A Robotic Platform for Optimizing Biomaterial Synthesis in Plants for Tissue Engineering

Introduction: As tissue engineered constructs translate into clinics worldwide, there will be a need to increase the production of important human recombinant proteins to develop the implanted tissue scaffolds. This includes extracellular matrix proteins to accurately mimic mechanical properties and growth factors for improved healing. While bacteria, yeast, and mammalian cells have frequently been used for this task, plants offer multiple benefits for biomaterial synthesis including higher scalability, reduced costs, and better sustainability. However, efficient methods for protein production in engineered plants still need to be discovered to make this feasible at larger scales. Robotic systems can allow for the high throughput experimentation necessary to identify the optimal growth parameters of these plants, as they are already leading towards new areas in automated and autonomous laboratories. Open-source robotic liquid handling systems are increasingly utilized in biological research for studying a wide range of organisms, but plants are understudied with these approaches. We report a fully open-source robotic plant phenotyping system that is capable of imaging over 100 plant cultures and performing fluid operations throughout 10-day experiments. To demonstrate these abilities, a small, fast growing aquatic plant with multiple applications in the production of food, fuel, and pharmaceuticals called duckweed (Lemna minor) was grown with both constant and time-varying concentrations of different nutrient solutions. These findings will be important in a variety of areas including tissue engineering, biorobotics, and biomechanics.

Materials and

Methods: Our robotic plant phenotyping system is based on the Opentrons OT-2 platform, an affordable, open-source robotic liquid handler widely used in life science research. Duckweed (Lemna minor) was grown with half-strength Schenk and Hildebrandt media with a 16-hour day, 8-hour night cycle in petri dishes. It was then transferred into 12-well plates inside the robot for an experiment where they were grown under continuous, low intensity light. To give the OT-2 system the ability to image plant cultures, a custom camera module was built using a Raspberry Pi 12-megapixel camera and was directly attached to the robot’s gantry for three axis positioning. Wireless commands to capture images and transfer them to an external computer for analysis were automatically sent to the camera every two hours for ten days to get time series data for each plant culture. Furthermore, to automate the data collection pipeline, we used the deep learning algorithm Cellpose to segment plant material within each image to monitor the total surface area and color of each culture as measures of plant growth and health. While Cellpose has typically been used for microscopy images of multiple cell types, we fine-tuned a custom model for analyzing macro-scale images of plants using only 40 images.

Results, Conclusions, and Discussions: Using this setup, duckweed was grown on a range of concentrations of media with and without sucrose, a known plant stimulant, to explore how reducing nutrients affected growth. For plants grown with constant nutrients, we observed increasing growth as the media concentration increased and with small concentrations of sucrose. However, we also saw faster growth when nearly no media was provided, which we think is a stress-based response by the plant to find more nutrients to continue growth. With the collected color data, we monitored the health of each plant. Cultures grown with higher nutrient concentrations maintained healthy, green fronds, while those with minimal nutrients gradually shifted to an unhealthy, yellow color over time. This time-based data could be directly interfaced with autonomous laboratories setups for biomaterial synthesis to identify when plants are dying or underperforming so robotic actions can be taken to mitigate production losses. Finally, we illustrated that by altering nutrients through time, growth can be maintained or even increased while reducing total nutrient utilization, providing opportunities for optimization. For example, by reducing media concentration by 70 % and adding a small concentration of sucrose, plant surface area increased by nearly 30 % after five days of growth. This showed that biomaterial production can be done efficiently and in a potentially cost-effective way. While duckweed has previously been used to produce recombinant human proteins, this robotic platform could also be used with other plant species to optimize the production of important compounds. In conclusion, our system can enable further advancements in the field of tissue engineering by performing high throughput experimentation for discovering new scientific insights into plant behavior and investigating new methods of sustainable biomanufacturing.

Acknowledgements (Optional):