron recurrent inhibitory postsynaptic potentials (IPSPs)?

Two sets of intracellular experiments were performed, namely those on motoneurons with fully blown spikes recorded with KCl- or KMeSO₄-filled electrodes (n = 22) and those on cells in which spikes were blocked by intracellularly applied QX-314 (n = 14). The mean resting potential (V_m) for the first set of cells was 69 ± 5 mV. One example of these experiments is given in Fig. 6 that shows simultaneous intracellular (bottom trace; rL₅-Mn) and extracellular recordings (top 3 traces with VR identification indicated alongside) of a single burst (comprising a series of oscillations) appearing at 5 min wash out of senktide. The intracellular record clearly shows that, at 60 mV membrane potential, the rL₅-Mn (recorded with a KMeSO₄-filled electrode) begun depolarizing 1.2 s earlier than the onset of the first oscillation recorded extracellularly. Each oscillatory waveform of this motoneuron displayed, at its peak, a cluster of 2-5 spikes, suggesting that action potentials were generated only when the positive phase of the oscillations brought the motoneuron membrane potential to firing threshold. Each oscillation comprised a hyperpolarization (4 mV below baseline) during which spontaneous activity was subdued.

View larger version (35K): [in this window] [in a new window]   Fig. 6. Intracellular and extracellular recording of a single burst. Simultaneous intracellular (bottom trace; rL5-Mn recorded with KMeSO4-filled electrode) and extracellular recording (top 3 traces from lL2, lL5, rL2 VRs) of a single burst appearing at 5 min senktide wash out. On the motoneuron the burst develops from Vm = 60 mV and comprises oscillations of T = 1.8 s (starting 1.2 s earlier than the onset of the first oscillation recorded extracellularly). A cluster of 2-5 spikes is present at the peak of each oscillation. Motoneuronal oscillations appear in antiphase (R = 0.71,  = 162°) with the extracellular activity, recorded synchronously from all 3 VRs. Large upward deflections appearing every 5 s are antidromic spikes (plus stimulus artifact).

In the example of Fig. 6, the oscillatory activity of the single motoneuron lasted longer than that recorded from the three VRs and was in antiphase (R = 0.71,  = 162°) with the extracellular activity appearing synchronously on all three VRs (despite a modest lag for lL₂ VR). Note that, unlike the pattern observed with the single-cell oscillations, the oscillation troughs detected extracellularly from VRs were associated with a noisy baseline, indicating a degree of asynchrony in motoneuron pool firing behavior at each segmental level.

The mean depolarization induced by senktide (100 nM), recorded from single motoneurons, was 12 ± 6 mV (n = 22) and was associated with a significant change in input resistance (108 ± 9%, n = 14; P < 0.05) measured by averaging at least five electrotonic potentials evoked every 10 s. In the presence of senktide, some motoneurons (n = 8) fired phasically with action potentials arising from oscillations while other motoneurons (n = 4) fired irregularly (average discharge rate = 5.4 ± 4.6 Hz) or did not reach firing threshold (n = 10). During 5-10 min wash out of senktide in coincidence with late bursting, the average baseline input resistance (measured after rejecting electrotonic potentials during bursts) was significantly increased (115 ± 16%, n = 14, P < 0.001) despite return of membrane potential to control level. At this stage phasic motoneurons (n = 8) fired bursts of action potentials (average rate = 8.8 ± 5.5 Hz) that were usually alternated with bursts recorded extracellularly from the contralateral VR. There was no difference in responses of cells impaled with KCl or KMeSO₄ electrodes.

We next examined the voltage dependence of bursting and oscillations. As indicated in Fig. 7A, membrane depolarization to 40 mV did not elicit bursting during senktide wash out, whereas membrane hyperpolarization (10 to 65 mV V_m) augmented the burst amplitude without suppressing it (Fig. 7, B and C). In both cases intraburst oscillations persisted with similar period. Similar observations were repeated on seven motoneurons (V_m = 66 ± 5 mV) depolarized by 10-20 mV or hyperpolarized by 5-15 mV. These results suggest that bursting was a network-dependent phenomenon and was not due to voltage-activated conductances intrinsic to the motoneuron membrane.

View larger version (52K): [in this window] [in a new window]   Fig. 7. Bursts do not depend on motoneuron membrane potential. A: senktide wash out phase of a rL3 motoneuron; note that spontaneous bursting is present, but cannot be elicited by manually depolarizing the cell to 40 mV. B: burst recorded intracellularly from a rL5 motoneuron at Vm = 55 mV. C: burst recorded at Vm = 65 mV from the same motoneuron. Oscillations are still present with 30% larger amplitude.

Intense spiking activity could actually mask the time course of oscillations and complicate membrane potential and input resistance changes evoked by senktide. To monitor oscillations more directly, five motoneurons were recorded with an intracellular solution containing QX-314 (in KMeSO₄ solution) to block Na⁺-dependent spikes and slow inward rectifiers. On these cells senktide induced 8 ± 4 mV depolarization associated with no significant change (3 ± 2%) in input resistance. The wash out stage was still characterized by bursting and 12 ± 7% rise (P < 0.05) in baseline input resistance. Figure 8C shows an example of raw data of electrotonic potentials recorded before, during, and after application of senktide. Figure 8A exemplifies recordings from one L₂ VR and a rL₄ motoneuron (impaled with QX-314-filled electrode) after 3 min wash out of senktide. Despite the fact that action potentials were suppressed, motoneuron baseline potential oscillated during bursts. In particular, the rL₄ motoneuron oscillated in phase with the homolateral rL₂. The trough of each oscillation did not become hyperpolarized with respect to membrane potential baseline, a phenomenon observed in all cells recorded with QX-314 electrodes (Fig. 8B).

View larger version (27K): [in this window] [in a new window]   Fig. 8. Bursting activity recorded intracellularly with a QX-314-filled electrode. A: recordings from rL2 VR and rL4 motoneuron (impaled with QX-314-filled electrode) after 3-min wash out of senktide. Motoneuron oscillations during bursts are present despite the fact that action potentials are suppressed. rL4 motoneuron oscillates in phase with the homolateral rL2. B: time scale expansion of a burst recorded from rL4 motoneuron. C: superimposed records of average electrotonic potentials used to measure motoneuron input resistance (same cell as in A and B) in control solution, during senktide application and its wash out (indicated by arrow). Resistance is similar in control and during senktide application, while it is increased during wash out. D: average recurrent inhibitory postsynaptic potential (IPSP) recorded from rL4 motoneuron (same as in A-C) in the absence of spike contamination (due to the presence of QX-314 in the recording electrode) in control (a), during senktide application (b), and during wash out (c). Note depression of IPSP during senktide wash out.

Recurrent IPSPs during senktide evoked bursting

On four cells impaled with QX-314-filled electrodes and stimulated antidromically, it was possible to observe a recurrent IPSP, the main component of which is believed to be mediated by glycine released by Renshaw cells (Werman et al. 1968), while the late component is GABAergic (Cullheim and Kellerth 1981). Suppression of the antidromic action potential avoided contamination of the IPSP by the spike afterhyperpolarization. The IPSP was recorded at 65 mV membrane potential and was depolarizing (amplitude = 11 ± 4 mV) regardless of the presence of KCl or KMeSO₄ in the electrode. The IPSP had average rise time of 1.4 ± 0.6 ms, and its decay could be fitted by a single exponential (time constant = 21 ± 9 ms). Since the peak amplitude was blocked (by 54 ± 8%) by 1 µM strychnine, it is confirmed that the IPSP was mainly mediated by glycine.

Recurrent IPSPs were studied in the presence of senktide (100 nM) or during the stage of senktide wash out. As shown in Fig. 8Db, during senktide application, IPSP area and peak amplitude were not significantly changed (on average they were 105 ± 8 and 104 ± 5%, respectively; n = 4). Nevertheless, during senktide wash out (see example in Fig. 8Dc), the IPSP area was significantly (P < 0.05) depressed (62 ± 20%), although the reduction in peak amplitude was not significant (40 ± 40%). On these preparations, bursting showed homolateral synchronicity (but left/right alternation).

Functional role for NK₃ receptors in the fictive locomotor network

The widespread distribution of NK₃ receptors does not preclude their presence also on CPG neurons responsible for generating fictive locomotor patterns. To explore this possibility, we studied whether the locomotor rhythm induced by standard neurochemicals (NMDA and 5-HT) (Kiehn and Kjaerulff 1998) could be perturbed by either activating or blocking NK₃ receptors. Figure 9A shows irregular rhythmic activity (recorded extracellularly from L₂ and L₅ VRs) in the presence of low concentrations of NMDA (2 µM) and 5-HT (3 µM), clearly under the threshold for fictive locomotion. After 2 min of senktide application in the continuous presence of NMDA and 5-HT (Fig. 9B), a regular, fictive locomotor-like rhythm (T = 1.25 ± 0.05 s) appeared with left-right alternation at segmental level and ipsilateral alternation between L₂ and L₅ VRs. The Rayleigh test gave significant coupling (P < 0.001) for all pairs of VRs (lL₂/rL₂: R = 0.93,  = 185°; lL₂/lL₅: R = 0.94,  = 183°; lL₅/rL₅: R = 0.91,  = 188°; rL₂/rL₅: R = 0.93,  = 186°).

View larger version (60K): [in this window] [in a new window]   Fig. 9. Facilitation of fictive locomotor patterns by activation of NK3 receptors. A: irregular rhythmic activity (recorded extracellularly from L2 and L5 VRs) evoked by low concentrations of N-methyl-D-aspartate (NMDA; 2 µM) and 5-hydroxytryptamine (5-HT; 3 µM), under threshold for fictive locomotion. B: regular, fictive locomotor-like rhythm (T = 1.25 ± 0.05 s) with left-right homosegmental alternation and ipsilateral alternation appears after 2 min of senktide co-application with NMDA and 5-HT. C: after senktide wash out and in the continuous presence of NMDA and 5-HT, episodes of fictive locomotion (T = 2.08 ± 0.09) appear during bursts (lasting 20 ± 7 s) and with typical homolateral and homosegmental alternation.

After senktide was washed out (Fig. 9C) and NMDA and 5-HT were still continuously applied, fictive locomotor patterns faded away. Nevertheless, even when the root polarization level had returned to the value before senktide application, episodes of fictive locomotion (T = 2.08 ± 0.09) with typical homolateral and homosegmental alternation appeared during bursts (lasting in this example 20 ± 7 s). Applying the Rayleigh test to these oscillations yielded the following values: lL₂/rL₂: R = 0.94,  = 183°; lL₂/lL₅: R = 0.95,  = 183°; lL₅/rL₅: R = 0.89,  = 190°; rL₂/rL₅: R = 0.92,  = 184° (P < 0.001), which are very similar to those observed during the application of senktide plus NMDA and 5-HT. This result was obtained in three of four preparations.

We also tested application of senktide during a stable fictive locomotor rhythm induced by NMDA and 5-HT. In 7/10 preparations, senktide (100 nM) raised oscillation period by 75 ± 15% with unchanged phase lags (in the other 3 cases period did not change).

Another test for the interaction between NK₃ receptor activity and the locomotor network was done by applying the NK₃ antagonist SR142801 (10 µM) to preparations exhibiting a stable fictive locomotor rhythm. An example of this is shown in Fig. 10A. In this preparation a stable rhythm (T = 1.8 s) was induced by 6 µM NMDA and 10 µM 5-HT. After superfusing the spinal cord with SR142801 (in the continuous presence of NMDA and 5-HT), the rhythm became irregular, with oscillations of smaller amplitude and slower period (T = 3.6 s; Fig. 10B). After 40 min wash of SR 142801, a regular rhythm was resumed, although the oscillations remained slow (T = 3.6 s; Fig. 10C). Analogous results were obtained in three other preparations, in which rhythmic activity was slowed down by 50 ± 20%.

View larger version (69K): [in this window] [in a new window]   Fig. 10. NK3 antagonist application disrupts fictive locomotor patterns. A: a stable, fictive locomotion rhythm (T = 1.8 s) is induced by 6 µM NMDA and 10 µM 5-HT, and it is recorded from pairs of left and right L2 and L5 VRs. B: after superfusing the spinal cord with 10 µM SR142801 for 35 min (in the continuous presence of NMDA and 5-HT), the rhythm becomes irregular, with oscillations of smaller amplitude and slower period (T = 3.6 s). C: after 40 min wash of SR 142801 a regular rhythm is resumed, although the oscillations remain slow (T = 3.6 s).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The main finding of the present study was the persistent oscillatory activity induced by NK₃ receptors in the neonatal rat spinal cord in vitro. In the majority of preparations, oscillations were alternating between left and right homosegmental motor pools but synchronous between different segments on the same side. This novel pattern thus differs from the fully alternating fictive locomotor rhythm or the fully synchronous discharges evoked by block of synaptic inhibition. Despite wash out of the NK₃ agonist, delayed bursts with superimposed fast oscillations emerged in coincidence with weakening of GABAergic transmission and of recurrent IPSPs of motoneurons.

Distribution and action of NK₃ receptors in the rat spinal cord

Tachykinin receptors comprise a heterogeneous family of G-protein-coupled receptors (Regoli et al. 1994). Recent studies of NK₃ receptors have indicated their widespread distribution in central and ventral areas of the spinal gray matter (Beresford et al. 1992; Linden et al. 2000; Mileusnic et al. 1999; Seybold et al. 1997). Interestingly, ultrastructural investigations have shown NK₃ receptors of the rat spinal cord to be closely associated with glomeruli of dendritic spines, making those receptors potentially capable of influencing synaptic transmission (Zerari et al. 1997).

While the precise mechanism of action mediating NK₃ receptor activity remains unclear, it seems probable that they operate on neonatal rat spinal neurons in a fashion similar to other tachykinin receptors, namely depression of a leak K⁺ conductance (Fisher and Nistri 1993), or of a Ca²⁺-dependent K⁺ conductance (Phenna et al. 1996), or activation of nonselective cationic channels (Inoue et al. 1995). Because intracellular recording from QX-314-injected motoneurons indicated no resistance change during application of senktide, this result suggests that senktide depolarizations were generated either remotely from motoneurons or were due to concurrent activation and depression of motoneuron intrinsic conductances. As TTX eliminates motoneuron responses to NK₃ receptor agonists, the second possibility seems most unlikely (Fisher et al. 1994).

The sparse location of NK₃ receptors implies activation of interneurons widely distributed within the spinal cord, a condition that makes it difficult to unravel the mechanism of action of such receptors merely through bath application of pharmacological substances known to activate or block intracellular second messengers or ion channels. Indeed, the bath superfusion method used for NK₃ agonists in the present study included widespread distribution of these substances rather than their discrete application to selected regions. This condition is also one of the limitations of using bath-applied neurochemicals to induce fictive locomotion. Further work aimed at clarifying the precise mechanisms responsible for the action of senktide should perhaps be based on simplified network preparations like organotypic slice cultures with focal drug application.

Oscillations and bursting due to NK₃ receptor activation

Although electrophysiological studies of NK₃ receptor activity at cellular level in the spinal cord are sparse, our laboratory briefly noted that NK₃ agonists surprisingly elicited tetrodotoxin-sensitive bursting activity much more intense than that observed with activation of other tachykinin receptors (Barbieri and Nistri 2001; Fisher et al. 1994). The present study examined in more detail the rhythmic activity induced by NK₃ receptors. The NK₃ agonists senktide or [MePhe7]neurokinin B evoked a slow depolarization comprising oscillations with relatively fast period and with phase alternation at segmental level but lacking, in most cases, synchronicity between L₂ and L₅ motor pools of the same side. In a minority of preparations, oscillations displayed the fully alternating pattern typically observed during fictive locomotion evoked by excitatory agents like NMDA and 5-HT. Even if oscillations often disappeared near the end of agonist application (see, for example, Fig. 1A), the wash out phase was accompanied by the emergence of late bursts with a complex oscillatory structure.

Factors controlling oscillations and delayed bursting

The senktide-evoked responses recorded from motoneurons or VRs were network-based phenomena as they are known to be suppressed by TTX (Fisher et al. 1994) and their periodicity was independent from motoneuron membrane potential. Detailed investigation into the nature of oscillations and bursting was limited by a major experimental difficulty, namely the desensitization of NK₃ receptors that develops rapidly and prevents, for 30 min, responses to further agonist application (Barbieri and Nistri 2001). The mechanism responsible for desensitization is not fully understood but, in the case of the endogenous ligand, namely substance P, it is known to involve rapid internalization of membrane NK₁ receptors (75% of receptors can be internalized within 8 min) with consequent loss of neuronal responsiveness for extended time (Honoré et al. 1999). Return of responses to tachykinins thus implies receptor trafficking to the neuronal membrane, an inherently slow phenomenon. These findings also indicate that it was unlikely that the delayed bursting emerging after senktide wash out, was caused by differential redistribution of this agent between superficial and deep layers of the spinal cord, because virtually all cells should have been desensitized by long-lasting application of a saturating concentration of senktide. Further evidence shows that delayed bursting did not involve sustained occupation and activation of NK₃ receptors in the spinal cord. In fact, the NK₃ antagonist SR 142801 did not block late bursting, which required intact glutamatergic transmission at network level as indicated by its full suppression by CNQX and APV. Further studies should address the issue of whether -amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) or NMDA receptor activity is preferential for supporting late bursting.

The ability of the neonatal spinal cord to generate persistent bursting after senktide wash out suggests that the excitability of the spinal network remained elevated long after the agonist had been removed and when the NK₃ receptors were still desensitized. Such an up-regulation of excitability, downstream of receptor occupancy, is certainly a novel phenomenon for the neonatal rat spinal cord. Nevertheless, it has been reported that hippocampal neurons after washing out metabotropic glutamate receptor agonists (Aniksztejn et al. 1995) or muscarinic receptor agonists (Chapman and Lacaille 1999; Kouznetsova and Nistri 2000) generate sustained rhythmic discharges. It is worth noting that, in the hippocampus, persistent spontaneous discharges are typical of immature postnatal neurons as they do not appear on adult cells and are proposed to be due to a particular homeostasis of intracellular free Ca²⁺ (Aniksztejn et al. 1995).

Despite the experimental constraints of the present study, it is possible to outline some general characteristics of oscillations and late bursting. Even though oscillations could first emerge during depolarization, they were indistinguishable from those present during late bursts. Indeed, in the majority of preparations, oscillations and late bursting, despite their different periodicity, had analogous phase coupling consisting of homosegmental alternation and homolateral synchronicity between upper and lower lumbar segments. The most parsimonious explanation is that the same network was responsible for these distinct forms of neuronal rhythmicity. Curiously though, this NK₃ agonist-induced bursting is somewhat intermediate between the full alternation expressed by the locomotor network (Kiehn and Kjaerulff 1998; Nishimaru and Kudo 2000) and the complete synchronicity typical of the disinhibited pattern (Bracci et al. 1996). This realization prompted further tests to study whether synaptic inhibition was fully operative as a consequence of NK₃ receptor activity.

Inhibitory synaptic mechanisms following NK₃ receptor activity

The onset and maintenance of late bursting were particularly prominent phenomena whose magnitude might have been partly due to tissue immaturity since at a later age bursting does not persist as much as found in the present study (Barbieri and Nistri 2001). One possible explanation for this phenomenon might have been some reversible depression of synaptic inhibition especially in view of the absence of ipsilateral rhythm alternation, which requires reciprocal synaptic inhibition (Beato and Nistri 1999). Thus synchronous activity observed after senktide application was reminiscent of the effects elicited by the GABA_A antagonist bicuculline (Cowley and Schmidt 1995; Kremer and Lev-Tov 1997). In our previous scheme for locomotor network operation, we envisaged variable strength of intersegmental excitatory connections (Beato and Nistri 1999). The present data therefore raised the possibility that the strength of inhibitory synaptic connections could also be varied.

To explore potential changes in synaptic inhibition strength three approaches were adopted: 1) testing the effectiveness of exogenous GABA or glycine; 2) monitoring the DR-DRP, which is the expression of GABAergic presynaptic inhibition (Curtis et al. 1971; Levy 1977; Nicoll and Alger 1979; Nistri 1983); and 3) examining the motoneuron recurrent IPSP that is due to Renshaw cell activity and mediated by glycine for its main (and early) component (Werman et al. 1968) and by GABA for its late phase (Cullheim and Kellerth 1981; Polc and Haefely 1982).

On most preparations, bath-applied GABA or glycine reversibly suppressed bursting. This result does not rule out concomitant depression of GABAergic or glycinergic inhibition, as, at least in the case of exogenous GABA, burst suppression could have been caused by widespread activation of extrasynaptic receptors or of additional receptor classes not modulated by NK₃ receptor activity. On a few preparations bursting was not blocked by exogenous GABA (or glycine) even though these amino acids always depressed VR reflexes induced by DR stimulation. The preferential location of NK₃ receptors in relation to the pathways mediating bursting rather than those involved in VR reflexes (note that the NK₃ antagonist SR 142801 does not affect VR reflexes) (Barbieri and Nistri 2001) might account for this discrepancy.

While data with exogenously applied transmitters were not conclusive, impairment of GABAergic transmission (as indicated by reduction in DR-DRPs) appeared after strong activation of NK₃ receptors. The depression of the recurrent IPSP is consistent with this interpretation especially because a reduction in IPSP area (to which the late, GABA-mediated phase contributes) rather than peak amplitude was manifested with consequent fall in the inhibitory charge across the cell membrane. Decrease in glycine-mediated inhibition might have also contributed to this phenomenon, although it was more difficult to demonstrate it experimentally. Of course, the efficacy of synaptic inhibition at various stations within the complex polysynaptic network mediating bursting remains untested and can only be inferred from studies on recurrent IPSPs or DR-DRPs. It is worth noting that intense sensory stimulation of spinal neurons has been found to be associated with reduction in the inhibition mediated by GABA or glycine (Lin et al. 1996), lending support to the notion that the strength of synaptic inhibition can change quite rapidly.

The reason why GABAergic inhibition seemed impaired remains uncertain. NK₃ receptors are G-protein-coupled units transducing the operation of intracellular second messengers, especially those belonging to the inositol triphosphate (IP₃) cycle (Buell et al. 1992). It is plausible that in the present experiments NK₃ receptor activity could have raised intracellular IP₃ and thus internal Ca²⁺ ([Ca²⁺]_i) (Pinnock et al. 1994) high enough to impair GABA_A receptor function, which crucially depends on [Ca²⁺]_i (Inoue et al. 1986). It should be noted that the broad spectrum tachykinin agonist substance P strongly depresses GABA_A receptor function via a Ca²⁺-dependent protein kinase C mechanism (Brandon et al. 2000; Yamada and Akasu 1996). Any reduction in GABA-mediated transmission would be expected to be a transitory phenomenon receding once [Ca²⁺]_i homeostasis is reestablished. The rise in motoneuron input resistance during delayed bursting might reflect depression of tonic GABAergic transmission and might have facilitated integration of inputs into bursting signals.

It seems also likely that "physiological" activation of NK₃ receptors by endogenous ligands may not necessarily involve late weakening of GABA-mediated inhibition. In fact, on a consistent number of preparations rhythmic activity with fully alternating patterns was preserved. Furthermore, activation of NK₃ receptors must have occurred during the fully alternating rhythms of fictive locomotion evoked by NMDA and 5-HT, as this pattern was readily disrupted by the NK₃ antagonist.

Can NK₃ receptors modulate locomotor rhythms?

Senktide transformed irregular oscillations induced by NMDA and 5-HT concentrations subthreshold for fictive locomotion into fully alternating, stable motor patterns. Interestingly, during senktide wash out and in the continuous presence of NMDA plus 5-HT, bursts developed with alternating (left/right and rostral/caudal) oscillations typical of fictive locomotion. Furthermore, senktide accelerated but did not disrupt fictive locomotor patterns. These data suggest that NK₃ receptor activity could concur to activate the locomotor CPG, without making its patterns prevail over those induced by NMDA and 5-HT. This result accords with recent work on the neonatal mouse spinal cord showing that alternating motor rhythms were of larger amplitude than the nonalternating ones, indicating that the operation of the locomotor network functionally switched off other rhythmic patterns (Whelan et al. 2000). Notwithstanding the resolution of the signal transduction processes mediating the action of NK₃ receptor agonists, tachykinin receptors are known to up-regulate glutamatergic transmission (Rusin et al. 1992; for review see Urban et al. 1994). This action of NK₃ receptors on rat spinal cord neurons (Cumberbatch et al. 1995) could account for the facilitation of fictive locomotor patterns evoked by NMDA and 5-HT. An alternative explanation is that NK₃ receptor activation (or the delayed bursting consequent to it) could generate sufficient interneuronal depolarization to facilitate fictive locomotion.

The ability of the NK₃ antagonist SR 142801 to disrupt the NMDA- and 5-HT-evoked patterns suggests that during the operation of the locomotor CPG there was endogenous release of ligands acting on NK₃ receptors.