Repository logo
 

A midbrain mechanism for computing escape decisions in the mouse


Type

Thesis

Change log

Authors

Evans, Dominic Andrew 

Abstract

Animals face frequent threats from predators and must generate appropriate behavioural responses to ensure their survival. To achieve this, they process sensory cues to correctly identify the presence and imminence of a predatory threat, and transform this information into defensive actions. However, despite much research in identifying the circuits that may be responsible for such transformations, little is known about how this occurs mechanistically.

We focus on how escape behaviour in the mouse is generated from visual predatory threats, and use a combination of behavioural, neurophysiological and anatomical methods to identify the relevant neurons and understand how they perform this computation.

In this work, we developed an innate decision making paradigm in which a mouse detects and assesses sensory stimuli of varying threat evidence during exploration, choosing whether to escape to a shelter, or not. The performance data in this task were best formalised with a drift-diffusion model of decision making, providing a framework to understand innate behavioural tasks in terms of evidence accumulation and boundaries.

Next, we performed calcium imaging in freely-moving mice to probe for neural correlates of decision elements and flight behaviour in brain areas that we show to be necessary for the flight responses: we found that VGluT2 neurons in the deeper medial superior colliculus (dmSC) increase their activity during a repeated threatening stimulus, while VGluT2 neurons of the dorsolateral periaqueductal gray (dPAG) are silent until just before the initiation of escape, and are maximally active during escape.

These results suggest that the dmSC accumulates evidence of threat which dPAG neurons threshold. This interpretation is supported by optogenetic activation of mSC-VGluT2 neurons in vivo, which recapitulates the statistics of escape probability evoked with a visual stimulus, while activation of VGluT2 neurons in the dPAG evokes an all-or-nothing escape response.

Finally, using channelrhodopsin-2-assisted circuit mapping and monosynaptic viral tracing, we reveal that over half of dPAG-VGluT2 neurons receive monosynaptic connections from mSC-VGluT2 neurons with a low probability of release, allowing this synapse to act as a high-pass filter and providing a mechanism for the computation of an escape decision. These findings advance our understanding of how defensive behaviours are generated at circuit and single-cell level, and of how neurons process information in a circuit critical for implementing basic behaviours.

Description

Date

2017-09-29

Advisors

Branco, Tiago
Jefferis, Gregory

Keywords

behaviour, innate, neuroscience, neural networks, synapse, escape, defensive, systems

Qualification

Doctor of Philosophy (PhD)

Awarding Institution

University of Cambridge