Animals
Animal experiments were permitted by the local government authorities (Regierungspräsidium Tübingen) and conducted in accordance with German federal law and Baden-Württemberg state law. Zebrafish larvae (Danio rerio) were reared on a 14/10 h light/dark cycle at 28°C in E3 medium (5 mM NaCl, 0.17 mM KCl, 0.33 CaCl, 0.33 MgSO4) with 0.01% methylene blue. Wildtype (nacre +/-) larvae were used for the mortality rate and behavioral experiments. Transgenic zebrafish expressing GCaMP6f [Tg(HuC:H2B-GCaMP6f)jf7Tg] were used for the two-photon calcium imaging experiments. The top 10–20% of larval zebrafish with high calcium indicator expression levels were selected from all fish at 3 dpf for in vivo calcium imaging experiments. All behavioral and calcium imaging experiments were performed on 4 dpf zebrafish larvae at room temperature. Supplementary Table 1 contains a list of experimental groups and animals used per group.
Exposure chamber for mortality rate experiments
To determine mortality rates of multiple free-swimming larvae, an exposure chamber covering a volume of 50x36x15 mm was designed and build out of polystyrene. The internal dimensions of the exposure chamber represent a trade-off between the space needed to safely exposure multiple larvae to an electric field, and the required voltage gradients. To protect against leakage currents, the programmable power source was connected to the exposure chamber by a residual-current circuit breaker (F200, B Type, ABB Ltd (Asea Brown Boveri), Zürich, Switzerland). This guaranteed safe operation for both AC and DC voltages. Two stainless steel electrodes (36x15 mm) with a thickness of 1 mm were placed inside the chamber at a distance of 50 mm. A custom-made 3D-printed electrode holder was used to hold the electrodes in place and to ensure the exact distance of 50 mm. This electrode holder also included a rigid net with an aperture of no more than 405 µm (PA-405/47, Eckert, Waldkirch, Germany), in which larvae could be placed during exposure. This ensured that larvae were always exposed to the center of the electric field (supplementary Fig. 5 and supplementary Fig. 7A). Electrodes were constructed by the precision mechanics workshop of Eberhard Karls University, Tuebingen, Germany. All polymer components were constructed and 3D-printed in the lab (Ultimaker3, Ultimaker B.V., Utrecht, Netherlands, ink: Ultimaker Material PLA Black, diameter 2.85 mm). For safety reasons, operation of the exposure chamber depends on a snap-action switch which require a closed lid. The lid and bottom of the chamber are transparent to enable diffuse back illumination at 920 nm from below and behavioral recording from above. An LED array containing 48 LEDs with a 50° emission angle and a diffuser (frosted glass) was used for illumination. Recording was done at up to 30 Hz using a camera (DMK23UV024, The Imaging Source Europe GmbH, Bremen, Germany) with a 6 mm objective and an IR longpass filter (SCHOTT RG780, Edmund Optics Inc., Barrington, USA). Both the power source and the camera recording were controlled by the custom written Python software “synchASR2000-control”. The software allowed to individually set voltage type (DC, PDC, AC), overall voltage, maximum current, frequency and the exposure duration.
Determining mortality rates
For each experiment, the exposure chamber was filled with 24 ml of room-temperature E3 medium (680 µS/cm conductivity). Five larvae were transferred into the chamber, followed by a 5 min acclimatization period. After that period, the temperature of the E3 medium inside the chamber was measured once using a digital thermometer (GMH 3200 Series, GREISINGER, Regenstauf, Germany) and the lid was closed. An electric field was then applied for a specific exposure duration. In case of controls, no electric field was applied. During the exposure, larvae were video-recorded. The recording was synchronized with the onset of the electric field via the Python software using a timestamp. Immediately after the exposure, the lid was opened, and temperature was measured again. Larvae were then moved into a dish of fresh E3 (room temperature). To determine mortality rates, all larvae were visually examined via a stereo microscope directly after exposure to an electrical field and a second time after a 30 min recovery period in fresh E3 (Fig. 2B). Mortality was defined by the following criteria: no activity of any kind, loss of equilibrium, no operculum movement, and no startle response (tested by a strong tactile stimulus). Together with muscle tone and cardiac failure (which were not assessed here), these criteria correspond to the fourth stage of anesthesia in fish, describing overdose 66. Muscle tone observation was excluded due to the ability of electrical fields to cause muscle tension. Cardiac arrest on the other hand is not a suitable criterion for death in larval zebrafish, because their small size ensures that oxygen diffusion into vital tissue is sufficient for survival even after the heart stops beating 10. Heartbeat may persist for up to 10 min after euthanasia with MS-222 and 40 min after hypothermic shock 10. Some studies even report that MS-222 treated larvae were able to recover their heartbeat when subsequently transferred to fresh water 10,11. Also, in our study, cardiac arrest rates differed from mortality rates for exposure durations between 2 and 32 s (supplementary Fig. 8). Larvae were only considered dead, if all tested criteria were met at both of the two observation points (directly after exposure and after the 30 min recovery period) (Fig. 2C). After the second observation point, all larvae were put into ice water to ensure proper euthanasia for potentially survived fish. In total 600 fish were used in 30 different groups (supplementary Table 1). Significant differences between different voltage types and between exposure durations within each voltage type were determined by z-test for proportions followed by Bonferroni correction. A pilot trial with 20 larvae anesthetized with MS-222 (168 mg/l) and 20 non-anesthetized larvae exposed to 50 V/cm for 32 s revealed no increased mortality rates for MS-222 and all subsequent experiments were carried out without anesthetic (z-test for proportions, p > 0.05., supplementary Fig. 1).
Determining loss of equilibrium
Loss of equilibrium was determined by hand using the video recordings from the mortality rate experiments and the image processing package Fiji 67. The time difference between the exposure onset of electrical stunning and the first frame where larvae departed from the upright position was measured. Not all recorded larvae could be analyzed, because they sometimes swam into the shadow cast by the electrode frame, which made it impossible to observe the time point of loss of equilibrium. Therefore, the number of larvae analyzed in this regard differed from the total number of larvae recorded. In total 295 fish were analyzed.
Swimming burst analysis
A high-speed recording (IDT iN8-S1 camera) with a sampling rate of 500 Hz was performed on a single agarose-embedded 4 dpf larva (nacre +/-) using the exposure chamber without the rigid net (Fig. 3A). Agarose was removed around the tail to observe exposure-related movements. Recording was performed for 10 seconds (exposure onset was approximately 1 second after the start of recording). Electrical stunning was performed with a sinusoidal AC field, 60 Hz, 50 V/cm, for 32 s.
To analyze the recordings, images were first smoothed using a median filter (filter size: 20x20 pixel) to remove noise. Afterwards, the gray values of all pixels were first inverted and then binarized. Areas smaller than 600 pixels in size were removed to clear the image, leaving only the silhouette of the fish. A skeleton of the silhouette representing the spine of the fish was computed to track tail deflections. Lateral displacement of the tail was defined as the distance (in pixels) of each skeleton pixel to the fish's original rostro-caudal axis, normed to the mean displacement during spontaneous swim bout and field exposure. Total tail angle (in degrees) was calculated for each frame by computing the angles between all neighboring pixels on the skeleton and then taking the sum along the entire tail. Tail length was normalized to the shortest skeleton computed using 1D interpolation. In total, two fish were recorded and analyzed.
Setup for simultaneous calcium imaging and electrical stunning
For calcium imaging experiments, a two-photon IR laser (Coherent Chameleon Vision S, Coherent Inc., Santa Clara, USA) at a wavelength of 920 nm was used and calcium-signals were recorded using a movable-objective microscope (MOM; Sutter Instruments, Novato, USA) with GaAs photomultiplier tubes (PMT), C7319 Hamamatsu preamplifiers and the MScan software (2016 version) by Sutter. All recordings were conducted in the optic tectum, scanning a single plane per fish approximately 40–70 µm deep with a 20x/1.0 objective (Zeiss W Plan-Apochromat 421452-9800, Jena, Germany) at a frequency of 2 Hz and a magnification of 1.8x. To conduct calcium imaging alongside simultaneous visual stimulation and exposure to an electric field, a setup containing a LED ring with 18x 650 nm LEDs and a centered 50 mm Petri dish with two small stainless-steel electrodes was developed. The electrodes (2x3 mm, 0.5 mm thick) were placed at a distance of 3 mm in the Petri dish using a custom-made 3D-printed frame. The frame enclosed the electrodes from three sides to avoid any unwanted possible contact with the microscope objective (supplementary Figs. 6 and 7C). To connect the electrodes to the power source, a 200 µm thin copper plate was welded to their back sides to allow soldering of cables. The power source was controlled by the custom written Python software ”synchASR2000-control”. To present a strong visual ON/OFF-stimulus, the LED ring was connected to an Arduino NANO, controlling power supply of the LEDs. To gate the LEDs during laser scanning, the Arduino NANO considered the fly-back signal provided by the MOM. ON/OFF-signals from the LEDs, fly-back signal from the MOM and the onset-signal of the electrical field were recorded at 100Hz via an Arduino UNO and another custom-written Python software called “recSignals-control”. The whole setup was electrically separated from the microscope objective by a modified SM1L05 lens tube containing an insulating plastic spacer.
Calcium imaging alongside electrical stunning and visual stimulation
For each experiment, a single zebrafish larva was embedded in 1.6% agarose gel (Biozym Sieve GeneticPure Agarose 850080 mixed with E3 medium) between the electrodes (Fig. 4A). Agarose is electrically neutral, so the electrical conductivity of the gel is determined by the liquid added 68. Therefore, it can be assumed that the conductivities of E3 medium and agarose gel were approximately the same and it was unlikely to affect the homogeneity of the electric field. Each experiment consisted of three subsequent recordings: baseline recording (6 min, of which 3 with visual stimulation), shock recording (6 min with visual stimulation) and recovery recording (12 min with visual stimulation) (Fig. 4B). Recordings were done in the optic tectum which is known to be a hub in visual processing and it contains different populations of neurons well responding to ON and OFF flashes 32,33. In total 7 fish were imaged.
Analysis of ROIs and Cross-Correlation Matrices
All data analysis was done in Python. Baseline and recovery recordings were registered to standard deviation (STD) z projection and regions of interest (ROIs) were segmented using suite2p 69. The calcium signal (ΔF/Fb, i.e. fluorescence changes relative to baseline) at time point t was calculated as in Leyden et al. (2022) 17. The calcium signal of each ROI was calculated for the second half of the baseline recording (visual stimulus present), and during the period preceding onset of a slow propagating wave for the recovery recording, which differed across trials. The overall number of ROIs for baseline and recovery recording differed, which could be explained a combination of morphological changes during electrical stunning, the overall decrease in calcium activity, as well as the increase of synchronized activity among ROIs which led to poor segmentation results in the recovery recordings.
Regression-based identification of stimulus-encoding neurons was used in this study to quantify alterations of vision-related activity during and after electrical stunning 70. For this purpose, an ON-regressor was computed by first resampling the recorded ON/OFF-signals of the visual stimulus to 2 Hz and then convolving them with a calcium impulse response function (CIRF) using an exponential decay time constant of 1.61 s for GCaMP6f. ROIs were then correlated with the ON-regressor to test how well their activities matched the expected calcium signal of a neuron responding to light onsets. Distributions of correlation coefficients for baseline and recovery recording were analyzed via a Levene-test.
To generate cross-correlation matrices, ROIs were sorted by their correlation coefficient (from positive to negative correlation with the stimulus) and a cross-correlation matrix of the 20% best correlated ROIs was computed. Matrices of all fish were brought to the same size (mean size of all matrices) using linear interpolation. A weighted average matrix was then computed, based on the total number of ROIs per fish.
Power spectral density analysis
Welch’s power spectral density (PSD) method was used for estimating the power density of fluorescence at different frequencies for all recordings (baseline, shock and recovery). Because recordings during exposure to an electrical field were affected by strong spatial drifts of the preparation and image registration and across-frames segmentation of ROIs were therefore not possible, PSD for each pixel in each recording for all fish was calculated, using the entire recording length of each raw image time series without any preprocessing. Calcium traces for all pixels were first detrended by subtracting the mean, then PSDs were calculated. Finally, the mean PSD spectrum of all pixels was calculated for all recordings. To prove that a large drift in the recording won’t significantly affect the mean PSD spectrum, and that the most prominent frequencies will still be visible, a 100-pixel drift over 720 frames (6 min) was simulated for a stable baseline recording and PSD was calculated subsequently (supplementary Fig. 4). PSD results were analyzed via 2-way repeated measures ANOVA with Tukey’s honest significant difference (HSD).