Postural Changes Occurring During One Month of Vestibular Compensation in Goldfish

The recovery of postural and locomotory control following a unilateral vestibular sensory deficit is the process of ‘vestibular compensation’ (von Bechterew, 1883; Schaefer & Meyer, 1974; Pfaltz, 1983; Precht, 1983; Galiana et al. 1984). The previous paper (Ott & Platt, 1988) described the use of goldfish as a ‘simplified’ vertebrate model to quantify some of the earliest aspects of the behavioural recovery. In this paper we report on the longer-term changes during compensation in goldfish, again to provide a time course for future investigations correlating behaviour to physiological or biochemical mechanisms for compen-sation.

In fishes and other vertebrates, a distinction is often made between ‘acute’ and ‘chronic’ stages of compensation (Werner, 1929; Schoen, 1950; Kolb, 1955; Schaefer & Meyer, 1974; Jensen, 1979; Pfaltz, 1983; Yagi & Markham, 1984). The acute stage lasts from minutes to a few days, depending on species, and consists of extreme torsional movements of the body accompanied by extreme eye deviation or spontaneous eye movements (nystagmus). There may be a rapid initial decline in the magnitude of these responses. The chronic stage may last for years and usually consists of some residual tilt and inducibility of abnormal nystagmus. These deficits may gradually return to normal or reach steady-state values that are not fully normal.

Often the deficits in the chronic stage are revealed only under experimental conditions. A useful behaviour for monitoring vestibular function is the dorsal light reflex (DLR) of many fishes. When a light is presented from one side of a fish, it rolls its back towards the light to a postural equilibrium position between the direction of gravity and the direction of the light source (von Holst, 1935). After a unilateral vestibular lesion a fish may appear to be compensated, swimming normally in vertical light, yet it may have a measurable change in its DLR from the preoperative equilibrium tilt when light is from one side (von Holst, 1950). Thus the change in the relative contribution of gravistatic input to posture can be followed postoperatively to establish a curve for compensation with time (Schoen, 1950).

The present paper describes the continuing changes in postural behaviour in goldfish during the compensation period for 30 days following the unilateral lesion. In contrast to the very abrupt recovery from ataxia seen in the first hour after a lesion (Ott & Platt, 1988), the recovery of symmetrical postural responses appears to require roughly 2 weeks, during which time the variability of responses also declines greatly.

Goldfish (Carassius auratus) of the comet variety were used, ranging in length from 50 to 70 mm from snout to base of tail. They were maintained in aquaria in the laboratory on a 12 h: 12h day: night cycle and fed three times per week.

Surgical details are given in Ott & Platt (1988). Briefly, the utriculus and semicircular canal organs of the right side of an anaesthetized fish were removed through a small opening in the skull, and the trailing nerves were cut. The opening was plugged with saline agar, and the fish revived. After measuring the acute behavioural parameters, the fish was placed in a special test chamber for periodic! measurement of the behaviour during the chronic period.

An experimental tank, in which postural behaviour could be recorded under controlled light levels and direction (Fig. 1), was set up in a small room that could be completely darkened. A translucent plastic cylinder (Nalgene bottle) was fitted inside the tank as a testing chamber. The fish to be tested was placed in a transparent plastic confining tube 25 mm in diameter suspended horizontally inside the translucent chamber. A clear plastic gate closed off the front end, with a hole punched in it for water flow, and netting part way down the tube also allowed water flow but prevented the fish from moving away from the focal point of the camera. In this tube the fish could breathe and tilt easily, but not turn round. The chamber was aligned so the nose of the fish faced a recording camera (Grass C4 using Linagraph Ortho Paper, Kodak 1930), placed roughly 15 cm in front of the fish, with a planoconvex lens on the front of the tank to enlarge the image of the fish. Three ‘pencil’ fluorescent lights were mounted longitudinally around the tank, one directly above and one to each side (at 90°), all equidistant (18-5cm) from the centre of the chamber. Cardboard shades directed the light only towards the tank and decreased the reflections of light around the chamber, and the translucent chamber prevented the fish from seeing objects around it.

After the room had been darkened, the camera and the light sequences were controlled automatically by an electronic circuit (designed on a Proto-Board Model 203; Continental Specialties Corp.) to minimize any observer interference during the experiment or observer bias in data collection. The lights cycled through an on-off sequence presenting the fish first with light from its right side, (then its left side, then from above, while the camera ran continuously in single-frame mode at a speed of 0·25 mm min-1, resulting in a picture frame every 18 s.

When a dark part of the cycle was added, a strobe flash (Grass PS-22) synchronized to the camera shutter provided light for photos without giving time for the fish to orient to light. We found that a given fish showed the same variance among 20 frames whether one picture was taken every 2 min during a 40-min period, or 4 frames min-1 were taken during a 5-min period, or whether the strobe was on or off during the light cycle. A timing rate of approximately 4frames min-1, light periods of 6 min each, and three light periods in a cycle, gave 23-24 frames for each light setting, and a total of 69-72 frames per test day for each fish.

The developed paper film was analysed using a film reader having a rotatable plastic disc with scribed lines extending over a fixed marginal ring of degree markings (Fig. 2). The paper film was run through a track to keep it aligned as it had been in the camera, with vertical set at 0°, and the tilt angle, a, of the fish’s dorsoventral axis was measured relative to vertical. To read data from a single frame the scribe lines were manually aligned with the dorsoventral axis of the head of the fish (centre of forehead to gular cleft), and the tilt angle read from the ring. Several pictures could not be used because the fish image was blurred by movement. Pictures were used for data only when the pupils of both eyes of the fish were visible, minimizing the perspective distortion of tilt angle measurements if the fish turned its nose away from the camera; we estimate this criterion required longitudinal alignment of the fish to within ±20° of the central axis of the tube. To prevent observer bias in selection of appropriate tilts, all photos that met these criteria were measured, usually representing about half the total number of pictures taken.

A later set of data was taken from three operated and two sham-operated fish using the same technique except that, instead of a film camera, a black and white video recorder was used and data were taken directly from the video monitor. A piece of polar-coordinate paper placed directly on the test chamber made tilt determinations relative to vertical accurate, regardless of video camera tilt. Measurements were made in 15-min sessions for each light condition. In the later set, a light was added below the fish so that four lights, above, below and to the right and left of the fish, were tested. Measurements from the two methods were very similar, so the data were combined for some analyses.

Three groups of five operated fish were tested. Measurements on the first group were made for 1 or 2 days before the operation to obtain preoperative control values, and on each weekday after the operation up to 30 days. Because of the rapidity of recovery, measurements on the second and third group were made only on days 0 (day of operation), 1, 2, 3, 8, 9, 15, 16, 22, 23, 29 and 30. Five preoperative control values were taken in the second group. Two sets of three sham-operated fish were tested in the same manner as the second operated group.

Several environmental factors were controlled throughout the experiments. Each operated fish was individually housed in a free-floating translucent cage in a tank in a dimly lit room when not being tested. All tests were run in the middle of the day to minimize circadian variability, and both sets of animals were tested in the spring or early summer to minimize seasonal variability, since some possible seasonal differences in responsiveness were seen (J. F. Ott, unpublished data). Throughout each test month all tank water temperatures were kept at 20·5 ± 1·5°C. Variable effects of hunger or satiety on visual-vestibular integration in goldfish (Traill & Mark, 1970) were minimized by consistently feeding the fish at the end of the day, after testing.

Each day’s test started with the fish in the chamber in front of the camera for 10 min undisturbed to allow the fish to become calm in the chamber before starting the test light cycle.

The mean and standard deviation of tilt angles for each light angle were determined for each fish on each test day. The mean control tilt angles (second control run for each fish, unless no second was available) were then subtracted from each of these postoperative tilt angles to obtain net tilts. Using a least-squares method of linear regression, a best-fit line was determined for these net tilts for days 1,2,3, 8, 9,15,16, 22, 23, 29 and 30. This line had a slope estimating overall rate of compensation for each light angle for each fish between day 1 and day 30. The Student’s t-test was used to test the null hypothesis that there was zero slope, or no compensation over time, for each light angle for each fish. For all analyses, the criterion for statistical significance was taken as P<0·05.

For each test day a grand mean net tilt was calculated using the mean net tilts for each light regime of the 10 fish, and the standard errors for these daily grand means were calculated. Daily values for the grand mean net tilts of the 10 fish were tested by t-test to determine if they were significantly different from zero (mean control tilts). A two-way analysis of variance (ANOVA) with repeated measures was made to determine whether the different light regimes, postoperative times, or interaction between these two factors had any effects on the means of the tilt angles.

For each fish, a total change in net tilt was determined by finding the difference between the net tilt immediately postoperatively and at 30 days.

Preoperative control values show that goldfish have a readily observable dorsal light response (DLR). When light was from the top (vertical), the control tilt was 0° (vertical). When light was from +90° or −90° (right or left side, respectively), preoperative control tilt angles were approximately +10° or −10°, respectively, thus towards the light. The control tilt angles when light was from the side were not always exactly symmetrical across 0°, showing variability between individuals that in some cases could produce rather large skews (see Fig. 3).

The chronic stage of vestibular compensation began immediately after the ataxic( stage had ended, approximately 10-20min postoperatively, and was characterized mainly by a lateral tilt of the fish that was biased towards the operated side at all light angles tested. The changes in tilt during the first few days are shown in Fig. 3, and the changes during a month are shown in Fig. 4.

Tilts were greatest immediately after the end of the ataxic stage. The first measurements showed a tilt towards the operated side at all light angles. Recovery back towards control values was very large during the day of operation (day 0), often within the first hour (Fig. 3). The percentage change of this shift in relation to the total change over 1 month is shown in Table 1. After the initial large shift, tilts continued to change more gradually over the next few weeks. Extended measurements on a few fish at 3, 4, 6 and 12 months after the operation showed that values for those fish had not changed further, suggesting that some steady state had been reached within the first month (J. F. Ott, unpublished data).

The rate of compensation was estimated by using linear regression over the 30 days tested. The differences among the mean slopes at each light angle are shown in Fig. 4. Only the slope of the compensation curve for light from the operated side was statistically significantly different from zero. The two-way analysis of variance with repeated measures showed significant differences among light treatments (F = 24·06, P<0·005), and over postoperative times (F = 3·18, P<0·005), but no significant interaction between the two variables. The effects of light and time can be considered independently.

For the first 2 weeks, greater tilts were produced by light from the operated side than by light from the other two angles (Figs 3,4). Table 1 shows that when light was from the operated side, tilts by the end of day 0 had shifted by only 19% of their total change over the month. But when light was from above or the unoperated side, tilts by the end of day 0 had shifted by 49% or 71%, respectively, of their total change.

During the month tilts did not recover to the preoperative control values for light from either side, but they did re-establish symmetry. Fig. 4 shows that tilts towards light from the operated side reached a steady state at roughly 1-5 times the control value. Tilts towards light from the intact side gradually increased to reach a steady state, again at roughly 1·5 times the control value.

There was a large variability in both the duration and magnitude of the chronic phase. The majority of the fish reached steady-state values of tilt angles in 2 weeks (day 15) (Fig. 4). However, some individuals appeared to reach steady state immediately, showing no further change in tilt angle after the first day, but others showed tilt angles still changing towards normal during the first 2 weeks, particularly when light was from the operated side (Fig. 5, fish 1,2). The standard error of the mean reflects both the variation of the tilts of an individual on a particular day, and the variation of the tilt means of several fish on a particular day. The S.E.M. values were largest, approximately ±30°, immediately after the ataxic stage, then they dropped to values close to the preoperative values of approximately ±3° by the second week (see Figs 3, 4). Means for tilts when light was from the intact side usually had the smallest S.E.M. values of the three groups (except for the one aberrant value on day 30 in Fig. 4), whereas S.E.M. values of tilt means when light was from the operated side were typically largest. Thus the behaviour was more variable early on, and when light was causing the fish to tilt towards the operated side.

Recovery of the five sham-operated fish was quite different from that seen in the labyrinthectomized fish. The ataxic stage ended within less than 3 min after respiratory recovery from anaesthesia in the sham-operated group (Ott & Platt, 1988). In fish tested immediately upon recovery from the ataxia, the mean tilts for the sham-operated group were not significantly different from preoperative tilts when light was from any of the three tilt angles (Fig. 6). The main effect of the sham operation appears to be an increase in the variability of the response for the first few days. The sham-operated fish remained at normal mean tilts, whereas the operated fish recovered to new steady-state values that were not a return to normal means.

Removal of the utricle with the canal organs was the standard operation we chose for several reasons. We did not try removing just the utricular otolith (leaving behind the macula), or removing just the utricle, because we felt we could not know what functions might remain in partially damaged tissues of utricle or canals. Cutting the utricular nerve alone was very difficult in goldfish because the branch is short, just medial to the organs, difficult to section without damaging the canal nerve branches where they join it at the brainstem, and difficult to confirm after a month has passed. The canal organs probably influence compensation, but they are impossible to avoid damaging in a dorsolateral approach to the utricle. As discussed in the preceding paper (Ott & Platt, 1988), we decided it was important to make the operation consistent, so all three ampullae and canals were always removed with the utricle, and the nerves cut. The entire vestibular system of one side was not removed, because actual removal of the saccule and lagena, or even just cutting their nerves cleanly, requires deep surgery right behind the gills and under the brain. The saccule and lagena are considered to be primarily auditory, not postural, organs in most teleosts (see Platt & Popper, 1981), and in goldfish their removal has not notably affected postural control (Manning, 1924). We do not know how much the utricular removal might affect the function of the saccule and lagena, but we presume their postural contribution is at most minor (Schoen & von Holst, 1950). For all these reasons, we operated to remove the pars superior entirely, leaving the pars inferior relatively intact.

Earlier measurements of compensatory behaviour in fishes were collected by direct observation of a moving fish (see Schoen, 1950, for an example). But our preliminary studies using direct observations of goldfish (J. F. Ott, unpublished data) seemed to be non-objective. It was easy to take selective data, ignoring periods when a fish appeared nervous rather than calm. We had two observers independently taking data on tilt angles by direct observation of the same fish on the same days. We confirmed a bias when we found that these data sets showed the respective control means fluctuated differently from day to day, and the standard deviations for the two observers were very different. To reduce this bias we tried mechanical methods to connect the fish to measurement devices for automated entry of data on tape or to computer, but restraint of the animal always resulted in a lack of ‘cooperation’ by the fish. The method described using the camera or video was chosen as the best alternative. Although data points must still be read by an observer, they can be checked later by an independent observer, and a rigid criterion set for accepting or rejecting a given photo or video frame for data use. Using sets of such film data, we found that the results from two observers were not different from one another, so we considered the observer bias eliminated.

Although the photographic and video camera methods were found to be the only feasible ones, they had disadvantages. (1) The animal had to be confined to a small space, though not otherwise restrained. Fish occasionally moved abruptly as if to escape from the tube. Such movements increased the variation in measured tilts. (2) The photographic method used almost 3 m of film per fish per day, making film processing a limiting factor and film cost a limiting expense. The change to reusable video cassettes decreased costs somewhat, but processing time was still a factor. We feel that faster sampling with the camera does not decrease the reliability of data. However, very fast sampling (several frames per second) over a duration of less than 1 min might not provide an accurate measure, since the fish’s normal periods of relative activity and quiescence often lasted several seconds.

Many different kinds of light conditions have been used to induce the dorsal light response in different fishes (von Holst, 1935, 1950; Schoen, 1950). We could not get goldfish to swim steadily into a water current as tetras and cichlids do (von Holst, 1950). Without being able to force fish to hold position as well as von Holst could, we could not reliably use a narrow beam, and we could not reliably measure the angle at which light entered the eye. In von Hoist’s highly quantitative work, he was able to show that the angle of the DLR is determined by the angle of gravitational stimulus to the otolith organs, the light intensity, and the angle at which light enters the eye, not just the angle of light with respect to the vertical (von Holst, 1950). We found that a large, uniform hemifield of light gave better responses in goldfish than an intense collimated beam. Since we were using the DLR simply as a monitor for the time course of vestibular compensation, we felt this light stimulus was quite adequate.

In spite of methodological and observational differences from earlier work (von Holst, 1950; Schoen, 1950), all the fish so far studied share similar compensatory phenomena. There is an immediate postoperative bias of tilt angles towards the operated side at all light conditions, and then a shift back to new steady-state values with time. For goldfish these new values were roughly 1-5 times larger than preoperative values. Thus the long-term compensated fish had a DLR that produced tilt angles greater than those for the normal DLR. This increased output can be considered a change in the gain of the DLR system, indicating that light had more influence on tilting the postoperative fish than on the preoperative fish. Since the influence of gravity and light are normally linked in the DLR, the relative increase in influence of the light component could also result from a decrease in the influence of the vestibular component. We did not try decreasing the light intensity, or changing the relationship of the contrast hemifield to eye position, so we cannot be more precise about the quantitative amount of the gain change.

These changes in relative contributions for light and gravity in goldfish under these conditions are different from those described for the tetra (family Characi-dae), one of the fish used by Schoen (1950). Under her conditions, preoperative tilts in tetra were very large (+60°) compared with those of goldfish, indicating a relatively greater normal influence of light. We could not increase a goldfish tilt to the degree found in tetra by increasing light intensity. The postoperative tilts in tetra were only about 10° more than normal, to 70°; the relative change in gain of the light influence seems to be far less in tetra than in goldfish. It may be that since the light has a greater influence on the DLR in tetras, the vestibular loss has less relative effect than in goldfish. Alternatively, it may be that there are locomotor limits that prevent extreme steady-state tilts. Galiana et al. (1984) suggested how asymmetrical gains across the vestibular commissures might be used in feedback loops to compensate for asymmetrical labyrinthine inputs, and later updated this model to include vestibular-visual interactions (Galiana, 1986). Such gain changes would be reflected in apparent gain changes of the influence of light, as seen here.

The compensation phenomenon is usually represented by a curve similar to exponential decay (see below). A curved recovery is not evident in our plots, but cannot be excluded. However, our simple approach of linear regression still led to some insights. In this analysis the slopes were averaged over the whole month, and none was very large; the three slopes were not significantly different from each other. But the slope of the line describing compensation was significantly different from zero for light presented from the operated side, whereas the slopes for light from above or from the intact side were not (see Fig. 4). Schoen’s (1950) curves of the tetra and angelfish response also showed a stronger effect of light from the operated side, a return to a steady state, and a restoration of symmetry. But the shape of the curve and the time courses were different (tetra taking a week and angelfish taking a month to compensate), and her findings were not statistically supported. Methodological differences may also account for some of these differences.

These behavioural changes are similar to those in other animals, but the parameters measured often differ. Head tilt, spine curvature or limb placement are frequently used measurements (see Bienhold & Flohr, 1980; Putkonen et al. 1977; Schaefer & Meyer, 1973,1974; Jensen, 1979). Although the characteristics of the acute phase, such as posture and eye deviation, end abruptly and early in fish, in other animals these may continue into the later chronic phase.

Many researchers consider a curve of compensation to have two aspects: the early, rapid change of tilt angle (the acute stage), and the later, slower shift of tilt angle (the chronic stage) (Schaefer & Meyer, 1974; Bienhold & Flohr, 1980). In goldfish, the rapid change of tilt on the day of operation (day 0) blended into the later more gradual change, and we did not separate these changes into different stages for two reasons. First, there was extreme variability on the day of operation, as shown by the standard errors of the means (S.E.M. values). These S.E.M. values are too large to support statistically the presence of two phases. Second, the degree of effect depended on light direction. Although rapid changes were seen when light was from the top and from the intact side, a less rapid change was seen when light was from the operated side. For these reasons, dividing the chronic stage into a two-phase recovery does not seem justifiable.

Neither von Holst (1950) nor Schoen (1950) mention or measure the variability of their fishes’ behaviour. Bienhold & Flohr (1980) show standard deviations in response after labyrinthectomy in a large number of frogs (more than 100), suggesting very large variability. In our fish, we found three interesting aspects of variability. First, the variability of the tilts was greater immediately postoperatively and then decreased to a variability similar to that around control tilt values. Second, the variability of the DLRs for light from the operated side was greater than the variability of DLRs for light from the unoperated side or from above. Similar patterns have been found in other vestibular studies (Bienhold & Flohr, 1980). Third, a particular individual may have low S.D. values one day and high ones the next (Fig. 5). On particular days, an individual may be calmer with less motor variability than on other days, possibly as a function of the animal’s emotional state, as suggested by von Holst (1950).

Some variability may be induced by the effects of the operation itself. The sham-operated fish showed S.E.M. values as great as those of the operated fish, also decreasing with time. In the sham-operated fish, the horizontal semicircular canal was consistently cut. We do not know how much local tissue damage, endolymph disturbance or anaesthesia may have contributed to the variability. It will be useful in the future to examine the role of the saccule and lagena in compensation, based on our current knowledge of the physiological responsiveness of the various otolith organs.

Individuals showed large differences in how quickly they compensated. Whereas one fish appeared to compensate immediately, with no tilt difference from steady state after day 0, another fish’s tilt remained at high values late into the test period. Differences in compensation rate might correlate with levels of some neurotransmitter, as suggested for recovery rates after nigrostriatal lesions (Marshall et al. 1979; Marshall & Drew, 1981). Variability among animals might also result from a preoperative asymmetry in the normal peripheral input, so that unilateral removal of a more dominant labyrinth could produce more severe postoperative symptoms than removal of the weaker labyrinth (Galiana et al. 1984).

There are three main results from this work on goldfish. First, there is a characteristic set of postural behaviours immediately following the operation, which lasts only 15-30 min, and in a given fish all these characteristics end together (Ott & Platt, 1988). We term this very brief period the ataxic stage, and while it lasts the fish shows no coordinated swimming, often simply lying on the operated side, and is thus untestable for the DLR. This period may correspond to the acute stage of some workers (Schaefer & Meyer, 1974; Precht, 1974; Jensen, 1979: Fisch, 1973) but is much briefer than the acute stage in humans (Pfaltz, 1983; Fisch, 1973).

Second, the results show that when postural behaviour is once again testable, the tilt towards the operated side may decline dramatically in the first day (the acute stage of many other studies) compared with later days. Based on the statistical evidence (see Fig. 3), we feel that the early (sometimes called acute) steeper part of the curve is not distinctly separable from the longer tail (sometimes called chronic). Instead we use the term ataxic only for the very brief stage when locomotion is incapacitated, and chronic to include the entire compensatory response curve, starting from the time locomotion first allows testing.

Third, these changes after unilateral lesion do not depend on any peripheral reorganization. The macula of the utricle on the intact side has been examined by scanning electron microscopy to see whether compensation is associated with changes in the remaining peripheral sensory system (Platt & Jew, 1978). The sensory surfaces of the remaining utricle and canals of a compensated goldfish showed no changes in the normal asymmetrical pattern of cellular receptor orientations, up to 1 month after unilateral lesion. Thus the compensation of symmetrical posture in goldfish does not depend on forming a new, peripherally symmetrical input from the normal asymmetric sensory array in a single utricle.

Our results are consistent with the other reports on compensatory behavioural phenomena, although goldfish show a time course faster than that of tetrapods. It remains important to find out if sensory feedback is necessary to re-establish the correct motor controls to restore upright posture. The behavioural baseline will now permit physiological and biochemical probing to clarify the mechanisms for the remarkable central plasticity in adult animals undergoing vestibular compen-sation.

This work was supported partly by a grant from Sigma Xi to JFO. We thank both Tom Roalkvam and Jody Watkins for their contributions in the laboratory, and Drs Neal Barmack, W. Lou Byerly, David Jensen, Werner Graf and Roger Reep for useful discussions.