Abstract
To perform the jamming avoidance response (JAR), the weakly electric fish Gymnarchus detects time disparities on the order of microseconds between electrosensory signals received by electroreceptors in different parts of the body surface. This paper describes time-disparity thresholds of output neurons of the electrosensory lateral line lobe (ELL), where the representation of timing information is converted from a time code to a firing-rate code. We recorded extracellular single-unit responses from pyramidal cells in the ELL to sinusoidally modulated time disparity with various depths (0-200 μs). Threshold sensitivity to time disparities measured in 123 units ranged from 0.5 to 100 μs and was ≤5 μs in 60% of the units. The units from pyramidal cells in the inner and outer cell layers of the ELL responded equally well to small time disparities. The neuronal thresholds to time disparities found in the ELL are comparable with those demonstrated in behavioral performance of the JAR. The sensitivity of ELL units to small time disparities was unaffected when the center of the cyclic time-disparity modulation was shifted over a wide range (up to 250 μs), indicating an adaptation mechanism for steady-state time disparities that preserves the sensitivity to small dynamic changes in time disparities. Phase-locked input neurons, which provide time information to the ELL by phase-locked firing of action potentials, did not adapt to steady-state time shifts of sensory signals. This suggests that the adaptation emerges within the ELL.
- electric fish
- binaural comparison
- coincidence detection
- electrosensory lateral line lobe
- jamming avoidance response
- temporal coding
Introduction
Many animal species are known to exhibit behaviors that require a sensory capability of detecting small time disparitiesa on the order of 10-8 to 10-5 s between sensory inputs (Knudsen et al., 1979; Mason et al., 2001; Simmons et al., 2003) (for review, see Carr and Friedman, 1999; Carr et al., 2001). Sensory organs in these animals inform the CNS of the moment of sensory events by the occurrence times of action potentials in the afferent nerve [for temporal coding, see Theunissen and Miller (1995)]. The behaviorally important time disparities in the afferent action potentials are then detected by central neurons that compare the times between inputs. These time comparator neurons inform higher centers of the brain of the time disparity by their firing rate (impulse-rate coding).
The weakly electric fish Gymnarchus and Eigenmannia generate quasi-sinusoidal electric organ discharges (EODs) from their tail at a few hundred hertz. When the fish's EODs are jammed by EODs from a neighboring fish, the fish shifts its discharge frequency away from that of the neighbor to avoid jamming of its electrolocation system (Watanabe and Takeda, 1963; Bullock et al., 1972, 1975). This jamming avoidance response (JAR) requires time disparity between sensory feedback signals of its own EODs that are differentially time modulated (or phase modulated) at different skin areas by the neighbor's EODs (Bastian and Heiligenberg, 1980; Heiligenberg and Bastian, 1980; Kawasaki, 1993). The time-disparity threshold for correct performance of the JAR is 10-7 to 10-6 s (Rose and Heiligenberg, 1985; Guo and Kawasaki, 1997). The occurrence times of feedback sensory signals from the EODs are sampled by a type of electroreceptor in the skin that sends phase-locked action potentials to the electrosensory lateral line lobe (ELL) in the brain via S-type afferents (here-after simply S-afferents) in Gymnarchus (Guo and Kawasaki, 1997). The time disparities between the S-afferents are detected by a neuronal circuit in the ELL (Fig. 1) (Kawasaki and Guo, 1996; Matsushita and Kawasaki, 2004), in which the output of pyramidal cells modulate their firing rates in response to time disparities between S-afferents from different skin regions.
A behavioral study in Eigenmannia suggested that the microsecond time sensitivity of the JAR requires spatial summation of sensory signals (Rose and Heiligenberg, 1985) and neurons showing comparable time sensitivity were found only in the last stage (diencephalon) of the sensory hierarchy (Kawasaki et al., 1988; Rose et al., 1988). Kawasaki and Guo (1996), however, reported that neurons in an early stage of time processing (hindbrain) may exhibit microsecond sensitivity in Gymnarchus. In the present study, we examined the threshold sensitivity of pyramidal cells in the ELL of Gymnarchus using a “phase chamber” in which stimuli presented to different skin areas were precisely controlled in time.
Materials and Methods
Animals. Approximately 50 Gymnarchus niloticus (total length, 12-21 cm) were used. Environmental conditions in the holding tanks were identical to those described previously (Kawasaki, 1994). After anesthesia with tricaine methanesulfonate (MS-222; 1:10000; Sigma, St. Louis, MO), we immobilized fish with an intramuscular injection of flaxedil (gallamine triethiodide; 8-15 μl of a 0.3 mg/ml solution; Sigma), which greatly attenuated EOD amplitude. The silenced EOD was replaced with an artificial electrosensory stimulus as described below. Activity of the EOD pacemaker command signal was recorded from the tail to monitor the fish's condition throughout experiments. Fish were placed inside a plastic chamber (phase chamber; 12 × 18 × 5 cm) set in a plastic tank (50 × 50 × 13 cm), gently held with a sponge-lined clamp, and submerged in water, except for a small area along the surface of the head. Water resistivity was set between 5 and 8 kΩ · cm. Oxygen-saturated water perfused the gills via a mouthpiece inserted in the mouth. After local application of 2% lidocaine, a small hole was opened in the skull above the hindbrain, exposing the dorsal surface of the ELL (see below). These procedures were approved by the Institutional Animal Care and Use Committee at the University of Virginia.
Electrosensory stimulation in the phase chamber. We applied electrosensory stimuli using a phase chamber, which electrically isolates the head and trunk into separate compartments using a high-resistance partition (Fig. 2 A). This partition was placed immediately behind the pectoral fin and was sealed against the fish's body using petroleum jelly. In the head compartment, a sinusoidal carrier signal that mimicked the silenced EOD was given between an electrode placed in the mouth and a pair of carbon rod electrodes straddling the head. In the trunk compartment, a sinusoidal stimulus with an identical frequency was given between a pin electrode inserted into the dorsal musculature near the anterior end of the dorsal fin and a pair of carbon rod electrodes straddling the trunk. The signals were delivered to each compartment via a homemade isolator with field effect transistors and adjusted to a stimulus amplitude of 1-2 mV/cm measured near the skin surfaces. Signal cross-talk between the two compartments was less than -30 dB, as measured outside the fish's body.
A phase synthesizer (Wavetek 650) generated an unmodulated sinusoidal signal for the trunk and a phase-modulated sinusoidal signal for the head, creating differential phase modulation (DPM) between the two compartments (Fig. 2 B, C). DPM was sinusoidal such that the phase of the head was delayed during the first half of the modulation cycle (0-180°) and advanced for the second half of the cycle (180-360°). The phase modulation was cyclic and continuous, and its frequency was typically 2 Hz. DPM was controlled by an analog sinusoidal signal that was generated by a digital-to-analog converter (DA3-4; Tucker-Davis Technology, Gainesville, FL) and applied to the differential phase modulation port of the phase synthesizer. The center of the sinusoidal phase modulation (center phase) was typically zero but could be changed ±180°. Various depths of DPM were used. DPM at a depth of n μs will be expressed as DPMn (e.g., DPM5 for DPM with a modulation depth of 5 μs). Various depths between DPM0 and DPM200 were applied in a random sequence to determine the response threshold for each unit. The threshold sensitivity of a unit to DPM was defined by the minimum depth of DPM with which the unit showed a significant response (see below, Data acquisition and analysis). DPM50 at 2 Hz was used as a search stimulus.
The accuracy and stability of the electrosensory stimulus generated by the phase synthesizer were measured by a time interval counter with absolute time accuracy of 1 ns (SR620; Stanford Research Systems, Sunnyvale, CA). The head and trunk sinusoidal signals generated by the phase synthesizer were directly fed into Schmitt trigger inputs of the time interval counter, and zero-crossing times of the signals were recorded over 104 cycles. The SD of period jitter of the unmodulated head signal was 178 ns. The SD of phase difference between the two signals at DPM0 was 288 ns. The smallest phase modulation that could be achieved by the analog input of the phase synthesizer via its analog-to-digital converter was 81 ns.
Electrophysiological recording. Extracellular recording was made with glass capillary electrodes with a carbon fiber (diameter, 10 μm) in most experiments [modified from Fu and Lorden (1996)]. A borosilicate glass capillary with no center filament was pulled on a Brown/Flaming electrode puller (P-97; Sutter Instruments, Novato, CA) after inserting a carbon fiber into the capillary. One side of the separated capillaries retained the carbon fiber that was subsequently cut by a scalpel and extruded by 5-10 μm from the glass end. Some units were recorded with glass capillary electrodes filled with 3 m NaCl (tip diameter, ∼8 μm) or with indium-filled and gold/platinum-plated glass capillary electrodes (tip diameter, ∼10 μm) (Dowben and Rose, 1953).
Extracellular single-unit activity (50 μV to 1 mV) and large field potentials (see below, Recording site) caused by phase-locked firing of primary afferents and giant cells (up to 10 mV) were amplified 10× with an amplifier (model 5A; Getting Instruments, Iowa City, IA). The large periodic field potentials (see below) were eliminated by a cancellation technique in which the raw signal was delayed by one period, inverted in sign, and added to the original signal with a digital signal processor (sampling frequency, 48 kHz; model DSP-4; QSC Audio Products, Costa Mesa, CA). The resulting signal, which contained only nonperiodic signals, was sent to a Schmitt Trigger circuit, which in turn triggered an event timer (ET1; Tucker-Davis Technologies) that generated time stamps for spiking at 1 μs resolution (Fig. 3 A, B).
Intracellular recordings from primary electrosensory phase-locking neurons (S-afferents and giant cells) were made in the deep fiber layer (DFL) of the ELL (see below, Recording site) using sharp glass-capillary electrodes (30-50 MΩ, filled with 3 m KCl). The electrode was lowered into the DFL, from which large field potentials can be recorded as determined by preliminary extracellular recordings (Kawasaki and Guo, 1996). Light vibration of the electrode established the penetration into a single neuron. When stable phase-locked action potentials were obtained, we started a program that lasted 30 s: the phase in the carrier signal was held at 0 μs for the first 5 s, then shifted to either + 100 or -100 μs for 20 s, and finally shifted back to 0 μs for 5 s (see Fig. 10, bottom traces). We recorded spike times and analyzed spike phase shifts.
Recording site. The ELL is a hindbrain structure that consists of bilateral lobes, the anterior half and medial edges of which are covered by the valvula cerebelli (Bass and Hopkins, 1982). The ELL consists of three distinctive zones: dorsal, ventral, and medial (MZ), each of which is layered. The DFL constitutes the core of the ELL and is shared by all zones. In the MZ, the DFL is covered with an inner cell layer (ICL), outer fiber layer, outer cell layer (OCL), and molecular layer, in this order from the core. The MZ occupies most of the caudal half of the ELL, and as the zone extends rostrally, the lateral extent of the MZ becomes smaller. We placed electrodes in the posterior half of the MZ, which corresponds to the region between panels G and H of Figure 1 in Kawasaki and Guo (1998). The recording site could be determined by monitoring the waveform of the large field potentials (up to 10 mV) generated by phase-locking neurons in the DFL (giant cells and S-afferent terminals) [as shown in Kawasaki and Guo (1996), their Fig. 2].
Data acquisition and analysis. Spike histograms were constructed relative to the stimulus modulation period (typically 500 ms) (Fig. 2C), and the following response parameters were computed for each histogram: average spike rate = (total number of spikes)/(total duration of stimulation); F1, the Fourier component of the histogram at the modulation frequency; F2, the Fourier component of the histogram at twice the modulation frequency (second harmonic) (F1 and F2 are expressed in the unit of spikes/s); and Va, the mean vector angle of spiking relative to the modulation cycle, which was computed as follows:
where xi is the spike count of the ith bin in a histogram. cosi = cos(i · 2π/n), sini = sin(i · 2π/n), i = 1,...,n. n is the total number of bins in the modulation histograms (Goldberg and Brown, 1969). Va takes a value between 0 and 360° and indicates “response center” where spikes are most likely to occur in a modulation cycle. Va = 90 and 270° correspond to the points with the largest phase delay and advance of the head signal with respect to the trunk signal, respectively. The vector strength of spiking relative to the carrier cycle, Vs, was computed as follows:
where xi is the spike count of the ith bin in a histogram constructed relative to the period of the carrier signal. i = 1,.., n. n is 1/f × 106 (f, frequency of the carrier signal). Vs expresses the degree of synchronization of spikes to the carrier signal and takes a value between 0 and 1. Values of Vs of 0 and 1 indicate, respectively, no and complete synchronization (phase locking) to the carrier signal.
Statistical significance of spike responses to small phase modulation was examined by the V test (α = 0.05) and Rayleigh z test (α = 0.05). The V test examines whether the observed circular distribution of spike times relative to the phase modulation cycle has a tendency to cluster around a hypothetical angle, which was obtained as the value of Va for a suprathreshold modulation depth for the unit (Batschelet, 1981). The z test examines the tendency of spiking to occur at any particular angle of phase modulation (Batschelet, 1981).
A computer program written in Matlab controlled the experiments. It controlled the digital-to-analog converters for phase and amplitude modulation, collected time stamps of spikes, and displayed a raster plot, histograms, and statistical data on-line.
Results
Single-unit recordings were made from ∼250 neurons in the medial zone of the ELL. All neurons showed a spontaneous discharge [spike rate, 9.82 ± 9.10 (mean ± SD)] when no differential phase modulation was given. Spike times of these units during spontaneous discharges were analyzed relative to the local field potential that reflects periodic phase-locked firing of the giant cells and afferent fibers (Fig. 3). Mean latency of the spike measured from the time point where the negative slope of the field potential is greatest was strongly clustered around 0.9 ms for a population of neurons. These units also highly clustered around Vs = 0.95, indicating strong phase locking to the local field potential (Fig. 3A,C). We regard these units as responses from the pyramidal cells in the ICL based on the previously reported strong correlation between Vs (Vs > 0.84) and morphology/location of the neuron (Kawasaki and Guo, 1998). Other units had longer and more distributed latencies with smaller Vs (Vs ≤ 0.84) (Fig. 3B,C). These units were regarded as pyramidal cells of the OCL, again based on the distribution of Vs. We call these two populations of units ICL and OCL units, respectively.
Responses to differential phase modulation
Units recorded in the medial zone of the ELL typically showed responses to sinusoidal modulation of differential phase by modulating their firing rate (e.g., more firing to phase advance and less firing to phase delay) (Fig. 4A). The units responded neither when a phase-modulated signal was applied to only one compartment (Fig. 4B), nor when two identical signals were applied to both compartments (Fig. 4C,D), confirming that the units responded to phase differences between the compartments (Kawasaki and Guo, 1996). The fundamental Fourier component of the spike histograms was used as a measure of response strength (Fig. 4A, dashed line). One hundred twenty-one units were tested with DPM50, and the response parameters are summarized in Figure 5. Eighty-two of these (36 of 57 from ICL and 46 of 64 from OCL) units responded well (F1/spontaneous discharge rate > 0.5) to DPM50 (Fig. 5A). These units showed maximum firing to either phase delay or advance of the head signal in reference to the trunk signal (Fig. 5B). The head delay- and head advance-preferred units occurred with an equal likelihood (ICL units, delay/advance = 20/16, p = 0.382; OCL units, delay/advance = 23/23, p = 0.5; sign test). Most of the units responded antagonistically to delay and advance, resulting in a weak second harmonic (F2/F1 ≈ 0) (Fig. 5C). A small number of the units, however, increased or decreased their firing rate both to phase delay and advance, resulting in a large F2/F1 ratio (Fig. 5C). The remaining 39 of the 121 units studied did not respond well to phase modulations even larger than DPM100, and some of them responded well to amplitude modulation. These units probably represent either differential phase-sensitive neurons with their receptive fields not arranged for head-trunk comparisons or amplitude-sensitive neurons in the OCL (Kawasaki and Guo, 1998). The mean spike rate to DMP50 was ∼1.6 times higher than spontaneous discharges (Fig. 5D), indicating that the overall responses of the pyramidal cells is excitatory. ICL and OCL units did not show any significant difference in response properties.
Sensitivity to differential phase modulation
Sensitivity of 123 ELL units was tested with various depths of phase modulation. Figure 6 shows response histograms of six representative units. Distribution of threshold DPMs is shown in Figure 7. Seventy-four units showed statistically significant responses to DPM5 (60%), 22 units showed statistically significant responses to DPM1 (18%), and 3 units showed statistically significant responses to DPM0.5 (2%). There was no significant difference in thresholds between ICL and OCL units (Fig. 7). These threshold sensitivities are determined by averaged responses to continuously applied sinusoidal phase modulation (∼60 cycles). To determine the number of cycles necessary for obtaining a significant response, increasingly larger numbers of cycles were resampled from the same data until a statistically significant response was achieved for each unit (Fig. 8). At DPM5, 24% of units (18 of 74) showed significant responses within 1 s (two cycles) (Fig. 8A). At DPM1, 59% (13 of 22) of units showed significant responses within 10 s (20 cycles) (Fig. 8B). No difference was noted when the resampling of data were performed with randomly chosen cycles rather than the experimental sequence of cycles. This indicates that no sensitization or desensitization occurred during the sequential presentation of phase modulations.
Adaptation to shifted center phases
A previous study showed that the differential phase-sensitive neurons in the ELL preserved their sensitivity to suprathreshold phase modulation even after the center value of the sinusoidal phase modulation (Fig. 2C, center phase) was shifted by a large amount (∼200 μs) (Kawasaki and Guo, 1996). In the current study, we examined near-threshold sensitivity of the neurons with shifted center phases. Seventeen units were tested with DPM5 using various center phases (eight ICL units and nine OCL units). After each shifting of the center phase, the units were allowed to adapt for 15-30 s before collecting data. In response to the shifts of center phase, units showed initially either facilitation or suppression of firing rate for a few seconds (Kawasaki and Guo, 1996). The responsiveness to DPM5 was preserved over a wide range of center phase. Figure 9 shows representative response histograms at various center phases of three units. The thresholds of these neurons were ≤5 μs (p < 0.001 with DPM5, z test) at a center phase of 0 μs. The neuron shown in Figure 9A remained sensitive to DPM5 with a preference for phase advance for a range between -100 and +200 μs of center phases. Sensitivity was lost beyond this range. The response center of this neuron remained constant across all histograms with significant responses (Va = 269 ± 16.4°; n = 10). The neuron shown in Figure 9B remained sensitive to DPM5 with a preference for phase advance between 0 and 200 μs of center phases (Va = 264 ± 12.1°; n = 7). The responsiveness was lost between -20 and -50 μs and reappeared at -100 μs with a 180° shifted Va (Va = 79.6°). This type of shift in Va was observed in only two units with DPM5 but was more frequently observed with larger DPMs (≥10 μs; data not shown). Figure 9C shows a unit that remained sensitive with a similar Va (Va = 301 ± 27.0°; n = 12) over an exceptionally wide range of center phases (-200 to 400 μs). On average, the range of center phases within which units kept the sensitivity to DPM5 was approximately ±125 μs. As shown in Figure 9B′, the peak of F1 was often slightly displaced from 0 μs, indicating that maximum response strength is not always at a center phase of zero. The displacement of this maximum was usually within ± 100 μs (n = 10, both ICL and OCL). When the center phase was shifted in a step without sinusoidal phase modulation, units initially showed facilitation or suppression similarly to cases with sinusoidal phase modulations presented above. After a few seconds, the discharge returned to the level of the spontaneous before the step change. The responses then reappeared when tested with small phase modulation. We conclude that sinusoidal phase modulation is not necessary for the adaptation.
Phase-locking neurons do not adapt to steady phase shift
The results above indicate that pyramidal cells in both ICL and OCL are insensitive to steady changes of the center phase and preserve the sensitivity to dynamic changes of small phase differences. Does this adaptation take place in the phase-locking neurons? The firing times of phase-locking neurons may adapt back to the original phase after a steady phase shift is introduced. To determine whether this adaptation occurs within the ELL, we recorded from phase-locking neurons (S-afferent fibers and giant cell axons) (Kawasaki and Guo, 1996; Matsushita and Kawasaki, 2004) while applying steady phase shifts in the electrosensory carrier signal to one compartment of the phase chamber. Figure 10 shows representative plots of the spike times of a phase-locking neuron in response to steady phase shifts. After an initial 5 s, the phase of the stimulus signal in the head compartment was shifted by either + 100 or -100 μs and held for 20 s and then shifted back to the original phase (0 μs). The spike times shifted in response to the shift in phase of the sensory signal remained shifted and showed no adaptation during the 20 s period. The amount of shift in spike times varied from neuron to neuron. This is probably because, in some neurons, a portion of the receptive field was exposed to the signal in the other chamber, which had no phase shift. When the phase of the carrier signal was shifted in the opposite direction, the spike times of the neuron showed an equally strong shift in the opposite direction (Fig. 10).
Discussion
Threshold of time disparity
The major findings of the current study are (1) 60% of recorded pyramidal cells in the ELL of Gymnarchus are sensitive to 5 μsof time disparity between sensory signals applied to the head and trunk, and (2) the sensitivity to sinusoidally modulated time disparity is preserved when a much larger steady-state time disparity is introduced between the head and trunk.
The time disparity between phase-locking neurons that provide time information to the time comparator in the ELL is expected to be smaller than the time disparity between the sensory signals applied to the head and trunk compartment. This is because giant cells, a type of phase-locking neurons, possess a large receptive field that spans across the head and trunk compartments, causing them to fire at an intermediate phase between the head and trunk signal (Matsushita and Kawasaki, 2004). The phase-locking neuron presented in Figure 10 exemplifies such a phase shift that is smaller than that of the stimulus. Therefore, the measured sensitivities of the ELL neurons are conservative. Behavioral and neuronal recording studies (Kawasaki, 1993; Kawasaki and Guo, 2002) demonstrate that the JAR occurs with phase differences between any body surfaces (head-trunk, left-right, etc.), and different ELL neurons are tuned to detect phase differences with different spatial orientations. In the current study, however, we examined time-disparity sensitivity of all encountered pyramidal cells exclusively with the head-trunk configuration alone. We assume that most pyramidal cells would respond to 5 μs of time disparity if the best orientation of the neurons for phase difference were used (Kawasaki and Guo, 2002). Rose and Heiligenberg (1986) found a comparable level of time-disparity sensitivity in midbrain neurons of the gymnoti-form electric fish Eigenmannia, which has evolved its electrosensory system and JAR independently of the mormyriform Gymnarchus (Lauder and Liem, 1983). They showed that some of the midbrain neurons were sensitive to time disparities of ∼10 μs using a relatively insensitive statistical method.
Rose and Heiligenberg (1985) discovered that the JAR occurs with time disparities at 10-7 to 10-6 sin Eigenmannia and argued that integration of sensory information from all body surfaces was required for this sensitivity. The current study in Gymnarchus demonstrates, however, that the individual pyramidal cells in the ELL, which are only one or two synapses away from the phase-locking neurons and receive no substantial spatial integration, are sensitive to 5 μs. Neuronal integration in higher brain levels (midbrain and diencephalon) to which the pyramidal cells project (Kawasaki and Guo, 1998) may result in even higher sensitivity, as in Eigenmannia (Kawasaki et al., 1988).
Exact neuronal mechanisms for the sensitivity of the pyramidal cells to these extraordinarily small time disparities are unknown. The following observations and theoretical considerations have been made in an attempt to understand the sensitivity. The SD of firing times of phase-locking neurons that project to the ICL is 5.6 μson average, and 95% of them show SDs <10 μs (Guo and Kawasaki, 1997). At their thresholds, pyramidal cells required several seconds of continuous stimulus presentation, during which thousands of accurately phase-locked action potentials arrive at the time comparator (Fig. 8). Statistically speaking, the sample SD is , where n is the number of action potentials given by a phase-locking neuron to the comparator during the several seconds. If a pyramidal cell receives action potentials at 400 Hz for 2 s, for example, the sample SD would be 0.2 μs (). This means that rather precise time information would be available to comparator neurons. Specialized anatomical structures have been found in the ICL. Terminals of the giant cells form a giant synapse that embraces ∼85% of the soma area of the postsynaptic neurons, termed ovoidal cells (Matsushita and Kawasaki, 2004). This morphology may enhance the rapidity of synaptic transmission due to ephaptic effects (Byzov and Shura-Bura, 1986; Voronin et al., 1999). No matter how accurate the firing times of input phase-locking neurons are, synaptic currents produced by these inputs are expected to have much longer time constants than the threshold sensitivity of pyramidal cells to time disparities (Trussell, 1999). Takagi and Kawasaki (2003) showed that the Hodgkin-Huxley equation may yield sensitivity to time disparity on the order of 10-5 s between two synaptic input currents with a much larger time constant (∼300 μs) due to the nonlinear dynamics of the equation.
Neuronal representation and processing of temporal information on the order of microseconds has been found not only in the electrosensory system (Heiligenberg and Altes, 1978; Rose and Heiligenberg, 1986; Kawasaki et al., 1988) but also in the auditory system of birds and mammals. Phase-locking neurons in the nucleus magnocellularis of the barn owl (Carr and Konishi, 1990), and those in the medial superior olive and the trapezoid body of mammals (Joris et al., 1994), show only a few tens of microseconds of jitter. Although statistical examinations have not been made for threshold determination, auditory neurons in the midbrain of birds and mammals appear to show considerable sensitivities to interaural time disparities of tens of microseconds (Goldberg and Brown, 1969; Yin and Kuwada, 1983; Takahashi and Konishi, 1986; Yin et al., 1986, 1987; Chan et al., 1987). Skottun et al. (2001) and Shackleton et al. (2003) applied a statistical method, the receiver operating characteristic analysis, to interaural time-disparity-sensitive neurons in guinea pig and demonstrated thresholds around 30 μs. The time-disparity thresholds demonstrated in the pyramidal cells of Gymnarchus in the current study are ∼5-10 times lower than those found in the auditory midbrain of birds and mammals.
We regard the units we recorded from as pyramidal cells in the ICL or OCL for the following reasons. First, the latencies of their action potentials from the peak of the phase-locking field potential were consistent with previous results using in vivo whole-cell recording and labeling, which confirmed the morphology of the recorded cell as pyramidal cells (Kawasaki and Guo, 1996, 1998), whereas the latency of the ovoidal cell is expected to be shorter than that of pyramidal cells. Second, the recording sites were shallower than the ovoidal cell layer, based on the waveform of the concurrently recorded field potential (Fig. 3) as well as actual depth in the ELL (Kawasaki and Guo, 1996). Third, the large action potentials are likely to be generated by the large apical dendrites of pyramidal cells (Bell et al., 1993; Kawasaki and Guo, 1998), which the ovoidal cells do not possess (Matsushita and Kawasaki, 2004).
It is remarkable that there was no significant difference in sensitivity to time disparity between ICL and OCL pyramidal cells given their different anatomical position: the pyramidal cells in the ICL receive phase-locked inputs, whereas those in the OCL do not. This anatomical difference reflects the difference in their vector strength, Vs, and firing latency. A possible explanation for the similarity of sensitivity is that the output of ovoidal cells may be given to pyramidal cells in the ICL via their dendrite and those in the OCL via their axon (Fig. 1) (Matsushita and Kawasaki, 2004).
Adaptation property of pyramidal cells
The pyramidal cells preserved their sensitivity to near-threshold DPM (5 μs) over a wide range of center phase (current study) as well as suprathreshold DPM (Kawasaki and Guo, 1996). The arrival times of action potentials of phase-locking neurons to the pyramidal cells depends on the conduction velocity of the neuronal path between the electroreceptors and pyramidal cells, which may easily vary in natural conditions because of temperature changes. Local temperature changes may occur if a fish stayed at a boundary between sun and shade (such as vegetation), resulting in a change of center phase experienced by a phase comparator. One degree difference of body temperature between head and trunk would result in a few hundred microseconds of conduction time difference along 10 cm of the body length of a fish (Nagy et al., 1978). The range of 250 μs of center phase in which the neuron retains sensitivity to threshold DPM5 is similar to the value in Eigenmannia (Rose and Heiligenberg, 1986) and may be reasonable to compensate such delays. Behavioral study in Eigenmannia showed that correct JAR occurred in response to a combination of small phase modulation (5 μs) and amplitude modulation (0.1%) with center phase shifted by >100 μs (our unpublished observation).
The phase-locking neurons in the DFL we recorded in this study were either S-afferents or giant cells. These neurons responded to the dynamic shifts of phase but did not show any adaptation over a period of 20 s, which is enough time for pyramidal cells to adapt (Kawasaki and Guo, 1996). The giant cell terminal gives a synaptic input to ovoidal cells, whereas S-afferents make synaptic contacts directly onto pyramidal cells in the ICL and ovoidal cells (Matsushita and Kawasaki, 2004). Thus, the adaptation may emerge at the level of pyramidal cells or ovoidal cells. The morphology of the pyramidal cells in both the ICL and OCL is similar to the Purkinje-like cells in mormyrid electric fish (Grant et al., 1996), whose large apical dendrites receive descending inputs from higher centers to cancel out predictable sensory inputs to the cell (Bell et al., 1993). Similar mechanisms for adaptively canceling steady-state phase shifts might be involved in Gymnarchus.
Footnotes
-
↵ a The terms “time” and “phase” will be used interchangeably in this paper following a convention in the electric fish literature. Although phase,φ, should formally be expressed by the unit of degrees, it will be often expressed in time, t.(φ/360° = f × t, where f is the frequency of the carrier signal.) “Time disparity” and “phase difference,” for example, are synonymous.
This work was sponsored by grants to M.K. from the National Science Foundation (IBN-0235533) and Precursory Research for Embryonic Science and Technology, Japan Science and Technology Corporation.
Correspondence should be addressed to Masashi Kawasaki, Department of Biology, University of Virginia, Gilmer Hall, P.O. Box 400328, Charlottesville, VA 22904-4328. E-mail: mk3u{at}virginia.edu.
Copyright © 2005 Society for Neuroscience 0270-6474/05/2511424-09$15.00/0