静的ではないニューロンのアイデンティティ：入力駆動型視点

Neuronal classification based on morphology, electrophysiology, and molecular markers is often considered static. Here, we challenge this view, showing that functional classification depends on input patterns. Using single-cell recordings from layer 2/3 barrel cortex neurons in mice, we compared responses to step-and-hold versus dynamic frozen noise inputs that mimic presynaptic activity. Action potential and waveform-based classifications varied significantly, highlighting the dynamic nature of neuronal identity. To assess the contribution of input versus neuronal attributes toward classification, we analyzed four attribute sets, namely action potential, passive biophysical, adaptation currents, and linear input filters derived via spike-triggered averages (STA). Our findings revealed that the STA, which captures a neuron's selective responsiveness to presynaptic activity, explained the most variance within the population. This highlights input-driven dynamics as key to functional identity, emphasizing the need for physiologically relevant inputs in defining neuronal classes and shifting the focus from static properties to dynamic functional diversity.