Reward learning deficits as a cross-species cognitive biomarker of negative affective states

Abstract

Animal welfare is an important topic of societal interest. Despite this, measuring animal affective states objectively and reliably is challenging. Animals are not able to directly tell us how they feel, so numerous methods have been developed to monitor affective state indirectly. Traditional physiological and behavioural welfare assessment methods have often been challenging to interpret. More recently, there has been interest in using affect-induced cognitive biases to examine affective state more specifically. However, these methods have also faced difficulties with reliable protocol design across species. This thesis investigates reward learning as novel method of assessing affect-induced cognitive biases. Reward learning deficits are a feature of human depression and may serve as a reliable cross-species biomarker of negative affect.

Previous work has demonstrated reward learning deficits in mice in putative negative states, using a novel associative learning task conducted in a T-maze. We progressed this by assessing reward learning in mice subjected to further affective state manipulations. Group-housed mice showed signs of reward learning deficits consistent with negative states, unlike singly-housed mice. However, when using repeated restraint as a stressor, neither control nor restraint animals could learn the task. Further experimental protocol optimisation and refinement is needed. To initiate investigation of reward learning and affective states in other species, a novel reward learning testing arena was also developed in chickens. Whether chickens can discriminate between different reward magnitudes in the arena still needs to be validated, before we can test whether reward learning deficits associate with negative affective states, in this species.

While previous work suggested that reward learning deficits could be used as a “cognitive” biomarker of negative affective states, this thesis highlighted some challenges with protocol design practicality and reliability. Future work needs to address this, if reward learning is to be widely applied for welfare assessment in practice.

Date of Award	5 Dec 2023
Original language	English
Awarding Institution	University of Bristol
Supervisor	Jennifer Davies (Supervisor) & Emma S J Robinson (Supervisor)