The mauthner-initiated startle response in teleost fish

Studies employing anatomical, physiological and behavioural techniques have provided strong evidence for the view that in teleost fish a startle response (the ‘tail-flip’ or ‘Mauthner reflex’) to acoustic or vibrational stimulation is mediated by the Mauthner cells (M-cells), a pair of giant intemeurones located at the level of the auditory-vestibular nerve in the hindbrain (see Diamond, 1971, for review; Eaton & Farley, 1975; Zottoli, 1976c). Kimmel, Patterson & Kimmel (1974) carried out the only previous cinematographic study which used both a fast frame rate (1 frame/ 6·7 msec) and vibrational stimulation to elicit startle behaviour probably initiated by the M-cell. Their study of the zebrafish (Brachydanio rerid) showed that the behaviour consists of an initial full flexion of the body followed by a weaker flexion in the opposite direction.

Other cinematographic studies (Hertel, 1966; Webb, 1975) have been done to develop mechanical models of rapid acceleration in trout (Salmo gatrdneri). However, these provide only limited insight into the underlying neurophysiological mechanisms because the filming rate of 1 frame/15·6 msec was not sufficient to adequately document a response mediated by the M-cells. That is, in larval zebrafish, the M-cell can fire within 6 msec from the start of a vibrational stimulus and the mechanical muscular response begins 2 msec later (Eaton & Farley, 1975). In adult goldfish, the Mauthner spike (M-spike) has a minimal latency of 2 msec following the acoustic stimulus, and the electrical response in the muscle begins 2 msec later (Zottoli, 1977). In addition, the use of an electrical stimulus in the studies on trout further complicates the physiological interpretation.

In the present study we used a fast frame rate (1 frame/5 msec) and replicate trials to quantitatively analyse the startle behaviour of goldfish and zebrafish to acoustical-vibrational stimulation. Our purpose was to develop insights into the neurophysiological basis and functional importance of the startle response. This analysis revealed the presence of a characteristic response pattern, probably mediated by the M-cell, and prompted a further analysis of examples of startle behaviour in 11 other teleost species to determine whether a characteristic behaviour exists among diverse fish with M-cells. Species studied are from seven orders, including both primitive and advanced groups (Gosline, 1971), a variety of body forms (Breder, 1926), and diverse habitats.

Experimental animals were obtained from laboratory stocks and were chosen to maximize phylogenetic, ecological and body-form diversity (Table 1). Previous studies established the presence of the M-cells in five species of this study (Table 1), and we used histological techniques (see below) to identify the M-cells in the remaining eight species. Both freshwater and marine fish are included. Several individual zebrafish, hatchetfish, catfish, garfish and calico rockfish were studied whereas only one individual was used in replicate trials for other species. Tropical specimens were identified according to the descriptions of Sterba (1962).

Experiments consisted of presenting a vibrational stimulus to the animal in a glass aquarium and photographing the response with high speed cinematographic equipment. The experimental aquarium contained 22·8 1 and had dimensions of 24·1 × 50 × 30·5 cm (width, length, height). Water temperature of the experimental aquarium (Table 2) was kept close to that of the home aquarium. Both the pre-test acclimation time in the experimental aquarium (30 min to 18 h) and interstimulus intervals (5–15 min) were adjusted to increase probability of obtaining a response. Experiments on each species were continued until several startle responses had been photographed.

The stimulus was produced by a 200 g rubber mallet which was hinged at the end and allowed to fall through an arc of 90° to strike the back centre of the aquarium. The fulcrum was in the plane of the back of the aquarium and the length of the mallet arm was 32·5 cm. The mallet was held suspended by an electromagnet and was silently released by turning off the current to the magnet. An electronic circuit synchronized filming and stimulus presentation. A touch-sensitive microswitch on the mallet signalled the stimulus onset by turning on, for 10 msec, a neon signal light in the photographic field. The subject was visually isolated from the stimulus and signal light. Control experiments, without stimulus presentation, verified that the subject did not respond to the falling of the magnet or the turning on of the camera. The signal light also provided a calibration for the film speed.

Photography was done with a Milliken DMB5-2, 16 mm movie camera with 1 msec exposure at 200 frames. sec⁻¹. These characteristics, and the 10 msec signal light, permitted latency measurements with an accuracy of ±4 msec. Four 300 watt floodlamps provided approximately 1000 footcandles of illumination (incident light measured at water surface with a Gossen Luna-Pro light meter). The floodlamps were turned off between trials to prevent heating the water. In separate experiments, filming was done from above and in front of the aquarium. In general, fish were allowed to swim freely in the tank and were usually filmed in midwater at various orientations.

The responses of goldfish to visual stimulation were studied by dropping a golf ball into the experimental aquarium from 17·5 cm above the water surface. A small magnet was glued to the ball and, as above, we used a timing circuit and electromagnet to release the ball and to synchronize filming. Only those startle responses which preceded the stimulus striking the water were counted as positive to visual stimulation. Latency measurements were not possible in these experiments because the time when the stimulus was first seen by the fish is unknown.

For single-frame data analysis we used an L-W Photo Inc., 224-A projector and made drawings and measurements from the projected images. To determine displacement speed we measured the position of the rostral-most extremity of the head at consecutive 5 msec intervals when the animal was viewed from above. A negligible error is introduced because the pathway of the head follows a very small arc. To measure angular velocities we determined the angle of the midline of the head in dorsal view sequences at consecutive 5 msec intervals. The error in this technique is estimated to be ± 5°.

For histological analysis, animals were anaesthetized with tricaine methanesulphonate and transcardially perfused with fish saline followed by 10% formalin. Brains were removed and stored in 10% formalin until they were embedded in paraffin and sectioned at 10 to 15 μm. Sections were stained according to either the Klûver-Barrera method for myelin and nerve cells (Luna, 1960), or Ziesmer’s (1952) modification of Bodian’s silver stain. Light microscopic examination of all species revealed the presence of a single pair of giant neurones which met the criteria for identification as M-cells: that is, location at the level of the eighth root, presence of an axon cap, conspicuously large ventral and lateral dendrites, and a decussating axon (see Zottoli, 1976 a).

The principal analysis of this study was carried out on six responses each from goldfish and zebrafish. Figs. 1,b, c and 9a show examples of responses of these species. All twelve responses from the two species appeared to have characteristic form and latency (Table 2) and are also similar to those previously described by Kimmel et al. (1974) for the adult zebrafish. As seen from Fig. 1b,c and further substantiated in Figs. 2–4 this startle response is divided into two phases. The initial phase consists of a short-latency, unilateral contraction of the trunk and tail. This contraction results in the fish assuming a C-like shape with both the head and tail displaced to one side. This is the most constant phase of the response and we name it the ‘fast-body-bend’ (Fig. 1b, c). The fast-body-bend is followed by the second phase which consists of a more variable movement pattern resulting from a straightening of the tail. During this phase the two species were propelled on the average 0·98 and 0·97 L (body lengths) from their initial position (Fig. 2). This variable, second phase we term the ‘return-flip’ (Fig. 1b, c).

Figs. 2(a1) and 2(b1) show the displacement pathways taken by the rostral-most part of the head during the response. The initial position of the fish (dotted silhouette) and direction of the fast-body-bend were standardized to facilitate comparisons. There is an initial period of about 15–20 msec during which the response pattern is stereotyped, and the head follows the same pathway from trial to trial. To further illustrate the stereotyped pathway, Fig. 3 shows in larger scale each of the data points in the first 35 msec of the six responses of the goldfish. The initial stereoytpic phase (first four points) corresponds to the fast-body-bend shown in Fig. 1,b, c. During the return-flip the fish turned at different rates so that after too msec the animals’ positions were quite variable with respect to the initial position (Fig. 2).

The biphasic nature of the response is clearly seen in graphs of displacement speed and angular velocity (Fig. 4). The fast-body-bend is characterized by a sudden increase in both displacement speed and angular velocity. During the fast-body-bend phase the animal achieves its greatest angular velocity and displacement speed (see Table 2). The end of the fast-body-bend occurs when the angular velocity returns to approximately o° per sec, whereas displacement speed, although decelerating, remains relatively high for a period of time as the fish executes its return-flip.

The above finding of a characteristic startle response behaviour in goldfish and zebrafish prompted us to determine whether similar behaviour is found in other diverse fish with M-cells (Table 1). We analysed from 1 to 4 startle responses for each species and found that the behaviours fell into three categories : (a) biphasic responses, such as those of the goldfish and zebrafish, in which the initial phase consisted of a maximal, unilateral contraction to one side followed by a variable flip of the tail in the opposite direction (Fig. 1,a, d–g) ; (b) sudden darts forward, the ‘fast-forwarddisplacement’ (Fig. 11) and; (c) in spiny eel only, a sudden bilateral contraction of the body musculature resulting in a retraction of the head (Fig. 1j). All these responses had similar latencies, but as discussed below, their angular velocities and displacement speed characteristics were quite different.

Biphasic responses were the predominant behaviour pattern observed, and were recorded in 36 out of a total of 38 responses analysed in fish other than calico rockfish and the spiny eel which did not display this form of startle behaviour. Examination of the displacement pathways (Fig. 5) reveals that the initial phase of this response was stereotyped among individuals within a given species, although in some cases we filmed only a few examples. In the trout, the initial pathways were similar in three examples during the first 15 msec (two of these responses are shown in Fig. 5 a) after which time the fish turned at different rates. Similar results were obtained in four examples from the garfish (two are shown in Fig. 5f, g) and two examples from the characid (Fig. 5c). As in goldfish and zebrafish, the initial stereotypic phase in other species also corresponds to the peak in the angular and displacement velocities (Fig. 6). The position of these fish after 100 msec varied from 0·5–1·5 L from the initial position.

Three species (black ghost, kelp bass and calico rockfish) exhibited a second category of response, the ‘fast-forward-displacement’, which consisted of a sudden dart forward. For the calico rockfish (Fig. 1i,5 k) the fast-forward-displacement was the only type of startle response observed. Fig. 5,b, h contrast displacement pathways of these darting movements (asterisks) with the biphasic startle response. These movements had the same latencies as the biphasic startle responses observed for the majority of species. However, the animals failed to achieve velocities comparable with those of their biphasic responses (Fig. 6f vs. i and h vs. k). In addition, there was no early peaking in the angular velocity or displacement speed (Fig. 6,i, k) and in the case of the response of the black ghost, the angular velocity maximum did not correspond to a single displacement maximum (Fig. 6i). These differences justify the distinction between the two categories of responses, and suggest that different neurophysiological processes underlie them.

All the responses of the spiny eel consisted of bilateral contractions of the body musculature (Fig. 1j) which resulted in a withdrawal of the head (Fig. 5d, e). The response is not biphasic, although there are simultaneous peaks in the angular velocity and displacement speed curves (Fig. 6e).

In some species, the animal’s position in the tank may have influenced the type of response that was obtained. In halibut, which we filmed from the side, only a slight movement of the fins and a withdrawal of the eyes could be obtained when the animal was resting at the bottom in five trials. Lack of responsiveness in these cases seems consistent with the adoption in this species of protective coloration which matches the background. However, a biphasic startle response was obtained when the animal was stimulated while swimming. Similar conditions may have been responsible for our obtaining only the fast-forward-displacement, or head retraction in calico rockfish and spiny eel, respectively. During these trials the animals were resting on the bottom of the tank.

In most species splaying of some or all fins occurred simultaneously with the initiation of the fast-body-bend. The number of fins which responded varied among trials. Fin movements are illustrated in Figs. 7 and 9 for the kelp bass and goldfish. In Fig. ga, b the goldfish respond by the second frame, and all the fins become splayed with the same latency. Fin movements are noted by the changes in ray patterns. These fin motions may determine the pathway of the animal’s movement, increase stability, and in spiny-finned species such as the kelp bass, serve a defensive function as well.Eye and opercular movements, probably involved in the response (Diamond, 1971) could not be clearly seen.

The pectoral fins play an important, though variable, part in the startle response of hatchetfish. In some cases the initiation of the fast-body-bend and flexion of these fins occurred simultaneously, whereas in other cases the pectorals were not involved in the initial phases of the response, but were activated later. Fig. 8 shows an example of a marbled hatchetfish startle response where pectoral fin movements precede (frames 2 and 3) and follow (frames 13 and 14) a fast-body-bend (frames 5–9). Following the second pectoral fin response the animal left the water (frame 15) for 175 msec and landed 8·4 cm away. The fish was not in view while above the water surface.

The startle responses to the visual stimulation elicited by a golf ball falling into the tank appeared identical to the biphasic startle responses elicited by the vibrational stimulation. Fig. 9 compares side views of goldfish responding to the two forms of stimulation. Out of eight trials, the goldfish responded five times before the golf ball hit the water. Other species were not tested for responses to visual stimulation.

Our data provide quantitative evidence that there is a characteristic biphasic startle response behaviour found in a wide variety of teleost fish having M-cell systems. This behaviour consists of a short latency (Table 2), unilateral contraction of the trunk and tail, such that the animal assumes a characteristic C-like shape with the head and tail displaced to one side (Fig. 1,a–h). The initial movement is followed by a less extensive contraction to the opposite side (Fig. 1,a–g). The result of the behaviour is that within 100 msec the animal is usually displaced from 0·5–1·5 L from its initial position (Figs. 2, 5). Two species did not display the characteristic response pattern under our conditions (Fig. 1i,j)

We propose that the biphasic startle behaviour we observed in the majority of 6 BXB 66 species is initiated by the M-cell. This hypothesis is strongly supported by a combination of previous physiological and anatomical data from goldfish and zebrafish. The M-cell systems in these two species are alike, and all evidence indicates that the system mediates a rapid reflexive movement. The M-cells receive excitatory electrical input from fibres of the eighth root and in turn directly activate the spinal motor neurones (Diamond, 1971; Eaton & Farley, 1975; Eaton, Farley, Kimmel & Schabtach, 1977).

Examination of species without M-cell systems would not provide a crucial test of this hypothesis because in the present study, some fish with M-cell systems did not display the characteristic behaviour and others displayed it only part of the time or under certain conditions. The best test of the hypothesis would be to record from the M-cell and simultaneously photograph the response with high-speed cinematographic equipment. Such a study has not yet been attempted.

However, the available data demonstrate that although other cells are involved, the M-cell appears to play an essential role in the production of most examples of startle responses studied. In restrained goldfish, Yasargil & Diamond (1968) recorded a strong contralateral contraction of the body musculature in response to electrical stimulation of the eighth root and activation of the M-cell. Eaton & Farley (1975) recorded a single M-spike in 116 out of 118 startle responses in restrained zebrafish larvae responding to vibrational stimulation. No M-spikes were recorded in 11 cases when the animal gave no response. More recently, Zottoli (1977) recorded from the M-cell and trunk musculature in free-swimming adult goldfish. Zottoli recorded single M-spikes in response to acoustic stimulation in 35 out of 36 cases when there was a maximal contraction of the contralateral body musculature, and the M-spike was absent when the ipsilateral musculature contracted in 38 cases. No M-spike was recorded in 52 cases when the goldfish failed to give a tail-flip. In the goldfish and zebrafish, a variety of neural units fire during the startle response, but the most constant feature of such records is the presence of an initial M-spike.

The physiological and anatomical data from common hatchetfish are also consistent with the idea that the M-cells initiate a rapid reflex movement (Auerbach & Bennett, 1969). In other species we studied there are no previous physiological experiments but the major anatomical features of the M-cell systems are similar to those of the goldfish and zebrafish. The anatomical and behavioural consistency support our hypothesis that the M-cells initiate a common type of behaviour in these species as well.

Therefore, we propose the name ‘Mauthner-initiated startle response’ for the characteristic startle behaviour described for the majority of species in this study. The previous designation ‘tail-flip’ is misleading because the most characteristic feature of the response is the initial contraction of the body which leads to a lateral displacement of both head and tail. The terminology ‘Mauthner reflex ‘or ‘Mauthner-mediated response’ is unjustified at present because the data are not sufficient to determine at what point the influence of the M-cell ends (see below).

Analysis of replicate examples clearly demonstrates that the Mauthner-initiated startle response is biphasic in zebrafish and goldfish (Fig. 2). This also appears to be the case in other species for which we studied fewer replicate examples (Fig. 5). The first phase, the fast-body bend, is a short latency stereotyped turning movement which appears to involve an extensive unilateral contraction of the body musculature (Fig. 3). These features lead to the proposition that the fast-body-bend phase is the direct result of the activation of a single M-cell and its spinal motor neurone pool. A test of this hypothesis is to cinematographically analyse startle behaviour following unilateral and bilateral lesions of the M-cells. This hypothesis predicts that the fastbody-bend to one side would be absent after destruction of the contralateral M-cell.

Webb (1975) did not find stereotyped acceleration responses in trout probably because he used a slower frame rate. Using similar measurement techniques our mean maximum displacement for zebrafish (42 L/sec, n = 6) is approximately the same as the maximum of 37 L/sec seen in the example of Kimmel et al. (1974) who used a frame rate comparable to ours (1 frame/6·7 msec vs. 1 frame/5 msec.) However, for trout we recorded a maximum of 24 L/sec, 12 msec after the start of the response (Table 2) whereas previous studies with slower frame rates (1 frame/15·6 msec) obtained maxima of 9·5 and 15·5 L/sec, 78 and 75 msec into the response (Hertel, 1966; Webb, 1975). In previous studies the average displacement speed, which is not as sensitive to frame rate, is close to the value we obtained. Over the first 100 msec of the response, we obtained values of 6–9 L/sec compared with 5–7 L/sec in earlier studies (Hertel, 1966; Webb, 1975). These comparisons show how slow frame rates can obscure important details of the startle response behaviour.

The second phase of the Mauthner-initiated startle response, the return-flip, is a variable contraction of the body in the direction opposite to that of the fast-body-bend. The return-flip appears to begin when the angular velocity returns to zero, and has an elevated displacement speed. In garfish, the return-flip appeared to be delayed or inhibited (Figs. 1 h, 5f,g). This suggests that the fast-body-bend and returnflip are controlled by separate, although perhaps interdependent, physiological mechanisms. The garfish examples support the idea that the return-flip is an active process and not simply a passive mechanical consequence of the fast-body-bend. The return-flip is probably not mediated by a second ipsilateral or contralateral M-spike because in the combined 151 examples reported for larval and adult fish, the startle response was always accompanied by one M-spike only (Eaton & Farley, 1975; Zottoli, 1977).

It is our view that other startle response behaviour, such as the fast-forwarddisplacement of the black ghost, kelp bass and rock fish, or the rapid contraction of the head of the spiny eel, are probably not initiated by the M-cell. In the case of the fast-forward-displacement, there is no maximal body contraction, as is expected for behaviour produced by the M-cell. In the species so far examined, the M-cell appears to predominantly innervate the contralateral motor neurones (Diamond, 1971), so it also seems unlikely that the M-cell mediates a behaviour involving an extensive bilateral contraction such as seen in the spiny eel.

In adult fish, the Mauthner-initiated behaviour is thought to be important in the avoidance of predators, such as birds, which strike the water surface from above (Diamond, 1971 ; Zottoli, 1976b). Our observation that the response is also activated by visual stimulation supports this idea. Diamond (1971) has argued that the Mauthner-initiated response is non-directional in terms of whether the left or right cell fires in the presence of an acoustic stimulus in the near field. Our data indicate that there is an additional source of unpredictability built into the system; between separate trials, 20 msec after the animal starts contracting, its position begins to become variable. Such variability is consistent with the strategy of escape, and could be useful to prevent a predator from compensating for the response. Unpredictability is also a feature of rapid escape responses of other animals such as the hawk moth (Roeder, 1962).

The fast-body-bend probably has two functions in predator avoidance. It may be a preparatory movement (see Webb, 1975; Weihs, 1973) in which the fish initiates a turn and prepares to swim away by a flip of the tail. At the same time, the most vulnerable part of the animal’s body, the head, is abruptly displaced and is no longer a stationary target. According to this view, the return-flip represents the main propulsive action (see Webb, 1975) which begins the fish’s escape from a threatening stimulus. This idea is supported by the response of the garfish, which under our conditions gave a fast-body-bend but not a return-flip, and therefore remained close to its original position (Fig. 5f,g).

The few examples of startle behaviour which did not appear to be Mauthner-initiated responses provide some insight into the possible significance of the M-cell system. Since those behaviours had latencies the same as the Mauthner-initiated responses, it appears unlikely that this giant fibre system was evolved simply to bring about a fast response. This is consistent with observations of Bullock & Horridge (1965) and Kennedy (1975) on giant fibre systems. It is likely that the importance of M-cell system lies in activating, roughly simultaneously, a large body of musculature to bring as strong a body bend as possible.

This research was made possible by the generous loan of cinematographic equipment from Dr John Hunter, National Marine Fisheries, La Jolla, CA. We thank Dr Sven O. E. Ebbesson for supplying the technical support for our histological analysis and Dr Theodore H. Bullock for providing facilities. Drs Theodore H. Bullock, Donald S. Faber, Charles B. Kimmel and Steven J. Zottoli provided valuable comments on the manuscript. Support was provided by Individual National Research Service Awards to R. C. Eaton and R. A. Bombardieri from the National Institutes of Health, by research grants to Dr Bullock from the National Science Foundation and the National Institutes of Health, by German Science Foundation grant Me520/1-4 and SFB33 to D. L. Meyer, and by grants to Dr Ebbesson from the National Aeronautics and Space Administration (NGR 47-005-186) and the National Institutes of Health (Ey 00154-05).