|Home | About | Journals | Submit | Contact Us | Français|
In the nervous system, spontaneous Ca2+ transients play important roles in many developmental processes. We previously found that altering the frequency of rhythmic spontaneous bursting episodes in embryonic chick spinal cords differentially perturbed the two main pathfinding decisions made by motoneurons, dorsal-ventral and pool-specific, depending on the sign of the frequency alteration. Here, we characterized the Ca2+ transients associated with these bursts, showing that at early stages while motoneurons are still migrating and extending axons to the base of the limb bud, they display spontaneous, highly rhythmic, and synchronized Ca2+ transients, as did some precursor cells in the ependymal layer. T-type Ca2+ channels and a persistent Na+ current were essential to initiate spontaneous bursts and associated transients, but subsequent propagation of activity throughout the cord resulted from network driven chemical transmission mediated presynaptically by Ca2+ entry through N-type Ca2+ channels and postsynaptically by acetylcholine acting on nicotinic receptors. The increased [Ca2+]i during transients depended primarily on L-type and T-type channels with a modest contribution from TRP channels and ryanodine sensitive internal stores. Significantly, the drugs used previously to produce pathfinding errors altered transient frequency but not duration or amplitude. These observations imply that the different transient frequencies may differentially modulate motoneuron pathfinding. In addition, the duration of the Ca2+ transients differed significantly between pools, potentially enabling additional distinct pool-specific downstream signaling. Many early events in spinal motor circuit formation are thus potentially sensitive to the rhythmic Ca2+ transients we have characterized and to any drugs that perturb them.
Spontaneous and relatively synchronous bursting activity occurs in many developing neural circuits (see Moody and Bosma, 2005 for review). Is such activity simply an epiphenomenon that is produced as circuits begin to assemble and become active, or does it play an essential role in circuit formation? In the visual system rhythmic waves of retinal activity are important for refining visual projections in multiple central targets (Stellwagen and Shatz, 2002; McLaughlin et al., 2003; Cang et al., 2005; Chandrasekaran et al., 2005). Spinal cords of chick and rodent embryos also exhibit rhythmic waves of propagating activity at early stages (E12.5 mouse; E3.5-5 chick) when motoneurons are still migrating and extending axons (Nishimaru et al., 1996; Milner and Landmesser, 1999; Branchereau et al., 2000; Hanson and Landmesser, 2004). However, this early activity differs from that characterized later in development (Sernagor et al., 1995). Rather than glutamatergic interneurons, the motoneurons themselves provide the main excitatory drive. This is achieved by the release of acetylcholine which activates nicotinic receptors. GABA and glycine, which are both excitatory at this stage, also contribute to the excitatory drive (Milner and Landmesser, 1999; Hanson and Landmesser, 2003; Hanson and Landmesser, 2004; Hanson and Landmesser, 2006).
In-ovo drug application was recently used to alter the frequency of this rhythmic activity during chick motor axon outgrowth. A modest decrease in frequency caused by the GABAA receptor antagonist picrotoxin, resulted in dorsal-ventral pathfinding errors and a reduction in the expression of several molecules required for this guidance decision (Hanson and Landmesser, 2004). In contrast, a modest increase in frequency caused by the glycine uptake inhibitor sarcosine, did not alter D-V pathfinding or guidance molecule expression, but completely disrupted the subsequent pool-specific pathfinding decision (Hanson and Landmesser, 2006). These results suggest that the precise frequency of the spontaneous bursts is a critical variable. Such depolarizing bursts would be expected to cause Ca2+ transients in the motoneuron somas and growth cones. However, within the spinal cord many transmitters, including acetylcholine (ACh), GABA, and glycine, will also be released during bursting episodes. These transmitters, acting via multiple receptor subtypes, could affect directly or indirectly the Ca2+ transients. Strong, spaced depolarizations can be highly effective in modulating gene activity (Dolmetsch et al., 1997; Wu et al., 2001), and the frequency, amplitude, and path of entry of the ensuing Ca2+ transients can determine which genes or signaling pathways are activated (Watt et al., 2000; Moody and Bosma, 2005; Greer and Greenberg, 2008).
We therefore characterized spontaneous Ca2+ transients in chick motoneurons during this early period under control conditions and following blockade of specific transmitter receptors, Ca2+ channel subtypes, or the release of Ca2+ from internal stores. We found that recently generated motoneurons near the ependymal layer exhibited network driven Ca2+ transients even before completing their migration to pool-specific locations. We also discovered novel roles of different Ca2+ channel subtypes and a persistent Na+ current in the actual generation of spontaneous rhythmic bursting activity. Therefore any maternally taken drugs that alter these transients could potentially affect a number of early developmental processes.
Experiments were performed on White Leghorn chick embryos at stage 25 of Hamburger and Hamilton unless indicated otherwise. The preparation was made as described previously (Milner and Landmesser, 1999). Briefly, embryos were decapitated, eviscerated and dissected in cool continuously oxygenated Tyrode's solution. A ventral laminectomy was performed to expose the spinal cord and muscle nerves and spinal nerves were exposed. After dissection, superfusion at a flow rate of 5 ml/min was performed continuously with oxygenated (95% O2 and 5% CO2) Tyrode's solution warmed to 30 ± 0.5 °C.
Extracellular recordings were made using suction electrodes pulled from polyethylene tubing (PE190). Spontaneous bursts of activity were recorded in a gap-free mode with AxoScope 9 (Axon Instruments, CA). The signals were recorded with GRASS P15 amplifiers with a bandwidth of 30 Hz - 3 kHz, digitized (DigiData 1200 series) and recorded on the computer using AxoScope 9 (Axon Instruments, CA). Cords and nerves were also stimulated as described in (Hanson and Landmesser, 2003).
Patch clamp recordings were made in standard (ruptured membrane patch) whole cell configuration at room temperature (22 - 24 °C). Pipettes (0.8 - 2 MΩ) were pulled from N51A Glass (Garner Glass, Claremont, CA), coated with Sylgard (Dow Corning, Midlan, MI), and fire-polished. Access resistance averaged 1.48 ± 0.12 MΩ (n = 4) and 1.25 ± 0.05 MΩ (n = 4), and membrane capacitance averaged 7.32 ± 1.28 pf (n = 4) and 5.3 ± 1.9 pf (n = 4) for the femorotibialis and sartorius cells respectively. Capacitative transients and series resistance were electronically compensated using an Axopatch 200B amplifier (Axon instruments, Sunnyvale, CA). Currents were activated, acquired, and leak-subtracted using a hyperpolarizing P/4 protocol under the control of pClamp 8.0 (Axon instruments, Sunnyvale, CA), running on a Pentium processor-based computer (Dell 200 Pro). Currents were four-pore Bessel-filtered at 5 kHz and digitized at 20 kHz. All values expressed are mean ± SEM.
The bathing solution was: 150 mM TEA-Cl, 5-10 mM BaCl2, 1 mM MgCl2, 10 mM glucose, 10 mM HEPES and 1 μM Tetrodotoxin, pH 7.3 (TEAOH). The pipette solution contained (in mM): 90 NMG-Cl, 30 CsCl, 30 TEA-Cl, 5 EGTA-NMG, 10 HEPES, 4 MgCl2, 4 creatine phosphate, 4 ATP-Na, leupeptine, creatine kinase, and 0.2 GTP, pH (CsOH). We used pharmacological agents and voltage-protocols to characterize the contribution of different Ca2+ channel subtypes (Rusin and Moises, 1995) by using whole cell recording of cells exposed to the following cumulative sequence of agents: 5 μM nifedipine (to block L-type), 1 μM ω-conotoxin GVIA (to block N-type), 20 nM ω-agatoxin IVA (to block P-type), 5 μM ω-conotoxin MVIIC (to block N + P/Q type), 2 μM LaCl3 (to block T-type (Mlinar and Enyeart, 1993) or 100 μM NiCl2 (to block T-R type). Although ω-conotoxin MVIIC and NiCl2 will block several types of Ca2+ channels, when applied in this cumulative sequence, sensitivity of the resistant current to these agents can be used to reveal Ca2+ channel subtypes. Furthermore we did not find any difference between the proportion of the total current carried by T-type Ca2+ channels using La3+ or Ni2+ as the antagonist (Mlinar and Enyeart, 1993). Cells were depolarized from -120 mV to +20 mV in 10 mV steps.
We utilized methods developed by O'Donovan and colleagues for Ca2+ imaging of motor circuits in isolated spinal cords of chick and mouse embryos (O'Donovan et al., 2008). Cords were prepared as described for physiological recording. After being exposed, the nerve was pulled into a suction electrode filled with calcium green-1 dextran 3,000 MW (30 mM, 5-10 μl, Molecular Probes) dissolved in the following solution (in mM: 10 NaCl, 130 K-gluconate, 10 HEPES, 1.1 EGTA, 0.1 CaCl2, 1 MgCl2, and 1 Na2ATP).
After retrograde labeling the cord was transferred to a recording chamber and continuously superfused with Tyrode's solution bubbled with 95% O2 and 5% CO2 at a flow rate of 5 ml/min. The solution was heated to 30 ± 0.5°C using an In-line Solution Heater (SH-27B, Harvard Apparatus Company, MA). In some cases we also labeled presumed precursor cells within the ependymal layer by pressure injection of Ca2+ green dextran into the central canal of isolated spinal cords in low Ca2+/high Mg2+ Tyrodes (in mM: 0.2 CaCl2 and 7 MgCl2), followed by electroporation using square voltage pulses (25 V, 50 ms duration, 5 pulses, 2 times) delivered by an ECM 830 electroporation system (BTX, Harvard Apparatus, MA).
Ca2+ transients from back-labeled cells were visualized with the Bio-Rad Radiance 2100™MP multi-photon Laser scanning microscope attached to an Olympus fixed-stage upright microscope (BX51WI, Olympus, Japan) with a 40× water immersion objective. Prior to two photon optical imaging, labeled neurons were visualized by illumination supplied by a 150 w Xenon bulb (1600 Power Supply, Opti Quip) and the focal plane for the subsequent laser-scanning was chosen. The laser was tuned to 780 nm to excite the tissue and emission was collected at 530 nm wavelength. A timed series of images from a field of view that contained labeled neurons was acquired at a 16-bit resolution with scan time between 350-520 msec per frame. To determine the position within the cord of the labeled neurons from which data were acquired, a transmission detector was applied for bright field transmission imaging allowing a transmission image to be acquired simultaneously with the confocal fluorescent images. The positions of labeled cells could be confirmed based on the merged images.
Data analysis of time-lapse images was made using MetaMorph7 (Molecular Devices, PA). Background subtraction was performed before measuring fluorescence intensity which could be in turn determined using this software. Normalized fluorescence changes were calculated as ΔF/F = (Ft − F0)/(F0), where Ft is the background corrected average fluorescence signal within the measurement box at any time point and F0 is the background corrected average intensity at the resting level (i.e. between bursting episodes). In some cases, Ca2+ transients were evoked by electrically stimulating the cord as described above. For cords treated with ω-conotoxin GVIA, Ca2+ transients were evoked by direct depolarization via bath application of 100 mM KCl for 1 min. The amplitude and duration of the evoked Ca2+ transients were then measured before and after ω-conotoxin GVIA.
St 25 embryonic chick lumbar motoneurons were retrogradely labeled with Ca2+ green dextran as described above. The cord was then fixed in 3.7% formaldehyde, frozen, sectioned and stained as described previously (Hanson and Landmesser, 2004).
All drugs were bath-applied for 15 min before assessing the effect on activity. Episode frequency was expressed as the interval between bursting episodes. At least 10 episodes of activity were measured for control and any drug treatment and expressed as the average. Each drug treatment was conducted in at least 3 spinal cords and data expressed as the mean ± S.E. To confirm that the effects were reversible, in the majority of cases, drugs were washed out until burst parameters returned to control levels.
A list of drugs used in our experiments includes the following: Ca2+ green-1 dextran 3000MW; the nicotinic cholinergic receptor blockers, mecamylamine, Dihydro-β-erythroidine hydrobromide (DHβE) and methyllycaconitine (MLA); the muscarinic cholinergic antagonist atropine; the GABAA receptor blocker picrotoxin; GABAB receptor blocker saclofen; the inhibitor of the GlyT1 glycine transporter sarcosine; the persistent Na+ current blocker riluzole; the L-type Ca2+ channel blocker nifedipine; the T-type Ca2+channel blocker nickel chloride and La3+; the P-type Ca2+ channel blocker ω-agatoxin IVA; the P/Q-type Ca2+ channel blocker ω-agatoxin TK; the N-type Ca2+ channel blocker ω-conotoxin GVIA; the N- and P/Q-type channel blocker ω-conotoxin MVIIC; the R-type Ca2+ channel blocker SNX-482; the Ca2+ internal store depleter cyclopiazonic acid (CPA); the ryanodine-sensitive Ca2+ store blocker ryanodine; the inositol (1,4,5)-trisphosphate (IP3)-sensitive Ca2+ store antagonist 2-aminoethoxydiphenyl borate (2-APB), the phospholipase C inhibitor U73122; the transient receptor potential (TRP) channel blocker SKF-96365. Ryanodine, SNX-482 and U73122 were purchased from Biomol and Ca2+ green-1 dextran (3000 MW) from Invitrogen and all other drugs from Sigma.
All the data are displayed as mean ± SE. Statistical analysis was made using one-way analysis of variance (ANOVA) or two-tailed Student's t test. A probability of <0.05 was accepted as significant.
By early St 25 (E4.5), many motoneurons in lumbar segments LS1-3 have projected axons to the crural plexus and have just made a binary dorsal-ventral pathfinding decision. Although muscles have not yet formed and motor axons are still growing toward their target regions, we were able to retrogradely label their somas by injecting Ca2+ green dextran into individual spinal nerves (Fig. 1). Retrograde labeling and Ca2+ imaging of chick motoneurons have been carried out at later developmental stages (E8-12) after they have migrated to their pool specific locations, established peripheral synapses, and acquired pool-specific bursting patterns (for review see O'Donovan et al., 2008). We have previously characterized the parameters of rhythmic spontaneous activity in much less mature cords (E4-5) by suction electrode recordings from spinal nerves or pool-specific fascicles (Milner and Landmesser, 1999). While these recordings show how the population of motor axons at the recording site are activated during spontaneous bursting episodes, they do not enable one to know how early motoneurons are activated as they withdraw from the cell cycle and migrate from the ependymal layer. They also do not reveal the patterns of activation compared to the spatial locations of the motoneuron somas or whether all motoneurons are activated during each bursting episode. In the present study we observed that large numbers of retrogradely labeled motoneuron somas could be clearly visualized along with their early processes. Since these were labeled from spinal nerves (Fig. 1), they will include motoneurons from the various pools located at that segmental level. At LS1 these would be predominantly sartorius and adductor motoneurons, while labeling from LS2 would include these plus the femorotibialis pool (Hanson and Landmesser, 2006). The cells labeled were clearly motoneurons based on their projection of axons into the limb and by their expression of the appropriate transcription factors (i.e. islet-1 for adductor motoneurons and Lim-1 for sartorius and femorotibialis motoneurons (Tsuchida et al., 1994).
Although motoneurons at this stage of development had projected to the proximal limb bud, most had not completed their migration to the lateral motor column (Figure 1) and their somas were elongated along the medial-lateral axis. Some were close to the ependymal layer (arrowheads, Fig. 1A), and many maintained central processes projecting toward or to (top arrowhead Fig. 1A) the ependymal layer. These are apparently relatively newly generated motoneurons. We have in fact observed that chick motoneurons appear to migrate by a process of somal translocation along their peripherally projecting axons (Kastanenka and Landmesser, unpublished observations). Other motoneurons had reached positions appropriate for their pool and had begun to extend dendritic processes.
At St 25, rhythmic episodes of bursting activity, occurring every 1-2 minutes, can be recorded from the motoneuron axons (Milner and Landmesser, 1999; Hanson and Landmesser, 2004), Such bursts are network driven waves of activity that can arise anywhere within the cord (but see (Momose-Sato et al., 2009) and which propagate at ~10 μm/msec. Following a burst, the network exhibits a period of depression during which bursts cannot be evoked even by exogenous stimulation (Fedirchuk et al., 1999; Hanson and Landmesser, 2003). Recovery from this depression determines the intervals between episodes and thus their frequency. To characterize the Ca2+ transients associated with such bursts, 2-photon Ca2+ imaging was performed in isolated cord preparations from groups of motoneurons within a single focal plane (Fig. 1B). Rhythmic Ca2+ transients occurred at 1-2 minute intervals (Fig. 1C), as expected from the frequency of the electrical bursting episodes that drive them. At this developmental stage each episode consists of 1 or 2 Ca2+ transients, reflecting the 1 or 2 bursts per episode. Within our fields of view, usually of a single spinal cord segment, most motoneurons were activated in a relatively synchronous manner. For example, the onset of the Ca2+ transients in cells 1-6 is shown on a faster time base in Figure 1D. Although the periods between episodes were long (1-2 minutes) the onset of the Ca2+ transients in each episode (compare with dashed line in Fig. 1D) differed by less than 500 msec, and thus can be considered to be synchronously activated.
Ca2+ transients were recorded from laterally located motoneurons that had completed their migration as well as from motoneurons located closer to the ependymal layer. Thus soon after leaving the ependymal layer and while still undergoing somal migration, motoneuron somas are subject to rhythmic Ca2+ transients. To address whether the Ca2+ transients in the migrating cells might differ in amplitude from those that had completed migration, we measured these in cells located close to or further from the ependymal layer. The mean peak intensity change (ΔF/F) was 76 ± 11% (n = 10 cells) in motoneurons within 30 - 70 μm of the ependymal layer and 73 ± 7% (n = 10 cells) for motoneurons 80 - 130 μm from the ependymal layer. There were also no statistically significant differences in duration of the Ca2+ transients from these two groups (6.2 ± 0.4 s for laterally located motoneurons versus 5.9 ± 0.2 s for medially located motoneurons, n = 10 for each group; also see as an example neurons 1 and 3 in Fig. 1B, 1D). Over the course of the experiments presented in Figures 1 and and22--55 we imaged motoneurons in all three segments that contribute to the crural plexus. Since these will include motoneurons belonging to different pools and we observed no differences in the inter-episode intervals of the spontaneous transients, we conclude that all motoneurons exhibit the same inter-episode intervals and thus frequency of Ca2+ transients.
Although the segregation of axons into pool-specific fascicles is just beginning to occur, we have previously shown that it is possible to selectively record from the adductor, sartorius, and femorotibialis fascicles (Milner and Landmesser, 1999) and that they exhibit pool-specific differences in burst duration. To assess whether individual motoneuron pools might exhibit differences in either amplitude or duration of the Ca2+ transients, we retrogradely labeled motoneurons projecting to two dorsal muscles, the sartorius and femorotibialis, or to the ventral adductor muscles by sucking these fascicles into suction electrodes containing Ca2+ green dextran. To confirm the selectivity of our retrograde labeling, cords were subsequently frozen, cut transversely, and immunostained with antibodies to the transcription factors Islet-1 and Lim-1. Motoneurons labeled from the sartorius nerve fascicle (Fig. 2A, top) all expressed high levels of Lim-1 (bottom), appropriate for this pool. Some of these neurons had migrated to the extreme lateral position of the sartorius pool (open arrow) while others (arrowheads) were still migrating. Motoneurons retrogradely labeled from both the sartorius and femorotibialis pools (LMC lateral motoneurons) were all Lim-1 positive and islet-1 negative, while the converse was true for those labeled from the obturator nerve (LMC medial motoneurons), as expected from earlier studies (Tsuchida et al., 1994; Hanson and Landmesser, 2004). There are no pool specific makers that are expressed this early in development. However the appropriate segmental location of the pools, which differs between the sartorius and femorotibialis (Hanson and Landmesser 2006) was also observed in the present study.
The spontaneous Ca2+ transients recorded from a femorotibialis motoneuron are shown in Figure 2B. Clear differences were observed for the duration of the Ca2+ transients of the sartorius, femorotibialis and adductor pools, respectively (Fig. 2C). Although the durations of the Ca2+ transients reflected the pool-specific differences in electrical burst duration, they were considerably longer. For example the duration of bursts versus transients was 500 msec versus 2.5 sec for the sartorius and 1-2 sec versus 8 sec for the femorotibialis. In summary, these observations demonstrate early pool specific differences in the duration of Ca2+ transients that could differentially affect downstream signaling pathways. However despite these differences in duration, there were no statistically significant differences in the inter-episode intervals of the Ca2+ transients from the three pools studied (121 ± 3, 119 ± 8 and 110 ± 10 s for sartorius, femorotibialis and adductor pools, respectively p > 0.05).
Given that Ca2+ transients could be recorded in apparently newly generated motoneurons close to the ependymal layer, we assessed whether presumed precursor cells within the ependymal layer might also exhibit transients by introducing calcium green dextran into these cells using electroporation. We were able to label many cells within the ependymal layer as well as more laterally located neurons (Fig. 3 top panel). Ca2+ transients (bottom panel) occurred every 1-2 minutes in ependymal cells (Trace 1) and these were synchronous with those observed in neurons that had migrated out of the ependymal layer (trace 2). These ependymal cell Ca2+ transients appeared to be dependent on the neurally generated bursts as they were blocked by nicotinic antagonists such as DHβE that blocked spontaneous motoneuron bursts (Fig. 3, bottom pair of traces).
Our previous results demonstrated that D-V and pool-specific pathfinding decisions were highly sensitive to drug induced alterations in the precise frequency of the bursting episodes. It was therefore important to determine if these drugs, sarcosine and picrotoxin, also affected the frequency of the Ca2+ transients and to assess whether they might in addition alter their duration or amplitude. Therefore 2-photon Ca2+ imaging was carried out in control and drug treated St 25 isolated cord preparations. In addition, since ACh drives the bursting episodes at these early stages and can activate a variety of nicotinic and muscarinic receptors, we also characterized the effects of cholinergic antagonists on the Ca2+ transients.
We first confirmed (Fig. 4A) that the glycine uptake inhibitor sarcosine (500 μM) decreased and the GABAA antagonist picrotoxin (50 μM) increased the interval between bursting episodes as previously reported (Hanson and Landmesser, 2004; Hanson and Landmesser, 2006). However in addition we show here that picrotoxin did not significantly alter burst amplitude, duration, or the number of bursts per episode (Fig. 4A). The GABAB antagonist saclofen also did not significantly affect any burst parameter (Fig. 4A). The α7 subunit containing nicotinic receptors have a high permeability to Ca2+ (Bertrand et al., 1993; Seguela et al., 1993) and are present in the lateral motor column (LMC) of embryonic chick spinal cords (Keiger et al., 2003). In addition, nicotinic signaling via these receptors regulates different aspects of neural development (Liu et al., 2006) including ACh mediated growth cone retraction of cultured motoneurons (Pugh and Berg, 1994). However we found that bath application of MLA, a specific antagonist of these receptors at 50 nM, had no effect on inter-episode interval, burst duration or amplitude and caused only a slight reduction in the number of bursts per episode (Fig. 4A). A broad antagonist of muscarinic receptors, atropine, was applied at 10 μM, and also had no effect on any burst parameter (Fig. 4A).
Activation of non-α7 nicotinic receptors is required for the generation of spontaneous rhythmic activity in both chick (Milner and Landmesser, 1999) and mouse embryos (Hanson and Landmesser, 2003) and such activity can be transiently blocked by the antagonist DHβE. We first confirmed (Fig. 4B) that spontaneous activity could be transiently blocked by 5 μM DHβE (n = 3). Transient blockade also occurred in cords treated with 5 μM mecamylamine (n = 3, data not shown), a non-competitive nicotinic antagonist (Papke et al., 2008) that is more selective for α3β4 subunits and at lower doses of mecamylamine (0.5 and 1 μM, n = 3 for each, data not shown) and DHβE (1 μM, n = 3, data not shown), which should be selective for α3β4 and α4β2 subunit containing receptors respectively. Thus both types of nicotinic receptors are normally required for generating spontaneous rhythmic activity at this developmental stage. However a burst could still be elicited in the presence of these blockers by initiating a propagating burst by stimulation of thoracic cord (T2) or by direct stimulation of the lumbar cord at LS3 (Fig. 4B). Thus the circuit driving bursting in the presence of non-α7 nicotinic blockers differs from that which normally drives the circuit. This fact will need to be considered when the effect of these nicotinic antagonists on Ca2+ transients or on downstream developmental processes is assessed.
Sarcosine and picrotoxin affected the frequency of the Ca2+ transients in the same way they had affected the frequency of spontaneous bursts; however neither drug affected the duration or amplitude of the transients (Fig. 4C). Transient amplitude and duration were also not affected by blocking muscarinic receptors with atropine, α7 nicotinic receptors with MLA, or GABAB receptors with saclofen (Fig. 4C). Taken together these results suggest that the effects of chronic sarcosine or picrotoxin treatment on axon pathfinding were most likely due to alterations in the frequency of spontaneous bursts/Ca2+ transients, and not to alterations in their amplitude or duration. They also suggest that alterations in the activation of GABAA receptors, glycine receptors, α7 nicotinic receptors, or muscarinic receptors probably did not contribute to the effects on pathfinding that we observed previously (Hanson and Landmesser, 2004; Hanson and Landmesser, 2006). If they did contribute, the mechanism most likely did not involve the frequency, amplitude or duration of the global somal Ca2+ transients. We also observed that 5 μM DHβE transiently blocked the spontaneous Ca2+ transients (n = 13 cells, data not shown). However, during this period of blockade, Ca2+ transients with control level amplitudes could still be evoked by stimulation of the thoracic cord.
In initial experiments, in which motoneurons were labeled from spinal nerve LS3, evoked Ca2+ transients with normal amplitude and duration could be recorded in some cells but were completely blocked by DHβE in other cells. Similar results were obtained with mecamylamine. To determine if the heterogenous responses to nicotinic blockers reflected differences in flexor versus extensor motoneurons, we next recorded from sartorius (flexor) and femorotibialis (extensor) motoneurons labeled from pool-specific fascicles. As shown in the bar graph (Fig. 4D) DHβE did not alter the amplitude or duration of the evoked Ca2+ transients in any of the motoneurons in either pool. While we do not know the identity of the cells that were blocked by nicotinic antagonists, differential sensitivity to such antagonists does not represent a general difference between flexor and extensor motoneurons. In addition we did not detect any heterogeneity in the response of cells within either pool. Additional studies will be needed to clarify the identity of the neurons, in which Ca2+ transients were blocked by non-α7 nicotinic antagonists.
Ca2+ signaling modulates many neuronal processes (Bootman et al., 2001; Berridge et al., 2003; Clapham, 2007). Both the pattern of Ca2+ transients and the mode of Ca2+ entry can affect downstream signaling pathways (Moody and Bosma, 2005; Greer and Greenberg, 2008). We therefore sought to determine the sources of Ca2+ contributing to the early motoneuron Ca2+ transients via antagonists that affected specific Ca2+ channels or Ca2+ release from internal stores. However the effect of such antagonists on the generation of spontaneous bursting activity itself had not been investigated, so this was studied first.
L-type Ca2+ channels increase in density on dissociated chick spinal cord motoneurons between E4 and E11 (McCobb et al., 1989), but it was not known if they contributed to the generation or modulation of spontaneous bursting. We found that 10 μM nifedipine, an L-type Ca2+ channel blocker, did not alter the burst frequency (p > 0.05) although it had a slight effect on burst duration and amplitude (p < 0.05, Fig. 5A). Clearly, acute blockade of L-type Ca2+ channels failed to abolish spontaneous bursts.
At mature mammalian neuromuscular junctions (Uchitel et al., 1992; Protti and Uchitel, 1993; Protti et al., 1996; Katz et al., 1997) and at synapses formed by E5-7 chick motoneurons in culture (Hata et al., 2007), P/Q-type Ca2+ channels are almost exclusively responsible for synaptic vesicle exocytosis and transmitter release. To determine whether the propagation of spontaneous bursting activity within the spinal network at much earlier stages was mediated presynaptically by one or both types of channels, two specific antagonists were bath-applied. When exposed to ω-agatoxin TK (100 nM), a specific P/Q-type Ca2+ channel blocker, the rhythmic spontaneous activity persisted without significant change in burst parameters (Fig. 5A). However bath application of ω-conotoxin GVIA (300 nM), a specific N-type Ca2+ channel blocker, abolished not only the rhythmic spontaneous activity but also the ability of electrical stimulation to evoke bursts (n = 5, Fig. 5B). Therefore, our data support the view that N-type Ca2+ channels, rather than P/Q-type, mediate the transmitter release responsible for network induced spontaneous bursting in very young chick cords.
Blockade of L-type channels with nifedipine (10 μM) caused only a slight reduction in burst amplitude and duration. However, it inhibited or reduced the accompanying Ca2+ transients in most cells, although there were a small number of cells that were only slightly affected. Since this data was derived from cells retrogradely labeled from spinal nerve LS3, the identity of the unaffected cells is unknown (Fig. 5C top). In contrast, blocking P/Q type channels with ω–agatoxin TK was without significant effect on any of the Ca2+ transient parameters (Fig. 5C bottom). To determine if any of the Ca2+ during a transient was entering via N-type channels, we directly activated motoneurons with high K+ (100 mM) to transiently depolarize them to levels required for activation of high voltage activated channels. In the presence of ω-conotoxin GVIA (300 nM), there was no significant change in mean amplitude or duration of high K+ evoked Ca2+ transients (Fig. 5C bottom).
Thus, L-type channels provide much of the intracellular Ca2+ responsible for the rhythmic spontaneous Ca2+ transients in the soma and the subsequent activation of downstream signaling pathways. In contrast, P/Q-type channels, which are expressed by motoneurons and mediate peripheral neuromuscular transmission, did not contribute. Blocking N-type Ca2+ channels prevented spontaneous bursting activity as well as the activation of the network by exogenous stimulation (Fig. 5B). This demonstrates that most of the transmitter release responsible for the network activity driving propagating bursts is due to N-type channels.
T-type Ca2+ channels are activated by relatively small depolarizations and generate low threshold Ca2+ spikes, which can in turn activate Na+-dependent action potentials and high-voltage-activated Ca2+ channels (Perez-Reyes, 2003). T-type channels can also contribute to the generation of spontaneous firing, including bursts, in central neurons (Chevalier et al., 2006; Yaari et al., 2007). T-type currents were previously shown to be present in dissociated E4.5 chick motoneurons (St 24-25), but they declined and became undetectable by E11 (McCobb et al., 1989). We found that blocking T-type channels with 100 μM NiCl2 completely abolished rhythmic spontaneous activity (Fig. 6A) in very young cords (St 24 - early St 25, n = 8). In addition NiCl2 greatly reduced the asynchronous firing of single motoneurons that occurs spontaneously between bursts (9 ± 1 unit events/minute in control versus 1 ± 1 unit events/minute in NiCl2, Fig. 6A). Since motoneurons provide the main excitatory drive to reach threshold for network burst initiation, especially at young stages (Hanson and Landmesser, 2003; see also Arai et al., 2007; O'Donovan et al., 2008), it would be expected that blockade of T-type channels would block spontaneous bursts.
However it was interesting to note that our ability to elicit bursts in lumbar motoneurons, by propagation of bursts arising from thoracic cord stimulation, was also severely impaired. Bursts were either blocked or only a small number of motoneurons were activated (Figure 6B). We also noticed that the response latency was greatly shortened. The long latency of control bursts following thoracic stimulation indicates a propagation rate of approximately 10 μm/msec, which is as least ten times slower than the estimated conduction velocity of even 1 μm unmyelinated axons. Similar rates of propagation were detected with optical imaging of propagating bursts in E8 chick cords (O'Donovan et al., 2008) and suggest that activity propagates by sequential activation of the network via chemical transmission. When this is blocked by NiCl2, the shorter latency but greatly reduced burst we observed may result from activation of descending fibers that can directly activate some motoneurons. NiCl2 also greatly reduced bursts in response to antidromic activation of motoneurons or to direct stimulation of the cord segments containing the motoneurons (Fig. 6B). NiCl2 failed to block spontaneous episodes at slightly later stages (late St 25–26), but significantly increased the inter-episode interval and greatly reduced burst amplitude and duration (Fig. 6C). Since these parameters were affected similarly in the adductor, sartorius and femorotibialis pools, the data from all has been combined in Figure 6C and expressed as a percent of the control value.
These observations suggest a critical role of T-type channels in very young spinal cord networks in providing sufficient depolarization to trigger network bursts. However, since 100 μM NiCl2 might also block some R-type channels (Tottene et al., 2000; Gasparini et al., 2001; Su et al., 2002), we treated cords with 100 nM SNX-482, a specific R-type Ca2+ channel blocker. This had no effect on spontaneous bursting (data not shown), but since SNX-482 blocks some but not all R-type channels we cannot exclude some contribution of R-type channels. Nevertheless our results suggest an essential role for NiCl2 sensitive low voltage activated channels, probably T-type, in the generation of early spontaneous network activity.
Application of 100 μM NiCl2 (Fig. 6D) blocked spontaneous Ca2+ transients as well as those in response to exogenous electrical stimulation at T4. However upon washout of NiCl2, spontaneous and evoked Ca2+ transients returned (n = 22 cells). It is likely that evoked Ca2+ transients were blocked because NiCl2 prevented normal size evoked bursts from propagating into the lumber cord. In many cases, only a small number of motoneurons were activated and the level of activation was likely to be below detection for a Ca2+ transient. Supporting this idea, we were able to detect asynchronized firing of single action potentials by motoneurons between bursts (Fig. 6A), but did not detect any associated Ca2+ transients. These observations suggest that we are unable to detect the increased Ca2+ resulting from a single motoneuron action potential.
To confirm that T-type currents made an important contribution to the overall Ca2+ current, we used patch-clamp recordings from St 27-29 dissociated femorotibilalis and sartorius neurons, identified by retrograde labeling with DI-I, to characterize the proportion of the total Ca2+ current carried by the major Ca2+ channel subtypes. Currents in response to voltage steps were recorded in cells exposed to a cumulative sequence of drugs to block the major Ca2+ channel subtypes (see Experimental Procedures for details). The recordings (Fig. 7A, B, C) revealed that the current blocked by either 100 μM Ni2+ (not shown) or 2 μM La3+ accounted for 62% of the total Ca2+ current in femorotibialis motoneurons and 49% in sartorius motoneurons. This is likely to be carried by T-type channels, although due to limitations in drug specificity we cannot exclude a contribution from R-type channels as well. In contrast L-type channels accounted for only 30% of the Ca2+ currents in both types of motoneurons, with N-type channels making even less of a contribution. We also observed that total Ca2+ current in sartorius neurons inactivated more rapidly than those in femorotibialis neurons. Thus, low voltage activated, NiCl2 sensitive currents, probably T-type, make the greatest contribution to overall Ca2+ influx at these early developmental stages.
Persistent Na+ currents (INaP) are important in modulating mammalian motoneuron bursting in mature locomotor circuits (Theiss et al., 2007), and INaP also contributes to locomotor pattern generation in isolated neonatal rat cords (Tazerart et al., 2007; Zhong et al., 2007; Tazerart et al., 2008; Ziskind-Conhaim et al., 2008). Inhibition of INaP with riluzole, also suppresses spontaneous bursting in disinhibited long term cultures from E14 rat cords (Darbon et al., 2004; Ziskind-Conhaim et al., 2008). However it was not known if this current was required for spontaneous activity in intact St 25 chick cords. We found that 10 μM riluzole completely blocked spontaneous activity in 4 St 25 cords (data not shown). In seven other cords (St 24-St 27) it did not block activity but significantly increased the inter-episode intervals (Fig.7D). Taken together, our data shows that both INaP and T-type Ca2+ channels are required for generating normal levels of spontaneous rhythmic activity during early stages when motoneurons are making important pathfinding decisions.
To determine if release of Ca2+ from internal stores affected spontaneous bursting or the associated Ca2+ transients, we treated cords with 10 μM CPA, to deplete Ca2+ from internal stores. This significantly increased the inter-episode interval and moderately decreased burst duration, but did not significantly alter burst amplitude or the number of bursts per episode (Fig. 8A). CPA also increased the intervals between Ca2+ transients and slightly reduced their amplitude (Fig. 8B). To test whether the release of Ca2+ from ryanodine sensitive stores affected bursting activity, the antagonist dantrolene (30 μM) was bath applied and was without effect on any burst parameter (n = 3, data not shown). To avoid contaminating the 2-photon perfusion system with the colored dantrolene solution, ryanodine was used instead to block Ca2+ release from ryanodine sensitive stores during the imaging of Ca2+ transients. Ryanodine (20 μM) produced a slight reduction in Ca2+ transient amplitude, similar to that observed with CPA (Fig. 8B).
We attempted to assess the importance of Ca2+ release via IP3 receptors. However, the IP3-sensitive store inhibitor 2-APB also blocks gap junctions (Juszczak and Swiergiel, 2009) which are required for spontaneous bursting (Milner and Landmesser, 1999). Since more selective IP3 receptor blockers that would be effective in intact cords were unavailable, we applied 10 μM U73122, an inhibitor of phospholipase C. This drug should inhibit production of IP3 and thus IP3-mediated release of Ca2+ from internal stores. However it had no effect on burst parameters (Fig. 8A) or on the amplitude or duration of Ca2+ transients (Fig. 8B). Taken together, our results show that release of Ca2+ from internal stores makes only a modest contribution to the spontaneous Ca2+ transients in chick motoneurons at these early developmental stages. The effects that were observed appeared to be mediated by ryanodine receptor rather than IP3 receptor mediated release.
Since TRP channels are present on young Xenopus spinal neurons and are required to mediate growth cone turning toward chemoattractants in culture (Wang and Poo, 2005; Li et al., 2005b; Jacques-Fricke et al., 2006), we assessed whether relatively selective TRP channel blockers would affect spontaneous bursts and the associated Ca2+ transients. SKF-96365 (10 μM), a blocker of a subset of TRP channels (Pena and Ordaz, 2008), tended to increase the intervals between bursting episodes and Ca2+ transients but not significantly (Fig. 8C). However, SKF-96365 significantly reduced the amplitude and the duration of the Ca2+ transients by approximately 20% (Fig. 8D). Although the identification of the channel subtypes is unknown, our results suggest that TRP channels contribute to the excitation that drives spontaneous bursting activity in young spinal cords and that they make a modest contribution to Ca2+ transients.
We found that the rhythmic spontaneous bursting episodes described previously (Milner and Landmesser, 1999), evoke Ca2+ transients in very young lumbar motoneurons, when their somas are still close to the ependymal layer and while they are making pathfinding decisions at the base of the limb. Signaling from these Ca2+ transients could thus potentially affect many early events in spinal motor circuit development, including pathfinding, as implied by our earlier studies (Hanson and Landmesser, 2004; Hanson and Landmesser, 2006).
Spontaneous bursting activity and the associated Ca2+ transients, are widespread in the developing nervous system. However the roles of such activity/transients, especially at early stages of circuit formation, are only beginning to be understood. It is important to recognize that these roles, and indeed the underlying mechanisms by which such activity is generated, may also differ between species and at different developmental stages (Spitzer, 2008; Root et al., 2008). Prior to network generated activity, different subtypes of Xenopus spinal neurons exhibit different Ca2+ transient frequencies and these are required to appropriately specify neurotransmitter phenotypes (Borodinsky et al., 2004). The present study showed that the propagating waves of activity in St 25 chick cords produce relatively synchronous Ca2+ transients in all motoneurons. Thus different motoneuron subtypes would be unable to use Ca2+ transient frequency differences to drive subtype-specific differences in downstream signaling. However our previous studies clearly showed that different pathfinding decisions were highly sensitive to modest alterations in the normal frequency, emphasizing the importance of maintaining the normal in-vivo frequency. Other aspects of early cord circuit formation may also be sensitive to modest alterations in frequency.
Given the strong homeostatic changes that occur in developing cord circuits when activity is perturbed, (Chub and O'Donovan, 1999; Milner and Landmesser, 19999; Gonzalez-Islas and Wenner, 2006; Wilhelm et al., 2009), it seems unlikely that complex circuits could form normally in the absence of activity. Birds and placental mammals undergo long periods of development (approximately 3 weeks for chicks and mice, much longer for primates) in the absence of extrinsic sensory stimuli that could activate developing spinal circuits. Thus spontaneous rhythmic activity may have evolved to supply such activity during this period. Different signaling systems regulating downstream developmental processes may have evolved to be tuned to the normal frequency of this activity. However, despite near synchronous activation on the minute time scale, we observed that the waves and associated Ca2+ transients propagate at speeds of 10 μm/msec. Thus near neighbors will be more synchronously activated than cells that are further apart and, as in the visual system, these differences could be used to drive spike timing dependent refinement of connections. The window for spike timing plasticity in the intact Xenopus tectum (Zhang et al., 1998) is very narrow (approximately 20 msec). Further studies will be needed to see if such mechanisms are used in developing spinal circuits and if they make use of information encoded in the spontaneous waves.
The frequency and pattern of Ca2+ transients have been shown to activate different patterns of gene transcription (Dolmetsch et al., 1998) and intracellular signaling pathways (Moody and Bosma, 2005; Greer and Greenberg, 2008). In cultured Xenopus spinal neurons GAD 67 expression and thus the GABAergic phenotype is regulated by the frequency of Ca2+ transients (Watt et al., 2000). Altering activity in intact spinal cords changed the proportion of neurons that expressed either excitatory or inhibitory neurotransmitters (Spitzer et al., 2004; Root et al., 2008). It will be important to now characterize the intracellular pathways by which alterations in the frequency of Ca2+ transients affected distinct D-V and pool-specific pathfinding decisions. To exclude alterations in GABAergic and glycinergic signaling in contributing to the pathfinding errors, we are using channelrhodopsin 2 (Li et al., 2005a) to alter frequency, independent of drugs that affect neurotransmitters (K. Kastanenka, unpublished observations). Although the Ca2+ transients we observed occurred relatively synchronously, they did exhibit pool-specific differences in duration. These differences could be used by downstream effectors to decode information and to generate different pool specific responses, as was previously shown in lymphocytes (Dolmetsch et al., 1997).
Another interesting finding was that Ca2+ transients occurred in presumed precursor cells in the ependymal layer/germinal epithelium. These were blocked by nicotinic antagonists that blocked the neurally generated activity. It will be important to determine the identity of these ependymal cells and how neural network activity causes Ca2+ transients. It will also be interesting to determine if blocking the transients has any effect on the proliferation or differentiation of cells that generate different subclasses of motoneurons and interneurons. Propagating Ca2+ transients have been observed in radial glial cells in the neocortex where they regulate cell proliferation (Weissman et al., 2004) and in the ventricular zone of the developing rabbit retina, where they were also dependent on neurally generated waves (Syed et al., 2004).
Neurotransmitters released during propagating bursts could also affect Ca2+ transients. As noted above, blocking GABAA receptors did not affect transient duration or amplitude and blockade of metabotropic GABAB receptors was without effect on any transient parameter. Nicotinic transmission and signaling via a variety of receptor subtypes, including α7 and α4β2, play important roles in neural development (Pugh and Berg, 1994; McLaughlin et al., 2003; Myers et al., 2005; Liu et al., 2006). However, while α7 receptors have a high Ca2+ permeability and can mediate transmission in ciliary neurons (Zhang et al., 1996; Ullian et al., 1997; Sargent, 2009), we found that blocking them had little effect on bursting episodes or Ca2+ transients. In contrast, blocking either α4β2 receptors with DHβE or α3β4 receptors with mecamylamine both transiently blocked spontaneous bursting. However these drugs did not affect the amplitude or duration of Ca2+ transients evoked in sartorius or femorotibialis neurons during nicotinic blockade. Since we were able to elicit propagating bursts via electrical stimulation during the period of blockade with either antagonist, these receptors also are not essential for bursts to propagate. However, ACh acting on them is what normally drives bursting activity. One possibility is that the waves evoked during blockade with these antagonists are driven by other neurotransmitters (Milner and Landmesser, 1999); alternatively ACh released by exogenous stimulation may act on other nicotininc receptor subtypes.
A major contribution of the present study was to characterize the sources of the Ca2+ that mediate bursts and their associated Ca2+ transients. These and the relationship between different currents, propagating bursts, and the Ca2+ transients are summarized in Fig. 9. A NiCl2 sensitive low voltage activated Ca2+ current, probably T-type, as well as a persistent Na+ current was essential for the generation of spontaneous activity at these early stages. The main contribution to intracellular Ca2+ associated with the bursts was Ca2+ influx via T-type and L-type channels rather than N-type or P/Q-type. While blocking P/Q or L-type channels had negligible effects on bursting activity, blocking N-type channels completely prevented bursting activity even with exogenous stimulation. This appeared to be because N-type channels mediate the release of the neurotransmitters required to initiate network driven bursts. This observation emphasizes the importance of chemical transmission in this process. Finally we found that TRP-like channels and ryanodine-sensitive Ca2+ stores also made a modest contribution to the Ca2+ transients.
Voltage-dependent Ca2+ channels were the main contributors to the Ca2+ transients that we observed, with L-type channels making a major contribution, as shown previously at later stages in both rat (Scamps et al., 1998; Dayanithi et al., 2006) and chick (McCobb et al., 1989) dissociated motoneurons. Although L-type channels were not essential for spontaneous bursting they may mediate downstream signaling events. We found that release from internal Ca2+ stores made only a modest contribution to the transients, which differs from the Ca2+ transients in Xenopus cords at earlier developmental stages. The release we observed was via ryanodine sensitive receptors similar to observations made in older rat motoneurons (Dayanithi et al., 2006).
T-type currents were previously shown to be important for the generation of spontaneous Ca2+ transients in Xenopus motoneurons (Gu and Spitzer, 1993; Gu et al., 1994). However such currents had not been shown to be essential for propagating wave-like activity in either the retina or spinal cord. We found that a NiCl2 sensitive current, probably T-type, was essential not only for the generation of spontaneous bursting activity, but also for the propagating spread of excitation via network mediated neurotransmission. Given that waves of activity at this stage appear to originate predominately in the upper cervical cord (Momose-Sato et al., 2009), excitation mediated by T-type channels would be essential for the normal bursting activity and Ca2+ transients experienced by lumbar motoneurons in-ovo. A second novel observation was that a riluzole sensitive INaP was also essential for the generation of normal levels of spontaneous activity at these early stages. Although known to be important for the activation of locomotor patterns in neonatal mammalian cords in response to neuromoduators (Tazerart et al., 2007; Zhong et al., 2007; Ziskind-Conhaim et al., 2008), INaP was not previously implicated in the generation of spontaneous rhythmic activity in young cord circuits.
Taken together our findings offer insights into the features of spinal motor circuits that mediate rhythmic spontaneous bursting activity and Ca2+ transients in very young chick spinal cords. It will now be important to determine how the precise frequency of Ca2+ transients is read out to influence both the dorsal-ventral and pool specific pathfinding decisions made by embryonic motoneurons and how early other spinal neuron subtypes (Goulding and Pfaff, 2005) exhibit Ca2+ transients and thus come under their potential influence.
We thank Maryanne Pendergast and Ben Strowbridge for their help with the 2-photon confocal imaging and gratefully acknowledge the use of the Neurosciences Imaging Center. We especially thank Michael O'Donovan and George Mentis for their initial instruction in 2-photon Ca2+ imaging in isolated chick cords and M. Gartz Hanson for participating in these experiments. We also thank Diana Kunze, William Schilling, Yuka Maeno-Hikichi, Ksenia Kastanenka, and Stefan Herlitze for their comments on the manuscript. Supported by NIH Grant NS19640 from NINDS