Mouse Behavior on the Trial-Unique Nonmatching-to-Location (TUNL) Touchscreen Task Reflects a Mixture of Distinct Working Memory Codes and Response Biases.

The trial-unique nonmatching to location (TUNL) touchscreen task shows promise as a translational assay of working memory (WM) deficits in rodent models of autism, ADHD, and schizophrenia. However, the low-level neurocognitive processes that drive behavior in the TUNL task have not been fully elucidated. In particular, it is commonly assumed that the TUNL task predominantly measures spatial WM dependent on hippocampal pattern separation, but this proposition has not previously been tested. In this project, we tested this question using computational modeling of behavior from male and female mice performing the TUNL task (N = 163 across three datasets; 158,843 trials). Using this approach, we empirically tested whether TUNL behavior solely measured retrospective WM, or whether it was possible to deconstruct behavior into additional neurocognitive subprocesses. Overall, contrary to common assumptions, modeling analyses revealed that behavior on the TUNL task did not primarily reflect retrospective spatial WM. Instead, behavior was best explained as a mixture of response strategies, including both retrospective WM (remembering the spatial location of a previous stimulus) and prospective WM (remembering an anticipated future behavioral response) as well as animal-specific response biases. These results suggest that retrospective spatial WM is just one of a number of cognitive subprocesses that contribute to choice behavior on the TUNL task. We suggest that findings can be understood within a resource-rational framework, and use computational model simulations to propose several task-design principles that we predict will maximize spatial WM and minimize alternative behavioral strategies in the TUNL task.SIGNIFICANCE STATEMENT Touchscreen tasks represent a paradigm shift for assessment of cognition in nonhuman animals by automating large-scale behavioral data collection. Their main relevance, however, depends on the assumption of functional equivalence to cognitive domains in humans. The trial-unique, delayed nonmatching to location (TUNL) touchscreen task has revolutionized the study of rodent spatial working memory. However, its assumption of functional equivalence to human spatial working memory is untested. We leveraged previously untapped single-trial TUNL data to uncover a novel set of hierarchically ordered cognitive processes that underlie mouse behavior on this task. The strategies used demonstrate multiple cognitive approaches to a single behavioral outcome and the requirement for more precise task design and sophisticated data analysis in interpreting rodent spatial working memory.

Maternal immune activation targeted to a window of parvalbumin interneuron development improves spatial working memory: Implications for autism.

Maternal immune activation (MIA) increases risk for neuropsychiatric disorders such as autism spectrum disorder (ASD) in offspring later in life through unknown causal mechanisms. Growing evidence implicates parvalbumin-containing GABAergic interneurons as a key target in rodent MIA models. We targeted a specific neurodevelopmental window of parvalbumin interneurons in a mouse MIA model to examine effects on spatial working memory, a key domain in ASD that can manifest as either impairments or improvements both clinically and in animal models. Pregnant dams received three consecutive intraperitoneal injections of Polyinosinic:polycytidylic acid (poly(I:C), 5 mg/kg) at gestational days 13, 14 and 15. Spatial working memory was assessed in young adult offspring using touchscreen operant chambers and the Trial-Unique Non-matching to Location (TUNL) task. Anxiety, novelty seeking and short-term memory were assessed using Elevated Plus Maze (EPM) and Y-maze novelty preference tasks. Fluorescent immunohistochemistry was used to assess hippocampal parvalbumin cell density, intensity and co-expression with perineuronal nets. qPCR was used to assess the expression of putatively implicated gene pathways. MIA targeting a window of parvalbumin interneuron development increased spatial working memory performance on the TUNL touchscreen task which was not influenced by anxiety or novelty seeking behaviour. The model reduced fetal mRNA levels of Gad1 and adult hippocampal mRNA levels of Pvalb and the distribution of low intensity parvalbumin interneurons was altered. We speculate a specific timing window for parvalbumin interneuron development underpins the apparently paradoxical improved spatial working memory phenotype found both across several rodent models of autism and clinically in ASD.

Raloxifene recovers effects of prenatal immune activation on cognitive task-induced gamma power.

There is currently no treatment available for the cognitive symptoms of schizophrenia, but evidence suggests that selective estrogen receptor modulators (SERMs) may provide relief. Our recent animal model data showed that a lack of female sex hormones in mice impairs the ability of hippocampal neurons to synchronise and generate oscillations within the frequency range of 30-80 Hz (gamma power) leading to cognitive impairment, while both estradiol and the SERM, raloxifene, recovered this. Given that cognitive impairment is accompanied by abnormal gamma power in schizophrenia, this study aimed to determine the effects of raloxifene on gamma power during spatial memory tasks in the prenatal immune challenged (poly-I:C) mouse model with relevance to schizophrenia. Pregnant dams received the viral mimetic poly-I:C (20 mg/kg, i.p.) at gestational day 17. Male and female offspring were treated with placebo or raloxifene implants at adulthood. Local field potentials from the CA1 hippocampus were simultaneously recorded during the Y-maze test of short term spatial memory and the cheeseboard maze test of long-term spatial learning and memory and cognitive flexibility. In female but not male mice, poly I:C exposure reduced gamma power during decision making and prolonged the time spent in the centre (decision making phase) during the Y-maze task. Female poly-I:C exposed mice also showed increased gamma power during acquisition of the cheeseboard long term memory task and perseverative behaviour. Treatment with raloxifene recovered gamma power and decision making deficits in the Y-maze and restored gamma power changes during the cheeseboard maze task as well as perseverative behaviour. Male mice showed no electrophysiological or behavioural effects of poly-I:C or raloxifene treatment. In summary, poly-I:C exposure induced female specific cognitive impairments accompanied by altered neural oscillations in the gamma frequency and raloxifene recovered these abnormalities.

The maternal immune activation model uncovers a role for the Arx gene in GABAergic dysfunction in schizophrenia.

A hallmark feature of schizophrenia is altered high frequency neural oscillations, including reduced auditory-evoked gamma oscillatory power, which is underpinned by parvalbumin (PV) interneuron dysfunction. Maternal immune activation (MIA) in rodents models an environmental risk factor for schizophrenia and recapitulates these PV interneuron changes. This study sought to link reduced PV expression in the MIA model with alterations to auditory-evoked gamma oscillations and transcript expression. We further aligned transcriptional findings from the animal model with human genome sequencing data. We show that MIA, induced by the viral mimetic, poly-I:C in C57Bl/6 mice, caused in adult offspring reduced auditory-evoked gamma and theta oscillatory power paralleled by reduced PV protein levels. We then showed the Arx gene, critical to healthy neurodevelopment of PV interneurons, is reduced in the forebrain of MIA exposed mice. Finally, in a whole-genome sequenced patient cohort, we identified a novel missense mutation of ARX in a patient with schizophrenia and in the Psychiatric Genomics Consortium 2 cohort, a nominal association of proximal ARX SNPs with the disorder. This suggests MIA, as a risk factor for schizophrenia, may be influencing Arx expression to induce the GABAergic dysfunction seen in schizophrenia and that the ARX gene may play a role in the prenatal origins of schizophrenia pathophysiology.

Estradiol and raloxifene modulate hippocampal gamma oscillations during a spatial memory task.

Previous work suggests that estradiol regulates the expression of hippocampal parvalbumin as well as hippocampus-dependent spatial memory in mice. Parvalbumin interneurons generate neuronal oscillatory activity in the gamma frequency range (30-80Hz) and gamma oscillations are closely linked with higher cognitive functions. Raloxifene, a selective estrogen receptor modulator, shows beneficial effects on human cognitive performance, and has few peripheral side effects unlike estradiol, but the biological mechanisms which underpin these benefits are not clear. This study aimed to investigate whether estradiol and raloxifene modulate hippocampal gamma-band oscillations during spatial memory performance. Prepubescent female mice were ovariectomized (OVX) and implanted with a subcutaneous pellet of either estradiol (E2), raloxifene or placebo. During adulthood, local field potentials were recorded from the dorsal hippocampus while mice were performing the Y-maze hippocampus-dependent spatial memory task. Ovariectomy caused deficits in spatial memory, accompanied by a significant reduction in hippocampal gamma oscillations, specifically during decision making. Estradiol as well as raloxifene rescued both behavioural and electrophysiological deficits. These data have significant implications for disorders of cognitive impairment where altered gamma oscillations are apparent, such as schizophrenia.