Associate Professor Worcester Polytechnic Institute, United States
Introduction:: Chemical modulators of neural excitability represent potential therapeutics for neuropsychological disorders. We developed a functional screening method of neural activity in whole, living organisms to identify and quantitate these modulators. The high through-put method allows for quick generation of large imaging data sets, yet manual data analysis is time consuming, not scalable, and can be error prone. This work focuses on automating data analysis to increase efficiency, decrease errors, and obtain drug classification results within hours instead of days or weeks. The increase in speed and accuracy should accelerate the drug discovery process and identify new modulators with therapeutic or research potential.
Materials and Methods:: C. elegans co-expressing the fluorescent calcium sensor GCaMP and optogenetic excitatory ion channel Chrimson in the AWA chemosensory neurons were immobilized by hydrogel fixation in 384-well plates (Lagoy, 2021). Compounds from the Library of Pharmacologically Active Compounds (LOPAC), were added to these wells and allowed to diffuse through the hydrogel to embedded, living animals. Every 6 hours, the neurological response to optogenetic activation of the AWA neuron (via 617 nm red light) was visualized via GCaMP emission at 470 nm. Over the course of an 18-hour experiment, a single plate produced over 60,000 neural responses from ~40 animals per drug tested. These data were then processed using a custom ImageJ script that isolated individual neurons based on their change in neural fluorescence. Once the location of each neuron was recorded, a second script extracted pixel intensity data to quantitate the optical response. 1% DMSO was used for negative controls and nemadipine-a, a known calcium channel blocker, was used for positive controls. Hits were determined to be compounds that showed a change in neural activity over the course of 18 hours by more than three standard deviations over the negative control.
Results, Conclusions, and Discussions:: The automated script processed data for one 384-well plate in 3.5 hours compared to 22 hours for manual processing, an 6-fold increase in process rate and requiring only 30 minutes of user touch time. The automated method correctly identified two manually-identified and validated modulators including antagonists of glutamate-gated chloride channels and voltage-gated calcium channels. In addition to these two compounds, the script identified several additional potential modulators which were not previously identified and are targets for validation.
This automated experimental method for drug screening created a bottleneck at the image segmentation and data analysis steps, requiring the manual selection of tens of thousands of neurons in images, an error prone and tedious process. I developed an automated pipeline that requires only 30 minutes of user touch time, and validated by identifying chemical modulators. Additional data filters are needed to decrease noise and variation, as well as to decrease computation time. This screening method currently only identifies intracellular neuromodulation targets only; future screens will target different neurotransmitters, and neuropeptides, by separating stimulation and reporter neurons.
