An Insula-Enriched Regulator of Retinoic Acid Marks a New Intersection in the Neural Circuitry of Mouse Social Behavior

Research in translational psychiatry seeks to link aspects of human psychopathology with biological mechanisms using model organisms such as laboratory mice. Disruptions in social behavior manifest in many psychiatric disorders, notably autism spectrum disorder and schizophrenia. Neuroimaging studies involving these patient populations reveal differences in insular cortex function associated with social tasks, such as emotion recognition (1). Beyond social cognition, insular abnormalities are among the most frequently identified in meta-analyses in psychiatry, positioning it as a putative common anatomical hub for pathophysiology (2). In search of molecules and signaling cascades that might explain pathophysiology in the insula, Kim et al. (3), reporting in this issue of the Journal, analyzed the genes of neurons in the mouse anterior agranular insular cortex, an area that plays a role in sociability and salience detection. By cross-referencing their results with genome-wide association study data obtained from postmortem brains of people with autism spectrum disorder, schizophrenia, or bipolar disorder (4), the retinoic acid–degrading enzyme cytochrome P450 26B1 (Cyp26B1) emerged as an intriguing candidate. The ensuing studies provide a compelling case that Cyp26B1 and retinoic acid in the insula are necessary for social novelty seeking and are anatomically situated to integrate the effects of the social neuropeptide oxytocin by a previously unknown serotonergic mechanism.

Retinoic acid is a growth factor derived from vitamin A (retinol) that regulates nuclear gene transcription. Retinoic acid was first understood for its role in cellular differentiation during neural development (5) and later for the maintenance of synaptic plasticity by regulation of postsynaptic glutamate receptors (6). Retinoic acid signaling is controlled by the cytochrome oxidase enzymes, of which Cyp26B1 is predominant in the brain. To investigate the relationship of retinoic acid and Cyp26B1, the experimenters selectively reduced or overexpressed Cyp26B1 using virus-mediated gene transfer techniques. Insular Cyp26B1 knockdown caused changes in genes associated with growth and maintenance of dendritic spines, decreased the number of spines on layer 5 insular pyramidal neurons, and reduced glutamatergic excitatory transmission. Importantly, all of these changes were reversed by overexpression of Cyp26B1.

Because the social lives of rodents are naturally quite nuanced, assessments of social behavior in translational neuroscience are often simplified to allow for objective quantification of the effects of neurobiological manipulations on social tendencies. The experimenters used a behavioral test consisting of two phases. At first, a test mouse was given a few minutes to interact with another mouse, which provided a measure of sociability, an endpoint that is altered in mouse models of autism spectrum disorder (7). This was followed by a second test in which the experimental mouse was presented with two mice, the original and a novel one. Under usual laboratory circumstances, mice will spend more time investigating the new mouse, a phenomenon reflecting an innate social novelty preference. Drawing a parallel to the vast human neuroimaging literature implicating the insula in social cognition, the experimenters first demonstrated that lesions to the anterior insula reduce the amount of time spent investigating the novel mouse without affecting sociability itself. Ensuing studies revealed that chronic retinoic acid administration and Cyp26B1 knockdown or knockout also reduced social investigation directed to the novel mouse but did not alter sociability itself. Furthermore, the deficit in social novelty interactions was rescued by Cyp26B1 overexpression or direct excitation of the insula.

Direct stimulation of insular neurons is not yet a viable therapy for humans, so next, Kim et al. turned to oxytocin, a neuromodulator that is well established as necessary for social novelty preference and many other social behaviors in rodents and primates (8). Surprisingly, while administration of oxytocin to the whole brain (via infusion to the ventricles) rescued social novelty preference in insula Cyp26B1 knockdown mice, direct infusion to the insula did not. This paradox generated a new question: By what mechanism did oxytocin restore behavior? Analysis of putative excitatory neural inputs to the agranular insular cortex revealed a strong input from the dorsal raphe nucleus, a brainstem region that provides the majority of serotonergic input to the forebrain (9), is a modulator of social behavior itself (10), and is innervated by oxytocinergic projections from the hypothalamus (11). The experimenters mimicked stimulation of insular serotonin receptors and found that only 5-HT_2C receptor agonists could rescue social novelty-seeking behaviors and the deficits in synaptic transmission present in the Cyp26B1 knockdown mice. Although not directly observed, the inference made from these results is that the social novelty test involves augmentation of the dorsal raphe serotonin neurons by oxytocin, leading to greater release of 5-HT with action at the insular 5-HT_2C receptors, which appear to modulate the insular synaptic transmission and circuit excitability necessary for directing social investigation toward the novel mouse. Consistent with this model, direct infusion of oxytocin to the dorsal raphe recapitulated the rescue of social deficits, suggesting that the raphe input to the insula may be an important mechanism by which systemic increases in oxytocin can influence brain network function (12).

These interesting results add to a long list of rodent behaviors that are influenced by manipulations of the insular cortex. A challenge to integrating these findings is identifying the unifying themes that may inform the sort of information processing that engages the insula. Indeed, insular neurons are involved in conditioned taste aversion learning (13), drug seeking and relapse (14), hunger and thirst (15), threat learning and expression of threat-associated behaviors (16), and several aspects of sociability (17). I suggest that the insula contributes something common to these behaviors, given evidence at the level of cellular and circuit physiology that the insula is a locus of multisensory integration (18) and internal state estimation (15). Looking more critically at the social behaviors used by Kim et al., interference with insula processing reduced the amount of time mice directed to social investigation of an unfamiliar mouse. Although it is appealing to conclude that this is a consequence of “recognizing” that the conspecific mouse is novel, recognition is a construct describing a set of processes by which sensory information about an individual is associated with declarative memory about their identity. While it is possible that the insular cortex contributes to the process of encoding a mismatch between the social sensory information and social memory needed to identify novelty, such processes are not directly observable in mice, which limits the inferences one can make about the motivation or psychological mechanisms that underlie a mouse’s preference for social novelty. Additionally, in the human brain, social recognition is centralized to occipital and temporal areas, such as the fusiform face area and the superior temporal sulcus (19), and the mouse insula is not homologous to these structures.

Alternatively, the effect of insula manipulations on social novelty-seeking behavior may not have to do with the discrimination between which target rat is familiar and which is new, but rather about what to do with that information. The decision to approach or avoid another individual is shaped by the attributes of the other, the affective state of the individual, and the environmental context for interaction. For example, humans tend to be prosocial in familiar, safe settings but avoid social interactions, especially with unfamiliar others, if they experience distress themselves or perceive the environment or others to be dangerous. And there is evidence that rodent social approach or avoidance behaviors are similarly influenced by self, other, and contextual variables (20). Usually mice are tested in safe and familiar settings conducive to social interactions, and it is likely the mouse uses environmental stimuli—such as the level of illumination, odors indicating the absence of predators, and social stimuli that may convey the conspecific’s sex, age, or affect (such as aggression or fear)—to determine whether approaching or avoiding the novel conspecific is appropriate. A model for how the insular cortex may shape the direction of behavior via its projections to areas of the “social brain” (21) unfolds as follows. In the safe and familiar setting, insular projections to appetitive areas like the ventral striatum may augment the encoding of possible reward associated with engaging in social interaction with the novel mouse. However, if the test is conducted in the presence of a predator odor, for example, or if the test subjects themselves experienced stress, then mouse behavior is likely to be directed away from social novelty and toward the familiar, which they know is safe. This may be mediated by insular projections to areas mediating defensive behavior, such as the bed nucleus of the stria terminalis or the central amygdala (17). Indeed, when considering the rodent insula literature more broadly, it becomes clear that the insula contributes to the direction of behavior as consummatory or hedonic in safe situations but also can cause behavioral inhibition if environmental stimuli or internal needs, such as hunger, are imminent (15).

From this perspective, the data of Kim et al. may expose the edge of what is ultimately a larger range of socioaffective processes that involve the insular Cyp26B1 neurons, which might apply to psychopathology beyond the social realm. It is intriguing to consider how targeting this circuit therapeutically might unfold clinically. For example, in light of the frenzy to identify fast-acting antidepressants, retinoic acid and synthetic agonists for its receptors have emerged as new targets. Like ketamine, which affects behavior by modulation of synaptic plasticity, retinoic acid and its receptor agonists themselves can, in the mature mouse brain, rapidly augment synaptic efficacy concomitant to antidepressant-like actions (22). Perhaps a boost to synaptic plasticity could offset aberrations in insular circuit function attributed to abnormalities in retinoic acid signaling and help individuals with disordered social behavior better incorporate social, environmental, and internal factors for more successful and meaningful social interactions.