A quantitative approach to breathing patterns across emotionally-salient states in mice

Breathing is an incredibly dynamic homeostatic behavior that is critical for survival and under both autonomic and voluntary control. Physiology, mediated by the autonomic nervous system, is classically considered part of the experience of emotional states, but whether breathing reliably differs across emotionally-salient tasks in mice is unknown. Here, we implement breathing recordings in mice using two different techniques, whole-body plethysmography and intranasal pressure recordings, to examine breathing across both innate and stimulus-dependent behaviors. With whole-body plethysmography, we demonstrate the limited separation of behavioral states with minimal breathing features. Using intranasal pressure, we expand evaluation of breathing to more commonly used behavioral tasks in fear and stress neuroscience. With the inclusion of lesser studied breathing features, we show four types of breathing patterns emerge in mouse breathing through unbiased k-means clustering. Importantly, visually similar behavioral immobility is accompanied by nuanced variations in breathing across the tasks implemented. Through two machine learning techniques, we then demonstrate the ability to reliably distinguish behavioral states from breathing features at a total accuracy of ~80%. We then show the value and ease of using this respiratory space as a framework to contextualize breathing from additional behaviors. Lastly via optogenetics, we show the ability of the central amygdala to increase the frequency of prolonged inhale pauses or apneas associated with predator odor exposure. We further show that these central amygdala apnea-associated neurons have higher population activity during investigative behaviors using fiber photometry. We examine the implications of these studies and suggest ways to uncover the physiological significance of these breathing patterns in the future.