Quantitative Tools for Phenotype‐Based Drug Discovery in Zebrafish Models of Neurological Disease

Abstract

Objective Transgenic zebrafish expressing human Tau protein in neurons develop rapid-onset motor phenotypes; the aim of the present work was to optimize quantitative analysis of complex datasets from high-throughput neurobehavioral assays, to facilitate identification of small molecules that rescue Tau-dependent neurological phenotypes in unbiased chemical screens. Hypothesis Analysis of multifactorial neurobehavioral data using principal component analysis (PCA) will enhance detection of phenotypic rescue and thereby increase the throughput and reliability of a chemical modifier assay. Methods Motor assays were carried out in 96-well plates, one zebrafish per well. Swimming movements evoked by changes in ambient illumination were measured using our published MATLAB program LSRtrack. The resulting dataset includes x, y coordinates of the zebrafish centroid in each frame of the video. A phenotypic profile for each zebrafish was generated, consisting of: mean swimming speed in the dark; mean swimming speed in the light; difference between light and dark swimming speed; acceleration in the dark; acceleration in the light; change in speed at dark-light and light-dark transitions; % time motile; and average contiguous rest and active periods. Since individual zebrafish responses are too variable to be useful as a screening tool, data from the control and Tau groups were separated and then averaged in randomly allocated groups of predetermined size. Each characteristic was analyzed statistically for separation between control and Tau zebrafish over a range of group sizes. Variables with the highest Z-factor were selected to create a PCA. Utility of the PCA for future screens was further analyzed using Z-factor analysis. Results At 8 zebrafish per group, there were statistically significant differences between control and Tau zebrafish for all measured characteristics except for mean active duration. Average swimming speed in the dark and % time motile were the parameters that allowed most reliable separation of control and Tau zebrafish, with individual Z-factors of 0.14 ± 0.01 and 0.12 ± 0.01, respectively. Combining these two parameters in a PCA produced a resultant Z-factor greater than the individual components at 0.18 ± 0.01, with a group size of 8 (with 1,000 iterations of random grouping). Conclusions The phenotypic profile describes the neurobehavioral changes in Tau transgenic zebrafish more completely than single-parameter measurements. In addition, compiling components of the phenotype into a PCA improved the assays metrics. The best single-parameter assay for this transgenic line requires a group size of 12 to achieve a useful Z-factor for screening, allowing 6 compounds plus controls to be evaluated on each 96-well plate. With the new PCA approach, screening is carried out in groups of 8 zebrafish, thereby expanding the assay throughout to 10 compounds per plate, a gain of 67%.