Core to the prospects of cellular engineering for disease treatment are facile, yet tunable, methods to load cells with a range of therapeutic cargo. The process ideally allows for delivery to many cells at once, and it maintains cargo and cell faculties after transport. Nanopore electroporation (NanoEP) offers a gentler, scalable, and more precise alternative for cell cargo manipulations. It utilizes an electrically insulating membrane (PCTE) with cylindrical, nanoscale pores whose size, shape, and distribution determine the membrane’s local electric field. This membrane sits on the electrode and localizes the applied voltage to permeate phospholipid bilayer membranes of cells growing on it. This process technology improves viability and delivery outcomes with a lower driving force (15-30 V DC pulses); however, reported efficacy and experimental parameters (i.e. cell density) in literature are system dependent, sometimes with theoretical inconsistencies. Thus, optimization of NanoEP process parameters for different cargo and cell types should be coupled with more thorough in-situ electrochemical measurements to gain data-driven insights of this complex electro-biological system.
In this work, we use a novel multiplex NanoEP device design to analyze the chemical flux through a combination of first principles modelling and high-dimensional, experimental data analysis. I utilize electrochemical impedance spectroscopy, in-situ fluorescent visualization, and flow cytometry to characterize the cellular response to NanoEP and establish correlations between the experimental parameters (voltage, concentration, resistance, etc.) to develop a data-driven model for predicting process outcomes. By using a potentiostat for conducting NanoEP experiments, we can measure >50 unique electrical parameters for an individual trial, which gives us unprecedented information during the electroporation process. We show that rather than simply having an optimum voltage, an optimum current density is more indicative of experimental success, as the voltage must be contextualized by the resistance of an individual device. Through these insights, we then train and compare several machine learning predictive models for depletion and delivery of cargo from cells, and we find good agreement between the model and measured flux, with the exception of plasmid transfection showing a discrepancy between plasmid delivery and plasmid expression.
