Walk-away automation of in-vitro patch-clamp electrophysiology
MetadataShow full item record
The complexity of the brain makes it a difficult target for systematized study. This is evidenced by the fact that to date, no unified taxonomy of cell type or connectivity pattern has emerged in the field of neuroscience. The ability to perform a complete census of cell types and connections in the brain would be a major step towards understanding the brain and treating its disorders. A gold-standard technique for performing neuronal classification is patch-clamp recording, which allows single-cell profiling of neuronal morphology, electrical activity, genetic expression, and connectivity patterns; however the technique is highly manual and laborious, making it unsuitable for large-scale studies that would be needed for neuronal classification efforts. A system that performs multiple recordings independently of human intervention, in a “walk-away” automated fashion, would be transformative. This work presents three techniques that enable full automation of the patch-clamp recording process. The first technique is the integration of pipette pressure control with trajectory planning which allows for reliable targeting of cells in brain slices. The second technique is automated pipette cleaning, which circumvents the need for a trained user to swap out pipettes between each patch-clamp trial. The third technique is machine vision which replaces a human operator in the final, most delicate aspects of patch-clamp recording. These techniques were combined to create a robotic system, called the “patcherBot”, enabling automated patch-clamp recording of 10 cells consecutively with no human intervention. The patcherBot was deployed to perform image-guided patch-clamp recording in adherent cells and neurons in brain slices. With one pipette, the system exhibits ~45 minutes of unattended operation, approximately an order of magnitude longer than previous automation efforts. The whole-cell success rate in both preparations was comparable to that of trained human operators (58 to 77%). The patcherBot was also modified to control two patch-clamp pipettes, which approximately doubled the throughput and enabled the study of inter-neuronal connections. Since unlike a human operator, the algorithm can control multiple pipettes simultaneously, adding more pipettes to the system could enable the patcherBot to surpass even the most skilled human operators and increase throughput. This system could thus serve as a tool for large-scale data collection for neuronal classification studies.