Modular Droplet-Based Microfluidic Platform for Functional Phenotypic Screening of Natural Killer Cells.

Natural Killer (NK) cells, as key effector cells of the innate immune system, display high heterogeneity in their ability to kill target cells. The underlying mechanisms remain poorly understood. Here, a droplet-based microfluidic platform is presented to identify and select NK cells with serial killing ability. To this end, primary human NK cells are encapsulated with several target cells using an efficient negative pressure-based droplet generator. Capitalizing on the large number of possible killing events due to quantization into droplets, a convolutional neural network analysis pipeline is developed to quantify the cytotoxicity and abundance of serial killing events with high accuracy. To physically select NK cells based on their serial killing ability, MultiCell-Sort - an advanced real-time image-based droplet sorting module - is presented. While conventional single-cell sorters mostly evaluate intrinsically-encoded properties, such as protein expression levels, MultiCell-Sort can select living NK cells based on complex functional phenotypes emerging from multiple cell-cell interactions within the droplet. This novel microfluidic phenotyping platform hereby allows the potential integration of complementary techniques to provide an understanding of regulators and markers underlying the heterogeneous nature of NK cell functional phenotypes.