Delineating Adaptive and Maladaptive Outcomes of Aggression in Mice

Authors: Barbara Benowitz, Meghan Flanigan

Aggression has evolved across species to serve a range of adaptive purposes, including defense, resource obtainment such as food and territory, and facilitation of mating (1). Although even intense aggression is a natural and necessary aspect of animal social behavior, aggression in humans is considered maladaptive under most circumstances as in our society, it does not often serve the purposes of protection, resource procurement, or fitness (2). Compared with adaptive aggressive displays, maladaptive aggressive displays in both humans and animals may be disproportionate, indiscriminate, or unnecessarily risky. Importantly, maladaptive aggression in humans has often been associated with a variety of neuropsychiatric disorders and has detrimental consequences for patients, their families, and their communities (3). As a result, a major objective of aggression research in animal models is to elucidate the neurobiology of maladaptive human aggressive behavior so that we may develop more effective psychiatric treatments. Critically, the neural circuitries engaged in adaptive versus maladaptive aggression are thought to be only partially overlapping (4), thus necessitating animal models that clearly distinguish between the two.SEE CORRESPONDING ARTICLE NO. 100399

Classically, rodent models of aggressive behavior use a dyadic model of territorial aggression called the resident intruder (RI) test. This test involves introducing a novel male intruder conspecific into the home cage of a male experimental mouse to induce territorial aggression in the resident (5). Some researchers have used both qualitative and quantitative measures to infer whether aggression is species-typical or species-atypical in the RI test, which has been proposed to loosely align with adaptive and maladaptive distinctions (5). Measures potentially indicating species-atypical aggression in the rodent RI test include low latencies to attack, continued attacks to smaller and/or subordinate individuals, attacks against females or juveniles, and attacks to vulnerable body parts such as the neck and abdomen (3). Despite these logical indications that an individual is engaging in aggression that is excessive for a given context, whether species-atypical aggressive displays are directly associated with maladaptive outcomes cannot be determined in the RI test. Therefore, the RI test is incapable of providing necessary information regarding the nature of aggressive displays and whether they serve a valuable purpose for the individual. In order for studies to determine whether rodent aggression is adaptive or maladaptive, which is critical for their utility in elucidating human aggression, they must employ holistic models in ethologically relevant environments that include multiple individuals; valuable resources to obtain such as food, water, and shelter; and naturally occurring, rather than forced, dyadic and nondyadic social interactions.

In a recent study in Biological Psychiatry: Global Open Science, Anpilov et al. (6) explored the adaptive versus maladaptive nature of aggression in two models of increased aggression in mice using an environmentally complex group housing apparatus called the social box (SB). Importantly, the two models of increased aggression characterized in this study were enhanced aggression was driven by early exposure to an enriched environment (EE) for one model and genetic oxytocin receptor deficiency (OxtR^−/−^) for the other. The large, physically varied SB apparatus was designed to allow spontaneous interactions, resource competition, and the formation of social hierarchies among the mice. A camera system captured video continuously during active hours, and each mouse, uniquely marked with a fur color, was automatically tracked to monitor specific individual and social behaviors. To analyze the rich dataset generated in this setting, the authors used advanced computational analyses, including machine learning–based models to parse out interaction patterns. They used a hidden Markov model to classify interactions based on distance and movement angles within and between mice, identifying behaviors such as approach, contact, and chase. Principal component analysis and logistic regression further distinguished specific individual and social behaviors by treatment group, providing insights into the differential use of aggression and social strategies between EE mice, OxtR^−/−^ mice, and their relevant control mice. This comprehensive methodological approach enabled the authors to translate raw behavioral data into meaningful indicators of adaptive versus maladaptive aggression, highlighting how each model’s aggressive behavior impacted group cohesion, resource access, and individual roles within social hierarchies.

The researchers first used early-life exposure to an EE to model enhanced aggression relative to mice reared in a standard cage (SC). For the EE condition, male mice were group housed after weaning with 16 individuals per spacious cage containing various forms of enrichment, such as tunnels, shelters, running wheels, and nesting boxes. This configuration allowed for both physical and social stimulation by providing ample space and opportunities for exploration, contrasting starkly with the more constrained standard cages housing 4 mice each without such enrichment. Following 6 weeks of housing in EE or SC, mice were separated into groups of 4 and continuously observed in the SB. EE mice displayed fewer overall social interactions and a generally lower level of aggression than SC mice. This observation highlights an important discrepancy between aggression measured in RI and aggression measured in SB because previous studies have shown EE mice to engage in increased aggression in RI (7). Interestingly, when social interactions did occur in the EE groups, there was a higher likelihood that these interactions would escalate into aggressive encounters compared with SC groups. This suggests that EE mice may use aggression more strategically. In other words, by limiting aggressive encounters to only those necessary for maintaining rank or securing resources, EE mice demonstrated a form of adaptive aggression that promotes group cohesion and minimizes injury risk within the colony. The observed stability of dominance hierarchies in the EE groups further supports the notion that selective aggression can function as a tool for establishing and reinforcing social order.

The second model, involving oxytocin receptor–deficient (OxtR^−/−^) mice, revealed a different dimension of aggression. These mice were more socially interactive and engaged in a higher number of reciprocal chases and social encounters than their OxtR^+/+^ counterparts. This elevated social interaction in the OxtR^−/−^ groups led to a higher absolute rate of aggressive interactions, raising the potential risk of injury due to the sheer number of conflicts. However, unlike the EE mice, where aggression was selectively used to maintain stability, OxtR^−/−^ mice displayed a pattern of frequent, widespread aggression that was less regulated. Despite this increase in aggressive behavior, the dominance hierarchies in OxtR^−/−^ groups were preserved. However, this stability came at the cost of more continuous aggressive exchanges rather than the occasional, strategically deployed aggression seen in EE mice. The OxtR^−/−^ model thereby likely exemplifies a form of maladaptive aggression that, while effective at maintaining social structure, is inefficient and potentially harmful to individual and group well-being.

While this study significantly advances our understanding of aggression dynamics, there are several limitations that highlight areas for future research as well as model refinement. One critical limitation stems from the lack of a detailed classification system for discrete aggression behaviors. While Anpilov et al. attempted to classify individual aggressive behaviors within the SB, their reliance on hidden Markov modeling to parse interactions proved limited. This shortcoming highlights the difficulty of classifying aggressive behaviors computationally with the current methods they described, especially in dynamic social environments. Future studies should attempt to use different or additional methods of analysis to remedy this problem. For example, using programs such as SimBA [Simple Behavioral Analysis (8)] or LabGym (9) may offer more advanced classification capabilities that could help future studies in the SB to identify and differentiate specific aggressive behaviors with greater accuracy. More detailed readouts of discrete aggressive and nonaggressive behaviors in future studies may also permit further classification of aggression subtypes beyond adaptive and maladaptive distinctions.

In addition to being adaptive and/or maladaptive, aggression can also be categorized into proactive and/or reactive. Proactive or goal-oriented aggression is characterized by low arousal and driven by positive reinforcement, noted by behaviors such as stalking, slow approach, and targeted attacks delivered with low latency (2). Such goal-oriented actions can be critical for adaptively establishing social rank if they are limited to interactions that confer specific survival or reproductive advantages. However, given the context and the intensity, proactive aggression may be adaptive or maladaptive. In humans, maladaptive proactive aggression is often observed in conditions such as antisocial personality disorder (3). In contrast, reactive, or defensive, aggression is more emotional, triggered by perceived threats, and is marked by high arousal without positive reinforcement, which can also be adaptive or maladaptive depending on context. In humans, maladaptive reactive aggression is often observed in conditions such as intermittent explosive disorder (3). The absence of measures such as low attack latency or targeted unprovoked attacks in the current study raises questions about whether the aggression observed is goal directed or defensive, which is critical for understanding distinct neural mechanisms and clinical implications. An alternative method to explore aggression motivation may involve tests such as aggression-conditioned place preference and self-administration, which have shown that proactive aggression is related to reward-seeking behavior (10). Combining these paradigms with the SB could clarify whether EE and OxtR^−/−^ mice exhibit proactive aggression tendencies in future studies.

Overall, the study by Anpilov et al. represents an important advancement in aggression research by employing a seminatural setup and novel behavioral analysis approaches to investigate complex individual and social dynamics in mice. By comparing the aggressive behaviors of EE and OxtR^−/−^ mice, the authors provide critical insights into how different models of enhanced aggression can influence social hierarchy and resource access, facilitating a distinction between adaptive and maladaptive aggressive patterns. Future research integrating the authors’ SB approach with in vivo physiological measurements (electrophysiology, fiber photometry, or 1P microscopy) could further illuminate the distinct neurobiological mechanisms underlying adaptive or maladaptive aggression, which would ultimately strengthen the translational relevance of animal models for understanding and treating aggression-related disorders in humans.