Accelerating Chemobehavioral Phenotypic Screening in Neurotoxicology Using a Living Embryo Array System

Zebrafish embryos and larvae can be effectively used for rapid screening of chemically induced behavioral phenotypes in neurotoxicology and neuropharmacology.^1,2 The test strategies implement in situ analysis of behavioral responses upon chemical treatment using short sequences of digital video and subsequent animal tracking. In this regard, a conventional multiwell plate environment is not conducive for effective embryo in-test positioning and shadow-free illumination that are very important during video-based chemobehavioral analysis. ³ This type of experiment relies also on very expensive commercial systems. Moreover, analysis of biometric data in popular embryo photomotor response (PMR) and larval simulated predator response (SPR) bioassays is not particularly well streamlined for high-throughput neurotoxicology applications. ⁴

In this study, we present an integrated bioanalytical workflow for rapid analysis of large cohorts of specimens in PMR and SPR neurotoxicity bioassays. It includes applications of custom embryo arrays in conjunction with an open-source and inexpensive infrared video imaging system coupled with a dedicated bioinformatic approach that facilitates high-throughput analysis of behavioral phenotypes.

The custom living embryo array was fabricated in a biocompatible polymethyl methacrylate plastic using rapid laser machining as described earlier. ⁵ The embryo array featured dimensions of 130 × 60 mm with 189 of individual embryo traps (ϕ 4 × 3 mm, volume 470 μL) (Fig. 1A and Supplementary Data). The design of the array facilitated manual loading and immobilization of up to 189 individual living embryos (Fig. 1A). The number of traps on the array was determined based on the smallest mean pixel intensity that could be accurately detected by the animal tracking algorithm. Uniform developmental staging of embryos immobilized across the array was performed upon keeping embryos on the array in the small individual chambers using microscopic analysis. The embryos were introduced into the chambers at 28 hpf and allowed to acclimatize for 15 min in the dark before the 30 s PMR assay. A normalized cumulative survival of 100% was achieved in negative controls.

OPEN IN VIEWER

FIG. 1.

High-throughput chemobehavioral phenotypic screening using a living embryo array system. (A) Living embryo array that allows for in-test positioning and simultaneous video imaging of 189 embryos. Inset: Magnified single embryo trap and video processing that includes grayscale conversion and background subtraction techniques. The embryo body coiling is calculated based on automated quantification of pixel intensity changes (highlighted in pink) when the embryo exhibits locomotory activity. (B) A schematic overview of the stage integrated with the orthogonal infrared illumination and photic stimuli modules. (C) A photograph of the integrated wide-field video imaging system that utilized 4K infrared video recording. (D) Fully programmable control of the photic stimuli enabled by an open-source microcontroller (Arduino Mega) and a dedicated PC user interface. (E) Design of the embryo PMR bioassay with analysis undertaken during the 10 s excitation phase (outlined in red) with the stimulus light (∼36,000 Lux) provided by the high intensity CREE LEDs 1 s prior (in yellow). (F) A time-resolved PMR response of embryos at 28–30 hpf exposed to different concentrations of OP. (G) The normalized startle activity in the excitation phase upon 6 h exposure to 10, 50, and 100 μM of OP, the graph bars represent the mean (±SEM) for activity (n = 6 chemical exposure replicates, 9 embryos per replicate). (H) Design of the larval SPR bioassay with analysis undertaken during 3 min response phase (outlined in red), with the stimulus light (∼9000 Lux) provided by the standard LEDs (in yellow). (I) Occupancy heatmaps demonstrating behavioral responses of larval zebrafish at 5 days postfertilization (dpf) stage upon 2.5 h exposure to OP (10 μM) and mixture of OP and antidote drugs atropine and pralidoxime (2-PAM) used at clinically relevant concentration ratios (A1 = 0.5 μM Atropine, 32.125 μM 2-PAM, A2 = 1 μM Atropine, 64.25 μM 2-PAM, A3 = 5 μM Atropine +321.25 μM 2-PAM). (J) The normalized data from SPR bioassays depicted in (I). The graph bars represent the mean (±SEM) for distance (n = 6 chemical exposure replicates, 1 larval zebrafish per replicate), The ANOVA revealed that there was a statistically significant difference between control group and OP10 (38.23 ± 18.31; mean ± SE, p = 0.028), OP10 and OP10+A1 (119.94 ± 35.75; mean ± SE, p = 0.032), and OP10 and OP10+A3 (102.61 ± 28.54; mean ± SE, p = 0.042). ANOVA, analysis of variance; LEDs, light emitting diodes; OP, organophosphate; PMR, photomotor response; SEM, standard error of the mean; SPR, simulated predator response.

The inexpensive wide-field imaging system was custom build and consisted of a BlackMagic Micro Studio 4K camera (BlackMagic Design, Australia) converted to the infrared spectrum and mounted on a vibration-less photographic column (Mini Repro, Firenze, Italy) in an isolation chamber (Fig. 1B, C). The digital video feed was acquired and saved on a PC computer through a HDMI interface using a dedicated BlackMagic 4K Decklink Mini card (BlackMagic Design). The system was equipped with an innovative orthogonal illumination system that used infrared light emitting diodes (LEDs, emission wavelength 850 nm) and was incorporated into the stage. It provided a uniform and artefact-free illumination of the entire embryo array. The photic stimuli were delivered using high-intensity white LED panels mounted above and below the stage (Fig. 1E–H). Independent control of both stimuli was programmable using a custom graphical interface, including the duration and number of ON/OFF cycles (Fig. 1D and Supplementary Data).

We demonstrated that this inexpensive and open-source platform can be effectively used for PMR bioassays in eco-neurotoxicity testing. PMR is a stereotypical nonvisual photomotor response of embryos that relies on opsin expressing neurons in the hindbrain. ¹ A strong light stimulus induces transient increase in the frequency of body twitching in zebrafish embryos at 28 hpf. Typically, PMR is performed in 96-well plates utilizing a microscope with a motorized stage. ¹ This is suboptimal because each embryo needs to be recorded for at least 30 s; hence, there will be a 48 min exposure time difference between the first and last wells in the plate. In this study, we demonstrated simultaneous video recording and analysis of PMR responses in up to 189 individual embryos.

We validated the system in a high-throughput chemobehavioral experiment utilizing the organophosphate (OP) insecticide Chlorpyrifos (up to 100 μM) as an example of a canonical neurotoxicant (Fig. 1E–G). Compared with the conventional approach as described in literature, the array-enabled PMR was significantly more efficient as only very small amounts of chemical solutions (38 μL) were required. ¹ Moreover, embryos were spatially separated and encapsulated in the traps, which ensured no positional changes during video imaging. The design allowed the meniscus of the medium to be kept slightly convex, and this alleviated any shadow and reflection artefacts. Importantly, the array dramatically simplified the setup of experiments and increased throughput of the analysis, whereas alleviating any temporal artefacts. The individual traps enabled filtered phenotypic analysis through matching identifiable embryo activity responses with data obtained at a later stage.

We also developed a high-throughput bioinformatic analysis workflow for the analysis of PMR that exploits detection of pixel intensity changes (Fig. 1A) that are detected only during embryo body flexions. Large-scale detection of pixel intensity changes in individual embryos was performed using a custom protocol established in Ethovision XT 15 software (Noldus Information Technology, The Netherlands).

The presented open-architecture platform is inexpensive and tremendously flexible. We demonstrated the latter by rapid repurposing it for larval SPR bioassays. As the Arduino controller signal runs at the same voltage to activate the supply circuit through the transistor, switching between assays involved simply switching the positive wire connector from one set of LEDs to the other. We used SPR as a second step of neurotoxicant risk assessment and, additionally, showed its potential to extend the analysis toward pharmacological rescue of the toxicant-altered behavioral phenotypes. In this regard, standard OP antidotes, such as atropine (A) and pralidoxime (2-PAM), when used at clinically relevant concentration ratios, counteracted abnormal OP-induced behavioral phenotypes (Fig. 1H–J).

We conclude that the low-cost open-source architecture and significant experimental flexibility of the presented system opens new avenues for accelerated high-throughput chemobehavioral assays in both predictive and environmental neurotoxicology.