Single-capsid identification of full and empty status of recombinant adeno-associated viruses via resistive pulse sensing

• Fabrication in a plastic using nano-injection molding a sensor with mixed-scale structures (nm → mm) with successful enclosure of the fluidic network via thermal fusion bonding. • A dual in-plane nanopore sensor consisting of two in-plane pores possessing an effective diameter of ∼100 nm and a 5 µm length flight tube serving as an electrophoresis channel (300 × 300 nm in width and depth). The two in-plane pores were used to elucidate the time-of-flight of different entities, which is related to their apparent electrophoretic mobility. • Ability to use the sensor to determine the fill status of adeno-associated viruses (∼20 nm diameter), which serve as gene therapy carrier vectors. • Using machine learning, showed the ability to determine the fill status of adeno-associated viruses with 100 % classification accuracy. • Due to the low-cost and high production fabrication capabilities, the sensor can potentially be used for near real-time measurements to monitor production fidelity of recombinant adeno-associated virus gene therapy construct. Recombinant adeno-associated viruses (rAAVs) are promising vectors for gene therapy, but their production is hampered by the presence of empty capsids, escalating costs and diminishing safety and efficacy. Traditional methods for assessing capsid content status lack adaptability to in-line production workflows. The lack of in-line methods to monitor rAAV production quality limits efforts seeking to improve production efficiency and subsequent in vivo performance. This study introduces a dual in-plane nanopore sensor fabricated in thermoplastics via replication, which can offer near real-time sensing for potential integration into gene therapy production lines and scalability for high-scale manufacturing at low cost. By leveraging differences in surface charge density and internal capsid content, coupled nanoscale electrophoresis and resistive pulse sensing enabled label-free identification of individual full and empty capsids using supervised machine learning with a neural network. This single-capsid analysis approach may also offer insights into rAAV-host cell interactions and population heterogeneity. We demonstrated using rAAV9 as an example due to its wide use in gene therapies, but establishment of the methodology as delineated in this manuscript will make the technology applicable to other vectors, such as lentiviruses and adenoviruses.