Cognitive-behavioural testing in preclinical models of Alzheimer’s disease has failed to capture deficits in goal-directed action control. Here we provide the first comprehensive investigation of goal-directed action in a transgenic mouse model of Alzheimer’s disease. Specifically, we tested outcome devaluation performance in male and female human amyloid precursor protein (hAPP)-J20 mice. Mice were first trained to press left and right levers for pellet and sucrose outcomes respectively (counterbalanced) over four days. On test, mice were pre-fed one of the outcomes to satiety and given a choice between levers. Devaluation performance was intact for 36-week-old wildtypes of both sexes, who responded more on the valued relative to the devalued lever (Valued > Devalued). By contrast, devaluation was impaired (Valued = Devalued) for J20 mice of both sexes, and for 52-week-old male mice regardless of genotype. After additional lever press training (i.e., 8 days lever pressing in total), devaluation was intact for all mice, demonstrating that the initial deficits were not a result of a non-specific impairment in reward processing, depression, or locomotor activity in J20 or aging mice. Follow up analyses revealed that microglial expression in the dorsal CA1 region of the hippocampus was associated with poorer outcome devaluation performance on initial, but not later tests. Together, these data demonstrate that goal-directed action is initially impaired in J20 mice of both sexes and in aging male mice regardless of genotype, and that this impairment is related to neuroinflammation in the dorsal CA1 hippocampal region.Significance statementTreatments for Alzheimer’s disease trialled in preclinical animal models have repeatedly failed to translate to the clinic. One potential reason for this could be that the cognitive-behavioural assays used in such models have been limited to one aspect of Alzheimer impairment: visuospatial memory. Here we demonstrate that male and female mice belonging to the transgenic hAPP-J20 mouse model also display consistent deficits in the initial acquisition of goal-directed action. This study therefore represents an important step towards the broader capture of Alzheimer-like cognitive deficits at a preclinical level which could improve translatability when used to more comprehensively test treatment efficacy prior to clinical trials.

Several lines of evidence accrued over the last 5–10 years have converged to suggest that the parafascicular nucleus of the thalamus and the lateral orbitofrontal cortex each represent or contribute to internal state/context representations that guide action selection in partially observable task situations. In rodents, inactivations of each structure have been found to selectively impair performance in paradigms testing goal-directed action selection, but only when that action selection relies on state representations. Electrophysiological evidence has suggested that each structure achieves this function via inputs onto cholinergic interneurons (CINs) in the dorsomedial striatum. Here, we briefly review these studies, then point to anatomical evidence regarding the afferents of each structure and what they suggest about the specific features that each contribute to internal state representations. Finally, we speculate as to whether this role might be achieved interdependently through direct PF→OFC projections, or through the convergence of independent direct orbitofrontal cortex (OFC) and parafascicular nucleus of the thalamus (PF) inputs onto striatal targets.