Automation-based 3D organoid culture workflow with deep-learning based label-free image analysis.

The 3D cell culture model system is increasingly popular because it recapitulates the in vivo microenvironment better than 2D cell cultures. 3D organoids are cellular aggregates derived from  pluripotent stem cells or adult stem cells and can self-organize into organ-like structures. These organoids have the capacity for stable differentiation and rapid growth and as such, the organoid model system offer huge potential in disease modeling, drug screening and precision therapy.

The technologies available for using organoids as a model system is still in its infancy compared to the more established 2d culture or animal models. More development is needed to address reproducibility of organoids between batches and to standardized and improve the process of organoid culture. In addition, current protocols for the generation and maintenance of organoids are complex, time consuming and requires extensive manual handling.

To overcome some of the challenges involved in the culture process, we sought to develop an end-to-end workflow that uses automation and deep learning analysis tools for the growth, maintenance and monitoring of organoids in culture. To automate the process for organoid culture, we set up a work cell that consists of an incubator, liquid handler, high-content imager, plate reader and centrifuge. All instruments in the work cell are accessible by a software-controlled robotic arm to transfer cell culture plates between instruments. Using pulmonary organoids as the 3D model, we developed protocols to seed and plate cells in Matrigel domes using the work cell’s liquid handler. Culture plates containing cells can be transferred from the incubator to the liquid handler for scheduled media changes. Because organoids are typically cultured for longer period of time than 2D cell cultures, it is important to be able to monitor the growth of organoids as a form of quality control. As such, the scheduling function in the automation software can be programmed to move the plates between the incubator and the high content imager. Bright-field images of the organoids are acquired and then analyzed using AI-based analysis tools which provides metrics related to organoid growth over time (e.g. size, shape, texture). Overall, our results demonstrate the feasibility of using an automated work cell for the culture and monitoring of 3D pulmonary organoids. These results provide the foundational workflow which can be adapted for other 3D cell models such intestinal, brain or patient-derived tissues and enable scaling-up of organoid production for other downstream applications.