Quantification of vestibular-induced eye movements in zebrafish larvae
© Mo et al. 2010
Received: 30 March 2010
Accepted: 3 September 2010
Published: 3 September 2010
Skip to main content
© Mo et al. 2010
Received: 30 March 2010
Accepted: 3 September 2010
Published: 3 September 2010
Vestibular reflexes coordinate movements or sensory input with changes in body or head position. Vestibular-evoked responses that involve the extraocular muscles include the vestibulo-ocular reflex (VOR), a compensatory eye movement to stabilize retinal images. Although an angular VOR attributable to semicircular canal stimulation was reported to be absent in free-swimming zebrafish larvae, recent studies reveal that vestibular-induced eye movements can be evoked in zebrafish larvae by both static tilts and dynamic rotations that tilt the head with respect to gravity.
We have determined herein the basis of sensitivity of the larval eye movements with respect to vestibular stimulus, developmental stage, and sensory receptors of the inner ear. For our experiments, video recordings of larvae rotated sinusoidally at 0.25 Hz were analyzed to quantitate eye movements under infrared illumination. We observed a robust response that appeared as early as 72 hours post fertilization (hpf), which increased in amplitude over time. Unlike rotation about an earth horizontal axis, rotation about an earth vertical axis at 0.25 Hz did not evoke eye movements. Moreover, vestibular-induced responses were absent in mutant cdh23 larvae and larvae lacking anterior otoliths.
Our results provide evidence for a functional vestibulo-oculomotor circuit in 72 hpf zebrafish larvae that relies upon sensory input from anterior/utricular otolith organs.
Vestibular-induced behaviors can be used to measure vestibular function. For example, the VOR is a simple reflex of eye movements used for assessment of semicircular canal function in human patients . This robust reflex has also been used to assess vestibular function in several different species, including monkeys  and rats . The VOR is characterized by compensatory eye movements in response to linear or angular accelerations, and to any changes in head position with respect to gravity. Rotation around the earth vertical axis is sensed by the semicircular canal system, which generates an angular VOR. Linear or translational acceleration drives the linear VOR, which is activated by macular organs with otoliths or otoconia. Rotations about an off-vertical axis also stimulate mainly otolith organs if performed at a constant speed . In addition to responses to dynamic stimuli, a roll-induced static tilt VOR, which is a static eye-to-head position change, has been characterized in many species including tadpoles  and fish larvae .
The vestibular system is highly conserved among vertebrates with respect to both anatomy and the genes required for function of the inner ear . The inner ear of zebrafish larvae contains three developing semicircular canals, as well as anterior and posterior maculae destined to become the utricular and saccular organs, respectively. Due to the imaging techniques and genetic tools available, the zebrafish is emerging as a popular model for studies of neurobiology and behavior, including the molecular basis of auditory and vestibular function. Many mutations affecting hearing and balance to differing degrees have been identified in zebrafish [8, 9], giving rise to the need to quantitate auditory/vestibular function in larvae. Analysis of vestibular-induced behavior in larvae is useful for determining the severity of balance deficits in auditory/vestibular mutants, and may ultimately yield insight into the in vivo function of the gene products.
Studies of vestibular-induced eye movements in zebrafish describe either static or dynamic responses. A static, linear VOR was examined by tilting the head down in embryonic or larval fish and observing changes in the angle of the eyes . The resting eye angle was dependent upon the presence of the anterior otolith during development. Experiments focusing on the dynamic VOR in zebrafish larvae have yielded conflicting results. Easter and Nicola examined the development of several behaviors in zebrafish including the optokinetic response (OKR), as well as the angular VOR, and reported that both reflexes were present at 3 days post fertilization (dpf) . In a later study, Beck et al. used a microscopic system with infrared illumination to measure the angular VOR, and found that zebrafish did not have an angular VOR until 35 dpf . The authors suggested that the angular VOR of larval zebrafish observed previously was probably due to the OKR, in which eye movements are driven by changes in visual cues.
In two recent studies, a larval response to vestibular stimulation was detectable upon rotation about an earth horizontal axis [13, 14]. We sought to expand on these findings by exploring the effects of varying the experimental parameters and genetic backgrounds of larvae on vestibular-induced eye movements. We confirmed that the larval response was vestibular and not visual, and we determined the ontogeny of the vestibular-evoked response. Our experiments provide evidence that the anterior otolith is required for sensing changes in linear acceleration and evoking the VOR observed during rotations in the vertical plane.
Animals used in this study were wild-type zebrafish larvae in the Tübingen or long fin background, and mutants identified in the present study (rock solo AN66 ) or previous studies (cdh23 1619ag and synj1 Q296X ; [14–16]). The rock solo mutant (recessive lesion) was identified from an ethylnitrosourea mutagenesis screen using a Tübingen background. Fish embryos and larvae were kept at 30°C in E3 embryo medium . If necessary, 20 μl pronase was added into the medium to help larvae hatch out of the chorion at 2 dpf, followed by a change of E3 medium. All of the behavioral tests were carried out at room temperature (22-25°C).
A customized microscopic system was constructed to monitor the eye movements during rotation. As shown in Figure 1A, this system was composed of a Mitutoyo 5 × long working distance lens and a digital eyepiece (DCM300; Hangzhou Scopetek Opto-Electric, Zhejiang, China). Two 45° mirrors were placed between the objective and the digital eyepiece to guide the light. The U-shaped setup makes efficient usage of the rotation plate space and balances the motor load. All components were mounted on an aluminum platform, which could be rotated by a motor system with the supporting structure. Since all of the parts were fixed on the platform, no relative motion existed between the specimen and the eyepiece during the rotation process. This guaranteed a consistent viewing area during the experiment, which avoided blurring due to the relative motion between the camera and the fish. The same area under constant illumination also warranted relatively constant image brightness on each frame during a single trial.
A servo motor (Model# BE231DJ-NPSN, Parker Hannifin, Cleveland, OH, USA) and servo controller (Model# GV6K-U3E, Parker Hannifin) were used to rotate the platform, which held the microscopic system. A motor gear head (Model # 23SP100, Parker Hannifin) with a gear ratio of 1:100 was attached to the servo motor to reduce the speed and increase the torque. This servo motor system can control the angular position of the platform with a precision of less than 0.2 degree. An Ethernet cable connects the controller and a computer, allowing the computer to program the motor rotation profile and read the motor position.
All experiments were conducted in the dark with a cover box, if not otherwise specified. Larvae were mounted on a transparent specimen plate, which was supported by a 3 D micromanipulator. Each fish was trans-illuminated by an infrared LED with emission wavelength around 820 nm. A dark background with infrared illumination was used to avoid stimulating the visually-evoked responses. The LED was approximately 10 mm away from the specimen plate. That distance, as well as the small size of the LED and a wide emitting angle, produced a relatively homogeneous illumination.
The digital eyepiece recorded a video with a resolution of 1024 × 768 pixels at a speed of about 7.8 frames per second. This speed produced more than 20 frames at each rotation cycle with a period of four seconds. The infrared filter inside the digital eyepiece was removed to increase the infrared sensitivity.
After the larvae were properly mounted and positioned, the video recording was performed with ScopePhoto, the software that accompanied the digital eyepiece. Before stimulation, the recorded frames were used to check the illumination. Ten seconds after the video started, the motor was turned on by the controller software Motion planner (Parker Hannifin). During the experiment, the motor moved the platform in a sinusoidal profile of amplitude ± 45°. After a one-minute recording, which included about 13 cycle rotations, the video was saved as a Windows Media Video (WMV) format file for analysis.
During the rotation, the motor controller was also used to control the infrared LED to synchronize the video with the rotation. In each rotation cycle, the controller sent out a 100 ms pulse to the infrared LED when the angle of the motor was at about + 28° in the clockwise direction. The illumination was turned off during the 100 ms pulse. This resulted in a dark frame in the video in every rotation cycle. This dark frame was detected by the image processing program and was used to synchronize the eye movements with the rotation angle changes (see Movie 1 in Additional file 1).
The second step was to define the eyes from the head image, and a grayscale threshold was then applied to invert the grayscale image (Figure 2C) into a black-and-white image (Figure 2D). The threshold was chosen using Otsu's method . This was implemented in MATLAB with the function graythresh. Alternatively, manual adjustment of the threshold was sometimes used due to the image intensity change resulting from the motion of the E3 media around the fish head. A scale factor was then applied to the threshold. With a scaled threshold, the area containing the eye was defined for analysis (Figure 2E). An area threshold was then applied to the black-and-white image to remove the small dark island formed by the lens (Figure 2F). The two eyes were then separated in order to calculate the parameters for quantifying the rotations. A similar process was introduced in Beck et al., 2004. Different from that method, here we modified that previous method by confining the eye area to the iris, which is darker than the rest of the eye. This definition of the eye area facilitated the detection of the eye rotation along the anterior-posterior axis, as will be discussed in the next section.
After defining the eye region, features were extracted and calculated from the eye to quantify the eye rotation and then to evaluate the reflex. During the experiments, eye rotations on two planes were observable: rotation about the dorsal-ventral axis and rotation about the anterior-posterior axis (see Movie 1 in Additional file 1). The former is on the image plane and the eye angle can be used to quantify it. To measure this angle, the extracted eye region was approximated by an ellipse, and the angle of the long axis was used to represent that of the eye, shown as θ in Figure 2G (outline of upper eye in panel 2F). The eye angle, together with the mass center of the eye, determined the long axis of the eye (red line in Figure 2G). Both the angle and the mass center coordinate were an output of a MATLAB function regionprop. The short axis (blue line in Figure 2G) was drawn perpendicular to the long axis. The length of the long axis and short axis were also determined by the function regionprop.
Because the eye rotation around the anterior-posterior axis was not in the image plane, direct measurement of this rotation was not practical. The videos showed that the anterior-posterior rotation resulted in a change in the shape of the eye (see Movie 1 in Additional file 1). By measuring the shape change of the eye in each image frame, we could quantify the rotation by examining changes in total area or ratio of the long and short axes.
where X is either the ratio or the area. As shown in both the time domain plots in Figure 3C and their spectra in Figure 3D, the amplitude of the eye axis ratio change was higher than the total area change. Also, the ratio change was less sensitive to the intensity variation, which influenced both the long axis and short axis in a similar manner. We therefore used ratio changes to determine the amplitude of the vestibular-induced eye movements. One drawback of the ratio method is that the ratio saturates when the rotation angle becomes more than ± 10-20 degrees. However, this limitation did not affect our ability to detect differences among various stages of development or genotypes as seen below.
Our experiments demonstrate that zebrafish larvae have robust eye movements in response to rotation around an earth horizontal axis. At larval stages, both vestibular and visual input may contribute to eye movements. Two lines of evidence support the notion that we are measuring vestibular function rather than visual function. Firstly, motion of the eyes occurred in the dark using infrared illumination. Secondly, vestibular mutants did not respond or had attenuated responses to rotation on the platform. The 1619ag mutation in cdh23 used in this study causes a premature truncation of the extracellular domain of Cdh23 . Larvae homozygous for this allele have severe balance defects and lack microphonics, suggesting that mechanotransduction is absent in hair cells [8, 16]. Mutant cdh23 1619ag larvae did not respond to the stimulus under infrared illumination, indicating that hair-cell function was required for movement of the eyes in our experiments. In contrast to experiments in the dark, cdh23 1619ag larvae exhibited an OKR in response to rotation in bright light, eliminating the possibility that OKRs occurred under infrared illumination. Mutant synj1 larvae present the opposite phenotype of cdh23 1619ag larvae in that synj1 mutants exhibit partial vestibular function , but vision is lost . We observed that the OKR was absent in synj1 Q269X mutants (data not shown), indicating that the remaining vestibular-evoked responses were driven by the partially functional vestibular system, and not the visual system. With respect to developmental onset, vestibular-induced eye movements were detectable by 72 hpf. At this stage, zebrafish begin to exhibit OKR responses [11, 12] and the auditory/vestibular nerve appears to be fully functional . Our data indicate that the vestibulo-oculomotor projections are operational at this early stage as well.
Testing rock solo mutants allowed us to identify which hair cells mediate vestibular-induced eye movements in zebrafish larvae. In every case, the anterior otolith was absent in rock solo mutants, whereas the posterior otolith was always present. Mutant rock solo larvae failed to respond to earth horizontal rotation of the body, indicating that the anterior utricular macula is required for the response in larvae. In teleosts, the utricular otolith has been previously implicated in vestibular function [6, 20] whereas the posterior saccular otolith is thought to be primarily for hearing . Larval zebrafish begin to maintain balance, keeping their dorsal side up, as early as 3 dpf . Following a startle involving sound, touch, or vision, they can coordinate their motor system to produce a forward movement, with an upright posture. Experiments in adult frogs have also shown that the utricular otolith is important for sensing linear acceleration and gravity . Our experiments with rock solo mutants support the notion that the anterior otolith acts as a detector of linear acceleration in developing larvae.
The rotation around the earth horizontal axis using our set up presents a complex stimulus to the larval vestibular system. The stimulus includes linear acceleration components of centripetal and tangential acceleration, as well as changes in linear acceleration due to head tilt with respect to gravity. The vestibular system typically uses combined semicircular canal and otolith information to distinguish between translational and roll tilt movements . Both types of inputs should be able to evoke compensatory eye movements . However, we did not observe any eye movements in fish larvae during rotations about an earth-vertical axis. One reason is that the vertical-axis rotation we delivered would primarily stimulate semicircular canals, which are not fully developed in our preparation . A second reason is that the centripetal and tangential accelerations due to the off-axis location of the preparation produced only negligible otolith stimulation (See Additional file 3: appendix A). In contrast, during earth horizontal-axis rotation, there was a large change in linear acceleration that provided a sufficient stimulus to the otoliths. Thus, we infer that the vestibular-induced eye movements we observed in larvae were due to otolith stimulation evoked by the change in head tilt of the specimen. This hypothesis is supported by our experiments with rock solo mutants.
The eye movements we observed in larvae included changes in eye position about the dorsal-ventral axis. These movements represent compensatory VOR responses. Other movements include skewed vertical eye movements (about the anterior-posterior axis) and are most likely related to the ocular tilt reaction (OTR) present in lateral-eyed animals such as fish or rabbits (reviewed in 27). In such animals, the OTR is thought to be an otolithic righting reflex. Our measurement of the changes in ratio of eye area included both VOR and OTR movements. Despite the complexity of the eye movement, the vestibular-evoked changes in eye position are sufficiently robust, permitting comparison of responses among mutants and experimental parameters.
Our results indicate that zebrafish larvae exhibit robust eye movements in response to changes in head tilt with respect to gravity. Our data also confirms that zebrafish larvae rely on the anterior/utricular otolith for maintaining an upright position and coordinating movements with respect to gravity. Measuring the robustness of vestibular-induced eye movements will be invaluable for genetic or pharmacological studies of vestibular function in larvae. In addition, the ability to test vestibular function at earlier stages is especially useful for early lethal phenotypes or accessing gene knockdown with morpholinos as their effectiveness normally decreases over time.
Windows Media Video
ocular tilt reaction.
The authors wish to thank Xubo Song for suggestions in developing image processing algorithm, Robert Peterka, Greta Glover, Katie Kindt and Josef Trapani for helpful discussion, Norman Edelen and Lieu Than for fish care. This study was supported by NIH/NIDCD R01 HD055303 and DC6880 funding to A.N. and T. N., respectively, and P30 DC005983.