DEMONSTRATING NEW APPROACHES TO SMALL-BATCH CELL THERAPY MANUFACTURE THROUGH A LABEL-FREE MICROFLUIDIC CELL SORTER

Abstract

Background & Aim

Automation of bioprocessing and analysis needs to be at the heart of advances in cell and gene manufacture for the transition to personalised therapies. Small-batch processes still require significant manual control and are limited to off-line quality analysis. Automated and decentralised production technologies, combined with in-line or at-line process analytical technologies, must be the route forward to realise the potential of personalised therapies. Integrating AI and machine learning (ML) into these approaches offers opportunities to reduce costs and access novel capabilities that have not yet been fully explored. Here, we demonstrate that simple and inexpensive devices can be constructed for specific applications, drawing on examples from the development of a microfluidic cell sorter. Driven by an ML algorithm, the device was designed to provide greater freedom to therapy developers without commitment to bulk manufacturing processes.

Methodology

The device was built from commercially available components. Fluid is driven by a syringe pump and controlled by a pair of rocker valves. The solution of mixed cells is held in a syringe. The mixed cells pass through a y-channel where they are imaged by an optics system. A ML algorithm identifies the cells and directs the valves to sort them into the correct channel. The total cost of all components is less than $3000.

Results

Morphologically dissimilar cells can be identified by optical imaging and separated without relying on extrinsic labelling. Cells do not require pre or post processing to add or remove fluorescent or magnetic labels. Labelling cells prior to sorting adds time and complexity to the process step and labelling physiologically similar cells for purification can be a barrier to the development of autologous cell therapies, where there is a risk of expanding and reintroducing diseased cells. The algorithm is trained to distinguish differences in the size and shape of cells. Once correctly trained, the device is applicable to any morphologically dissimilar mixture of cells delivered into the channel.

Conclusion

Increasing automation while reducing process complexity is essential for the development of personalised therapies, particularly autologous cell therapies. Automating small-batch production need not require high investment in manufacturing facilities, while leveraging new technologies and alternative approaches to manufacture can lower production times and costs, providing better outcomes for patients and developers.