Time disparity sensitive behavior and its neural substrates of a pulse-type gymnotiform electric fish, Brachyhypopomus gauderio
- First Online:
- Cite this article as:
- Matsushita, A., Pyon, G. & Kawasaki, M. J Comp Physiol A (2013) 199: 583. doi:10.1007/s00359-012-0784-4
- 326 Views
Roles of the time coding electrosensory system in the novelty responses of a pulse-type gymnotiform electric fish, Brachyhypopomus, were examined behaviorally, physiologically, and anatomically. Brachyhypopomus responded with the novelty responses to small changes (100 μs) in time difference between electrosensory stimulus pulses applied to different parts of the body, as long as these pulses were given within a time period of ~500 μs. Physiological recording revealed neurons in the hindbrain and midbrain that fire action potentials time-locked to stimulus pulses with short latency (500–900 μs). These time-locked neurons, along with other types of neurons, were labeled with intracellular and extracellular marker injection techniques. Light and electron microscopy of the labeled materials revealed neural connectivity within the time coding system. Two types of time-locked neurons, the pear-shaped cells and the large cells converge onto the small cells in a hypertrophied structure, the mesencephalic magnocellular nucleus. The small cells receive a calyx synapse from a large cell at their somata and an input from a pear-shaped cell at the tip of their dendrites via synaptic islands. The small cells project to the torus semicircularis. We hypothesized that the time-locked neural signals conveyed by the pear-shaped cells and the large cells are decoded by the small cells for detection of time shifts occurring across body areas.
KeywordsElectrosensory systemPhase comparisonPhase-locked neuronsTemporal codingTime coding
Electrosensory lateral line lobe
Electron microscopy or electron micrograph
Electric organ discharge
Light microscopy or light micrograph
Mesencephalic magnocellular nucleus
Phosphate buffer saline
Time and other features of sensory signals are often processed by separate parallel sensory subsystems, e.g., the magnocellular and parvocellular visual pathways of the primates (Livingstone and Hubel 1988) and independent auditory pathways in the owl (Takahashi et al. 1984). Clear separation of pathways for coding of time and amplitude information is well known in the electrosensory system of weakly electric fishes (Heiligenberg 1991; Kawasaki and Guo 1998). In the time coding pathway, each electric organ discharge (EOD) of a fish’s own or of neighboring individuals’ triggers a single action potential in the electroreceptors in the skin. These ‘time-locked’ action potentials are propagated to a central mechanism via fast conducting axons and synapses with microsecond accuracy, representing local stimulus times that carry behaviorally relevant temporal information (Carr et al. 1986a; Guo and Kawasaki 1997). The time differences between signals from different electroreceptor locations are known to be utilized for important behavioral functions in majority of electric fish species (Kawasaki 2009). The time coding pathways are known to serve species and sex recognition in pulse-type mormyrid electric fishes (Hopkins and Bass 1981; Xu-Friedman and Hopkins 1999), electrolocation and the jamming avoidance responses in wave-type gymnotiform electric fishes and in Gymnarchus (Heiligenberg and Bastian 1980; Carr et al. 1986a; Kawasaki 1993). Behavioral roles of the time coding pathway in pulse-type gymnotiform electric fishes, however, are poorly understood—time information, which is essential for the jamming avoidance responses of wave-type fishes is not required for the jamming avoidance responses of pulse-type gymnotiform fishes (Baker 1980; Heiligenberg 1980). No behavior of pulse-type gymnotiform fishes has been shown to be driven solely by the time coding pathway (Heiligenberg and Altes 1978).
In the present study, we first tested whether a pulse-type gymnotiform fish, Brachyhypopomus, responds to microsecond time differences between sensory signals from different body areas. The novelty response, which is a temporal rise in frequency of the EODs in response to novel sensory stimuli, was observed as an indicator of the fish’s ability to detect time differences. Secondly, we physiologically recorded from neurons and anatomically labeled them in the mesencephalic magnocellular nucleus (MMN), an unpaired hypertrophied midbrain structure unique to pulse-type gymnotiform fish. The MMN consists of large neurons, thick axons, and smaller neurons (Réthelyi and Szabo 1973a) that are reminiscent of the neurons found in the time coding pathways in other weakly electric fishes (Carr et al. 1986b; Castelló et al. 1998; Friedman and Hopkins 1998; Matsushita and Kawasaki 2004). One of the large neurons with thick axons is projected from the electrosensory lateral line lobe (ELL) in the hindbrain, called pear-shaped cell after the nomenclature of Réthelyi and Szabo (1973a). Light and electron microscopic observation of labeled materials elucidated the connectivity of these neurons implicating the MMN’s function as a time comparator.
Materials and methods
We used approximately 40 Brachyhypopomus gauderio (formerly Brachyhypopomus pinnicaudatus) (11–17 cm in total body length). They were bred and maintained in the laboratory at 25–28 °C, 12:12 L:D cycle, and water conductivity at 50–100 μS/cm. Water conductivity for the behavioral and physiological experiments was between 60 and 80 μS/cm.
Novelty responses in the isolation chamber
We first anesthetized fish with MS-222 (1:10000) and then curarized them by intramuscular injection of gallamine triethiodide (10–15 μg). They were fitted in a plastic isolation chamber (10 × 20 × 6 cm) in which the head and trunk regions of the fish were electrically isolated and given independent signals (Heiligenberg and Bastian 1980; Carlson and Kawasaki 2006). This isolation was achieved by a pair of ‘sliding doors’ positioned at the base of the pectoral fins (approximately 15 % of the total body length) and sealed with petroleum jelly (vaseline). Electrosensory stimulation signals for the head and trunk compartments were generated by a PC computer. These signals were isolated from the ground by field effect transistor-based isolation circuits. Positive outputs from the two isolators were tied and connected to a gold-plated electrode (0.8 mm diameter, World Precision Instruments, 5482) which was inserted into the pharynx along with a water breathing tube. The negative output of the isolators was connected to a gold-plated wire electrode at the bottom of each compartment. Signal isolation between the two compartments was approximately −40 dB.
Simultaneous pulses at 25 Hz (see below for the waveform) were given to the head and trunk compartments and maintained throughout experiments. We call this as the background stimulus. We replaced this constant background stimulus with a test stimulus for novelty responses for 1 s. The pulse times of the test stimulus were delayed or advanced by a constant magnitude (typically 100 μs) in reference to otherwise constant pulses (Fig. 1b). Novelty responses were quantified as the mean increase of frequency of the pacemaker command pulses during the test stimulus period with respect to the mean frequency during the 2-s period before the onset of the test stimulus (Fig. 1a). The 1-s long test stimuli (25 pulses) were separated by 60 s and semi-randomly presented over a few hours to minimize the effects of long-term habituation. Curarized Brachyhypopomus showed stereotypical frequency rises that reached 30–70 Hz over several seconds. These frequency rises were apparently spontaneous and had no clear relation to stimulus presentation. Thus, pacemaker frequency data that contained these large spontaneous rises were eliminated from the data analysis.
The background and test stimulus pulses were numerically generated by a custom written software (MATLAB, Mathworks) and transferred to sequential memory buffers in an arbitrary function generator (1 μs resolution; Agilent 33522A). The background-test stimulus sequences were read out seamlessly without any noise generated at their boundaries.
Transdermal potential in response to the fish’s own EODs was measured for calibration of stimulus amplitude for behavioral and physiological experiments. Transdermal potential on the gill cover showed inside positive monophasic pulse which corresponded in time to the first head positive phase of EOD pulses recorded between the head and tail. The amplitude was 11.5 ± 1.08 mV (data from three males and two females, 120–160 mm long). At anterior body locations between the tip of the snout and one-third of the body length, the waveform remained monophasic (second peak less than 10 % of peak-to-peak amplitude) and the amplitudes were constant (standard error 0.3 mV). At the midpoint of the fish length, the second negative peak appeared as strong as in the head–tail recorded EODs (Hopkins et al. 1990). Since pulse marker electroreceptors appear on the anterior 25 % of the body length (Yager and Hopkins 1993), we sampled the monophasic waveform of the natural EOD pulse at the gill cover and used it in the computer-generated stimulus pulses (Fig. 5 bottom). The amplitude of the computer-generated pulses was set to generate a transdermal voltage of 11.5 mV at the gill cover in the curarized preparation.
Electrophysiology and anatomy
Field and intracellular potentials from hindbrain and midbrain were recorded with an amplifier (Getting 5A) and standard electrophysiology equipment. Glass-capillary electrodes with tip diameter 10 μm were used for field potential recordings. For simultaneous intracellular recording and labeling, we inserted sharp glass-capillary electrodes filled with 2 % biocytin (Sigma B4261) or 2 % neurobiotin (Vector Laboratories SP-1120) in 1 M KCl into the MMN. The tracers were intracellularly iontophoresed at +0.1 nA for ~10 min. Extracellular injection and simultaneous field potential recording were performed with glass-capillary electrodes with the tip trimmed to 20 μm filled with 2 % biocytin or 2–4 % neurobiotin in 1 M KCl. The tracer was iontophoresed at +2 to 4 μA for 10–50 min. The survival time for intracellular and extracellular injections was, respectively, 1 h and 4–6 h. Fish were transcardially perfused first with saline and then fixative (2 % paraformaldehyde (PFA) and 2 % glutaraldehyde (GA)) in 0.1 M phosphate buffer (PB, pH = 7.4). Brain was removed and subjected to post fixation in the same fixative at 4 °C overnight. Transverse sections at 50 μm were sliced with a vibratome (Leica model VT1000S). Floating sections were first rinsed in 0.02 M PB with 0.9 % NaCl (PBS), treated with 0.5 % hydrogen peroxide (H2O2) in PBS, and soaked in PBS containing avidin–biotin complex (ABC) (Vector Laboratories PK-6100) and 0.3 % Triton X-100 for 12–20 h at room temperature. After being rinsed with 0.02 M PB, the sections were presoaked in 0.1 M PB containing 0.05 % diaminobenzidine (DAB) and 0.04 % nickel ammonium sulfate for 10 min, and then reacted with 0.002 % H2O2 for 2 min.
For LM, slide-mounted sections were counterstained with neutral red, dehydrated with an ethanol series, cleared with HemoDe (Fisher Scientific), and sealed with Permount (Fisher). The sections were observed with a light microscope (BX-51 or BH-2), equipped with a DP71 camera system. Composite photomicrographs were created with Adobe Photoshop CS3 extended and a focus-composition software (e-Tiling, Mitani Corporation) was used for z axis projection imaging.
For EM, 0.05 % TritonX-100 was used for the ABC reaction described above. The DAB reacted sections were sandwiched with filter paper (Whatman No. 54), rinsed in 0.1 M PB, and then soaked in 1 % OsO4 (osmium tetroxide) in 0.1 M PB for 1 h. Then the sections were dehydrated through an ethanol series, infiltrated with propylene oxide, and flat-embedded in Quetol 812 between two sheets of Aclar film. Ultrathin sections, double stained with uranyl acetate and lead citrate, were examined with a transmission electron microscope (Hitachi, H7650).
Since the ultrastructure of labeled preparation was partially damaged by Triton X detergent, we also prepared samples for conventional electron microscopy. We followed the protocol used in Matsushita and Kawasaki (2004). Briefly, after anesthetization, fish was perfused with 2 % PFA and 2 % GA in 0.1 M PB, and the brain was subsequently postfixed in the same fixative for 4–16 h. Fifty micrometer sections of the midbrain were sandwiched with filter paper, rinsed, and postfixed with 1 % OsO4 in 0.1 M PB for 1 h. The sections were then processed as above and embedded in Quetol 812.
Neutral red stained transverse sections (50–100 μm thick) were observed also in specimens of other pulse-type gymnotiforms—Hypopygus, Microsternarchus, Hypopomus, Steatogenys, Rhamphichthys, Gymnorhamphichthys, Gymnotus, and Electrophorus. These materials were prepared by late Walter Heiligenberg at University of California at San Diego.
Novelty responses to pulse time differences between the head and trunk
While these results indicate the sensitivity to time differences between the head and trunk stimulus pulses, a possibility remains that the novelty responses could be a result of small amplitude reduction due to the signal crosstalk between the experimental compartments—numerical simulation indicated that the peak amplitude of the pulses are reduced by ~5 % when two pulses with a 100 μs time difference interacted with 100 % crosstalk (i.e., averaging of the two time-shifted pulses). The following control experiments examined if observed novelty responses were due to amplitude modulation resulting from the signal crosstalk. During the background stimulus period, the head and trunk stimuli were delayed and advanced, respectively by 50 μs. Conversely, the head and trunk stimuli were advanced and delayed, respectively by 50 μs during the test stimulus period. These stimulus pulses, if interacted through crosstalk, would not result in amplitude modulation at the swap of the relative delay–advance relations. Good novelty responses still occurred to this stimulus arrangement (Fig. 2h). Another control experiment against amplitude-caused novelty responses was performed with pulse amplitude eight times larger than natural EOD feedback signal. The novelty responses still occurred to 100 μs time shifts between the head and trunk signals (1.6 ± 0.3 Hz). At this high amplitude, the amplitude coding electroreceptor afferent neurons, the burst duration coders (Bastian 1976), saturated their number of output action potentials, whereas the time coding afferent, pulse marker continued to fire an action potential to a stimulus pulse normally.
Electrophysiological responses of the time coding system
By intracellular recording and following labeling, morphologically two types of time-locked neurons were strongly labeled: pear-shaped cells whose somata are in the electrosensory lateral line lobe (ELL) and large cells intrinsic to the MMN. Another type of neuron, small cells, is retrogradely labeled by extracellular injection into the torus. These are identified in unlabeled materials in a gymnotiform pulse fish Gymnotus carapo by Réthelyi and Szabo (1973a). We adopt their nomenclature in the following description.
After passing the decussation to the contralateral sides, the axons of the pair-shaped cells run anteriorly to the ventral margin of the torus semicircularis keeping a thick and dense bundle (~80 μm in diameter) (Fig. 7g). There, the bundle turns dorsally and spread into multiple bundles each of which contains approximately 20 axons (Fig. 7h). These thinner bundles penetrate the torus semicircularis dorsally and reach the MMN (Fig. 7i). The axon terminals of the pear-shaped cells fill the entire extent of the MMN with a dense matrix of fibers (Fig. 7i). The axon diameter (~3.5 μm) remained constant from the level of the ELL through the decussation to the MMN. The number of pear-shaped cells, approximately 1000, was estimated by counting the thick axons at the decussation in a sagittal semi-thin section. In a few cases, when unilateral injection of biocytin was made in one side of the MMN, strongly labeled somata of the pear-shaped cells were found exclusively in the contralateral side of the ELL (e.g., Fig. 7i). No intracellularly labeled axons of the pear-shaped cells were found to bifurcate along their passage to the midbrain. Thus, we conclude that the axons of the pear-shaped cells decussate completely with no ipsilateral projection.
Figure 9 shows morphology of the large cells. Somata of the large cells were found in the ventral two-third of the MMN and had diameter of ~20 μm (Fig. 9a). Approximately 300 large cells were estimated in the MMN based on soma counts in neutral red stained series sections. The somata have spherical shape with slight elongation toward the initial segment. The cell is characterized by its typical round nucleus and a nucleolus (Fig. 9b). Conventional EM of the large cells revealed that the soma was myelinated except for the surface where club endings were attached (Fig. 9b, e). The contacts are mostly located at the opposite side of the initial segment. The synaptic interfaces at the club ending are alternately arranged gap junctions and attachment plaques (Fig. 9c; Sotelo et al., 1975). Figure 9d1 and d2 shows Camera Lucida drawings of intracellularly labeled large cells. The proximal axon has a diameter of ~5 μm and extends over hundreds of microns without major branching (Fig. 9f). The distal axons repeatedly ramify to thinner collaterals that extend in all three dimensions within the MMN. While the large cells shown in Fig. 9d1 and 9d2 had one axon coming from the soma, other cells were often observed to possess more than one (Fig. 9e). The terminals of the collaterals are calyx endings with two to three fingers embracing the soma of a smaller neuron (Fig. 9g, h). EM observation of a serially cut calyx ending showed that the soma embraced by the calyx has a dendrite (Fig. 9i).
The MMN in other pulse-type gymnotiforms
The MMN, a large unpaired structure above the torus semicircularis that contained large somata and thick fibers, could be recognized in neutral red stained sections from the materials of genera Hypopygus, Microsternarchus, Hypopomus, Steatogenys, Rhamphichthys, Gymnorhamphichthys, Gymnotus, and Electrophorus.
We found in this study: (1) Brachyhypopomus exhibits a behavioral response to microsecond temporal disparities between sensory stimulus pulses in different body areas, (2) the occurrence times of stimulus pulses at the periphery are carried by a hypertrophied fast electrosensory pathway to the midbrain as the form of time-locked action potentials, (3) the fast electrosensory pathway does not adapt to high-frequency stimulation, and (4) the small cells in the MMN are the most likely convergent site for detection of the time disparities because they receive both inputs from the time-locked components of the fast electrosensory pathway, the pear-shaped cells and the large cells.
Behavioral sensitivity to time disparities
Behavioral roles of temporal signals are well established in pulse-type mormyrid fishes for their sex and species recognition (Hopkins 1986) and in the jamming avoidance responses of a gymnotiform wave-type fish, Eigenmannia (Heiligenberg and Bastian 1980; Heiligenberg 1991) and a mormyriform wave-type fish, Gymnarchus (Kawasaki 1993). No behavioral responses of pulse-type gymnotiform fishes, however, have been known to require temporal disparities between electrosensory signals from different body areas (Heiligenberg et al. 1978; Heiligenberg 1980), though joint changes in temporal and amplitude signals may evoke behavioral responses (Heiligenberg and Altes 1978; von der Emde 1999). The results shown in Fig. 2 are the first demonstration that microsecond changes in stimulus pulse times are detected as increased time disparities between signals from different body areas in a pulse-type gymnotiform fish (Fig. 2a–d). No novelty responses occurred when identical time shifts occurred in the stimulus pulses in the entire body areas (Fig. 2e) or when time shifts were given in one part of the body without a reference signal in other parts of the body (Fig. 2f, g). Thus, the time interval changes alone at the beginning of the test stimulus period in a body area (Fig. 1 dotted arrow) had no behavioral effects. The importance of relative times is clear from the comparable magnitudes of responses when relative time shifts are identical (cf. Fig. 2a, d or b, c). These results suggest that the time shifts are detected by coincidence or anti-coincidence detector neurons (Mugnaini and Maler 1987; Carr and Friedman 1999) that are sensitive to time differences between input neuronal signals (i.e., postsynaptic potentials) originating from different body areas. Figure 4 shows that the novelty responses were observed only when the head and trunk pulses were within ~±500 μs apart suggesting that the postsynaptic potentials in the detector neurons are required to occur within a few hundred microseconds and the duration of the interacting potentials are within this range.
Pulse-type gymnotiform fishes, including Brachyhypopomus, show novelty responses to small changes in amplitude of stimulus pulses (Heiligenberg 1974). Since the signal isolation between the head and trunk compartments in our experiment was imperfect (−40 dB), time-shifted pulses may have interacted across the head-trunk partition of the isolation chamber to reduce their peak amplitude. The selectivity for the base time shift in Fig. 4 could also be explained by this artifact amplitude reduction since the mutual reduction of stimulus amplitude occurs only when head-trunk pulses are overlapped in time. The positive control experiment in Fig. 2h, in which the artifact amplitude reduction at the switching between background and test stimulus periods was eliminated, resulted in similar novelty responses as shown in Fig. 2d. Thus, we conclude that novelty responses shown in this study were driven mainly by relative time shifts between signals from the head and trunk compartments.
The magnitude of time shifts used in this study, 100 μs, is somewhat larger compared to values expected from naturally occurring electrolocation targets or capacity values for which behavioral responses were demonstrated in other species of electric fishes (von der Emde 1993, 1998, 1999; von der Emde and Zelick 1995). Correspondingly, we observed strong and consistent novelty responses (up to an increase of 10 Hz with a resting pacemaker frequency of ~20 Hz) (Figs. 2, 3). We expect much smaller values for the threshold. In fact, our preliminary data show that some Brachyhypopomus individuals exhibited good novelty responses to 10 μs of time shifts (Kawasaki and Matsushita 2009). These values are comparable to threshold values found in wave-type electric fishes (Rose and Heiligenberg 1985; Guo and Kawasaki 1997).
Time-locked firing of the pear-shaped cell and the large cell
We have recorded field and intracellular potentials strictly time-locked to the inward electromotive force by the sensory stimulus pulses. The large amplitude of the field potential reflects synchronous firing of thick axons of the pear-shaped cells and large cells that exhibit one-to-one firing of action potentials (Fig. 5). The extraordinarily short latency of the potential, 600–900 μs, correlates with rapid conduction through the large-diameter axons and electric synapses in these neurons (Figs. 7, 8, 9). Our results confirmed morphological and electrophysiological features found in the fast electrosensory pathway of a pulse-type gymnotiform fish, Gymnotus carapo (Sotelo et al. 1975; Castelló et al. 2008). The time-locked neurons of the midbrain in Brachyhypopomus, however, fired one-to-one to repeated stimulus pulses at high frequencies (Fig. 6) unlike midbrain neurons in Gymnotus that adapted to repetitive pulses with interval shorter than ~20 ms (Fig. 5 in Castelló et al. 2008). We conclude that synaptic transmission at the somata of the pear-shaped cells and the large cells of Brachyhypopomus does not adapt up to 100 Hz. The adaptation seen at even higher frequencies in Fig. 6c appears to occur at the electroreceptor because the degree of adaptation depends on the amplitude of the stimulus pulses. Castelló et al. (1998) suggested the dendrites of the pear-shaped cells as one of possible mechanisms for the adaptation. Correspondingly, the somata of the pear-shaped cells of Brachyhypopomus were found to be adendritic. While the normal discharge frequency of Brachyhypopomus is less than 60 Hz, males of Brachyhypopomus do exhibit high-frequency discharges, the chirps, during their courtship (Kawasaki and Heiligenberg 1989). Among various types of chirps of Brachyhypopomus, the type-M chirps show relatively high amplitude and the discharge frequency as high as 500 Hz (Perrone et al. 2009). This indicates that the time-locked MMN neurons of a female are expected to fire one-to-one, representing chirp’s temporal structure when a male’s tail is positioned close to her body which often happens during their courtship.
The time comparator circuit in the MMN
Pulse delay and advance with the same magnitude of 100 μs often resulted in novelty responses with different magnitudes (cf. Fig. 2a, c). The large receptive field of the large cells proposed above explains this asymmetry. Among the multiple inputs to a large cell from pear-shaped cells, earlier inputs, rather than delayed ones, are expected to trigger firing of the large cell. Assume that the receptive field of a large cell extends into both head and trunk areas. The firing of this large cell is expected to fully advance when one of the receptive field areas is given advanced pulses relative to the other. When delayed pulses are given to one of the areas, however, the firing of this large cell may not be as delayed as the delayed sensory pulses because non-delayed pulses from the other part of the receptive field fire this large cell earlier. If a small cell that receives an input from a large cell with a large receptive field covering both head and trunk areas, and an input from a pear-shaped cell with a small receptive field in the trunk area, the small cell would experience a larger time difference between the two inputs when the head area is advanced than delayed, because the delayed pulse in the head tends to be ignored by the large cell which had already been activated by an earlier (i.e., non-delayed) pulse from the trunk area. If we assume that this hypothetical small cell is an anti-coincidence detector, then its output action potentials should follow the behavioral results in Fig. 2a and c. We assume the receptive fields of large cells distribute across the head and trunk compartments in the isolation chamber and may or may not be symmetric against the electrical partition depending on individuals. The above argument made with a single small cell holds at the population level if the average receptive field of the pear-shaped cells inclines toward the trunk compartment. If it inclines toward the head compartment, we expect, an opposite outcome, larger novelty responses when the head signal is delayed than advanced, which we indeed observed in some individuals (Fig. 3).
The present study revealed the topology of time coding neurons within the time comparison circuit in a gymnotiform pulse-type fish, Brachyhypopomus. The discovered topology is that the axon of time-locked input neurons (the pear-shaped cell) entering the circuit bifurcates and synapses on to two types of neurons (1) time-locked neurons (the large cell) intrinsic to the circuit and to (2) the small cell, which also receive time-locked input via the intrinsic time-locked neuron. Since a small cell receives different types of time-locked inputs, this topology suggests that the time comparison by a small cell occurs between the input and intrinsic time-locked neurons, but not between the same types (Fig. 11). It is remarkable that the identical topology has been known from the time coding circuits in all other major groups of weakly electric fishes so far examined, wave-type gymnotiform Eigenmannia (Carr et al. 1986b), wave-type mormyrid Gymnarchus (Matsushita and Kawasaki 2004), and pulse-type mormyrids Gnathonemus and Brienomyrus (Mugnaini and Maler 1987; Friedman and Hopkins 1995, 1998; Amagai et al. 1998). The sizes of component cells are also remarkably similar. Time-locked neurons have a large adendritic soma and a thick axon; the time comparator neurons are small.
Major differences between the different types of electric fishes appear to exist in the modes of synaptic transmission to the comparator neurons. In Brachyhypopomus, both inputs seem excitatory—the somatic input to the small cell from the large cell is via large gap junctions, and the connection at the dendritic tip of the small cell is predominated by chemical synapse with round synaptic vesicles. On the contrary, both excitatory and inhibitory inputs appear to be involved in other electric fishes. In Gymnarchus, the inputs of S-afferents, an electrosensory neuron, to the dendrite of an ovoidal cell would be excitatory because it is direct inputs of a sensory neuron and the round synaptic vesicles in the terminal are thought to be excitatory. On the other hand, the somatic inputs to an ovoidal cell via chemical synapse by a giant cell have been suggested as inhibitory because the terminal is immunoreactive to glycine (Zhang and Kawasaki 2007). The somata of the small cells in Gnathonemus receive both excitatory and inhibitory inputs (Mugnaini and Maler 1987). The long dendrite of the small cells in Brachyhypopomus is noteworthy in comparison to non-existent or short dendrites in time comparator cells in Gnathonemus, Gymnarchus, and Eigenmannia.
Myelination of somata and initial segments of time-locked neurons (present study; Sotelo et al. 1975) is unique to gymnotiform pulse-type fishes among electric fishes, but is found in other vertebrate sensory systems that require fast conduction and time critical processing (goldfish, Rosenbluth and Palay 1961; hachetfish, Model et al. 1972; chick, Sun et al. 1996). The myelination of pear-shaped cell with one input may be effective on fast and secure relaying of information, whereas that of the large cell may be not only for the fast conducting but also for controlling the locations of integration of multiple inputs and spike initiating zone (Carr and Boudreau 1993; Kuba et al. 2006).
The time comparison circuits occur in different brain structures in different groups of electric fishes—the anterior midbrain in pulse and wave-type gymnotiform fishes, the posterior midbrain in pulse-type mormyrids, and the hindbrain in wave-type mormyriform fish Gymnarchus. In wave-type gymnotiforms such as Eigenmannia, Apteronotus, and Sternopygus, the time comparison circuits are confined within a layer embedded in the torus semicircularis. In pulse-type gymnotiforms, however, the layer is displaced out of the torus semicircularis and forms a separate dorsal structure, the MMN. The current study confirmed that the MMN occurs in major genera of pulse-type gymnotiforms, Electrophorus, Gymnotus, Rhamphichthys, Gymnorhamphichthys, Hypopygus, Hypopomus, Microsternarchus, Steatogenys, and Brachyhypopomus (Alves-Gomes et al. 1995; Crampton and Albert 2006). The MMN represents another example of a brain structure for a specific neural function (time comparison) that is uniquely shared with members of major phyletic groups (Carlson et al. 2011).
We thank two anonymous referees for critical reading of the manuscript. This work was supported by a grant from National Science Foundation IOS-0723356 to MK, and by the JSPS (Japan Society for the Promotion of Science) Grants-in-Aid for Scientific Research no. 22570073, and by a grant from the CPIS (Sokendai Center for the Promotion of Integrated Sciences) to AM. We also thank to Kentaro Arikawa for providing us with experimental space and equipment for anatomical studies. The histological materials other than those of Brachyhypopomus were prepared by Grace Kennedy and other members of Walter Heiligenberg’s laboratory in early 1990 s and donated to MK. All experiments were approved by the University of Virginia Animal Care Committee.