ABSTRACT
The response characteristics of auditory neurons in the multimodal midbrain torus semicircularis of rainbow trout were analyzed to examine their topography and functional differences. This analysis included the localization of recording sites, the measurement of spontaneous activity, the ratio of transient/sustained activity, and the synchronization, latency, preferred direction and directional range of the response. On average, units with a directionally selective (DS) response are positioned 60 μm more dorsally than other auditory units. Directionally selective units usually have a higher response rate, a higher transient/sustained activity ratio and are better synchronized. Auditory units encountered within the same electrode track tend to be either all DS or all non-DS. Within a track, there is no uniformity of the response characteristics observed except that the preferred direction of DS units appears to be the same. The anatomical stratification of the torus, containing 66 000 somata (5–10 μm in diameter), does not match the electrophysiologically observed vertical distribution of functionally distinct units. On the basis of the topographical distribution of response characteristics, two types of well-synchronized DS units can be distinguished, hypothetically representing separate channels for the processing of acoustic motion and (mainly) pressure information. A third type of DS unit which receives input from both these channels and actually encodes the source direction uniquely for all directions is postulated.
Introduction
Neurophysiological research has resulted in some beautiful examples of topographical arrangements of neurons processing sensory information in the brain. A classic example is the organization of the visual cortex of mammals in relation to visual stimulus parameters (Hubel and Wiesel, 1962, 1968). In the mammalian cochlea, frequency selectivity is a function of the position of the hair cells along the basilar membrane. This topographic representation of stimulus frequency, or tonotopy, is preserved in the cochlear nucleus and throughout the central auditory pathways. In the inferior colliculus of the cat, responses to stimulus frequency, to the sharpness of frequency tuning and to the modulation of an amplitude-modulated tone are topographically organized (Langner and Schreiner, 1988; Schreiner and Langner, 1988). In the nucleus mesencephalicus lateralis dorsalis of the owl (homologous to the mammalian inferior colliculus), there is not only a tonotopic region but also a region where auditory space is represented (Knudsen and Konishi, 1978). In both birds and mammals, the midbrain plays a role in the interpretation of the external auditory space (for a review, see Irvine, 1992). Apparently, however, the fine-tuning of auditory localization depends on the visual system, which functions as a reference (Knudsen and Knudsen, 1985). By analogy with other vertebrates (e.g. the cat), frequency and temporal analyses also take place in the fish midbrain (Echteler, 1985; Nederstigt and Schellart, 1986; Schellart et al. 1987; Lu and Fay, 1993; Crawford, 1993). For some species, indications of a tonotopic organization have been found (Cyprinus carpio, Echteler, 1985; Oncorhynchus mykiss, Schellart et al. 1987). However, in the fish species that have been studied, tonotopy is much less pronounced than in higher vertebrates and has been described as ‘somewhat diffuse, at best’ (Carassius auratus, Lu and Fay, 1993). The midbrain torus semicircularis (TS; homologous to the inferior colliculus) of fish is the first area along the auditory pathway that has considerable binaural input (e.g. de Wolf et al. 1983). It also receives input from the visual system (Schellart and Rikkert, 1989; Coughlin and Hawryshyn, 1994; Wubbels et al. 1995) and from the mechanosensory lateral line system (Schellart and Kroese, 1989). The lateral line system is distributed over the entire body of the fish and is, therefore, well suited to provide spatial information (Denton and Gray, 1983; Wubbels, 1991; Bleckmann and Zelick, 1993). In the TS of electric fish of the genus Eigenmannia, amplitude and phase modulations of electric organ discharges are analyzed (in a manner resembling the operations underlying directional hearing in terrestrial vertebrates) in order to locate moving objects (Heiligenberg and Rose, 1985). By analogy with other vertebrates, this binaural and multimodal input suggested a role for the TS in spatial hearing. This area was, therefore, studied electrophysiologically and its involvement in directional hearing was confirmed (Wubbels et al. 1995; Wubbels and Schellart, 1997).
In many higher vertebrate species, there is psychophysical evidence that they can perceive the direction of a sound source, and behavioural observations support this conclusion. Similar psychophysical evidence that fish can perceive the direction of sound is limited to a few species (e.g. Leuciscus idus, Schuijf et al. 1977; Gadus morhua, Buwalda et al. 1983). Evidence that trout have directional hearing is circumstantial (Schellart and Buwalda, 1990). However, the similarity between its peripheral auditory system and, for instance, that of cod, which is able to distinguish sound direction (Buwalda et al. 1983), suggests that the trout should have comparable abilities.
There is still much to be learned about the structural and organizational basis of sensory analysis and integration in the midbrain of higher vertebrates, and data on the fish midbrain are even more scarce. We have reanalyzed measurements from units in the trout TS, previously used in our directional hearing studies (see Wubbels et al. 1995; Wubbels and Schellart, 1997), to extract further information about the activity of neurons in this region. Data from 189 auditory units for which the location, spontaneous activity and at least one response characteristic had been determined have been reanalyzed and statistically examined. Since our aim in this study was to reveal auditory response characteristics, special attention has been paid to the topographic and temporal properties of two functional types of neurons: those whose response depends on stimulus direction (DS units) and those with a non-directionally selective response (non-DS units).
Materials and methods
The methods used in the electrophysiological experiments, including surgical procedures, recording and stimulus generation, have already been described in detail (Goossens et al. 1995; Schellart et al. 1995; Wubbels and Schellart, 1997) and will, therefore, only be briefly outlined in this paper.
Preparation and recording
Rainbow trout, Oncorhynchus mykiss (Walbaum), 25–30 cm in body length, were initially anaesthetized with 250 mg l−1 MS 222 (Sandoz, Switzerland). The upper jaw was fixed between a mouthpiece (providing water for respiration containing 50 mg l−1 MS 222) and a nosepiece. The trunk was fastened to the fish tank with Velcro strips. The skull was opened with a dental drill. After surgery, the fish was immobilized with 0.45 mg kg−1 pancuronium bromide (Pavulon, Organon, Netherlands). Wounds were treated with Lidocaine ointment, and circulating water was replaced to remove MS 222. Animals were treated in accordance with Dutch law and were killed at the end of the experiment.
Glass microelectrodes (40–80 MΩ) were advanced at an angle of 70–75 ° through the optic tectum and midbrain ventricle to record from the torus semicircularis (TS). When the electrode tip arrived at the TS surface, a jump in the d.c. signal thus recorded could be seen on the oscilloscope, and this served as a recording depth reference. From the penetration point on the tectum, and the penetration depth and angle, the recording position was calculated with an accuracy of ±80 μm in the horizontal plane (visual estimate of penetration position) and ±20 μm for the depth from the TS surface (estimated on the basis of the variance in stepping motor position for units that were encountered during penetration of the TS and re-encountered during retraction). In the horizontal plane, recording position was expressed in relative coordinates (i.e. as a percentage of the tectal length and width). Since the fish were approximately equal in size, recording depth was not corrected for differences in brain size. The occurrence of spikes and marker signals for the temporal properties of the stimulus were stored on disk, together with the amplitude and phase of three accelerometer signals (see below).
Stimulus generation
In most fish species, the wavelengths for audible frequencies are much longer than body length. This means that the body of the fish, which has approximately the same density as water, moves in concert with the water, and the directionally sensitive hair cells are displaced relative to the otoliths, which have a higher inertia than the surrounding tissue (de Vries, 1950). The motion vector of diametrically opposed (monopole) sound sources is directed alternately towards and then away from the fish and this means that the fish has to deal with a directional ambiguity of 180 °. This ambiguity, however, can be overcome with an additional cue provided by the gas-filled swimbladder, which functions as a sound-pressure-to-displacement transducer, indirectly stimulating the inner ears from a body-fixed angle (Schuijf, 1976; Schellart and de Munck, 1987).
A vibrating platform was used in our experiments to apply a well-defined directional motion stimulus in the horizontal plane directly to the inner ear (Schellart et al. 1995). The contribution of the swimbladder to the stimulation of the inner ear was eliminated by keeping the water level below the body of the fish (Wubbels and Schellart, 1997). This also prevented the lateral line system from being stimulated, which simplified the interpretation of the response since the lateral line also projects to the TS (Schellart and Kroese, 1989).
The fish tank was mounted on a platform driven independently in the x-and y-directions by electromagnetic coils. The mechanical properties and performance of the platform are described in Schellart et al. (1995). The tank and platform are mounted on top of an air-damped experimental table (noise level at the position of the fish, between 50 and 300 Hz: 0.36 mm s−2 RMS). Prior to an experiment, the motion of the skull of the fish was measured using a miniature three-dimensional accelerometer attached to the skull with small screws (Wubbels and Schellart, 1997). This allowed us to compensate for the fact that the motion of the fish head was not an exact copy of the motion of the platform (Goossens et al. 1995). Consecutive tone bursts of rectilinear motion at a frequency of 172 Hz, lasting 720 ms and alternating with silent intervals of an equal duration, were applied. During an experiment, the motion vector was continuously rotated anti-clockwise in the horizontal plane, with each consecutive tone burst 7.2 ° apart. Amplitudes ranged from 0.6 to 90 mm s−2 (for more details about stimulus generation, see Wubbels and Schellart, 1997).
Data analyses
Dot displays, representing spike activity (see Fig. 1), were split into two parts. The first part shows the response during the tone burst. The second part shows spike activity during the silent interval. The spike density histogram, as a function of the stimulus direction, was calculated (shown on the right of each dot display). After smoothing (by eye), the directional selectivity range was determined by taking the average of the −3 dB width of the peaks in the spike density histogram. Maxima occur in the spike density histogram when the stimulus is ‘located’ at the preferred direction of the unit. When −3 dB points are absent, a unit is not directionally selective.
Post-stimulus time histograms below each dot display (PSTHs) show the global transient/sustained nature of the response. To allow differentiation of response types, a qualitative classification for the transient/sustained nature of the response was introduced. The criterion for a transient response was that it had to be at least twice as large as the sustained response component during the second half of the tone burst or, in the case of a sustained decrease, twice as large as the spontaneous activity. In an alternative approach, this response has been quantified by calculating the spike activity during the first and the last 360 ms of the tone burst.
Spikes were registered with respect to the positive zero-crossings of the sinusoidal stimulus (resolution 40 μs). From the Fourier-transformed spike density of the period histograms (obtained from sections of the dot display), the synchronization index (SI) and the phase (ϕ) of the response were calculated. The SI, calculated using the formula SI=0.5F1/F0 (where F0 is the mean activity and F1 is the first harmonic), ranged from 0 to 1 (Kroese et al. 1978). A χ2 version of the Rayleigh test (Mardia, 1972) was used to establish whether the spikes deviated from a uniform distribution, which was our criterion for synchronization. For auditory units of the TS, SI appears to be relatively independent of stimulus amplitude (Wubbels and Schellart, 1997).
Histology
A trout brain was stained with Cresyl Violet. From serial sagittal sections (20 μm), the total area containing cell bodies was measured using a VIDAS image-processing system equipped with a light-pen (Kontron Elektroniks, Zeiss, Eching/Munich, Germany) while the cell body density in the TS was estimated using a ‘counting-grid’. These two measurements, cell body density and total area, enabled the number of neurons to be calculated.
Biocytin (10 % in water) was injected into the TS of eight anaesthetized trout using a 3 μA current passed for 20–30 min through a platinum filament inside the glass electrode. Fish regained consciousness, were anaesthetized again after 20 h and then perfused with 0.1 % heparin in phosphate-buffered saline (PBS, pH 7.4) followed by 5 % glutaraldehyde in PBS. Dissected brains were immersed in fresh fixative for several hours, kept overnight in 15 % sucrose in PBS, and sagittal sections (50–60 μm) cut on a cryostat were collected in 0.5 % Triton X-100 in PBS. These were incubated (1 h, 20 °C) in peroxidase-labelled avidin D (Vector) diluted (1/200) in PBS with 0.5 % Triton X-100, washed in PBS, rinsed in 0.1 mol l−1 Tris–HCl buffer (pH 7.6), and incubated in 0.05 % w/v diaminobenzidine in Tris–HCl buffer supplemented with 0.01 % H2O2 (5–10 min). The reaction was stopped by rinsing in fresh buffer solution.
Results
One hundred and eighty-nine auditory units were included in the analysis of the relationship between response characteristics and topography in the rainbow trout TS. The location and spontaneous activity of all the units are known, together with at least one of the eight response characteristics. Table 1 lists the characteristics that have been examined, the number of units that were used for statistical analysis and the number of tracks of the electrode in which two or more units were encountered.
Among these 189 auditory units, we observed 92 DS units, 73 non-DS units and 24 units that could not be classified. In our search for auditory units, we also encountered units that responded to an object moving in the visual field of the animal, and the location of these visual units was recorded. All units were examined to see whether there was a relationship between their location, projected on three cross-sectional planes (i.e. from a dorsal, lateral and frontal point of view), and any of the other experimentally established characteristics.
Topography in the horizontal plane
The spontaneous activity of DS units was 2.0±5.5 spikes s−1 (mean±S..) and that of non-DS units was 3.8±6.6 spikes s−1, although most DS units are ‘silent’, i.e. show spontaneous activity of less than 1 spike s−1 (see Fig. 4 in Wubbels and Schellart, 1997). No topographic organization was apparent in the spontaneous activity of either DS or non-DS units.
A transient increase (+) in the spike rate at the start of the tone burst was quite often observed. This was sometimes followed by a transient (T) decrease (±) and, occasionally, a transient decrease without a preceding transient increase (−) was found. The sustained (S) response component could be an increase (+) or a decrease (−) in spike rate relative to spontaneous activity, or it could be absent (0). The numbers of units showing the different responses are given in Table 2. Examples of some of the response types are shown in Fig. 1A–H (see also Fig. 2A-F in Wubbels and Schellart, 1997). The results show that the response could be purely sustained (S+/S−; N=51), purely transient (T+/T−/T±; N=45) or could contain both components (N=84).
A unit that responds with an initial transient increase in the spike rate followed by a transient decrease (T±; N=33) has been called a ‘pauser’ (Rhode and Greenberg, 1992). This transient decrease can be relative either to the sustained response component or to the spontaneous activity. The numbers of DS and non-DS units for five response types (i.e. S−, T±, T+/T± and S0/S−, S+ and T0, T+/T± and S+) are given in Table 3, and the units have also been arranged with respect to their medio-lateral position in the TS (lateral, y<47 %; criterion based upon the observed distribution of the preferred direction) and with respect to their rostro-caudal position (caudal, x<44.5 %; the criterion is the median of x-coordinates of 269 units with spike activity).
A χ2-test was used to ascertain whether the response types mentioned above occurred more often in combination with the DS or non-DS characteristic. It appears that most S− units are non-DS (P<0.0005), whereas most transient/sustained (i.e. T+/T± and S+) units are DS (P<0.0005). The other response types showed no significant distribution (criterion P=0.05). The combination of response type and recording site was also examined but no significant distribution along the medio-lateral axis (medial units, y⩾47 %, N=72; lateral units, y<47 %, N=117) or the rostro-caudal axis was present for any of the response types. Thus, no topographic organization was found for these response types.
The difference between DS and non-DS responses with respect to transient/sustained components can also be inferred from the relationship between spike activity during the first and the second half of the tone burst (Fig. 2). For both DS and non-DS units, the regression was calculated twice, with either variable as the independent one. The bisector of these two regression lines is drawn for each type of unit (Fig. 2). A comparison of the regression (pairs) of both types of units reveals that they are significantly different (P⪡0.001, for the slopes and/or the intersection of the axes).
On average, the spontaneous activity of DS units is less than that of non-DS units (see above), but the situation is reversed for the response rate (Fig. 2). We have chosen an arbitrary criterion of 5 spikes s−1 increase relative to the spontaneous level of activity (either in the first or in the second half of the tone burst, or both). Only 10 % of the non-DS units exceed this response criterion when a stimulus of standard amplitude (30 mm s−2) is applied so that the response rate of non-DS units hardly increases with respect to spontaneous activity. For DS units, the response criterion is exceeded by 88 % of the units. The units for which directional selectivity is unknown gave an intermediate value of 42 %. In the case of spontaneous activity, response rate, SI, the absolute value of ϕ and the response latency, τ, no apparent relationship with topography was found.
When the medial and lateral parts of the TS are compared, a topographic organization of the preferred direction of DS units can be discerned (Fig. 3). The medial units are largely orientated along the rostro-caudal axis of the fish, whereas all preferred directions occur in the lateral part of the TS. Line segments, the length of which is a measure of the directional selectivity range, appear to be longer (signifying sharper spatial tuning) in the lateral part of the TS (Fig. 3). The distribution of preferred direction is shown in Fig. 4. In contrast to results presented previously (Wubbels et al. 1995), the distribution of preferred directions shown here, of which approximately 50 % is new data, deviates significantly from a rectangular distribution (χ2; P<0.005). Units with a preferred direction parallel to the long axis of the fish appear to be more numerous (see Discussion). In addition, a topographic differentiation appears to be present for the shift of ϕ as a function of stimulus direction (Wubbels and Schellart, 1997). There also appears to be some topographical differentiation of visual and auditory units along the rostro-caudal axis, since most of the rostral recordings appear to be from visual units (Fig. 5). Even when numbers are corrected for visual units located below the TS (see below), the distributions of auditory and visual units along the rostro-caudal axis are not identical (χ2; P<0.0005). Along the medio-lateral axis, the ratio of visual to auditory units increases medially.
Topography in the vertical direction
It has been noted previously that auditory units occur in the upper 800 μm of the TS while visual units occur in the upper part and also at greater depths (Wubbels et al. 1995). With the supplementary data from this study included, 95 % of the auditory units occur 107–724 μm below the TS surface (mean 333 μm), while for visual units the 95 % interval ranges from 100 to 1293 μm (mean 584 μm) below the surface. Inspection of the vertical distribution of DS and non-DS units shows that, on average, DS units are located slightly more superficially in the TS. The 95 % intervals range from 102 to 733 μm (mean 306 μm) in depth for DS units and from 127 to 825 μm (mean 364 μm) in depth for non-DS units, which is a significant difference (t-test, P<0.05). Some units responded to both visual and auditory stimuli. The distribution in the TS of these bimodal units was similar to that of visual units. Only three out of 22 bimodal units showed a DS auditory response compared with 11 non-DS auditory responses. No topographic differentiation in the vertical direction was observed for any other auditory neuronal characteristic.
The lower border of the TS has been reconstructed from microphotographs of sections of the TS (dashed line in Fig. 5). It can be concluded that, at its maximum thickness, the TS is approximately 1000 μm deep and it is much less deep at the periphery. At least 22 % of the visual units were located in the tectum opticum (Fig. 5B). In the deepest part of the TS, recordings of auditory units were rare, whereas visual units were quite frequently encountered there.
‘Columnar’ organization
Some characteristics of neurons recorded in the TS of the rainbow trout tend to be similar when they are encountered along the same electrode track. Response modality, i.e. auditory versus visual (Schellart et al. 1987), and directional selectivity, i.e. DS versus non-DS (Wubbels et al. 1995), appear to be organized in columns (but see Discussion). In the present study, we have checked whether the quantitative neuronal characteristics (i.e. spontaneous activity, spike activity during the first and the last 360 ms of the tone burst, SI, ϕ, τ, preferred direction and directional selectivity range) are similarly organized. For every track, the mean and standard deviation of each of these characteristics were calculated directly from the experimental data. The measured values (including values from tracks where just one unit was encountered) were then redistributed at random, and mean values and standard deviations were calculated as before. (It should be emphasized that in this procedure the number of units per track remains the same as for the original set of data.) This Monte Carlo realization was repeated a thousand times. Finally, the distribution of the track standard deviations for the experimentally established data was compared (χ2) with the distribution of the standard deviations of the Monte Carlo realizations.
For directional data, in this case ϕ and preferred direction, the standard deviation within a track has to be minimized both for the measured data and after each Monte Carlo realization (by allowing a shift of 360 or 180 ° respectively) before a relevant χ2-test can be performed. With ϕ1=4 ° and ϕ2=355 °, for instance, the relevant phase difference is 9 ° and thus one of these value would have been shifted 360 ° before calculating the standard deviation.
There was no significant difference between these distributions for spontaneous activity, spike activity during the first and the last 360 ms of the tone burst, SI, ϕ and τ. The measured track standard deviations for the directional selectivity range were significantly smaller than the track standard deviations after redistribution (P<0.01). However, this includes DS and non-DS units and is consistent with the previous observation that DS and non-DS units tend to be organized in columns (Wubbels et al. 1995). When this procedure was repeated for DS units only, no significant difference remained. The only response characteristic with smaller track standard deviations for the actual data than for the redistributed data is the preferred direction (P<0.01), as shown in Fig. 6. This is in agreement with our previous observation (Wubbels et al. 1995), although this had not been statistically confirmed until now. Apparently, the columnar organization of response characteristics is limited.
Histology
The Cresyl-Violet-stained sections of the TS show that the main contribution to the total number of neurons, which was estimated to be approximately 66 000 (unilaterally), comes from neurons with small (diameter 5–10 μm) unipolar cell bodies. Almost all these neurons are arranged in three patchy layers, 1–5 somata wide, parallel to the ventricular surface. In the central part of the TS (dorsal view), these layers are found 100, 200 and 400 μm from the TS surface. At the periphery of the TS, the layers approach the ventricular surface and tend to fuse with one another. From Fig. 3A in a report by Ito (1974), it can be inferred that these layers were found 80, 150 and 220 μm below the TS surface. A layered structure of neurons with small somata has been reported in the TS of other fish species (Ito, 1974; Cuadrado, 1987; McCormick, 1989; McCormick and Hernandez, 1996). The biocytin-stained sections show neurons with a main dendritic shaft, in some cases more than 100 μm long, which is usually orientated away from the TS surface and with axons that usually run parallel to the TS surface.
Discussion
Of 189 auditory midbrain TS units, 92 were DS units, 73 were non-DS units and 24 units could not be classified. On the basis of the response rate increase criterion of 5 spikes s−1, it is concluded that this population of ‘unknown’ units is a mixture of both DS and non-DS units. These numbers deviate somewhat from the numbers documented previously (Wubbels and Schellart, 1997) because different criteria were used in the classification of types. The number of non-DS units reported in the present study is smaller (and, therefore, an underestimate), because their response is generally much weaker than that of DS units, and thus the response characteristics are less reliable.
It is possible that the classification of a unit as DS or non-DS may depend on either subthreshold stimulation or saturation. The stimulus amplitudes applied ranged from 0.6 to 90 mm s−2 and recordings usually started at 30 mm s−2, which is approximately 56 dB above the psychophysical threshold determined by Abbott (1973). The response of DS units is relatively independent of stimulus amplitude (Wubbels and Schellart, 1997). If the directional response disappeared at lower amplitudes, all response vanished. Neither subthreshold stimulation nor saturation is, therefore, considered to have influenced the number of DS units that were found. Furthermore, it was shown that the populations of DS and of non-DS units also differ with respect to other response characteristics (see below). Thus, even if our classification of DS and non-DS neurons should turn out to be sometimes incorrect, the distinction between two types of auditory units would still be valid.
In the following paragraphs, the main results of this study and of our previous reports on directional hearing in fish are summarized to provide a complete picture of what is known about the topographic and/or functional relationships between the neuronal characteristics of the units recorded in the trout TS. Some remarks and conclusions are added, and our results are interpreted in the context of what is known about the neuronal organization of the TS of fish.
Visual (and bimodal) units versus auditory units
Auditory units are located in the upper 725 μm, and are thus restricted to the TS (Fig. 5). Visual units can be found both in the TS and in deeper layers (Fig. 5 and Wubbels et al. 1995), i.e. in the optic tectum which extends below the TS. Bimodal units are found at the same depth as visual units. Visual units appear to be arranged in columns (Wubbels et al. 1995).
Our principal experimental goal was the characterization of responses in terms of auditory directional selectivity. As this study progressed, experience taught us that auditory units were not to be found at very great depths. Therefore, the number of visual units will be underestimated, especially in the deeper layers. We found remarkably few bimodal units (22 versus 189 auditory units) compared with a previous study on the TS of the same experimental animal (36 versus 68 auditory units; Schellart et al. 1987). This may be due to our rather unsystematic procedure for visual stimulation (object size 0.10 m×0.15 m; distance 0.2–1.5 m; velocity 0.2–2 m s−1) or it may be that bimodal units require a combination of visual and auditory stimuli to produce an explicit response (Schellart and Rikkert, 1989).
DS units versus non-DS units
Both DS and non-DS units occur at all levels in the upper 725 μm, but on average DS units are located a little higher in the TS than non-DS units. The spontaneous activity of non-DS units is greater than that of DS units (Wubbels and Schellart, 1997), but for the response rate the situation is reversed (Fig. 2). Only 23 % of the non-DS units have a response synchronized to the stimulus frequency while 75 % of the responses are synchronized for DS units (Wubbels and Schellart, 1997). The phase angle of this synchronized response was constant for non-DS units, but for many DS units it was a function of stimulus direction (Wubbels and Schellart, 1997). Most units that respond with a sustained decrease of spike activity (S−) are non-DS units, whereas most units with both a transient and a sustained excitatory response component are DS units (Table 3). DS and non-DS units are encountered in separate columns (Wubbels et al. 1995; the present study). DS and non-DS units represent two separate populations of midbrain neurons with different functions in the analysis of the auditory input signal. It is remarkable that this difference is reflected in all the temporal response properties (except latency) analyzed, i.e. the qualitative characterization of the transient/sustained nature of the response, the response rate, SI and the shift in the angle of synchronization. It appears that much more temporal information is present in the response of DS units than in the response of non-DS units. However, none of these temporal properties has any relationship with topography in the TS, except the shift of the angle of synchronization (Wubbels and Schellart, 1997).
Although, the distribution of recording sites in our study does not seem to match the clearly layered structure demonstrated histologically (Ito, 1974; Cuadrado, 1987; McCormick, 1989), the small neurons observed in the Cresyl-Violet-stained preparation are probably identical to the neurons we identified as auditory, since electrophysiological studies indicate that they have to be small (Wubbels and Schellart, 1997). Although, we have been unable to demonstrate a relationship between electrophysiological activity and the layered structure of the TS, it may nevertheless exist. Our data are pooled from many trout and it is likely that some of our recordings were not from the cell bodies of the neurons, but from their extensions, e.g. the main dendritic shaft. In addition, the accuracy of recording site coordinates is relatively poor (see remarks at the end of the next subsection). However, the possibility remains that some of our recordings originate from scattered neurons that occur ventral to the layered structure (McCormick and Hernandez, 1996).
Topography of DS units
DS units occur in columns with the same preferred direction (Wubbels et al. 1995; the present study). Units with a preferred direction parallel to the long axis of the fish are more numerous than other units (Fig. 4). In the medial TS, the preferred direction is parallel to the long axis of the fish, while in the lateral part of the TS all possible preferred directions have been found (Wubbels et al. 1995; Wubbels and Schellart, 1997). At the micro-level of columns, no significant topography for the directional selectivity range of DS units could be demonstrated (see Results). Nevertheless, it has previously been shown that, in the medial TS, all DS units have a directional selectivity range of 60 ° or greater, whereas in the lateral part approximately 25 % show spatial tuning better than 60 ° (Wubbels and Schellart, 1997; see also Fig. 3). A pronounced shift (as a function of stimulus direction) in the phase angle of synchronization was only observed for lateral DS units (Wubbels and Schellart, 1997).
Thus, within the population of DS units, two subpopulations can be distinguished that differ with respect to both their location and their response characteristics. When we first observed that two such populations might exist (Wubbels et al. 1995), the focus of our experiments shifted somewhat to the medial part of the TS. Relatively more recordings were made from this region, thus increasing the frequency of occurrence of units with a preferred direction parallel to the long axis of the fish (compare Fig. 3 of Wubbels et al. 1995 with Fig. 3 of this report).
Our study shows that it is difficult to establish the topographical fine structure of the trout TS with respect to stimulus parameters. It may be that the accuracy of localization of recording sites (80 μm in the horizontal plane) was insufficient and that the number of units per fish too limited. In addition, penetration of the TS deviated, on average, by 5–10 ° from the perpendicular, which is not optimal for the detection of a columnar structure. It also appears that the structure of the TS, consisting of a few relatively thin layers (1–5 somata deep) spread far apart, is not very suited to a columnar organization. This is in sharp contrast with, for instance, the structure of the visual cortex of the monkey, where many somata are closely piled on each other (Hubel and Wiesel, 1968). It is possible, therefore, that such a fine structure does not exist for most of the stimulus parameters investigated. Nevertheless, our results show a clear differentiation between some of the response characteristics of auditory units and, in some cases, it could be demonstrated that this functional differentiation is correlated with a topographical differentiation. A relationship between a representation of auditory space and a topographic differentiation of the midbrain has been demonstrated in higher vertebrates (Knudsen and Konishi, 1978; Middlebrooks and Knudsen, 1984; Irvine, 1992). Our study of the trout TS is the first to show that such a relationship also occurs in the midbrain of fish.
Physiological considerations with respect to the neuronal organization of directional analysis in the TS of fish
As shown in Fig. 1, auditory responses in the TS are very diverse with respect to the form of their PSTHs. This is consistent with results obtained in the TS of other teleost species (Lu and Fay, 1993; Crawford, 1993). A sustained increase in spike activity occurs quite often (60 %, Table 2), but a transient increase at the start of the stimulus occurs even more frequently (70 %), which is very different from the observations of medullary units (Wubbels et al. 1993). Probably only part of the diversity can be related to the functional differences (i.e. DS versus non-DS) investigated in this study (see below). The lack of uniformity suggests that this same population of neurons must play a role in the processing of other stimulus features as well. This idea finds some support in the relatively small numbers of neurons that appear to be present in the multimodal TS (66 000) compared with the number of hair cells in the inner ear of trout (tens of thousands), the many hair cells of the lateral line system and the numerous rods and cones of the retina (unilaterally approximately 3×106). For the latter sensory system, however, much of the visual processing occurs in the tectum opticum, which contains at least an order of magnitude more neurons than the TS (Meek and Schellart, 1978).
The lagenar hair cells and more than half of the saccular hair cells are oriented in the vertical direction, with the other hair cells on the saccule oriented along a line in the horizontal plane (Dale, 1976; Popper, 1977). The utricle has a fan-like hair cell pattern almost in the horizontal plane (Dale, 1976). Our experiments were performed with stimuli in the horizontal plane. Since hair cells (Shotwell et al. 1981) and auditory afferents (Fay, 1984) have a directional selectivity range of 90 °, non-DS units probably receive their input from populations of utricular hair cells with different orientations.
Because the acoustic motion vector can be decomposed instantaneously along the various orientation axes of the hair cells, the inner ear of fish, unlike that of terrestrial vertebrates, is intrinsically a motion detector. However, an unambiguous interpretation of sound direction is possible only when an additional indirect component originating from the gas-filled swimbladder (functioning as pressure-to-displacement transducer) is also taken into account (Schuijf, 1976; Schellart and de Munck, 1987). Our directional hearing experiments strongly suggest that the two subpopulations of DS units represent two different systems, one for the processing of pressure information (the medial units) and the other for the processing of motion information (the lateral units). In the TS of goldfish, where the specialized hearing system has Weberian ossicles connecting the swimbladder to the saccule (Fay and Popper, 1980), the differentiation is reversed, with the lateral part being more pressure-sensitive and the medial part relatively more responsive to vibration (Fay et al. 1982).
The concepts of segregation of pressure and motion information in the fish auditory system and possible underlying mechanisms have been considered before (Buwalda et al. 1983). Central subtraction of bilateral input (common mode rejection) enhances binaural differences (Fig. 7). For bilaterally symmetrical input, the contribution of the swimbladder can be eliminated, thus providing a ‘motion’ channel (Fig. 7 upper panels; Buwalda et al. 1983) for which all preferred directions are possible. Addition of bilateral input from the saccules could be the mechanism for a (mainly) ‘pressure’ channel (Fig. 7 lower panels). For fish with specialized hearing systems, such as the goldfish, input for a ‘pressure’ system would probably come from the saccule. For fish without auxilliary hearing structures, such as the trout, the subject of pressure detection and processing remains more speculative, although the proximity of the swimbladder suggests that the saccule may play a role. Under natural conditions, most stimulation of the saccule will be from the pressure component (Schellart and de Munck, 1987). In our experimental arrangement, higher-order neurons of such a hypothetical ‘pressure’ channel would show a preferred direction along the rostro-caudal axis of the fish. Whether this kind of analysis along separate ‘motion’ and ‘pressure’ channels really takes place, however, remains a matter for speculation.
Thus far, we have distinguished two types of DS units. First, the medial DS units with preferred directions parallel to the long axis of the fish and a relatively broad directional selectivity range of 102±21 ° (mean ± S.D., N=31). These units show little shift of the phase of synchronization with changing stimulus direction (see example in Fig. 7 lower graph) and were therefore thought to be part of the ‘pressure’ system (Wubbels and Schellart, 1997). Second, the lateral DS units, which have a variety of preferred directions and display a relatively large phase shift (>30 ° when the stimulus direction is rotated by 90 °; see example in Fig. 7 upper graph), suggesting an interaction of inputs from different hair cell populations in the horizontal plane, probably from the utricle. Since they encode all possible particle motion directions in the horizontal plane, they may be part of the ‘motion’ system. These two types of DS units are well-synchronized and can, therefore, provide an unambiguous temporal code (i.e. the phase difference between them) for stimulus direction (Fig. 7; see also Wubbels and Schellart, 1997).
Some DS units, which do not belong to these two types, show various preferred directions and have a narrow directional selectivity range. However, synchronization is either poor or, if their response is synchronized, does not change phase with rotating stimulus direction. It is possible that this third type of unit, which tends to be located in the medio-lateral part of the TS and which hypothetically receives input from both ‘pressure’ and ‘motion’ units (Fig. 7), uniquely (i.e. over 360 °) encodes auditory space in the TS of fish in terms of a spike rate code.
The different biophysical constraints in an aquatic habitat require that the way in which auditory space is represented in the midbrain of fish differs considerably from the representation that has been demonstrated in land vertebrates (Knudsen and Konishi, 1978; Middlebrooks and Knudsen, 1984; Irvine, 1992). As a result, it is difficult to predict how auditory and visual space will be related in the fish brain. Schellart and Rikkert (1989) have shown that, in the TS of the rainbow trout, the size of the receptive fields of visual units tends to be large, retinopy is rather poor and responses to moving objects are unspecific. We have shown that visual and auditory units tend to be organized in separate columns and that only three out of 22 bimodal units showed a DS auditory response. Our findings lend little support to the assumption that auditory and visual spatial information are well integrated in the TS.
ACKNOWLEDGEMENTS
We are grateful for the contribution of M. Prins to the histological work. This study was supported by the Netherlands Organization for Scientific Research.