Episodic memory is thought to be preferentially encoded by sparsely distributed memory-eligible "primed" neurons with high excitability in memory-related regions. Based on in vivo calcium imaging on freely behaving mice, we developed an analytical method to determine the neuronal activity hierarchy and establish hippocampal primed neurons. Neurons with high activity and memory-associated burst synchronization are identified as primed neurons. When a trace fear memory is being formed or retrieved, the major pattern of the calcium dynamics is predominantly mediated by primed neurons and highly correlated with mouse freezing behaviors. In cilia knockout mice that exhibit severe learning deficits, the percentage of their primed neurons is drastically reduced, and any burst synchronization is strongly suppressed. Consistently, the first principal pattern of cilia knockout neurons does not fully distinguish itself from other minor components or correlate with mouse freezing behaviors. To reveal how a portion of neurons get primed, we developed a numerical model of a neural network that incorporates simulations of linear and non-linear weighting synaptic components, modeling AMPAR- and NMDAR-mediated conductances respectively. Moderate NMDAR to AMPAR ratios can naturally lead to the emergence of primed neurons. In such cases, the neuronal firing averages show a right-skewed log-distribution, similar to the distributions of hippocampal c-Fos expression and the activity levels measured by in vivo calcium imaging. In addition, High basal neuronal activity levels speed up the development of activity hierarchy during iterative computation. Together, this study reveals a novel method to measure neuronal activity hierarchy. Our simulation suggests that the accumulation of biased synaptic transmission mediated by the non-linear weighting synaptic component represents an important mechanism for neuronal priming.
