Response Properties of Lateral Line and Auditory Units in the Medulla Oblongata of the Rainbow Trout (Oncorhynchus Mykiss)

The lateral line is a mechanosensory system consisting of many small sense organs distributed over a large area of the fish body. The morphology, biomechanics and transduction properties of the peripheral lateral line system have been extensively studied (for reviews, see Dijkgraaf, 1963; Cahn, 1967; Coombs et al. 1989). Recently, the projection of lateral line nerves to the brainstem of several teleost species has been investigated (Claas and Münz, 1981; McCormick, 1983, 1989; Meredith, 1984; Meredith et al. 1987; Puzdrowski, 1989; Schellart et al. 1992). The region of the medulla oblongata that receives most of the primary afferent lateral line input is called the medial nucleus. Less is known about the central processing of the lateral line information in teleosts: single-unit activity from medullary neurones which receive primary lateral line input has been recorded in only two teleost species, the eel (Alnæs, 1973) and the goldfish (Russell, 1976; Caird, 1978).

The otolith systems of the teleost inner ear are involved in hearing (Popper etal. 1988). The afferents innervating the sensory epithelium project to a number of nuclei located ventral to the medial nucleus (McCormick, 1983; Nieuwenhuys and Pouwels, 1983; Meredith and Butler, 1983; Meredith et al. 1987). Auditory responses in the medulla oblongata have been studied in the herring (Enger, 1967) and the goldfish (Page, 1970; Sawa, 1976). For the rainbow trout, the following data have been measured: (1) primary lateral line afferent activity (Kroese et al. 1988, 1989; A. B. A. Kroese, M. Prins and N. A. M. Schellart, in preparation; Kroese and Schellart, 1992); (2) afferent and efferent projections of the trunk lateral line system to the medulla (Schellart et al. 1990, 1992);

(3) projections of acousticolateral units in the torus semicircularis to the medulla (de Wolf et al. 1983); and (4) the activity of acousticolateral units in the torus semicircularis (Nederstigt and Schellart, 1986; Schellart et al. 1987).

In the present study, the neural activity of acousticolateral units in the medulla oblongata of the rainbow trout was investigated. The main objective was to examine their response characteristics and frequency sensitivity to lateral line stimulation and to combined lateral line and auditory stimulation. The response properties of the medullary units are compared with the response properties of trout primary lateral line afferents (Kroese et al. 1988, 1989; A. B. A. Kroese, M. Prins and N. A. M. Schellart, in preparation; Kroese and Schellart, 1992) and with the response properties of acousticolateral units in the midbrain of the trout (Nederstigt and Schellart, 1986; Schellart et al. 1987). Another objective of this study was to investigate to what extent convergence of lateral line input and input from other sensory systems, especially the auditory system, occurred in the medulla. An abstract of the results of this study has been presented previously (Wubbels et al. 1991).

Rainbow trout [Oncorhynchus mykiss (Walbaum)] 25–30cm in body length were initially anaesthetized with 250mg l⁻¹MS 222 (Sandoz, Switzerland), and then immobilized with 0.15mg of pancuronium bromide (Pavulon, Organon, The Netherlands) injected intraperitoneally. They were transferred to the experimental tank which measured 40cm×20cm×10cm (see Schellart and Rikkert, 1989) and fixed down with clamps on the nose and two Velcro strips, leaving most of the trunk lateral line uncovered. Artificial respiration was provided by water saturated with fresh air (0.4 l min;⁻¹) containing 50mg l;^{− 1} MS 222 during the surgical procedures. A dental drill was used to make an opening in the skull to expose the rhombencephalon. The water level in the experimental tank was just below the opening in the skull with the trunk lateral line canal submerged. Water temperature was 8–14°C, depending on the season. To prevent lateral line organs and central nervous system units being inactivated by MS 222 (Späth and Schweickert, 1977), recordings started more than 2h after removal of this anaesthetihc. Treatment of the animals was in accordance with the Dutch law on experimental animals (January, 1987).

In order to obtain access to the medulla, the cerebellum was carefully pushed aside. It proved impossible to expose the entire medulla without causing injury. For this reason, recordings were carried out in the caudal half (Fig. 1) and thus no representative impression of the processing in the entire acousticolateral area of the medulla was obtained. All (single-unit) recordings were unilateral and took place in the lateral part of the right side of the medulla. Glass microelectrodes filled with 3mol l⁻¹ KCl (50–100 MΩ) were advanced in 5 μm steps into the medulla at an angle of 75° to the horizontal plane. Electrode position in the horizontal plane and electrode depth were stereotactically measured with the caudal tip of the fourth ventricle and the dorsal surface of the medulla, respectively, as stereotactic references. In the horizontal plane, the accuracy was about ±0.1mm. These data were not corrected for brain size differences, because the fish were about equal in size. Electrode tip positions were checked in four cases by injection of horseradish peroxidase (HRP) at the recording site.

Spikes (2–40mV), which were recorded either extracellularly or intracellularly, were converted to standard TTL pulses and fed into a computer (resolution of the counter timer was 10 μs) along with signals marking the start of the tone burst and the zero crossings of the stimulus. Frequency response characterisics were measured for 20 medullary units in 15 trout. In addition, the spontaneous activity and/or response over a small frequency range was obtained for another 80 mechanically sensitive units. In total 61 trout were used.

Sinusoidal stimulation was provided as tone bursts, either from an oscillating membrane or from an oscillating glass sphere (diameter 4.7mm). Prior to stimulation with the glass sphere, the receptive field of the medullary unit was localized with a small water jet, and a subjective impression of the receptive field was obtained. Linear frequency sweeps were applied, with frequency increasing from 20Hz to 400Hz in 5–15min. This long-lasting tone was gated, yielding tone bursts of 720ms alternated with silent intervals of 720ms. Rise and fall times for the tone bursts were about 30ms. A visual stimulus could be applied by means of a flashlight. The oscillating membrane was mounted in the rear wall of the experimental tank and generated a stimulus field in the entire tank (also stimulating the fish inner ear). When stimulating with the membrane, the sound pressure was measured with a hydrophone (Brüel and Kjaer, type 8103) inserted in the trout’s stomach. The sound pressure was kept constant for all frequencies during a single frequency sweep by adjusting the amplitude of the sine wave by parametric feedback (Smit et al. 1986; Nederstigt and Schellart, 1986). The amplitude of the membrane stimulus was usually 10.3Pa. However, below 80Hz, feedback was not optimal, resulting in a 6–20dB lower stimulus amplitude at 25Hz. Occasionally, recordings were maintained long enough to allow us to measure the response to more than one frequency sweep. In such cases, the membrane stimulus was attenuated by 10 or 20dB for the second frequency sweep. From the results of these experiments it was concluded that, for small stimuli, the response increases linearly with stimulus amplitude in medullary units.

The water displacement component of the field generated by the membrane was estimated by considering the stimulus source as intermediate between a monopole and a dipole (Denton and Gray, 1983). For this purpose, the near-field approximation was taken of the amplitude relationship between pressure (p) and particle motion (Kalmijn, 1988): v=1.5ρ/ρwr, where v denotes the velocity of the water particles, ω the angular frequency, ρ the water density and r the distance from the centre of the source. In this expression, the constant is 1 for a monopole and 2 when a dipole is involved. At the location of the hydrophone and at 100Hz, a pressure of 10.3Pa corresponds to 0.2 μm of water displacement. The glass sphere, vibrating parallel to the canal axis, was used to provide a local stimulus field. The glass sphere was mounted on either a piezoelectric element (providing a maximal displacement of 33 μm) or, for larger stimuli, on a pen-motor. Since the latter stimulus device had a resonance peak in the frequency range of interest, the acceleration was controlled at 19 ms⁻² for all frequencies by parametric feedback (corresponding to 48 μm at 100Hz). The phase of the response was corrected for the frequency-dependent phase shift of the pen-motor. The sensitivity of the recorded units was expressed as the calculated (Harris and van Bergeijk, 1962) reciprocal particle displacement (nm⁻¹) at the lateral line receptors that would be necessary to evoke 20% modulation of the spontaneous activity at F_max (Kroese and Schellart, 1992). F_max is defined as the frequency at which the largest response was obtained during constant-amplitude stimulation. The particle displacement at 20% modulation was obtained by extrapolation, in the assumed linear amplitude response range, from the measured response (i.e. the modulation of the spontaneous activity) and the estimated particle displacement that caused such a response.

The response amplitude is proportional to the synchronization index (Wubbels, 1990, p. 41). The frequency response characteristics of 20 units were obtained in this way. They were categorized into two classes. Units which responded at frequencies below 50Hz and whose gain did not increase at frequencies above 120Hz were classified as ‘low-frequency’ units. Units whose gain increased at frequencies above 120Hz and which responded up to at least 150Hz were classified as ‘high-frequency’ units. When the response phase was plotted on a linear frequency axis, the phase lag increased approximately linearly with frequency for most units (see Fig. 5). Latency was calculated directly from these data without correcting for the dynamic part of the phase shift. The variance of the measured latencies may be somewhat increased by the temperature range (6°C), but no correlation was found between the actual temperature during an experiment and the measured latencies. High-frequency and low-frequency units were evenly distributed over this temperature range. The latency of responses in afferent neurones from the posterior lateral line nerve has been measured just behind the operculum (Kroese and Schellart, 1992) and ranged from 6 to 12ms (see Table 1). The conduction velocity of these fibres is greater than 19 ms⁻¹ (Kroese et al. 1989; A. B. A. Kroese, M. Prins and N. A. M. Schellart, in preparation), so the latency of responses in primary afferents in the medulla should range from 8 to 14ms. Low-frequency units with longer latencies were considered to be higher-order lateral line neurones. The latency of afferent neurones from the inner ear will be less (e.g. Furukawa and Ishii, 1967) because of the proximity of these organs. High-frequency units with latencies greater than 10ms may safely be considered to be higher-order auditory neurones.

The spontaneous activity of 80 units which were sensitive to mechanical stimulation was 27±20spikes s ⁻¹ (mean ± S.D.). Inspection of the interval histograms and the serial correlation coefficients of the interspike intervals occasionally indicated a response to unintended mechanical background noise, e.g. movements of the experimental table (Wubbels, 1990; Wubbels et al. 1990). No relationship was found between the spontaneous activity and the absolute sensitivity of the units.

Attempts to localize receptive fields of medullary units (278 in total) with a small water jet succeeded in only 16% of all cases. In Fig. 3 the location of these receptive fields is indicated. Most of them appeared to be small, but some were considerably larger. Occasionally, the water jet was applied contralaterally or on the head. No contralateral receptive fields were found, but receptive fields did occur on the head. Units whose receptive field could not be found were often sensitive to mechanical stimulation with the membrane, which also stimulates the inner ear. Of 133 units stimulated by the membrane, 57% displayed a response. Almost all sensitive units (95%) showed a sustained response, phase-locked to the fundamental frequency of the stimulus. 25% of the units displayed an additional transient response component at the onset and/or offset of the tone.

The area where afferent projections from the trout lateral line system have been found (Schellart et al. 1990, 1992) is indicated by the dashed contour in Fig. 1. Somata with projections to the acousticolateral area of the torus semicircularis are located in the medial nucleus (de Wolf et al. 1983). The area where acousticolateral activity was recorded is indicated by the closed contour (Fig. 1). In four cases, HRP was injected to check the stereotactically established electrode tip positions. Excellent agreement was found in two cases, with tip positions at a depth of 700 and 460 μm (stereotactically) and 650 and 500 μm (HRP) respectively. In the other two cases, the HRP marks (700 and 500 μm) were found at about 60% of the depth estimated stereotactically (1220 and 795 μm). We conclude that a substantial number of our recordings were obtained outside the medial nucleus.

Neurones recorded in the relatively small part of the medulla that we investigated had receptive fields scattered along the entire trunk canal and also on the head. No topographical representation was observed in this part of the medulla. This is in agreement with earlier physiological studies of the eel and the goldfish (Alnæs, 1973; Caird, 1978) and anatomical studies of the projection of lateral line afferents (Claas and Münz, 1981; McCormick, 1983, 1989; Meredith, 1984; Meredith et al. 1987; Puzdrowski, 1989). No topographical relationship could be detected with respect to transient response characteristics or latency. The spontaneous activity increased in the rostral direction with a slope of 0.36±0.33spikes s ^−l% ^−l of the medulla length (95% confidence interval for a t-distribution). There appeared to be no relationship between recording depth and spontaneous activity.

The frequency response characteristics of a unit stimulated by the oscillating sphere at three different distances, and by the membrane, are shown in Fig. 4. The gain curves, relative to acceleration, are similar. Phase values proved to be consistent over the entire frequency range measured, although the phase lag tended to increase less rapidly at higher frequencies when stimulated by the oscillating sphere. The shape of the gain curves suggests that this unit obtained input from canal neuromasts (Kroese and Schellart, 1992).

According to the near-field relationship (Harris and van Bergeijk, 1962), the gain at 12.35mm distance should be 27dB less than at 4.35mm. However, the gain measured for this unit decreased considerably less than expected with increasing distance. Stimulation of another unit with the oscillating sphere at different distances produced similar results. It appears that the sensitivity of these units, after correction for the near-field distance effect, increased with increasing distance from the stimulus source. Of course, the range of this effect will be limited.

For 20 units the recording time was long enough to permit the measurement of the response over a wide frequency range, while a stimulus was applied by the membrane (18 units), the sphere (1 unit) or both (as for the low-frequency unit shown in Fig. 4). Ten units responded to relatively low frequencies (<50Hz) and responded maximally to frequencies between 70 and 120Hz. The other units responded to higher frequencies, with maximal responses to frequencies above 150Hz. The phase of four low-frequency units and three high-frequency units is shown on a linear frequency scale in Fig. 5. The mean spontaneous activity of both classes of units was the same. The response of these two types is shown in Fig. 6. The two classes of units also showed a significant difference (x²-test, P<0.05) with respect to their location in the medulla. The stereotactically measured recording sites are indicated in Fig. 1. Low-frequency units were mainly encountered in the dorsal part of the medulla, i.e. in the medial nucleus or nearby. High-frequency units were found ventral to the medial nucleus, mainly in the nucleus octavus descendens (de Wolf et al. 1983; Nieuwenhuys and Pouwels, 1983).

The latency of 59 units was 18±6ms (mean ± S.D.; Fig. 5). There was a significant difference of 4.2ms (t-test for difference between two means, P<0.05) between the mean latency of high-frequency units and that of low-frequency units (Fig. 5). Furthermore, high-frequency units were more sensitive (16±10nm) than low-frequency units (40±19nm; t-test for difference between two means, P<0.01). It appeared that 14 other lateral line units stimulated by the oscillating sphere were much less sensitive than the 20 units above (see Discussion).

All 10 attempts to localize the receptive fields of the high-frequency units were unsuccessful, but the receptive fields of low-frequency units were successfully localized in five out of six attempts. The localized unit whose responses are shown in Fig. 4 is a typical low-frequency unit. The recordings that were lost during a frequency scan are indicated in Fig. 6 by an asterisk. No low-frequency units (except possibly the one that was lost) responded to higher frequencies. Seven out of ten high-frequency units were lost during the frequency scan, which means that they may have been responsive to still higher frequencies. It is concluded that the low-frequency units were lateral line units and that the high-frequency units received their input from the inner ear.

Units displaying a transient response component (N=21) had lower resting activities, ranging from 1 to 33spikes s⁻¹ (13±11spikes s ⁻¹) than the entire population (27±20spikes s ⁻¹) of responding units (t-test, P<0.01).

Transient responses generally consisted of an increase in spike activity at tone onset and a decrease at tone offset. A number of units exhibited exceptional response characteristics. The responses of three units showed a delay, with respect to the start of the tone burst, of about 100ms. One unit displayed a transient response consisting of an increase in spike activity at tone onset and offset. Two other units showed a decrease at the onset and an increase at the offset of the tone. One of the latter units also responded to visual stimuli applied with a flashlight. One more vision-sensitive medullary unit was found, but whether this unit was also sensitive to mechanical stimulation was not determined. It is concluded from these exceptional responses that interaction of inputs from the different sensory systems occurs in the medulla.

Primary lateral line neurones have their medullary termination sites predominantly within the medial nucleus (Claas and Münz, 1981; McCormick, 1983, 1989; Meredith, 1984; Meredith et al. 1987; Puzdrowski, 1989; Schellart et al. 1992). We studied single-unit activity of rainbow trout neurones that responded to a mechanical stimulus. In Table 1 some characteristics of the medullary units are compared with those of primary afferents of the trunk lateral line (Kroese and Schellart, 1992) and with the characteristics of units sensitive to acousticolateral stimulation in a midbrain centre, the torus semicircularis, which show more variable responses (Nederstigt and Schellart, 1986; Schellart and Kroese, 1989).

It is inferred from the measured properties that the low-frequency units of this study were higher-order lateral line neurones. Most of the low-frequency units had longer latencies than would be expected for primary afferents (see Analysis section of Materials and methods; Fig. 5). Some responded transiently, which is very uncharacteristic for primary afferents. Also, when stimulated by an oscillating sphere, the sensitivity of these units appeared to be less than that reported for primary afferents (Kroese and Schellart, 1992). A possible explanation for the observed sensitivity differences in relation to the stimulus source is given below. Attempts to localize the receptive fields succeeded in 16% of all cases. However, this is a low estimate of the percentage of medullary units receiving lateral line input, because only a fraction of the total surface of the fish could be scanned by means of the water jet. Moreover, the head region was not scanned systematically. Although only a small area of the medulla was investigated, the neurones encountered had receptive fields which were distributed over the entire fish body (Fig. 3). This is thought to indicate that there is, at most, a very global somatotopic map of the lateral line sensory surface in the medulla. This is in agreement with earlier physiological studies of the eel and the goldfish (Alnæs, 1973; Caird, 1978) and anatomical studies of the projection of lateral line afferents (Claas and Münz, 1981; McCormick, 1983, 1989; Meredith, 1984; Meredith et al. 1987; Puzdrowski, 1989).

The oscillating membrane was used to test whether units responded to mechanical stimuli, but this also stimulated the fish’s inner ear. Because the eighth nerve projects to an area just ventral to the medial nucleus (McCormick, 1983, 1989; Nieuwenhuys and Pouwels, 1983; Meredith, 1984; Meredith et al. 1987), it was not initially clear to what extent each sensory system contributed to the measured response. Auditory responses in the medulla oblongata have been studied in the herring (Enger, 1967) and in the goldfish (Page, 1970; Sawa, 1976). Although interspecific differences exist, the high-frequency units of the rainbow trout show qualitatively similar responses to the auditory units of these other species.

The response characteristics of 20 units were measured over a wide enough frequency range to allow classification as low-frequency units or high-frequency units. High-frequency units were recorded ventral to the low-frequency units, which supports the conclusion that these units were predominantly auditory units and lateral line units, respectively. HRP stains suggest that the depth estimated by the stereotactic method is often too great (see Results), probably because the electrode compressed the tissue. This should affect the stereotactic estimate of deeper-lying cells more than that of superficial cells. In spite of the unreliability of the stereotactic estimates, the difference in location of high-and low-frequency units (Fig. 1) appears to be significant. Although this evidence alone is not very strong, it supports our conclusions.

Also, the difference in mean latency agrees with a longer propagation time needed to traverse the lateral line nerve. A similar latency difference was observed in the torus semicircularis of the trout (Schellart and Kroese, 1989). All units to which a receptive field could be assigned were found to be low-frequency units. Therefore, it appears that lateral line input and auditory input are mainly processed separately. In contrast, convergence of lateral line and auditory input is a common feature in the torus (Schellart and Kroese, 1989).

Auditory units in the medulla of the trout appear to respond up to a maximum frequency of about 400Hz (Fig. 6B). This is in agreement with the psychophysically determined audiograms (Abbott, 1973; Hawkins and Johnstone, 1978) and with the frequency range of acousticolateral units in the torus semicircularis (Nederstigt and Schellart, 1986; Schellart and Kroese, 1989).

Surprisingly, we encountered two units which proved to be visually sensitive, at least one of which also responded to mechanical stimulation. On the basis of the recording sites, the visual sensitivity is considered not to originate from the Mauthner neurone (Zottoli and van Horne, 1983). Recently, it has been shown that, in the toadfish primary lateral line, afferent activity can be modulated by visual stimuli (Tricas and Highstein, 1990, 1991). The other exceptional responses described in the Results section also support the idea of sensory integration taking place in some medullary units.

From the observed phase locking (always a unimodal period histogram) it is conjectured that the input from the two populations of hair cells (with opposite directional sensitivities) in the neuromasts does not converge at this level (Wubbels, 1991). This may be a general property of the medullary lateral line units of fish (Alnæs, 1973; Caird, 1978; Bleckmann et al. 1989). However, lateral line input from hair cells with the same directional sensitivity in several neuromasts may converge on medullary units, as concluded from the sizes of the receptive fields (Fig. 3).

Evidence for convergence of lateral line input also comes from the measured sensitivity of medullary units. The water displacement at the skin of the fish necessary to evoke 20% modulation of primary afferent fibres at F_max appeared to be about 12nm (Kroese and Schellart, 1992). The 10 lateral line units were, on average, 10dB less sensitive. In general, medullary units stimulated by the membrane are about as sensitive as the primary lateral line afferents. However, 16 localized medullary lateral line units which were stimulated with the oscillating sphere were 15–65dB less sensitive than primary afferents. Because the F_max of 14 of these medullary units could not be measured, the estimated sensitivity will be less than the actual sensitivity, but probably by no more than about 10dB. In addition, however, a small stimulus field generated by the oscillating sphere does not evoke a response in all neuromasts of a large receptive field. Thus, the measured sensitivity of medullary units may be expected to be lower than that of primary afferent fibres. A poor sensitivity of medullary lateral line units has been reported before (Paul and Roberts, 1977; Caird, 1978). The apparent increase in the sensitivity of the units with increasing distance from a small stimulus source, as was observed for two medullary units stimulated by the oscillating sphere, might be related to the large size of the receptive field. Of course, the estimation of the sensitivity is limited because the presence of the fish body was not taken into account in this calculation (Harris and van Bergeijk, 1962). Caird (1978) found that the response of second-order lateral line units in the medulla of the goldfish decreased faster than the primary afferent response when an oscillating sphere was moved away. If, in the medial nucleus, spatial summation of several trunk neuromasts takes place from hair cells with the same directional sensitivity, as was conjectured before in this Discussion, then our results agree with those of Caird (1978).

When a dipole source, vibrating parallel to the fish, is moved directly away from the surface (as in our experiments) the primary afferent response will decrease. Projected on the skin of the fish, the sites where directional reversal of the dipole field occurs (Sand, 1981; Wubbels, 1991) will diverge with increasing distance, thereby increasing the area of the skin stimulated in the same direction. The response of a medullary unit with a large receptive field will decrease less than expected (proportional to distance⁻³), because at greater distances more neuromasts within its receptive field will be stimulated. When the same dipole source is moved away from the centre of the receptive field, parallel to the skin (Caird, 1978), the site of the reversal of displacement direction in the lateral line system, caused by this dipole field, will move with respect to the receptive field (Sand, 1981; Wubbels, 1991). Therefore, in this case, the response of a medullary unit will decrease because of the increasing distance, but also because fewer neuromasts in its receptive field will be stimulated in the same direction. The functional organization of the trunk lateral line system, suggested here, appears to be well suited for the detection of relatively large and/or distant vibration sources.

We would like to thank Dr B. L. Roberts for his comments and an anonymous referee whose suggestions helped to improve the manuscript. This work was supported by the Netherlands Organization for Scientific Research (NWO-BION).