Background

Hair cells function as evolutionarily conserved sensory receptors within the auditory and vestibular systems of the vertebrate inner ear. These same cells act as sensory receptors for detecting water flow in the lateral line systems of amphibians and fishes [2–4]. In the lateral line system, bundles of hair cells, supporting cells, and innervating neurons located near the surface of the skin along the head and the body constitute functional units, called neuromasts, which detect hydrodynamic stimuli (changes in water velocity and turbulence) and sense low-frequency water flow. The hydrodynamic stimuli are transduced by hair cells from a mechanical stimulus into electrochemical signals. Hair cells release the excitatory transmitter glutamate onto afferent neurons of the lateral line ganglia, and the directional flow information is integrated in the hindbrain (lateral line nucleus and cerebellum), along with other sensory modalities interpretable by the central nervous system [3]. These features of sensory hair cell function are highly evolutionarily conserved, and their fundamental mechanisms for transducing sensory stimuli are similar across vertebrate species.

The zebrafish lateral line is a well-established model system for studying hair cell development and function, including studies evaluating the effects of mutations linked to deafness or environmental insults that damage hair cells [5–11]. In larval zebrafish, lateral line hair cells are located superficially along the skin, making them more accessible for experimentation than those of the mammalian ear, which are enclosed in the skull. This accessibility enables the use of various pharmacological treatments and imaging techniques, allowing researchers to apply reagents and visualize cellular processes in an intact larval fish. Yet despite the widespread experimental use of the zebrafish lateral line system in studies of hair cell sensory organ development, damage, and regeneration, there is currently no standard behavioral assay used to evaluate zebrafish lateral line function.

Rheotaxis—the ability to orient into and maintain position in water flow against oncoming current—is a behavior mediated by multiple sensory modalities, including visual, vestibular-tactile, chemotactile, and proprioceptive [12–14], and it includes the lateral line system. Rheotaxis helps fish maintain their position in the water flow and align their body upstream (i.e., positive rheotaxis [2,15]). A specific class of neuromasts, termed superficial neuromasts, is located on the surface of the body and has been shown to mediate rheotaxis [13]. Researchers have developed various methods to study rheotaxis in larval zebrafish, including measuring alignment in a rectangular flume [15], using computer vision to analyze swim patterns in radial flow [14], testing lateral line responses in a suction chamber [16], and employing a line actuator in a horizontal tube to reveal the role of the lateral line in detecting direction [17]. However, none of these approaches has been adopted for general use in laboratories.

Here, we describe a protocol and outline the essential components to reproduce our experimental rheotaxis assay, providing a validated behavior setup to study lateral line–mediated rheotaxis behavior in larval zebrafish. We sought to develop a behavioral assay and analytical methodology that is sensitive, spatiotemporally scalable, and robust enough to be adapted for use on a variety of species. The protocol incorporates computer vision and machine learning tools to precisely quantify the movement, velocity, and acceleration of the fish during rheotaxis. This assay provides the sensitivity needed to quantitatively assess the functional impact of both moderate and severe genetic or environmental disruptions of the lateral line system. For example, our assay has been used in prior studies to define lateral line–mediated behavior deficits in larval zebrafish with a mutation that impairs synaptic neurotransmission at the hair cell synapse [18] and in fish treated with both moderate and high doses of the ototoxic chemotherapy drug cisplatin [1,19].

Our behavioral setup consists of a translucent 3D-printed microflume with flow collimators providing consistent water flow that is delivered by an Arduino-controlled water pump to a small arena containing a single larval fish. The behavioral arena is illuminated with infrared light from below to eliminate visual guidance cues. The apparatus was developed in-house [1] and has subsequently been adapted and optimized at the Neurotech Hub at WashU (https://neurotechhub.wustl.edu/). We provide the method for printing and assembling the microflume and fish arena/basket (Figure 1) as well as for connecting the hardware and electronics (Figure 2). Furthermore, the flow rate can be parameterized and assayed using our flow velocity assay via visualization with methylene blue under bright light.

This protocol also describes how to process behavioral video recordings and track the fish’s body positions with DeepLabCut. DeepLabCut is a powerful standardized algorithm developed by Mathis et al. [20] that has been extensively used for behavioral tracking of a large variety of animal species, including zebrafish. It is a supervised deep learning tool based on residual networks (ResNet) for feature extraction that requires manual labeling of key body points of the fish, followed by training on a large image dataset (ImageNet), after which it generalizes to automatically track features in datasets of interest.

In addition, we describe the methodological use of SimBA [21], coupled with DeepLabCut, to algorithmically define, extract, and classify rheotaxis behavior. The extracted features are analyzed using customizable R scripts (provided in the GitHub companion to this protocol) to perform spatial, temporal, and spectral analysis, as well as measurements of the fish’s orientation (mean angle vector), velocity, and acceleration relative to the flow. Additional scripts generate heatmaps to visualize the average position occupied by the fish in the chamber arena over time. Time series analysis allows one to quantify both the distance travelled by the fish and the amplitude of oscillations of movements performed to maintain rheotaxis. Spectral analysis further reveals specific patterns of fluctuations in the swimming behavior of the fish to maintain rheotaxis under flow conditions (e.g., the overall trend and fluctuations in linear and angular movement.)

This protocol can be generalized to study other swimming behaviors in fish from different species by adapting the feature extraction script to another behavior of interest, like the C-start or more specific features like the tail posture, or changes in body movement and velocity. Furthermore, by integrating 3D tracking with multiple cameras and extending the size of the microflume, our protocol could be adapted for studying collective behavior, such as shoaling or schooling in multiple fishes.