Online High Throughput Microfluidic Single Cell Analysis for Feed-Back Experimentation

Unraveling heterogeneity remains a major challenge to overcome when advancing our understanding of bio(techno)logical processes. Microfluidic single cell cultivation combined with live cell imaging has become a powerful technology to elucidate cellular heterogeneities temporally resolved on the single cell level. In this context, within recent years, versatile cultivation devices as well as image analysis tools have been developed, tailored to microbial single-cell studies. At the current state, acquired image data are typically analyzed after experimentation has been finished, resulting in static offline approaches and long cycles of insight generation. To shorten these cycles and to allow novel types of experiments extended by interacting with the biological system, the experimental setup needs live analysis with joint automated experimental control. This work develops software-based online capable techniques for feed-back experimentation: The application and development of image analysis pipelines suitable therefor, as well as the establishment of a new automated experimental control platform, HiMiCs, for microfluidic live cell experiments: An existing image analysis pipeline is described, aiming at one dimensional microfluidic growth channels. It is applied and connected into the platform. A new image analysis pipeline specifically aiming at filamentous growing microorganisms in cultivation chambers was developed and applied. Furthermore, a machine learning based approach for segmenting rod-shaped bacteria has been investigated. The experimental control platform connects high-level microscope control specifically tailored to the necessities of microfluidic experimentation, that is, hardware control of the microscope and additional peripherals, image acquisition and remote storage in centralized infrastructure, jointly with connectivity protocols for image analysis routines. This allows for direct, automated analysis of the image data, as it is acquired, presenting the user with results and remote control. By passing the results to a simple yet powerful scripting interface, model based steering of the experiment at a high abstraction level becomes possible. Furthermore, a simulation based image generation system is described, for in silico end-to-end testing scenarios of the platform, as well to gather insight towards microcolony growth by modeling. As a proof-of-concept experiment facilitating feed-back control, growth of the biotechnologically relevant organism Corynebacterium glutamicum is automatically controlled via nutrient availability. With that, the experimental control platform HiMiCs lays the foundation to new classes of experiments while reducing the “time to insight”: Based upon biological behavior, experimental conditions can be adapted automatically, thereby closing the cycle of data generation and experiment parameter decision within a single experiment