Response properties of electrosensory units in the midbrain tectum of the paddlefish (Polyodon spathula Walbaum)

Electroreception is one of the ancestral vertebrate senses and is used by a variety of aquatic animals. It is important for prey detection and orientation in nearly all non-teleost bony and cartilaginous fishes, in some groups of teleosts and in urodele amphibians(Wilkens and Hofmann, 2005). Whereas some teleosts actively emit electric fields to probe their environment, most electrosensory vertebrates only sense the weak ambient fields that originate from other living organisms. This passive electrosensory system has been well investigated at the level of the primary afferent fibers and the hindbrain (dorsal octavolateral nucleus, DON). However, less is known about the processing of electrosensory information at higher levels in the brain.

Anatomical descriptions of ascending projections have shown that the major target of hindbrain electrosensory neurons is the midbrain. There are three major targets depending on the species. In teleosts, the most important target is the electrosensory part of the torus semicircularis(von der Emde, 1998). In elasmobranchs, most fibers innervate the lateral mesencephalic nucleus. However, the relationship between lateral mesencephalic nucleus and torus semicircularis is not clear and they may be homologous. The tectum is also innervated directly by DON fibers in elasmobranchs(Boord and Northcutt, 1982; Boord and Northcutt, 1988; Northcutt and Boord, 1984). In the paddlefish, all three mesencepalic targets receive direct input from the hindbrain (Hofmann et al.,2002).

There have been relatively few physiological investigations of the passive electrosensory system in the midbrain. In elasmobranchs, Platt et al.(Platt et al., 1974) and Bullock (Bullock, 1979; Bullock, 1984) described evoked potentials in various midbrain areas in response to direct nerve shock and electric field stimulations. Northcutt and Bodznick(Northcutt and Bodznick, 1983)identified the lateral mesencephalic nucleus in the dogfish as the main target of ascending electrosensory information. Later, Bodznick(Bodznick, 1991) found a topographic map of electrosensory receptive fields in the tectum of a skate. Single unit responses have been recorded in the midbrain of elasmobranchs(Andrianov et al., 1984; Schweitzer, 1986), the torus semicircularis of catfish (Knudsen,1976a; Knudsen,1976b; Knudsen,1978), and the tectum of a urodele, the axolotl(Bartels et al., 1990). However, these studies were mainly aimed at demonstrating the presence of electrosensory information in the midbrain and its topographic representation in the tectum. A detailed analysis of response properties to a variety of stimulus waveforms and frequencies was not performed.

Paddlefish have the largest number of electroreceptors of any passive electrosensory animal (Fig. 1). The physiology of the electrosensory system has been extensively studied in primary afferent fibers (Wojtenek et al.,2001), and at the level of the hindbrain DON(Hofmann et al., 2004; Hofmann et al., 2005). In this study, we analyzed the response properties of tectal units for comparison with primary afferents and DON units.

Paddlefish (Polyodon spathula) were obtained from the Honeywell Hatchery, Missouri Department of Conservation, and kept in large bio-filtered and aerated tanks containing dechlorinated water that was raised to a salinity of 2‰ by the addition of stock salt. Before surgery the fish were anesthetized with MS-222 (1:10 000). Local anesthesia (0.4% lidocaine, Sigma,St Louis, MO, USA) was applied at the dorsal surface of the skull, and the brain was exposed. Fish were then placed in the recording tank, immobilized with 5–20 μl tubocuraine (Apothecon, Princeton, NJ, USA) and the gills were irrigated through the mouth with fresh, aerated water. Occasionally, local anesthesia with lidocaine was refreshed throughout the experiment.

Electrical stimuli were delivered via silver wires at each end of the tank,anterior and posterior to the fish, providing a quasi-homogeneous field across the tank. In the figures the stimulus is displayed with respect to the anterior electrode, i.e. if the stimulus goes positive, the anterior electrode is positive with respect to the posterior electrode and vice versa. Due to the geometry of the tank and the presence of the fish, the field was probably not completely homogeneous. However, it allows stimulation of all receptors simultaneously and avoids the necessity of locating the receptive field for each unit with small local dipole electrodes. Stimuli were produced by the computer and delivered through a USB audio interface (Burr Brown PCM2902, Tucson, AZ, USA) that allowed DC output. The sampling rate was 22 050 s^–1 and resolution was 16 bit. The signal was galvanically isolated and converted to constant current by a linear stimulus isolator (A 395, WPI, Sarasota, FL, USA).

Electric fields applied to the tank were calibrated by measuring the voltage drop between two points in the middle of the tank parallel to the field lines. A sinusoidal stimulus at 5 Hz and different amplitudes was generated and the resulting field strength measured. We also changed the frequency to test for non-linearity between computer signal output and actual field strength in the tank. In the frequency and amplitude ranges used in this study, no non-linearity was present. While recording from higher multimodal areas, it was important that the experimental conditions excluded the stimulation of other modalities. Fortunately, stationary electrodes do not emit any physical signals other than electricity.

Tectal and DON units were recorded with tungsten electrodes (5–20 MΩ) or glass capillaries (1–10 MΩ), filled with 3% lithium chloride. Since tectal units showed no or little spontaneous activity, a search stimulus was applied. This was usually a 5 Hz sinusoidal wave with 10μV cm^–1 peak-to-peak amplitude. Recordings were amplified×1000 and filtered between 100 Hz and 5 kHz (AM Systems, Model 1700,Carlsberg, WA, USA). The signal was digitized by a USB sound interface (Burr Brown) and stored on a computer. The data were analyzed with Igor 5(Wavemetrics, Portland, OR, USA), and standard dot plots, peri-stimulus time histograms (PSTHs) and phase plots were computed.

For the spontaneous rate, we calculated the mean spike rate and the coefficient of variation, which is the standard deviation of interspike intervals divided by the mean interval. A value of 1 means that spikes are generated randomly and lower values indicate that the spike train is more regular.

To evaluate the response magnitude at different frequencies of a sinusoidal electric field, two parameters were calculated. The first parameter was the average spike rate during stimulation and the second parameter was the degree of phase coupling. The latter was obtained by computing the normalized period histogram of phase angles of spikes relative to the sine wave cycle with a bin width of 10 degrees. Then, a value D was calculated(Kajikawa and Hackett, 2005),which is based on an estimate of entropy of the period histogram. This value is 0 for random data and 1 if all spikes are in the same bin of the period histogram. This method gives reliable results even if there is more than one spike per period, which is often the case. The usual vector strength calculations and Rayleigh tests were not used here since they are not reliable if there is more than one spike per cycle or if spikes accumulate at two different phase angles. The probability P for the significance of synchrony was calculated according to Kajikawa and Hackett(Kajikawa and Hackett, 2005). Instead of calculating 1000 surrogate spike trains, we calculated 100, which gives a probability resolution of 100, sufficient to detect significance levels of P<0.01.

The mean phase angle of the spike trains was calculated according to Goldberg and Brown (Goldberg and Brown,1969). Phase angles were only calculated if the Rayleigh test showed a significant phase coupling (z>4.6)(Batschelet, 1981).

Upon entering the tectal surface with the microelectrode, multiunit activity (hash) was recorded in response to both visual and electrosensory stimuli. Electrosensory hash was greatest at depths of 250–500 μm below the tectal surface. In this region single units could be isolated and tested for their response to a variety of electrosensory stimuli. Many of them also responded to visual stimuli, e.g. lights switched on and off or objects moving in the visual field contralateral to the recording site. However,visual responses were not studied systematically.

The spontaneous spike rate of tectal units was always very low(2.20±2.22 Hz, N=48) compared with the ongoing spike rate of DON units (27.6±11.7 Hz, N=71). Many units produced only a few spikes and the interspike interval was highly variable. There was no indication of an intrinsic rhythm. The mean coefficient of variation of tectal units was 1.12±0.29, which means that the interspike interval was random. In contrast, DON units showed a mean coefficient of variation of 0.21±0.206. This indicates that the spike trains were generated by a regular mechanism. The difference between DON and tectum units was even more apparent if the coefficient of variation was plotted against the spike rate(Fig. 2A). This also shows that it is unlikely that we recorded from afferent fibers in the tectum. There is also strong evidence that we recorded from secondary units in the DON. It is known that the spike trains of primary afferent fibers are driven by two oscillators (Neiman and Russell,2004). A fast Fourier transform (FFT) of their spike trains showed two peaks, one corresponding to the spike generator and another at around 25 Hz that probably originates from the hair cells(Fig. 2B). This 25 Hz oscillation was absent in DON units (Fig. 2C) (see also Hofmann et al.,2005).

The response of DON units to electrical stimulation has been described in detail elsewhere (Hofmann et al.,2005). Here, we focused on DON activity relevant for comparison with that of tectal units. A 5 Hz sine wave caused the DON units to modulate their firing rate around their spontaneous rate(Fig. 3A). During the positive phase of the sine wave, the ongoing activity was reduced at low stimulus intensities and suppressed at higher intensities; during the negative phase the spike rate was increased. In some units the phase relationship was reversed and an increase in spike rate was observed during the positive phase of the stimulus. However, there was no significant difference in overall spike rate before and during stimulation (P>0.05, Wilcoxon paired-rank test, N=37). In contrast, tectal units exhibited little spontaneous activity and responses to a sine wave were mostly excitatory, i.e. the spike rate was increased (P<0.0001, Wilcoxon paired-rank test, N=42; Fig. 3B). Most tectal units produced one to three spikes per cycle of the sine wave. However,response activity was quite variable with some cycles failing to trigger spikes whereas other cycles triggered two or three spikes. In most tectal units, the response was uniform during prolonged stimulation(Fig. 3B), although some units had a delayed response starting a few seconds after stimulus onset(Fig. 3C).

Phase plots of DON units showed that most produced spikes during the negative half-cycle of the sine wave (Fig. 4A–C); a few cells were excited during the positive half-cycle (Fig. 4D). In the DON, most units showed a mean phase angle of around 310 degrees. Two units had phase angles between 100 and 130 degrees, and two units had phase angles around 250 degrees. On the other hand, tectal units were highly variable, with phase lags anywhere between 0 and 360 degrees(Fig. 4E–H) and spikes sometimes occurred at two different phase angles. Most tectal units had mean phase angles between 0 and 180 degrees, but there was more variability than in DON units.

In 20 DON units and 19 tectal units, the response to different amplitudes of a 5 Hz sinusoidal stimulus was measured. The phase angle was independent of sine wave amplitude over a wide range of intensities in the DON(Fig. 5A). Tectal units showed more variation in mean phase angle, but there was little systematic change over a wide range of amplitudes in most units(Fig. 5B).

As a measure of the response magnitude at different amplitudes, we calculated the overall rate change. This is the mean spike rate during stimulation minus the rate before stimulation. Furthermore, we calculated D as a measure of phase locking. In the DON, the spike rate did not change for amplitudes below ca 3 μV cm^–1(Fig. 6A). Above this, the rate change was highly variable with most units showing an increase in rate. Phase locking, as expressed by the D value, increased rapidly for amplitudes between 0.3 and 3 μV cm^–1(Fig. 6B), a range at which no overall spike rate change was observed. Above 3–10 μV cm^–1, D values seemed to be saturated.

In the tectum, spike rate change was more variable at higher amplitudes but, as in the DON, it was not a linear function of stimulus amplitude. Phase locking increased more systematically from the lowest amplitudes tested up to tenths of microvolts per centimeter, where it decreased again in some of the units. However, phase locking was much better (more than one order of magnitude) at amplitudes below 1 μV cm^–1.

Tectal and DON units were also tested at different frequencies of a uniform sinusoidal field at 5–20 μV cm^–1 peak-to-peak amplitude. Since DON units had a spontaneous rate, any modulation that was symmetrical around the spontaneous rate did not lead to an increase in overall spike rate. Consequently, low stimulus frequencies that caused only minor modulations resulted in only small changes in overall spike rate(Fig. 7A). At frequencies above ca 1 Hz some units increased their rate up to 10 Hz but some decreased their rate. If the response was measured in terms of phase coupling(D), DON units showed a linear relationship between frequency and response magnitude below ca 5 Hz(Fig. 7B). Above 10–20 Hz, modulation decreased steeply. The frequency tuning curves of DON units were remarkably uniform with all units showing a similar frequency response.

In the tectum, units were more variable in their frequency response, in terms of both mean spike rate and phase coupling(Fig. 7C,D). As in the DON,phase coupling decreased above 20 Hz, but many units showed no or little attenuation towards lower frequencies. Around 0.1 Hz, phase coupling was about one order of magnitude better than in the DON.

This study addressed the question of how paddlefish tectal units respond to uniform electric fields and how their response differs from those of DON units. Although uniform fields such as those produced by geoelectric sources could have some functional significance for the fish(Wilkens and Hofmann, 2007),they are probably not relevant for detecting small planktonic animals, the primary food of the paddlefish. Here we used uniform stimulus fields, not with the intention of simulating natural stimuli but for comparison with those used in previous studies on the response properties of primary afferents and hindbrain neurons. Furthermore, uniform fields are spatially uniform, i.e. the field strength at a given point in time is the same around the animal.

The first obvious difference between tectal and DON units is that tectal units had little or no spontaneous activity. This has implications for neuronal coding. DON units with a spontaneous rate respond to electric field changes by modulating their spike rate. The mean spontaneous rate is higher(27.6 Hz) than the highest stimulus frequency the cells respond to (up to 20 Hz), which allows for accurate coding in the signal bandwidth. As a consequence, a sinusoidal signal causes modulation of the DON spike rate with no or little change in overall rate. This would be equivalent to a frequency modulation where the spontaneous rate is the carrier frequency. Tectal cells with low spontaneous activity cannot apply a rate code. Furthermore, the coefficient of variation is around 1, which suggests that spikes are generated at random intervals. Without stimulation, tectal units either generate spikes due to an internal Poisson-like process or they respond to the noise that is present in their synaptic input. If the latter is true, the spike train of tectal cells would represent the demodulated signal of the frequency modulated input from the DON units.

Like DON units, tectal units produced spikes phase locked to the sine wave. However, the mean phase angle varied from unit to unit. In the DON, spikes occurred during either the negative or the positive half-cycle, with most units responding to the negative half-cycle. In all passive electrosensory fishes with the exception of teleosts, primary afferent fibers respond with an increase in spike rate when the electric field at the pore is negative relative to the internal reference potential of the fish. Due to the high skin impedance of freshwater fish, this reference potential is the average potential of the fish's body (or head)(Kalmijn, 1974). Recordings from primary afferents of the paddlefish showed that fibers innervating the rostrum respond with an increased rate if stimulated with a uniform field with the cathode in front of the animal. Fibers innervating the gill cover,however, respond to the same stimulus with a decrease in spike rate (M.H.H. and L.A.W., unpublished observations). In this case, the opercular pore openings are closer to the stimulus electrode behind the fish and the center of the internal reference, which is somewhere around the mouth, is closer to the electrode in front of the animal. This explains the two populations of units in the DON with mean phase angles that differ by roughly 180 degrees. As in the DON, tectal units were also highly phase coupled, but the mean phase angle was variable from unit to unit and did not fall into two classes. Each unit, however, had a characteristic phase angle that was mostly independent of stimulus amplitude (see Fig. 5).

In contrast to the phase angle, the degree of phase coupling changed systematically with stimulus amplitude in both the DON and tectum. In the DON,phase coupling increased steeply between 0.3 and 3 μV cm^–1, a range over which overall spike rate was unchanged. This suggests that in DON units, amplitude information is frequency modulated with the spontaneous rate being the carrier frequency. Tectal units appear to be more sensitive than DON units. Phase coupling was at least one order of magnitude higher at amplitudes below 1 μV cm^–1.

Tectal and DON units were also different in their frequency response. DON units were remarkably similar when tested with sinusoidal waves at different frequencies. Although there were differences in spontaneous rate, the shape of the frequency response curve was nearly identical from unit to unit and the phase locking decreased proportionally with frequency below ca 5 Hz. Tectal units were more heterogeneous and did not show much attenuation at lower frequencies.

Due to the frequency response properties of primary afferents and DON units that center on 10 Hz, the passive electrosensory system was thought to be tuned to this frequency band. However, behavioral studies showed that best frequencies are always lower (Peters and Evers, 1985; Peters et al.,1988). We have shown here that higher brain centers such as the tectum have units with a frequency response tuning that agrees better with the results of the behavioral studies. But, why do peripheral neurons attenuate frequencies that are obviously relevant for the animal? It is known that the filter curve for peripheral units can be described as a first derivative with an additional low-pass filter (Hofmann et al., 2004). In the special case of a sine wave, a derivative filter would attenuate lower frequencies and shift the phase of the signal,but would not change the waveform. However, if any other signal was applied,its wave form would change. If we consider for example a monopolar DC electric field that passes the animal, the electric field at a receptor would increase while the source approaches and decrease afterwards. This is not a periodic waveform, but just a transient increase in field strength that slowly goes back to zero. Depending on stimulus polarity, this would result in either a transient increase or a transient decrease in spike rate. A derivative filter would convert this signal into a bipolar stimulus that would cause an increase in spike rate followed by a decrease or vice versa. Thus, regardless of stimulus polarity, DON units would always show a period of increased firing in response to a monopolar moving stimulus. Furthermore, from the derivative it is possible to calculate the distance of the source from the skin surface(Hofmann and Wilkens, 2005). We think that a major function of the derivative filter is to change the waveform of such non-periodic events rather than to high-pass filter sinusoidal signals. However, the side effect that low frequencies are attenuated in the hindbrain is apparently compensated for in the tectum. This is supported by the fact that tectal units respond better to lower frequencies than DON units, which agrees better with behavioral studies(Peters and Evers, 1985; Peters et al., 1988). We can only speculate about the mechanism involved, but we now know where it takes place and can investigate whether perhaps a temporal integration could reverse the effect of the peripheral differentiation.

The experiments complied with the `Principles of animal care', publication no. 86-23, revised 1985, of the National Institutes of Health. This research was supported by grants from the NSF/DAAD, the National Science Foundation(IOB-0524869), and a University of Missouri Research Board grant. Paddlefish were kindly supplied by the Missouri Department of Conservation.