Scientists speed up artificial organoid growth and selection


Current method of stem cell-derived tissue production is very inefficient. By semi-automating tissue differentiation, researchers from MIPT and Harvard have made the process nearly four times faster, retaining the same quality of the tissue. The new algorithm is also useful for analyzing the factors that affect cell specialization.

Robots facilitate stem cell production. Credit: Daria Sokol/MIPT Press Office

Retina is an eye tissue, consisting of chained layers of neurons. It senses light and pre-processes visual information before sending it to the brain. Due to the lack of retinal cells regeneration potential, their loss leads to permanent blindness.

Today about 15 million people in the U.S. alone have retinal degenerative diseases. This number is growing, mainly because of population aging.

Several methods are being developed to address retinal health issues. Some of the options are neuroprotection, gene therapy, and cell replacement. Their development requires massive amounts of retinal cells for research purposes.

There is a method of growing retinal tissue in vitro. It involves placing stem cell clusters into a special medium, which produces undeveloped neurons that can be differentiated into retinal cells.

The method however has its limitations. It takes up to 30 days for the artificial retina of a mouse to develop correctly, and about one year for human organoids. The MIPT-Harvard team made an attempt to address these problems by increasing the number of cells produced and improving their quality.

The researchers compared the quality of the robot- and human-grown cells by producing several thousand retinal tissue samples for automatic processing and the same amount for manual handling. The biologists scanned the welles housing the tissue samples from the first group and analyzed the resulting images with a Python script they wrote to perform that specific task. The program determines the areas in the photos where the glow of the fluorescent protein is the strongest. Since this protein is only produced in developing retinal cells, the high fluorescence intensity points to the parts of the sample with the right tissue. That way the software can determine the amount of developing retina in each organoid.

The automation algorithm was able to optimize cell production by simultaneously testing many systems — without any adverse effect on tissue quality. The approach reduced the time researchers needed for cell processing from two hours to just 34 minutes.

“We implemented automated liquid change during retinal differentiation and showed it had no negative effect on cell specialization,” commented Evgenii Kegeles of the Genome Engineering Lab at MIPT. “We also developed a tool for automatic retina identification and organoid classification, which we showed in action, optimizing cell specialization conditions and monitoring tissue quality.”

“One of our goals in this research has been to scale up cell differentiation to enable a high-throughput tissue production for drug tests and cell transplantation experiments. Automated sample handling makes it possible to reduce the effort on the part of the personnel and produce several times more cells in the same period of time. With some slight modifications, the algorithm would be applicable to other organoids, not just the retina,” Kegeles added.

“Numbers matter: The automation empowered us to produce trillions of retinal neurons for transplantation and we are excited to see the translation of our approach into routine cell manufacture,” commented Petr Baranov from the Schepens Eye Research Institute of Massachusetts Eye and Ear.

Full research can be found in Translational Vision Science & Technology.

Originally published at