Background

Cognitive flexibility is the ability to shift behavioral strategy based on context to receive a desired outcome. This process depends on the prefrontal cortex (PFC); lesion studies, optogenetic studies, and humans with damaged PFCs support a critical role for the PFC in cognitive flexibility [1–3]. Deficits in cognitive flexibility are associated with frontal lobe dysfunction, particularly in a range of neuropsychiatric disorders including schizophrenia [4], attention deficit/hyperactivity disorder [5], Alzheimer’s disease [6], major depressive disorder [7], and addiction [8]. Moreover, age-related cognitive deficits are predicted to rise dramatically given the predicted demographic trends in the coming decades [9]. No current treatment for these disorders specifically targets these deficits, in part because the neural circuitry of cognitive flexibility is not yet fully understood. Thus, it is necessary to incorporate techniques that directly manipulate or record specific neural populations to help distinguish the neural mechanisms and circuitry underlying cognitive flexibility. Circuit-specific investigations in rodent models will play a critical role in finding future treatments.

Rodent cognitive flexibility tasks involve animals learning a decision rule, typically motivated by consummatory reward, most frequently food or water. After dozens to hundreds of trials, the experimenter changes the task rule without cueing the animal, requiring the animal to change its strategy to continue to receive the same reward. These rodent paradigms are inspired by human neuropsychological tests. A prime example is the Wisconsin Card Sorting Task, which involves a deck of cards depicting objects that differ by three dimensions: color, shape, and number [1,9]. Participants are instructed to sort the cards into piles in one of these dimensions. Participants are never explicitly told the active rule, but rather, are given feedback after each trial (correct/incorrect). Once the participants learn which dimension is “active” and reach criterion accuracy, the task rule (active dimension) covertly switches. Thus, participants must adapt their strategy and apply a new rule to maintain positive feedback.

To identify relevant functional circuitry in cognitive flexibility, animal cognitive paradigms can be beneficial for enabling detailed in vivo imaging, electrophysiology, pharmacology, and optogenetic manipulation of neurons in relevant brain regions. Much of the foundational work in this area used free-roam versions of an attentional set-shifting (ASST) paradigm, which involved learning and alternating between two task rules associated with sensorimotor pairings, most frequently involving olfaction and somatosensation. Training for ASST involves multiple stages, with each building upon the previous stage [10–12]. A well-established example of the rodent ASST involves a testing chamber with multiple bowls containing digging media that vary by texture and odor. One bowl contains a food pellet on each trial [13,14], and the animal must switch between odor and texture rules to locate the food reward. The training begins with the animal learning simple discriminations between stimuli within one modality (e.g., gravel vs. sandy digging medium). As the training progresses, stimuli from the distractor modality are included (e.g., citrus vs. clove-scented bowls), eventually building up to the modality rule-switching ASST task. While this free-roam task incorporates a naturalistic foraging model, it carries constraints that include limits on trial numbers, extensive manual intervention by the experimenter, and limits on the neural recording techniques that can be employed. Some head-fixed paradigms have been used to study cognitive flexibility, which particularly incorporate virtual reality environment navigation (i.e., associating reward with a particular location within a virtual environment) [15] or environments depicting different visual gratings that are consequently associated with a reward [17]. It is also possible to model cognitive flexibility in head-restrained mice without requiring explicit stimulus discrimination by employing a 2-armed bandit or reversal paradigm. While these alternative approaches test cognitive flexibility, the cross-modal attentional set-shifting task presented here affords the unique advantage of building on an extensive experimental literature that spans basic, translational, and clinical work. Head-restrained experiments can also be designed to be temporally specific by having fixed times between stimulus, behavior, and reward—an advantage over free-roam navigation tasks. Trade-offs of the technique presented here include sacrificing naturalistic behavior (i.e., foraging) as well as the potential for stress associated with head-restraint [17]. However, proper and repeated habituation and handling help animals adjust to temporary restrictions.

This protocol includes instructions on behavioral training to prepare animals to learn the Serial Extradimensional Set-Shifting Task (SEDS). The SEDS task involves two modalities: whisker vibrations (e.g., 35 Hz vs. 155 Hz) and odorants (e.g., almond oil vs. nutmeg oil) [2]. Water-restricted mice are placed on a treadmill facing two lick spouts and must use one of these sensory stimuli to determine which spout releases water during each trial; both sensory modalities are presented concurrently, but within each trial block, only one of the two modalities signals the correct spout for each trial. In the paradigm presented here, as in the WCST and other rodent rule-switching paradigms [17], the rule switches each time the animal reaches criterion accuracy (80% correct within a 30-trial moving window). However, the switch frequency can also be set to a fixed trial number [18] or pseudo-randomized [19], depending on the constraints of the experiment.

Through modifications to the FlexRig hardware and Arduino software, the behavioral protocol can be edited for application to other tasks, including but not limited to delayed non-match to sample tasks, serial pattern learning tasks, concurrent schedule learning, multi-armed bandit tasks, and other conditioning paradigms.