Neural constructs of neurons (green) and nuclei (blue) resembling the human brain are developed in P.E.G. hydrogels prior to experimentation.

Neural constructs of neurons (green) and nuclei (blue) resembling the human brain are developed in P.E.G. hydrogels prior to experimentation.

Traditional methods for testing neural toxicity have often been costly and, at times, failed to detect certain compounds toxic to the brain. However, researchers from University of Wisconsin-Madison, UC Santa Barbara and Morgridge Institute for Research in Madison have developed an accurate and cost-effective method in screening compounds for neurotoxicity.

The research team first utilized stem cells to develop the various components of the brain during its various stages of early development. Stem cells are undifferentiated cells that have an enormous potential to differentiate into any cell type. Another remarkable aspect of stem cells is their ability to self-renew and multiply while maintaining the ability to differentiate. The team first cultured stem cells into neural-progenitor cells, vascular cells, and microglia on polyethylene glycol (PEG) hydrogels; these differentiated cells all have a connection to the composition and function of the brain. These constructs would then fold into three-dimensional structures call organoids, in this instance, organoids that mimic the human brain in early development. Through these methods the research teams were able to synthesize accurate models of the developing human brain at different stages of development.

Michael Shwartz, an assistant scientist in biomedical engineering at UW-Madison explained the advantage of using PEG hydrogels over matrigels to culture stem cells on to create models of the brain.

“The traditional method of replicating a model for the human brain involved growing cultures on matrigel gels, which was not only difficult to process but also contained many components which made experimental control very difficult,” Shwartz said. “With PEG hyrdrogel, the only bioactive components are MMP peptide crosslinks and peptide pendant which greatly reduces unwanted component interactions and thereby increase control, uniformity and reproducibility.”

The team then determined compound neurotoxicity by analyzing 19,000 genes responsible for the functions within the various brain tissues using a machine-learning algorithm. This experiment was achieved by first taking two substances – one toxic, the other non-toxic – and exposing them to one sample of synthesized neural construct. The program would then analyze gene expression in each sample, with the genes showing a change in gene expression, which strongly indicated toxicity. Thirty-nine known toxic compounds and 21 known non-toxic compounds were exposed to RNA-sequencing data from neural tissue constructs to produce an accurate predictive model using the machine’s algorithm.

The next step was to test the predictive model by subjecting each sample to a blind trial test using ten compounds, five toxic and five non-toxic, but were otherwise unknown to the predictive algorithm. The results demonstrated that the model correctly identified nine compounds for toxicity and non-toxicity.

“To put that into perspective, animal testing has an accuracy of 70% while this model surpasses that by 20%,” Shwartz said. “The accuracy of our model could further be improved if we exposed our machine-learning algorithm to more known compounds.”

The results not only demonstrated the accuracy of the predictive model and the algorithm from which it was derived from, but also demonstrated the samples’ accurate representation of the human brain and therefore the utility and advantage of the method used in this experiment to produce those samples. The research teams have therefore pioneered a method incorporating the production of complex biological structures in a manner that is highly reproducible.

“Our model incorporates the complexity of cells and that added complexity provides accurate representation of the brain,” Shwartz said, “Our model’s reproducibility makes studying tissue more quantitative allowing scientist to be more confident that the changes observed in their neural experiments using this method are real changes.”

The National Institutes of Health as well as 10 universities from the Tissue Chip for Drug Screening program funded this study. A list of toxic and non-toxic chemicals was provided by the Environmental Protection Agency.

A version of this story appeared on p. 18 of the Thursday, Oct. 8 print edition of the Daily Nexus.