Abstracts

Rationale: Our understanding of how brain networks change during seizure onset has been hampered by an inability to record neuronal activity at a seizure focus with sufficient resolution and sensitivity. Recent advancements in voltage-sensitive fluorescent proteins have made them brighter, more stable, and more sensitive to voltage changes. Technical and financial constraints have limited the broad application of voltage imaging in neuroscience, including epilepsy research. Here we present a low-cost, highly capable, hardware and software pipeline for imaging voltage in large populations of neurons, with sensitivity to both action potentials and subthreshold deflections.


Methods: Organotypic hippocampal slice cultures were prepared from P7 mice that had been injected on P0 intracerebroventricularly with adeno-associated virus encoding for the red voltage-sensitive fluorescent protein VARNAM2. Images were acquired on an inverted widefield fluorescence microscope using a relatively low-cost, high quality CMOS camera (FLIR GS3-U3-51S5M, less than $2K at time of writing). To capture the entire field of view at frame rates sufficient to detect action potentials, we used 2x2 decimation and 8-bit digitization to obtain 1280x1024 frames at a rate of 200Hz. While these settings generate coarse raw images, averaging the 50-100 pixels per cell produced voltage traces with clear single-trial action potential detection in >1000 neurons.

We tested two methods for measuring subthreshold post-synaptic voltage deflections. First, we computed spike-triggered voltage traces in the population of cells using each individually recorded neuron as the “trigger” source. Second, we used the recently developed SUPPORT, a self-supervised learning method for denoising voltage imaging data.


Results: Image quality in raw data was sufficient to resolve action potentials independently in immediately adjacent neurons without averaging. Using thresholds determined with pharmacological controls in which action potentials were blocked, we could robustly measure spiking for >4 minutes in populations of >1000 neurons. Spike-triggered averaging of population activity that immediately followed spiking of a given index cell produced an “output map” for that index cell. Some index cells produced predominantly excitatory responses in the network, while others produced predominantly inhibitory responses. By repeating this process for each cell, we generated complete maps of inhibitory and excitatory network connectivity. Finally, SUPPORT denoising of the data improved the signal-to-noise ratio to the degree that excitatory and inhibitory post-synaptic potentials could be imaged without averaging.


Conclusions: The latest generation of voltage sensors can be imaged using low-cost acquisition hardware and open source analysis tools to provide unprecedentedly high resolution and high sensitivity measures of neuronal electrical activity underlying epileptiform discharges. Ongoing pharmacological and electrophysiological experiments will be used to validate and extend these findings.


Funding: NIH R01NS112538