Automation of 3D organoid culture and assays powered by machine learning and advanced image analysis.

Abstract: Majority of developing drugs fail at the later stages of the drug development pipeline and in clinical trials. This high failure rate is partly due to insufficient predictive models being used to screen drug candidates. In recent years, organoids have emerged as a game-changer for disease modelling and drug screening. Studies show that patients and their derived organoids respond similarly to drugs, suggesting the therapeutic value of using organoids to improve therapeutic outcomes. However, challenges commonly associated with using organoids, such as assay reproducibility, ability to scale up, and cost have limited their widespread adoption as a primary screening method in drug discovery.

To overcome these limitations, we have developed CellXpress.ai, a highly integrated organoid generation and culture instrument. The system incorporates several modern technologies from hardware, software, and biological science with the aim to automate 2D and 3D tissue culture. System contains several functional components including automated imager, liquid handler, and incubator, connected by AI- powered software. Monitoring cell cultures and 3D objects by periodic imaging and analysis allows process control and triggering automatic decisions-making to initiate passaging, end-point assay, or troubleshooting steps. As a proof of concept, we demonstrated successful culture of iPSC and human intestinal organoid workflows over multiple passages. The workflows included cell/organoid seeding, feeding, and passaging, together with an in-line monitoring, and machine learning imaging analysis and classification.

We present the results from automation and prolonged culture of 3D organoids in Matrigel domes. Intestinal organoids were plated and cultured, passaged, and expanded in Matrigel domes (24 well). Automated media exchanges and monitoring in transmitted light by imaging was done every 24h. Organoids self-organized and developed complex crypt structures and expected phenotype. Machine learning-based image analysis allowed to determine organoids number, size (by area) and complexity. After 4-5 days organoids were automatically passaged: collected, purified from Matrigel and dispersed, then mixed with fresh Matrigel and re-pated. For endpoint assay (96well), organoids were stained for viability, nuclei, cytoskeleton, and other markers. We monitored concentration and time-dependent effects of the panel of anti-cancer compounds on healthy intestinal organoids for toxicity evaluation.
As an example of other complex protocols, we also demonstrated iPSC culture and passaging triggered by imaging-based automated decision -making; as well as automated formation and culture of colorectal cancer spheroids in U-shape low attachment plates, including compound treatment, staining and imaging.

Cell culture process automation powered by imaging and AI -based culture control has a great potential to bring 3D biology into another level, allowing to increase productivity, throughput and reproducibility, enabling variety of high throughput drug discovery, and precision medicine applications.