A hydrodynamic topographic map in the midbrain of goldfish Carassius auratus

Fish use the lateral line to detect weak water motions(Dijkgraaf, 1963). The peripheral lateral line of fish is separated into two subsystems: the superficial neuromasts (SNs) located on the skin, and the canal neuromasts(CNs) located in subepidermal canals (e.g. Coombs et al., 1988). Lateral line neuromasts are low-frequency (<150 Hz)hydrodynamic receptors, which may contain up to three thousand mechanosensory hair cells(Bleckmann et al., 1989a). The ciliary bundles of lateral line hair cells project into a jelly-like cupula,which is deflected by minute water movements. While SNs are sensitive to water velocity, CNs are more sensitive to water acceleration (e.g. Münz, 1989; Kroese and Schellart, 1992; Engelmann et al., 2000).

Lateral line information reaches the brain via lateral line nerves, which terminate throughout the medial octavolateral nucleus (MON) of the medulla (Puzdrowski,1989). Cells of the MON project bilaterally to the midbrain torus semicircularis (TS), which relays information to the tectum opticum and to the diencephalon (Wullimann,1998). From the diencephalon lateral line information finally reaches the telencephalon (e.g. Wullimann,1998). To study the physiology of the peripheral and central lateral line, earlier investigators used stationary vibrating spheres(dipoles) as a stimulus source (e.g. Harris and van Bergeijk, 1962; Münz, 1985; Bleckmann and Bullock, 1989;Coombs et al., 1996, 1998). Using dipole stimuli applied in still water to a fish rigidly fixed in a holder, researchers uncovered many physiological properties of the lateral line, such as vibration amplitude thresholds, dynamic amplitude ranges, and the phase coupling of peripheral and central lateral line units to sinusoidal water motions (e.g. Kroese and van Netten, 1989; Münz, 1985; Plachta et al., 1999; Engelmann et al., 2000).

Natural sources of lateral line stimuli are usually not stationary, and they rarely oscillate with constant frequency and amplitude. Instead they move around and produce low-frequency transient water motions with irregular amplitude and frequency fluctuations(Enger et al., 1989; Bleckmann et al., 1991) and pressure waves (e.g. Kalmijn,1988; Hassan,1993). In addition, swimming aquatic animals like fish leave a vortex trail in the water (Blickhan et al., 1992), which may persist for more than 1 min (Hanke et al.,2000) and may be used by piscivorous animals to track prey fish(Pohlmann et al., 2001; Dehnhardt et al., 2001).

To stimulate the lateral line with complex water motions, researchers moved small objects along the side of the fish (e.g. Bleckmann and Zelick, 1993; Müller et al., 1996; Montgomery and Coombs, 1998; Wojtenek et al., 1998). Using such stimuli, peripheral and central lateral line units revealed specific response properties that would not have been discovered with stationary vibrating spheres. Examples include central lateral line units responding to a moving object with a short single peak of excitation(Mogdans and Goenechea, 1999),units respondingu with multi-peaked, long-lasting excitations(Mogdans and Bleckmann, 1998),and units responding with inhibition(Mogdans and Bleckmann, 1998). Moving object stimuli also revealed that certain central lateral line units are highly sensitive to the direction of object motion(Bleckmann and Zelick, 1993; Müller et al., 1996; Wojtenek et al., 1998).

The lateral line does not respond to a moving object but to the water motions caused by that object. Therefore it is desirable to measure and compare water motions caused by the object with the neural responses caused by these water motions. To do so, we combined particle image velocimetry (PIV)with physiological recordings in the midbrain TS (TS for torus semicircularis). Our experiments revealed three types of toral lateral line units. Type TS1 units responded maximally while water velocities across the surface of the fish's body were increased due to sphere movement. It is most likely that units of this type received their input from SNs. Type TS2 units responded only while the sphere passed a certain location on the fish's head or body, but responses did not positively correlate with the water motions at this location. TS2 units were topographically organized. A third type of units(TS3) discharged only after the sphere had passed the fish. The responses of these units could not be attributed to either sphere-caused water motions or the pressure gradients caused by the moving sphere.

Data were collected from a total of 29 goldfish Carassius auratusL. ranging in length (snout to tail) from 10-12.5 cm. Animals were acquired from local suppliers and maintained in large communal tanks at 18-22°C on a daily 10 h:14 h L:D cycle.

For surgery fish were anesthetized in iced water. In addition the skin at the preparation site on the head was infused with the local anesthetic Xylocaine® (ASTRA-Chemicals). A small area of skin was then carefully removed. An opening (about 4 mm×4 mm) was made with a drill in the skull over the contralateral midbrain. Excess fatty tissue and fluids were aspirated to reveal the optic tectum contralateral to the stimulation side. Following surgery, Pancuronium Bromide® (0.1-0.2 mg g^-1 i.m.; Organon Teknika, Oberschleissheim, Germany) was administered to immobilize the animal. The immobilized fish was then transferred to the experimental tank (40 cm×48 cm×25 cm) filled with standing tapwater at room temperature(20±2°C). To reduce background vibrations the tank rested on a vibration-free table (Micro-g, TMC®; Peabody, MA, USA).

For electrophysiological recordings fish were positioned on a styrofoam support. The fish was tipped up at about 20° with the dorsal surface of the head just above the water surface. The exposed brain was kept moist with Ringer's solution. To keep the animal in a fixed position, its head was glued with Histoacryl® (Braun Melsungen, Melsungen, Germany) to the tip of a small Plexiglas rod attached to a micro-drive. Aerated freshwater was pumped at a rate of 70-100 ml min^-1 over the gills through polyethylene tubing inserted into the fish's mouth.

Hydrodynamic stimuli were generated with a sphere (diameter 8 mm) moved on a linear path along the side of the fish from anterior to posterior (AP) or from posterior to anterior (PA). The path of the sphere was guided by a ball-bearing rail system, positioned at the level of the trunk lateral line canal, which enabled us to move the sphere without recoil or vibration. A total length of 23.8 cm on the rail system was marked by trigger contacts. The extremes of movement exceeded the length marked by the trigger contacts by about 5 cm, to allow deceleration of the sphere. During each trial the sphere velocity was held as constant as possible. We changed the mean sphere speed from trial to trial gradually from slow to fast (3 to 36 cm s^-1,respectively). In all experiments the mean velocity of the sphere was calculated from the time the sphere needed to pass the two trigger contacts. Moves were interrupted by a 5-15 s pause to allow the water in the experimental tank to calm down after each sphere run. The manual control of sphere velocity proved to be sufficient for a first classification. Independent of sphere velocity, all units recorded consistently fell into one of three classes.

A dipole stimulus (50 Hz, 100 Hz, 160 μm peak-to-peak amplitude) was generated by a vibrator (V 106; Ling Dynamic Systems, Royston, USA) at the holder of the sphere. For each unit the sphere was placed along the side of the fish (from snout to tail) in 10 mm steps to examine the unit's responses to dipole stimuli.

To measure the water movements across the surface of the fish's body, the water in the experimental tank was seeded with white neutrally buoyant particles (Vestosint 1101, donated by Hüls AG, Marl, Germany; mean diameter 100 μm). Two laser diode modules (output <10 mW) with cylindrical lenses were mounted on opposite sides of the tank to generate two light sheets of about 1 mm thickness. One laser was installed to cast the light sheet vertically between the moving sphere and the fish, the other laser produced a horizontal light sheet 1-3 mm below the sphere to avoid shadows(Fig. 1). The light reflected by the particles was recorded with two CCD cameras. The pictures were controlled online on a LCD monitor (Citizen LCDTV; Irvine, CA, USA), stored on a VCR (Panasonic NVF70 HQ) and digitized off-line with a video capture card(Miro Video DC 30; Pinnacle Systems, Mountain View, CA, USA). The card was set to a compression factor of 1:10, which does not affect the quality of the PIV analysis (Freek et al., 1999). Off-line analysis of the pictures was performed with a computer (IBM compatible AMD K6) and MatLab 5.1 (cf. Hanke et al., 2000). A custom-made script file correlated subimages of succeeding frames according to the principles of digital PIV,first introduced by Willert and Gharib(1991). The error of this calculation depends on various parameters, including particle velocity. For high particle density, slow water velocities result in smaller errors, e.g. 5%at 2 cm s^-1, whereas larger water velocities result in larger errors, e.g. 10% at 10 cm s^-1. However, suboptimal particle density and small-scale velocity gradients resulted in a bias of the measured velocities towards lower values. The calculations were converted into a vector plot, showing the calculated and interpolated water flow within the camera view (e.g. Fig. 3D). The audio channel of the video recorder was fed with the signals from the two trigger contacts that marked the position of the sphere.

Single unit and few-unit recordings were made in the contralateral TS,using either indium electrodes (impedance <1 MΩ; Dowben and Rose, 1953) or quartz glass filaments (outer diameter 80 μm), whose 30 μm cores were filled with platinum tungsten (impedance <4 MΩ at 1000 Hz) (Thomas Recordings, Marburg, Germany). The tip of the indium electrode was plated with gold and then platinum to reach a diameter of 4-10 μm. The six platinum-tungsten electrodes were arranged in a Reitboeck microdrive (Eckhorn System®, Thomas Recording, Marburg, Germany), which allowed a computer controlled recoil-free thrust for each electrode with a precision of ±2μm. The six electrodes of the system were arranged in a rectangle consisting of two rows. The distance between two electrodes was fixed by guiding tubes at 250 μm, resulting in a rectangle size of 250μm×500 μm.

Action potentials were amplified (DAM 80®, WPI or Eckhorn System®,Thomas Recordings), bandpass filtered (300-3000 Hz), displayed on two oscilloscopes (DL 1300A®; Yokogawa, Atlanta, GA, USA) and stored for off-line analysis on an 8-channel digital tape recorder (DTR 1800®;Biologic, Science Products, Hofheim, Germany). Units were isolated from background noise using window discriminators (WPI, Model 121® or Eckhorn System®, Thomas Recording). These delivered TTL (transient transistor level) pulses for each action potential above a selected level or within a level window. A sweep of action potentials as well as the discriminator level selected was displayed on the oscilloscopes during data aquisition. The spike waveforms were visually inspected to discriminate single unit recordings from few-unit clusters. If not otherwise stated this report presents data from single unit recordings.

For final analysis, TTL pulses of the window discriminators were digitized(Instrunet® and SuperScope II®, GWI®) and stored on a computer(Apple Macintosh, Power PC 7300®). The times of occurence of TTL pulses relative to stimulus onset were calculated with a precision of 100 μs. Ongoing activity was calculated for at least 100 ms prior to stimulation and expressed in spikes s^-1. Raster plots and peristimulus-time histograms (PSTHs) were computed across 10 stimulus repetitions. Peak spike rates were determined from the bin (binwidth 50 or 100 ms) in the PSTH with the greatest number of spikes and expressed in spikes s^-1.

In order to distinguish lateral line units from visual, auditory, and vestibular units the following stimuli were also applied.

Sound stimuli. Acoustic stimulation was presented by a loudspeaker(see materials in Plachta et al.,1999), by human voice and handclapping. Units that responded to any of these stimuli were considered auditory units.

Vibratory stimuli were applied by slightly tapping the edge of the experimental tank with the tip or rubber ball of a pasteur pipette. Units that responded to this stimulus were assumed to receive vibratory input.

Photic stimuli were applied by switching on and off the light of the binocular used to monitor the position of the recording electrodes. Units that responded to changes in illumination or failed to respond to the moving object in complete darkness were assumed to receive visual input.

Unimodal lateral line units were distinguished from all other units in that they responded even in the dark to a sphere (diameter 8 mm) vibrating close by the fish or to a sphere that was moved along the side of the fish. Unimodal lateral line units were classified without any doubt. However, units responsive to acoustic and/or to vibratory stimuli often also responded to the vibrating or the moving sphere. We did not attempt to learn whether these units were unimodal acoustic or received additional lateral line input.

In 15 cases the location of the recording site was marked with a small electrolytic lesion. If indium electrodes were used, lesions were obtained by passing a current of 1-7 μA d.c. for 2-15 min through the electrode. If platinum tungsten electrodes were used, a high frequency a.c. current(frequency >500 kHz, I_eff>7 μA) was applied for 10-15 min. Fish were deeply anesthetized with MS 222® and perfused intracardially with freshwater teleost Ringer's solution followed by 5%glutaraldehyde solution in 0.1 mol l^-1 phosphate buffer (pH 7.4). Brains were removed, postfixed and cut at 15 μm in a transverse plane parallel to the electrode penetrations. Sections were stained with Cresyl Violet and analyzed microscopically. Digital images of slices with lesions were stored.

We recorded 120 toral lateral line units from 29 fish, of which 57 responded reliably when the sphere was moved along the side of the fish from anterior-to-posterior (AP) or from posterior-to-anterior (PA) but did not respond to acoustic, optic or vibratory stimuli. Only these 57 units were further analyzed. In 15 cases the recording site was verified with a lesion. All lesions were located in the nucleus ventrolateralis of the TS (e.g. Fig. 2).

Of the 57 units, 48 (83%) did not respond to a stationary vibrating sphere(stimulus frequencies 50 and 100 Hz, peak-to-peak displacement amplitude 160μm), while 9 units (17%) responded to both the moving sphere and to the stationary vibrating sphere. 17 of the 57 units, called TS1 units (TS = torus semicircularis), responded with a reproducible single burst (duration 200-800 ms; end of burst defined by a gap of neural activity of at least 500 ms) while the object passed the fish and, in most trials, continued to fire for up to 20 s afterwards. The late spike activity of TS1 units was not predictable across stimulus presentations (e.g. see raster plots in Fig. 3A). 35 of the 57 units,called TS2 units, responded only with a reproducible single spike or with a single short (<500 ms) burst of spikes while the object passed the fish(e.g. Figs 3B,right, 5A). Five out of the 57 units,called TS3 units, barely responded while the object passed the fish. However,300-500 ms after object stop these units fired strong, long-lasting bursts(e.g. Fig. 3C). PIV measurements acquired while recording the responses of TS3 units showed that at burst onset (after end of sphere movement), peak velocity and the water velocity averaged over the fish's body were far below the maximal values in a given trial (for one example see Fig. 3C,D).

In TS1, TS2 and TS3 units, sphere positions at response onset were different for motion in the AP and PA directions. For instance, if the sphere,moving AP, caused a reproducible burst of a TS1 unit while passing a distinct point P along the longitudinal axis of the fish, sphere movements in the opposite direction could cause a response before or after the sphere reached P. In contrast, the ill-defined late response component of a TS1 unit only occurred after the sphere had passed a distinct rostro-caudal location of the fish. It was obvious that at the times of neuronal responses of TS1 units,water velocities across the surface of the fish were enhanced and that the flow patterns were not reproducible from trial to trial. In contrast to TS1 units, the first stimulus that evoked responses of TS2 units (N=35)always occurred before the sphere had crossed a distinct location (P) along the longitudinal axis of the fish. Although not analyzed quantitatively it was apparent from the particle movements in the video frames that water velocities at response onset were never pronounced at this location, i.e. TS2 units did not respond in proportion to water velocity at location P.

A further analysis of TS2 unit responses revealed a midbrain lateral line map (Figs 4, 5, 6). TS2 units recorded in the anterior TS responded while the sphere passed the head or the anterior body of the fish, whereas TS2 units recorded in the caudal TS responded while the sphere passed the posterior body of the fish. Responses from TS2 units in the medio-lateral part of the TS occurred while the object passed the medial part of the body (Figs 4, 5, 6). Fig. 4 shows the main response areas for all TS2 and 13 TS1 units. Main response areas of TS2 units map topographically in the midbrain. In five instances two TS2 units were recorded simultaneously with electrodes spaced 250 μm apart (for an example see Fig. 5A,B). Action potentials were usually recorded first with the electrode placed more rostral if the direction of object motion was AP. If the direction of object motion was PA,the opposite was true (Figs 5, 6).

Units that responded exclusively to the moving sphere were, on average,more ventrally located in the lateral TS than the units that responded exclusively to the vibrating sphere. Although the vertical separation was not absolute (Fig. 7), the difference in depth distribution was significant (Wilcoxon-Mann-Whitney U-test, Z=2.326, U=489.0625, P<0.01).

A moving object causes both a transient pressure pulse and transient water motions, followed by a long-lasting, ill-defined wake that contains high water velocities but almost no pressure fluctuations(Mogdans and Bleckmann,1998).

Consequently type I primary lateral line afferents, i.e. velocity-sensitive afferents receiving input from SNs(Mogdans and Bleckmann, 1998;Engelmann et al., 2000, 2002), continue to respond after an object has passed the fish(Mogdans and Bleckmann, 1998). Like type I primary lateral line afferents, TS1 units of Carassiusalso responded to the transient water motions generated by the passing sphere and to the water motions in the wake of the moving sphere. Therefore it is conceivable that TS1 units received their input mainly, or even exclusively,from the SN system. The same may be true for type MB units found in the TS of the catfish Ancistrus, which also responded to the wake of a moving object (Müller et al.,1996).

The responses of TS2 units resemble those of type II primary lateral line afferents, known to receive CN inputs(Mogdans and Bleckmann, 1998),and it is conceivable that TS2 units of goldfish received input from CNs.

If that interpretation is correct, that part of the torus semicircularis which processes the hydrodynamic information caused by a moving object might contain at least two subsystems: one for the analysis of SN information (TS1 units), and one for the analysis of CN information (TS2 units). A similar physiological subdivision is present at the level of the MON(Kröther et al., 2002). Separation of SN and CN information at the level of the midbrain does not exclude the possibility that TS1 unit responses can be modulated by CN input or vice versa. A TS3 unit was first found in the midbrain of the catfish Ancistrus sp.(Müller, 1996). TS3 units responded with excitation only after the object had passed the fish, and they may belong to a third subsystem whose functional properties are not yet understood.

Fish have a visual topographical map in the optic tectum (for a review, see Northcutt and Wullimann, 1988)and a computed topogaphical acoustic map in the nucleus centralis of the torus semicircularis (e.g. Schellart et al.,1987). Computed lateral line maps have been found in the clawed frog Xenopus laevis (Zittlau et al., 1986) and in the Axolotl Ambystoma mexicanum(Bartels et al., 1990). In these animals the direction of water surface waves is represented systematically in the optic tectum. Using a stationary vibrating sphere as a stimulus, Knudsen (1977) and Bleckmann et al. (1989b) found a topographical lateral line map in the midbrain of catfish and rays,respectively. Whether these units received input from the canal system or from SNs was not investigated. In the present work, we found in the goldfish that units which respond to a moving object with a short burst topographically map in the torus semicircularis. If our interpretation that these units received inputs from CNs is correct, the canal system preserves the information about the spatial distribution of CNs. Our physiological data support conclusions drawn from behavioral studies on the mottled sculpin Cottus bairdii. The initial orienting and approach behavior of this fish to a dipole source relies on canal rather than SN input(Coombs et al., 2001). It is most likely that C. bairdii uses the point-to-point spatial representation of a source location along the sensory surface of the lateral line system to localize a prey object(Conley and Coombs, 1998).

We do not know whether the neuronal units described here were recorded from neuron somata or fibres. The spike width of the 57 midbrain units analyzed in this study showed a unimodal distribution (0.8-2.4 ms), i.e. distinct spike populations were not apparent. Therefore it is unlikely that the classification into TS1, TS2 or TS3 units corresponds to a difference between somata recordings and, for instance, incoming MON fibre recordings. In any case further investigations are necessary to reveal more details about midbrain lateral line subsystems and lateral line maps.

We thank Arthur N. Popper (Maryland University, USA), Hermann Wagner (RWTH Aachen, Germany) and Jacob Engelmann (University of Bonn, Germany) for discussion and suggestions and Catherine McCormick (Oberlin College, Ohio,USA) for help with anatomical analysis. This work was supported by a grant of the Graduiertenförderungsgesetz NRW to D.P. and the Deutsche Forschungsgemeinschaft to H.B.