In the hindbrain of zebrafish and goldfish, reticulospinal (RS) neurons are arranged in seven segments, with segmental homologs in adjacent segments. The Mauthner cell (M-cell) in the fourth segment (r4) is known to trigger fast escape behavior. Its serial homologs, MiD2cm in r5 and MiD3cm in r6, are predicted to contribute to this behavior, which can be evoked by head-tap stimuli. However, little is known about their input–output properties. Therefore, we studied afferent projections from the auditory posterior eighth nerve (pVIIIn) and firing properties of MiD2cm and MiD3cm for comparison with the M-cell in adult goldfish. Labeling of RS neurons and the pVIIIn afferents with fluorescent tracers showed that the pVIIIn projected to r4–r6. Tone burst and electrical stimulation of the pVIIIn evoked EPSPs in the M-cell, MiD2cm, and MiD3cm. Stepwise depolarization typically elicited a single spike at the onset in the M-cell but repetitive spiking in MiD2cm and MiD3cm. This atypical property of the M-cell was mediated by dendrotoxin-I (DTX-I)-sensitive voltage-gated potassium channels together with recurrent inhibition, because combined application of DTX-I, strychnine, and bicuculline led to continuous repetitive firing in M-cells. The M-cell but not MiD2cm or MiD3cm expressed Kv1.2, a DTX-I-sensitive potassium channel subunit. Thus, the M-cell and its segmental homologs may sense common auditory information but send different outputs to the spinal circuits to control adaptive escape behavior.
- Mauthner cell
- reticulospinal neurons
- segmental homologs
- eighth nerve afferents
- recurrent inhibition
Vertebrate hindbrains are segmented structures. Seven clusters of reticulospinal (RS) neurons are periodically arranged along the neuraxis in the hindbrain of zebrafish and goldfish (Kimmel et al., 1982; Metcalfe et al., 1986; Lee and Eaton, 1991; Lee et al., 1993a) (see Fig. 1A1). RS neurons sharing a common morphology in adjacent segments are referred to as segmental homologs. The Mauthner cell (M-cell), MiD2cm, and MiD3cm are bilaterally paired RS neurons, located in r4–r6, respectively. These segmental homologs are collectively called the “Mauthner series” (M-series) (see Fig. 1A2). Each M-series neuron has two major dendrites and sends an axon to the contralateral spinal cord, with M-cells having the largest cell body. In addition to their morphological similarities, the homologous neurons share common developmental and neurochemical properties (Mendelson, 1986a,b; Hanneman et al., 1988; Hanneman and Westerfield, 1989).
The M-cell is known to play an important role in escape behaviors from predators. During escape, an abrupt stimulus to the fish leads to a fast C-shaped body bend toward the opposite side [C-start (Eaton et al., 1981)]. There is a tight link between M-cell firing and the initiation of the C-start (Zottoli, 1977; Eaton et al., 1982, 1988; Hackett and Greenfield, 1986; Oda et al., 1998). M-cells receive excitatory inputs from the auditory [posterior (p)] and vestibular (anterior) branches of the eighth nerve (pVIIIn), the posterior lateral line nerve, and the optic tectum (Faber et al., 1991). The pVIIIn afferent inputs directly excite the M-cell through mixed electrical and chemical synapses (Furshpan, 1964; Nakajima, 1974; Tuttle et al., 1986; Lin and Faber, 1988). The M-cell axon synapses on the contralateral spinal motoneurons innervating the trunk muscles (Fetcho and Faber, 1988). Thus, the principal circuit for producing fast escape is composed of pVIIIn, an M-cell, and contralateral spinal motoneurons. However, activation of the M-cell produces only the initial phase of the C-start (Nissanov et al., 1990), suggesting involvement of other RS neurons in the control of the subsequent body movement (Foreman and Eaton, 1993). Calcium imaging analysis in larval zebrafish shows that the M-cell homologs are activated by head-tap stimuli, presumably through activation of the auditory system (O'Malley et al., 1996). However, very little is known about the sensory inputs and electrophysiological properties of MiD2cm or MiD3cm.
To understand how the M-series neurons sense auditory inputs and transform them into output firing, we examined the pVIIIn afferent projection, synaptic responses evoked by auditory stimulation, and the firing patterns of these neurons in adult goldfish. Here, we demonstrate morphologically that the pVIIIn afferents project to r4–r6. All M-series neurons receive excitatory inputs from the pVIIIn afferents but exhibit different firing properties. While the M-cell shows single spiking mediated by both dendrotoxin-I (DTX-I)-sensitive potassium channels and the recurrent inhibition, MiD2cm and MiD3cm show repetitive firing in response to stepwise depolarization. These data suggest that all three M-series neurons participate in sound-induced escape behavior by carrying different aspects of the inputs to the spinal cord.
Materials and Methods
Intracellular recordings. Electrophysiological experiments were performed on adult goldfish (Carassius auratus, 7–9 cm in standard length) in air at 20°C. All procedures were performed in compliance with the guidelines stipulated by the Osaka University Committee on Animal Research. Fish were anesthetized with 0.015% 3-aminobenzoic acid ethyl ester (MS222; Sigma, St. Louis, MO), immobilized with d-tubocurarine chloride (2 μg/gm body weight; Sigma), and held horizontally in a recording chamber with tapered rods at the head and body, as shown previously (Furshpan and Furukawa, 1962). During the experiment, fish were artificially respirated by gill perfusion with aerated water containing anesthetic (0.007% MS222). The skull was opened, and the cerebellum was retracted rostrally to expose the medullar surface. Experimental set-up is illustrated in Figure 1 B. Thin bipolar tungsten electrodes, insulated except at the very tips and separated by ∼100 μm, were placed on the pVIIIn. Silver bipolar electrodes were placed just above the vertebral column for antidromic (AD) activation of the axons of RS neurons. Sound stimuli were applied from a loudspeaker positioned 60 cm to the left of the fish. Intracellular recordings were obtained from the M-series neurons with an Axoprobe-1A dual-channel preamplifier (Axon Instruments, Foster City, CA). Recording micropipettes were filled with 4 m potassium acetate (7–15 MΩ at 60 Hz) containing 3–5% neurobiotin (Vector Laboratories, Burlingame, CA). We first located the M-cell axon cap, a glial structure surrounding the axon hillock, with the help of the antidromically evoked large negative field potential (see Fig. 4 A, bottom) (Furshpan and Furukawa, 1962). The locus of the maximum field potential of M-cell spikes (>30 mV) in the axon cap was used as a landmark for finding the RS neurons in r5 and r6. RS neurons were located with the help of field potentials of AD spikes evoked by spinal cord stimulation. All of the recorded RS neurons were labeled with neurobiotin and reconstructed by camera lucida drawings for morphological identification. In this study, ∼90% of the recorded neurons were successfully labeled with neurobiotin iontophoretically injected through the recoding micropipette. Four types of RS neurons (MiD-cm, MiD-cl, MiD-i, and MiV) were identified in r5 and r6, as shown previously (Metcalfe et al., 1986; Lee et al., 1993a). Among them, the RS neurons with axons that crossed the midline and descended along with the M-cell axon in the medial longitudinal fasciculus (mlf) toward the contralateral spinal cord were identified as MiD2cm or MiD3cm. Electrophysiological data from morphologically identified MiD2cm and MiD3cm were analyzed. To assess membrane excitability, stepwise depolarizing currents were injected through the recording micropipettes, and the firing responses were observed. For the M-cell, θ-style microelectrodes (TGC200-10; Harvard Apparatus, Edenbridge, UK) were used, because very large currents were needed to fire the M-cell because of its low input resistance [70–200 kΩ (Faber and Korn, 1978)], with one channel for passing current and the other for recording the voltage responses. Firing patterns of M-cells but not fast voltage responses within an action potential were compared under several conditions, because the latter might be affected by compensation of the coupling capacitance between the two channels of the θ-style micropipette. Blockers of receptors and channels were obtained from Sigma, except 6,7-dinitroquinoxaline-2,3-dione (DNQX) and d-(–)-2-amino-5-phosphonopentanoic acid (d-APV), which were obtained from Tocris Cookson (Bristol, UK). Electrophysiological data from RS neurons with resting membrane potentials (Erest) ranging from –70 to –85 mV were analyzed. Data represent means ± SEM (number of cells). The Mann–Whitney U test was used to assess statistical significance.
Intracellular labeling of RS neurons. Recorded neurons were identified morphologically. We injected neurobiotin from the micropipette by passing iontophoretic anodal current pulses (10–20 nA) of 700 msec duration every 2 sec for 15 min. Action potentials and membrane potentials were monitored during the tracer injection. After a survival time of 1–3 hr, the fish were perfused transcardially with 0.9% saline followed by a solution of 4% paraformaldehyde (PFA) in PBS, pH 7.4. Their brains were removed from the skulls, postfixed overnight at 4°C, and then immersed overnight at 4°C in PBS containing 20% sucrose for cryoprotection. The brains were embedded in 5% agarose (type IX; Sigma) containing 20% sucrose in PBS and quickly frozen with –50°C n-hexane. Serial coronal or horizontal sections were cut at 50 μm on a cryostat (CM1850; Leica, Nussloch, Germany). Sections were mounted on gelatin-coated slides, dried for 1 hr, and then treated with a 0.3% H2O2 solution in methanol for 15 min to block endogenous peroxidase activity. After washing thoroughly, the sections were incubated with ABC solution (ABC elite kit, 1:100; Vector Laboratories) for 3 hr at room temperature. The sections were rinsed and reacted with a 0.05% 3′,3′-diaminobenzidine (DAB; Dojin, Kumamoto, Japan) solution containing 0.01% H2O2 and 0.04% nickel ammonium sulfate. The sections were then dehydrated through ethanol series, cleaned in xylene, and cover-slipped. The labeled neurons were photographed with a digital CCD camera (Axiocam HRc; Zeiss, Oberkochen, Germany), mounted on a light microscope (Axioskop; Zeiss), and reconstructed by camera lucida.
Retrograde labeling of RS neurons. Hindbrain RS neurons were retrogradely labeled to investigate their rostrocaudal arrangement in adult goldfish, as shown previously in juvenile goldfish (Lee et al., 1993a). Fish were anesthetized with MS222 and mounted on the chamber as mentioned above. The spinal cord was exposed at the cervical level, and a minimal crystal of biocytin (Sigma) was inserted into the spinal cord using a fine needle. After injection, the dorsal opening was sealed with a waterproof sheet. After a survival period of 3 d in a tank, the fish were fixed by transcardial perfusion. The subsequent sectioning, reaction, and staining procedures were the same as those used for intracellular labeling of RS neurons.
Afferent projection from pVIIIn to RS neurons. To examine the morphology of the projection of pVIIIn afferents to hindbrain RS neurons, we labeled the pVIIIn afferents and RS neurons with different fluorescent tracers. To label the RS neurons retrogradely, Oregon Green 488-conjugated dextran (molecular weight of 10,000; Molecular Probes, Eugene, OR), dissolved in a very small quantity of distilled water and then recrystallized, was injected into the spinal cord at the cervical level. Fish were maintained postoperatively in a dark tank for 3 d. Fish were then deeply anesthetized by 0.05% MS222 solution, perfused intracardially with chilled saline followed by chilled 4% PFA solution, and postfixed in the same fresh fixative at 4°C for 24 hr. Cranial nerves were removed, and the brain was circumfused with 5% gelatin (type A; Sigma) to prevent neuronal tracers from scattering. To label the pVIIIn afferents, a small crystal of 1,1-dilinoleyl-3,3,3′,3′-tetramethylindocarbocyanine (DiI; Molecular Probes) was inserted unilaterally into the peripheral part of both the saccular and lagenar nerves. The insertion site was covered with 5% gelatin, and the brains were stored in fresh fixative in darkness at 37°C for 2 weeks. The brains were embedded in 7% gelatin, fixed for 3 d, and sectioned horizontally with a Vibratome (VT1000S; Leica) at 80 μm. The serial sections were mounted on slides, observed with an epifluorescence microscope (Axioskop; Zeiss), and photographed with a digital CCD camera.
Immunohistochemistry for Kv1.2 α-subunit. Fish were perfused transcardially with chilled 0.9% saline followed by a 4% PFA solution. Serial cryostat sections (50 μm) were cut and mounted on gelatin-coated slides. The sections were blocked by incubating them in PBS containing 3% normal goat serum (NGS; Vector Laboratories) and 0.03% Triton X-100 for 1 hr. Next, polyclonal anti-rat Kv1.2 antibodies raised from rabbits (1:100; Alomone Labs, Jerusalem, Israel) were applied overnight at room temperature at a dilution of 1:100 containing 2% NGS and 0.03% Triton X-100. The sections were then incubated with a biotin-labeled secondary antibody (goat anti-rabbit, 1:200; Vector Laboratories) for 3 hr. After that, the sections were incubated in ABC solution and reacted with nickel–DAB solution. The secondary antibody was diluted in a solution containing 2% normal fish serum (Seablock; Pierce, Rockford, IL) to prevent background signal from nonspecific binding of the secondary antibody. Control sections were prepared both by omitting the primary antibodies and by preabsorbing the antibodies with the fusion proteins that were used as the antigen (Alomone Labs) at room temperature for 2 hr. A monoclonal anti-rat Kv1.2 antibody, raised from mice (1:100; Upstate Biotechnology, Lake Placid, NY), was also tested.
Segmental arrangement of RS neurons in adult goldfish
Retrograde labeling of RS neurons from the spinal cord of an adult goldfish showed that the RS neurons were clustered in seven discrete groups along the hindbrain neuraxis (Fig. 1A1), as described previously in juveniles (Lee et al., 1993a). By convention, these clusters are designated r1–r7 from rostral to caudal. The clusters appeared periodically with a spacing of ∼400 μm for r1–r5 and ∼600 μm for r5–r7. Thus, the segmental organization of the hindbrain RS neurons in goldfish is maintained in the adult, as demonstrated in zebrafish (Kimmel et al., 1982; Metcalfe et al., 1986; Lee and Eaton, 1991). The M-cell and its segmentally homologous neurons, MiD2cm and MiD3cm, are bilaterally located in the middle segments, r4–r6, respectively (Fig. 1A2).
Morphological characteristics of the Mauthner series
The morphology of each of the M-series neurons was examined by camera lucida reconstruction from serial sections containing intracellularly labeled neurons. Images of horizontal sections that contain M-series neurons in a fish are shown in Figure 2, and the camera lucida reconstruction is shown in Figure 3A. The reconstruction of coronal sections containing M-series neurons in different fish is shown in Figure 3B–D. The distinguishing features of the M-cell were the large size of the soma (Fig. 2A2, asterisks) and two thick principal dendrites, namely the lateral and ventral dendrites (Figs. 2A1–A4, 3A,B, filled and open arrowheads, respectively). The Mauthner axon (M-axon) decussated and descended in the dorsal part of the mlf (mlfD) toward the spinal cord (Fig. 2A1,A2, thick arrows). The soma of MiD2cm (Fig. 2B3, asterisk), like the M-cell, was located in the dorsal cluster of hindbrain RS neurons, ∼350 μm caudal to the M-cell axon cap (Fig. 3A, Table 1). The axon of MiD2cm crossed the midline and descended in the contralateral mlfD, along with the M-axon in the medulla oblongata (Figs. 2B1, open arrows, 3A). The MiD2cm had two principal dendrites. One extended caudolaterally and slightly dorsally, and the other extended rostroventrally (Figs. 2B2–B4, 3A,C, filled and open arrowheads, respectively). Both dendrites branched into higher-ordered thin branches and covered larger fields than those of the M-cell. In 7 of 16 cells examined, we observed a third type of dendrite. It extended ventromedially, and in three cases, it partially crossed the midline into the contralateral side (Fig. 3A,C, gray arrowheads). The soma of MiD3cm (Fig. 2C2, asterisk) was also located in the dorsal clusters, ∼650 μm caudal to MiD2cm (Fig. 3A, Table 1). The axon of the MiD3cm decussated in the hindbrain (Fig. 2C1, gray arrow) and descended along with the M-axon in the mlfD into the contralateral spinal cord (Fig. 3A). Typically, MiD3cm also had two principal dendrites, one projecting dorsolaterally and the other projecting rostroventrally (Figs. 2C1,C3,C4, 3A,D, filled and open arrowheads, respectively). Although there are differences in their higher-order and widely spread dendritic arborizations, MiD2cm and MiD3cm share the morphological motif of the M-cell: dorsally located soma, a contralaterally projecting axon descending in the mlfD, and principal lateral and ventral dendrites.
Antidromic responses of the M-series neurons
Electrical stimulation of the spinal cord elicited an AD action potential in the M-cell soma in an all-or-none manner at the threshold intensity (Fig. 4A). The onset latency ranged from 0.20 to 0.42 msec, with an average of 0.26 ± 0.01 msec (30 cells) (Fig. 4C). The amplitude of AD spikes measured from baseline in the M-cell soma was always <50 mV (Fig. 4A, top). This is attributable to a large voltage drop extracellularly within the axon cap (Fig. 4A, bottom) and the electrical inexcitability of the somadendritic membrane of the M-cell (Furshpan and Furukawa, 1962). In MiD2cm, the onset latency of AD spikes ranged from 0.42 to 0.92 msec, with an average of 0.58 ± 0.03 msec (17 cells) (Fig. 4B,C), and in MiD3cm, the onset latency of AD spikes ranged from 0.40 to 0.90 msec, with an average of 0.57 ± 0.03 msec (17 cells) (Fig. 4C). The latencies of MiD2cm or MiD3cm were significantly longer than that of the M-cell (p < 0.001), indicating slower conduction velocity of their thinner axons (∼15 vs ∼50 μm) (Fig. 3). The spike amplitudes from baseline of MiD2cm and MiD3cm appeared larger than that of the M-cell (Fig. 4A,B, top). This could be explained by the fact that we did not observe a large extracellular field potential near MiD2cm or MiD3cm (Fig. 4B, bottom), suggesting that neither MiD2cm nor MiD3cm has an axon cap-like structure.
Inputs from pVIIIn afferents to RS neurons
One of the dominant excitatory inputs to the M-cell that triggers the fast escape comes from the pVIIIn afferents. It has been established morphologically as well as electrophysiologically that the pVIIIn afferents make mixed electrical and chemical synapses on the distal lateral dendrite of the ipsilateral M-cell (Furshpan, 1964; Nakajima, 1974; Faber and Korn, 1978; Tuttle et al., 1986; Lin and Faber; 1988; Faber et al., 1991). Projection areas of the pVIIIn afferents within the brain have been described previously in goldfish (McCormick and Braford, 1994). However, it was not known whether the afferents connect to RS neurons other than M-cells. In the present study, we first examined this issue anatomically by labeling the pVIIIn afferents and the RS neurons with DiI and Oregon Green 488, respectively. Projections from the DiI-labeled pVIIIn afferents to the first-order octaval nuclei were observed as described previously (McCormick and Braford, 1994). Interestingly, the pVIIIn afferents formed four discrete bundles and extended medially to the vicinity where the lateral dendrites of dorsally located RS neurons, including the M-series neurons, are probably located (nine fish) (Fig. 5A, compare Fig. 3). The most rostral bundle appeared to contact the distal part of the M-cell lateral dendrite (Fig. 5A,B), as shown previously (Lin et al., 1983; Pereda et al., 1995). The second rostral and the fourth bundles of pVIIIn appeared to project to r5 and r6, respectively (Fig. 5A,C). The pVIIIn afferents did not enter more ventral regions where ventral RS neurons (MiV cells) were located. These results suggest that the projections of pVIIIn afferents are segmentally organized and target the dorsally located RS neurons in r4–r6.
Next, we assessed electrophysiologically whether pVIIIn forms synapses onto MiD2cm and MiD3cm as it does for the M-cell. Electrical stimulation of pVIIIn evoked initial fast EPSPs, followed by slow EPSPs in the M-cell (Fig. 6A, open and filled arrowheads, respectively) with a short latency (0.19 ± 0.00 msec; 16 cells) (Fig. 6A). It has been demonstrated that those fast and slow EPSPs are mediated by electrical and chemical synapses, respectively (Furshpan, 1964; Faber and Korn, 1978; Lin and Faber; 1988; Faber et al., 1991). Similar biphasic, fast, and slow EPSPs with constant onset latencies were observed in MiD2cm in response to pVIIIn stimulation (Fig. 6B). The latency (0.46 ± 0.03 msec; nine cells) was longer than that of the M-cell (p < 0.001). Stimulation of pVIIIn at intensities that were subthreshold to fire the M-cell evoked action potentials in seven of nine MiD2cm cells in the same fish (data not shown). Stimulation of pVIIIn also evoked EPSPs in MiD3cm (Fig. 6C) with an onset latency of 0.67 ± 0.03 msec (nine cells), also longer than that of the M-cell (p < 0.001). The EPSPs elicited in MiD3cm consisted of a fast potential (Fig. 6C, open arrowhead) with little onset fluctuation and subsequent fast but fluctuating potentials (Fig. 6C, asterisks). Most MiD3cm cells had a higher firing threshold in response to pVIIIn stimulation than that of the M-cell (eight of nine cells).
Furthermore, we explored whether natural auditory stimulation could evoke synaptic responses in M-series neurons. Tone bursts of 500 Hz with intensities of 95–100 dB were applied. The M-series neurons showed EPSPs in response to the tone bursts. Sharp depolarizing potentials in response to both condensation and rarefaction phases of sound were evoked in M-cells (Fig. 6D). The response started 0.94 ± 0.02 msec (21 cells) after the onset of sound. The sharp potentials seem to be the coupling potentials through the electrical synapses, associated with activities of presynaptic pVIIIn afferents synchronized to the sound (Fay, 1995). The sound consists of two components, particle motion and pressure change. The failure in firing M-cells might be attributable to reduction of the relative movement of otic particles in fish sustained in the air compared with that of freely swimming fish in the water (Canfield and Eaton, 1990; Casagrand et al., 1999). MiD2cm cells generated action potentials in response to the same stimuli (11 of 12 cells) (Fig. 6E1). Sharp EPSPs similar to those in the M-cell were observed when the cell was hyperpolarized (Fig. 6E2). The onset latency of MiD2cm was 1.20 ± 0.03 msec (12 cells), slightly longer than that of the M-cell (p < 0.01). Sound stimulation also evoked EPSPs in the MiD3cm (Fig. 6F). However, the onset in MiD3cm (1.46 ± 0.09 msec, 10 cells) was the most delayed among the members of the M-series neurons. These data show that MiD2cm and MiD3cm, like the M-cell, are excited by auditory inputs. However, their latencies are longer, and their firing thresholds are different.
Firing properties of the M-series neurons
Although the M-series neurons all received auditory inputs, they showed different firing properties in response to stepwise membrane depolarization. A much larger current (105.8 ± 7.09 nA; 17 cells) was needed to fire the M-cell than MiD2cm and MiD3cm (MiD2cm, 5.1 ± 0.98 nA, 10 cells; MiD3cm, 12.7 ± 3.34 nA, 6 cells). The M-cell produced only a single action potential with a stepwise depolarizing current of up to 1.5 times threshold (1.5T). The action potential of the M-cell was followed by recurrent IPSPs (Fig. 7A, asterisk), which are mediated mainly by glycinergic interneurons and in part by GABAergic interneurons in the collateral network of the M-cell (Furukawa et al., 1964; Diamond and Huxley, 1968; Triller and Korn, 1981; Lee et al., 1993b). When much larger currents (2T) (Fig. 7A) were applied, an additional spike was elicited after the recurrent IPSPs, probably as a result of postinhibitory rebound excitation. The lack of IPSPs after the second spike may be explained by the frequency-dependent depression of the collateral inhibitory network (Furukawa and Furshpan, 1963; Waldeck et al., 2000) or failure of synaptic transmission caused by the small amplitude of the second spike. In contrast to the M-cell, MiD2cm and MiD3cm exhibited tonic, repetitive firing at constant frequencies almost proportional to the amplitude of the injected current (Fig. 7B1,B2). The average rate of rise in frequency was 26.2 ± 3.03 Hz/nA (10 cells) in MiD2cm and 28.5 ± 5.33 Hz/nA (six cells) in MiD3cm. Neither MiD2cm nor MiD3cm seemed to receive recurrent inhibition, because no IPSPs were recorded after the spikes in these experiments, and depolarizing IPSPs were not observed when high Cl– was loaded intracellularly from the recording micropipette (data not shown). Together, the M-series neurons possess distinct firing properties; only the M-cell showed a propensity for generating single spikes among the M-series neurons.
Mechanisms for the transient firing property of the M-cell
The differences in the firing patterns of the M-cell and its homologs indicate heterogeneity in the composition of their voltage-gated channels or local circuits. We investigated possible mechanisms that could underlie the single-spiking nature of the M-cell. We first assessed the involvement of the DTX-sensitive potassium conductance, which has been shown to contribute to single spiking in central auditory neurons of mammals and birds (Brew and Forsythe, 1995; Rathouz and Trussell, 1998; Bal and Oertel, 2001; Dodson et al., 2002). DTX is a selective blocker of voltage-gated potassium channels composed of subunits from the Kv1 family (Hopkins et al., 1994; Harvey, 1997, 2001). During application of DTX-I (1 μm) to the brain surface, depolarization of the M-cell resulted in repetitive firing at frequencies increased with the amplitude of the injected current (Fig. 8A1 vs B1) (10 cells). The time-to-spike peak of each action potential from the onset of the stepwise depolarization in the absence or presence of DTX-I was plotted in Figure 8, A2 and B2, respectively. During DTX-I application, the M-cell showed initial spiking followed by repetitive firing after a brief pause (Fig. 8B1,B2).
The timing of this pause in firing corresponded to the duration of the recurrent IPSP (Fig. 8A3,B3), estimated from its shunting effect on the AD spike, as shown previously (Furukawa and Furshpan, 1963; Faber and Korn, 1982; Oda et al., 1995). The accuracy of this estimate is supported by the fact that the AD spike of the M-cell propagates passively into the soma and dendrites (Furshpan and Furukawa, 1962). The data suggest that the inhibitory conductance suppresses action potentials after the initial spiking of M-cell. Therefore, we examined the effect of blocking the recurrent inhibition on M-cell firing (Fig. 9). When strychnine (100 μm) and bicuculline (100 μm), antagonists of glycine and GABA receptors, respectively, were applied to the brain surface, the recurrent inhibition was abolished (Fig. 9A3), and M-cells produced a short, transient burst at the onset of the depolarization (Fig. 9A1,A2) (five cells). The burst adapted within 10 msec, and no additional spike followed. The short burst was probably a result of intrinsic membrane properties of the M-cell, because similar strong adaptation was observed when most synaptic transmission was blocked by extracellular application of strychnine, bicuculline, DNQX, d-APV, d-3-(2-carboxypiperazin-4-yl)-propyl-1-phosphonic acid, and d-tubocurarine (data not shown) (five cells). Finally, we blocked both the recurrent inhibition and the DTX-I-sensitive potassium channels by applying strychnine, bicuculline, and DTX-I. Under this condition, the M-cell generated uninterrupted repetitive firing with an initial adaptation followed by regular spiking (Fig. 9B1,B2) (10 cells). The frequency of this regular spiking increased as the cell was more depolarized (Fig. 9B2). Together, these data indicate that in the M-cell, DTX-I-sensitive potassium conductance prevents tonic repetitive firing, and that the recurrent inhibition suppresses the initial burst, permitting only a single spike at the onset of depolarization.
Expression of Kv1.2 α-subunit proteins on the M-cell
The differences in firing properties between the M-cell and its homologs, MiD2cm and MiD3cm, imply that they express different types of voltage-gated potassium channels. Because voltage-gated potassium channels containing a Kv1.2 α-subunit are sensitive to DTX-I (Hopkins et al., 1994; Harvey, 2001), we investigated the expression of the Kv1.2 α-subunit in the goldfish hindbrain with immunohistochemistry. As shown in Figure 10, the soma and dendrites of M-cells were immunoreactive to the anti-Kv1.2 antibody (Fig. 10A–F) (12 fish), although its axon was labeled only faintly. In control experiments in which the primary antibodies were omitted or preabsorbed with antigenic protein, all of the labeling was abolished. In contrast to the M-cell, dorsal RS neurons in r4–r6, including MiD2cm and MiD3cm, were not immunopositive for Kv1.2 (Fig. 10G–I). These observations suggest that voltage-gated potassium channels containing Kv1.2 α-subunits are expressed in the M-cell but not in the homologs of the M-cell. In addition, afferent fibers apposed to distal lateral dendrite of the M-cells (Fig. 10A,B, asterisks) and some axons in the mlf were also labeled (Fig. 10G–I).
We investigated how the M-series neurons obtain different electrophysiological properties and are incorporated into the hindbrain neuronal circuits in the adult teleost. The functional similarities and differences are schematically summarized in Figure 11. Although all of the M-series neurons received excitatory inputs from the pVIIIn afferents in parallel, as demonstrated both morphologically and electrophysiologically, they showed different firing patterns in response to stepwise depolarization. Both MiD2cm and MiD3cm produced tonic firings, whereas the M-cell showed typically phasic onset firing with a single spike. This unique feature of the M-cell was controlled by a combination of DTX-I-sensitive potassium conductance and recurrent inhibition. The present study has clarified the sensory afferent projections and output firing properties of the M-series, which have been suggested to work as a functional group during escape (Foreman and Eaton, 1993; O'Malley et al., 1996; Liu and Fetcho, 1999).
The M-series neurons receive inputs from pVIIIn afferents in parallel
The present study shows segmental organization of the pVIIIn afferents in the hindbrain (Fig. 5A–C), as observed previously in the anterior VIIIn afferents in frogs projecting into r5–r7 (Straka et al., 2001). It remains unclear from the morphological analysis (Fig. 5A) whether the pVIIIn afferents directly contact the MiD2cm and MiD3cm, because their lateral dendrites could not be traced, presumably because of the small diameters of their dendrites. However, the fast EPSPs with short and constant latencies (0.5 msec) elicited in the MiD2cm by electrical stimulation of the pVIIIn (Fig. 6B) as well as the fast depolarization recorded in response to tone bursts (Fig. 6E2) imply a direct, electrical connection from the pVIIIn afferents rather than indirect connections mediated by the unknown interneurons. Notably, the onset latency of the MiD2cm was longer than that of the M-cell (0.2 msec) (Fig. 6A). The longer latency of the MiD2cm indicates that it received connections from slower-conducting afferent fibers than those of the M-cell, because the conduction lengths to the M-cell and to MiD2cm from the stimulation site were approximately the same (Fig. 5A). The pVIIIn contains two types of fibers, called S1 and S2 fibers, the diameters of which are ∼15 and 5 μm (Furukawa and Ishii, 1967), respectively; S1 fibers are thought to synapse on the M-cell lateral dendrite (Furukawa and Ishii, 1967; Nakajima, 1974; Tuttle et al., 1986). Assuming that the conduction velocity of the afferent fiber is proportional to its diameter, the longer onset latency of the EPSPs evoked in the MiD2cm (0.5 vs 0.2 msec) could be caused by electrical synapses onto the MiD2cm by S2 fibers. A difference in the types of pVIIIn afferents contacting the M-cell and its homologs is also indicated from the expression pattern of Kv1.2 (Fig. 10). The Kv1.2-immunopositive fibers terminating on the M-cell distal lateral dendrite (Fig. 10A,B) are likely to be pVIIIn afferents on the basis of their location (Fig. 5A,B) (Faber and Korn, 1978; Lin et al., 1983; Pereda et al., 1995). In contrast, those projecting into r5 or r6 (Fig. 5A) did not appear to be labeled by the Kv1.2 antibody. Thus, the M-cell and its homologs may receive inputs from different types of pVIIIn afferents.
Mechanisms underlying onset, single spiking of the M-cell, and its function
There is a striking difference in the firing properties of the M-series neurons. Especially obvious is the phasic spiking of the M-cell. Some neurons in the auditory pathway of mammals and birds also generate a single spike at the onset of depolarization. Blocking low-threshold potassium channels with DTX transforms the firing pattern of these neurons from phasic to tonic (Brew and Forsythe, 1995; Rathouz and Trussell, 1998; Bal and Oertel, 2001; Dodson et al., 2002). Similarly under DTX-I application, the M-cell fired repetitively during current injection, except for a pause caused by recurrent inhibition (Fig. 8B–B3). Thus, the single onset firing observed in controls is at least partly attributable to the activation of the DTX-I-sensitive potassium channels rather than the inactivation of the sodium channels that was observed in the teleost thalamic neurons (Tsutsui and Oka, 2002). We can exclude the possibility that the repetitive firing during DTX-I application is a result of increased circuit excitability for the following reasons. First, recurrent inhibition of the M-cell appeared normal (Fig. 8B3). Second, the DTX-I-induced repetitive firing was also observed in the presence of blockers for most excitatory synapses (DNQX and APV; data not shown) or for both excitatory and inhibitory synapses (Fig. 9B1,B2). These observations indicate that the locus of the DTX-I effect is in the M-cell itself. If the DTX-I-sensitive potassium conductance in the M-cell was activated around resting membrane potentials, it could also contribute to the high threshold of the M-cell by repolarizing membrane potential and lowering input resistance.
DTX-I is a selective blocker of potassium channels that contain mammalian Kv1.1, Kv1.2, or Kv.1.6 α-subunits (Hopkins et al., 1994; Harvey, 1997). Little is known about the molecular structure and tissue expression of potassium channels in the fish brain. However, α-subunit 1 of Shaker-related trout CNS potassium channel (Tsha1) has been shown recently to be a fish homolog of rat Kv1.2 α-subunit and forms low threshold, delayed rectifier potassium channels sensitive to DTX (Nguyen and Jeserich, 1998; Nguyen et al., 2000). The C-terminal region of rat Kv1.2 used for generating the antibody shares high sequence similarity with that of Tsha1 (Nguyen et al., 2000), indicating a likelihood of cross-reactivity. In this study, we demonstrated that the M-cell was immunoreactive for an anti-rat Kv1.2 antibody (Fig. 10). These results suggest that the DTX-I-sensitive potassium channels, responsible for the phasic spiking of the M-cell, may contain Tsha1 subunits.
Single spiking of the M-cell seems sufficient or critical for initiating normal fast C-starts from the following observations. First, a single spike of an M-cell is associated with the initiation of a C-start in goldfish (Zottoli, 1977; Eaton et al., 1982) and is sufficient to induce the initial part of the behavior (Nissanov et al., 1990). Second, repetitive firing of M-cells was not observed in zebrafish exhibiting matured escape (Kimmel et al., 1974; Eaton and Farley, 1975), whereas more than one spike was observed in young embryos performing immature behavior (Eaton et al., 1977). Third, spurious fast double turns were observed in zebrafish space cadet mutants in which the M-cell excitability was assumed to be increased (Lorent et al., 2001).
Physiological importance of the segmentally homologous neurons
This study supports the idea that the M-series contributes to fish escape behaviors (Foreman and Eaton, 1993; O'Malley et al., 1996; Liu and Fetcho, 1999) evoked by auditory stimuli. In addition, our study indicates that different M-series neurons convey different aspects of sensory inputs to the spinal cord, because they show different excitabilities and postsynaptic responses to auditory stimuli. Recent morphological observations in zebrafish showed a diversity of arborization patterns for axon collaterals of the M-series neurons in the spinal cord (Gahtan and O'Malley, 2003). The M-cell has short knob-like collaterals restricted to the side ipsilateral to its stem axon, whereas MiD2cm and MiD3cm have axon collaterals with extensive branching in the spinal cord. Furthermore, the collaterals of MiD2cm terminate bilaterally. These distinctive arborization patterns suggest that they synapse on different types of neurons. These physiological and morphological differences among the M-series neurons indicate that they play different roles in initiating and controlling the escape behavior. The lower firing threshold of MiD2cm to auditory stimuli than that of the M-cell may imply that MiD2cm can transmit subthreshold sound information to preshape or modify the spinal activity before escape.
It is believed that the M-cell plays a key role in initiating fast escape responses (Zottoli, 1977; Eaton et al., 1981; Hackett and Greenfield, 1986, 1988; Nissanov et al., 1990). However, it has been demonstrated that fish still perform escape after M-cells have been abolished, although in many cases, the onset is delayed (Eaton et al., 1982; Kimmel et al., 1982; DiDomenico et al., 1988; Liu and Fetcho, 1999; Zottoli et al., 1999). This observation suggested the presence of alternative circuits to initiate the escape independent of the M-cells. MiD2cm and MiD3cm have been proposed to play a dedicated role in the alternative network (Liu and Fetcho, 1999) as well as in the intact escape network (O'Malley et al., 1996), although many other neurons might also be involved (Bosch et al., 2001; Gahtan et al., 2002). It would be a fascinating idea if the alternative network were suppressed by the M-cell-initiated escape networks, as proposed previously (Nissanov et al., 1990). If that is the case, it is possible that MiD2cm and MiD3cm contribute differently to the escape behavior in the presence and absence of the M-cells. Therefore, the next step for understanding the physiological importance of the M-series neurons in escape networks would be to study mutual connections among the M-series neurons in both intact and M-cell-ablated animals.
This work was supported by grants-in-aid for scientific research to Y.O. We are grateful to Drs. E. S. Ruthazer (Cold Spring Harbor Laboratory, Cold Spring Harbor, NY), W.-J. Song, Y. Zhu, F. Murakami (Osaka University, Osaka, Japan), and N. Yamamoto (Nippon Medical School, Tokyo, Japan) for their helpful comments on this manuscript. We also thank N. Yamamoto for helpful advice on morphological analysis and H. Matsui, Y. Murakami, and Y. Kobashi for their collaboration in the early electrophysiological study.
Correspondence should be addressed to Dr. Yoichi Oda, Neuroscience Laboratories, Graduate School of Frontier Biosciences, Osaka University at Machikaneyama 1-3, Toyonaka, Osaka 560-8531, Japan. E-mail:.
Copyright © 2004 Society for Neuroscience 0270-6474/04/243199-11$15.00/0