These probes capture incredible detail in the spatial structure of extracellular action potentials.

Due to improved SNR and resolution, NP Ultra finds 3x more visually-responsive neurons in cortex. We developed a new sorting algorithm suited to NP Ultra called DARTsort with Paninski lab. The number of well-isolated neurons passing strict QCs was also improved by ~1.6x.

We recorded across mouse brain regions and in three other species (monkey, lizard, and fish), finding everywhere a proportion of units with really small spatial “footprints”. These small units are especially enriched in areas densely packed with small neurons or axon fibers.

Do small footprint units correspond to small interneuron cell types? In mouse visual cortex, no: we optotagged PV, SST, and VIP neurons but none had small footprints. And, the small units could be either excitatory or inhibitory in cross-correlations. They are likely axons there!

Finally, can high-res NP Ultra recordings tell us more about neuron identity? Yes! We can decode cell type among the three optotagged classes ~75% correct, significantly better with NP Ultra than with NP 1.0.

So, is this the device for your experiments? It’s ideal for experiments that: target dense nuclei or layers; need to detect small fields; or need cell type identification. The tradeoff is span: with all 384 sites packed so tightly, it covers ~300 µm compared to ~3 mm for NP 2.0.

I want to thank the huge team who participated - leading efforts from Zhiwen Ye and Andrew Shelton, major contributions from Jordan Shaker, Julien Boussard, and Jennifer Colonell, and fantastic work from the whole team!