Aggressive behavioral phenotypes in mice

https://doi.org/10.1016/S0166-4328(01)00298-4Get rights and content

Abstract

Aggressive behavior in male and female mice occurs in conflicts with intruding rivals, most often for the purpose of suppressing the reproductive success of the opponent. The behavioral repertoire of fighting is composed of intricately sequenced bursts of species-typical elements, with the resident displaying offensive and the intruder defensive acts and postures. The probability of occurrence as well as the frequency, duration, temporal and sequential patterns of aggressive behavior can be quantified with ethological methods. Classic selection and strain comparisons show the heritability of aggressive behavior, and point to the influence of several genes, including some of them on the Y chromosome. However, genetic effects on aggressive behavior critically depend upon the background strain, maternal environment and the intruder. These factors are equally important in determining changes in aggressive behavior in mice with a specific gene deletion. While changes in aggression characterize mutant mice involving a variety of genes, no pattern has emerged that links particular gene products (i.e. enzyme, peptide, receptor) to either an increase or a decrease in aggressive behavior, but rather emphasizes polygenic influences. A potentially common mechanism may be some components of the serotonin system, since alterations in 5-HT neurotransmission have been found in several of the KO mice that display unusual aggressive behavior.

Introduction

The behavioral biology of mouse aggression offers insights to understanding the neurobiological and molecular mechanisms mediating behavior in social conflict, the critical developmental and adult determinants, and the functional significance of conflict in males and females [69]. In an effort to trace the behavioral phenotypes of aggressive behavior to chromosomal candidate mechanisms, classic genetic analyses proceeded via strain comparisons and selective breeding to quantitative trait loci (i.e. ‘top-down’ genetics) [63], [102]. More recently, the ‘bottom-up’ genetic approach has attracted an audience in scientific and popular journals finding that single gene mutations can engender unusual aggressive behavior [37], [87].

The term ‘aggression’ is scientifically useful when it specifies the particular type of aggressive behavior, and differentiates the motivation to fight from the actual behavior [6]. More than 50 years ago, the first summary of the early research on rodent aggression emphasized the importance of antecedent aversive stimuli that provoke aggression, and how the consequences of aggression reinforce and suppress this behavior [101]. However, a prerequisite to understanding the causes of aggression in mice is to appreciate the social organization of this species (Mus musculus). Several ethological analyses have characterized the features of social intercourse among wild house mice in different ecological niches highlighting individual variation in frequency, intensity and probability of aggression in this rather pugnacious species [7], [17].

House mice most often develop in social units that have been described as demes (breeding units) or ‘Grossfamilie’. They are composed of the parental generation, pre-pubertal juveniles and pups [24]. When males begin to leave their parental deme, these young adults and subadults form an itinerant population [98], [123]. Breeding males mark, patrol and defend their territories [17], and exclude other males (‘exclusive territory’) or dominate other males (‘dominance territory’) [18], [42], [91]. More than 90% of all offspring are sired by these dominant males, while subdominant and subordinate males are rarely seen mating and are subject to frequent attacks [58]. Dominant females fight in defense of food and safe nesting sites against invading males and females [42]. The predominant forms of biting attacks in mice occur in situations of social conflict, when a male defends against a territorial intruder (e.g. territorial aggression, intermale aggression) or when a female defends the pups (e.g. maternal aggression). Breeding success, social status, and access to resources—indices of the relative fitness of the individual—are enhanced by prevailing in such confrontations.

Housing conditions of laboratory mice rarely allow for the formation of demes, but rather force several adult males or females to cohabitate [78]—a setting that is at variance with the social organization of this species [41]. In the daytime, laboratory male mice are often seen to huddle together, indicating sleep behavior for this nocturnal species [9]. Most male house mice require considerable experience with other members of their species before they display aggressive behavior as adults [48]. In group-housed Swiss mice despotic relationships often develop, with one male dominant and the remaining males subordinate [5], [90], [118], whereas males from many other strains of laboratory mice fail to show any aggressive behavior.

The most frequently used methodological approach in the laboratory to induce mice to fight is to isolate males for some time, ranging from 24 h to 8 weeks and subsequently, to confront the isolate with a group-housed stranger in an unfamiliar test arena or in the isolate's home cage (isolation-induced aggression; [59], [73]). Depending on strain, a considerable proportion of isolated male mice do engage in non-agonistic social interactions or respond defensively toward a stimulus mouse rather than deliver attack bites (isolation-induced timidity; [9], [61]). Functionally, aggressive behavior by isolated laboratory male mice has been interpreted as pathological and abnormal, representing a cardinal symptom of the ‘isolation syndrome’ [119]. By contrast, in the wild, dominant male mice that exclude other males from their territory are behaviorally isolated and their aggression may be an adaptive behavior that increases their relative fitness [5].

Several experimental attempts to induce aggressive behavior in mice apply noxious and painful stimuli. The so-called dangling method consisted of holding a stimulus mouse by its tail with forceps and delivering repeated tactile contact to the experimental mouse with the dangling stimulus mouse, until biting attacks were seen [99]. Further methods scheduled electric shock pulses that were delivered to the tail of two mice in an experimental chamber, to initiate fighting [93] or in order to induce biting of an inanimate object [124]. Brain et al. [7] evaluated the problems with validity of these rarely used methods that proved effective in generating some forms of aggressive behavior in several strains of mice that usually did not display aggressive behavior.

Aggressive behavior in female mice typically is studied during the post-partum period when the lactating female attacks male and female opponents (maternal aggression) [114], [115]. The attacks are more frequent and intense shortly after having given birth and decline during the next 2 weeks. Dominant females effectively protect their litter, in contrast to subordinate females that rarely succeed in raising their young. Anogential investigation or threats rarely precede attacks by females, though threats are typically exhibited once the sequence of aggression has been initiated. Attack bites are directed towards the head and snout of the opponent [42], [115].

Species-typical aggressive behavior in situations of social conflict can serve to provide normative values for evaluating intense, injurious and excessive aggressive behavior [69]. This latter pattern of aggressive behavior attracts attention, because it may be relevant as model system for pathological aggression in species other than mice, including humans [119]. Models of excessive aggressive behavior in mice may comprise such protocols as brief social provocations or instigations prior to a confrontation with a stimulus animal [26] or administration of a moderate dose of alcohol [70]. Deletion of a specific gene may provide the conditions under which aggressive behavior emerges in otherwise placid strains of mice, and this type of aggression will be discussed in detail (see below).

The repertoire of aggressive behavior and the reactions to aggression—referred to in toto as agonistic behavior—is well delineated and catalogued in murine species [1], [31], [100] and methods for its reliable measurement have become adequate [68]. A valuable comparative account of the behavioral elements of aggression in mice with those in rats, hamsters and guinea pigs was provided initially by Grant and MacKintosh [32], and this resource has been the basis for many subsequent behavioral analysis [2], [3], [51], [66], [74], [88], [89], [92], [106].

The attack bite is the most salient and conspicuous element of aggression in mice; it is directed preferentially at the back or flanks of the opponent, and bites toward the back are accompanied by kicking movements of the rear legs. This rapid behavioral act is usually preceded by a sideways threat, which is a lateral rotation of the body, accompanied by piloerection and short steps, directed toward an opponent. Although sideways threats and attack bites follow each other often during an aggressive encounter, sideways threats may also be displayed without subsequent bites. The tail rattle, a rapid vibration of the tail, typically occurs shortly before and during an attack flurry. Tail rattles may reflect a state of high arousal, and are preferentially seen in dominant mice. The pattern of aggression in male mice consists of a series of bursts or flurries that are separated by periods of relative inactivity. During these gaps, mice may engage in non-aggressive behavioral acts such as walking, rearing, grooming, and anogential contact. Social interactions between two individuals are readily quantified by anogenital contact behavior [57]. The most prominent reaction to aggressive acts is an escape leap that becomes full flight. Fleeing opponents are pursued in a rapid chase. When cornered, mice rear up on their hindlegs and assume a defensive upright posture, rotating the upper torso toward the aggressive opponent. Eventually, frequently attacked mice display a defeat posture with limp forelimbs, upward angled head and retracted ears [31], [63], [69], [74], [75].

Measurement of aggressive behavior begins with simply tallying: (1) the proportion of animals fighting, and proceeds to tracking; (2) the latency for the first attack bite; and (3) the duration of attack bouts or flurries. These first three measures are sufficiently basic so that minimal training and operating simple timers and counters can accomplish a reasonable level of reliability. It is convenient to arrive at a single quantitative index of aggression, and the development of scales that allow ratings fulfills this goal [107], [121]. However, lumping several aggressive acts and postures together assumes common mechanisms mediating distinctive elements in an elaborate behavioral repertoire.

In order to differentiate species-typical aggressive behavior from unusual behavior patterns, including those that are injurious and extreme in intensity and rate, a more detailed analysis of videorecorded encounters is required. Aggressive behavior in mice is very fast, and red light-sensitive videocameras can produce records that are suitable for analysis. This analysis is aided by various software programs, some of which are commercially available. In order to achieve acceptable levels of inter- and intra-observer reliability, it is advisable to become very proficient before initiating the experimental work. The first level of measurements requires the encoding of each occurrence of operationally defined acts and postures. This approach provides the total frequency and duration of these behavioral elements during an encounter of fixed duration, usually beginning with the first attack. A second, more detailed level of measurements generates information on the pattern of aggressive behavior such as to reveal the sequential structure of the behavior and its burst-like character [72], [122].

The analysis of aggressive behavior in laboratory mice is often constrained by rules, regulations and constraints that are imposed by the research facility, particularly when focusing on mutant mice. Even in a species that is flexible and adaps readily, it is beneficial and instructive to study aggressive behavior in mice under adequate breeding and housing conditions, in confrontations with standard opponents, using appropriate recording and analysis equipment.

Behavior genetics studies of aggressive behavior in mice have successfully demonstrated that the propensity for these types of behavior in inherited. Earlier studies relied on comparison of different strains in terms of aggressive behavior [102], [110], but more persuasive evidence emerged from breeding studies in which the selection criterion was a specific level of fighting as a behavioral phenotype.

Three selective breeding programs for offensive aggression in male mice began to be established in the 1960s and 70s. These founded the Turku aggressive (TA) and non-aggressive (TNS) strains in Finland [52], [96], the NC900 and NC100 strains in North Carolina [30], and the short attack latency (SAL) and long attack latency (LAL) strains in The Netherlands [111].

The base population for the TA and TNA strains was an outbred colony of Swiss albino mice. In agonistic encounters with a standard male opponent at 60 days of age in a neutral cage, individuals were rated on a seven-point scale, and an aggression score was based on this rating. For the TA line, high scoring males were mated to the sisters of high scoring males. Conversely, for the NTA line, low scoring males were mated to the sisters of low scoring males. For both lines, brother by sister mating was avoided. A randomly bred control line was also maintained. There were significant differences between the lines at the second generations of selection. A slightly larger difference occurred in most of the succeeding generations. Cross-fostering studies indicate that this strain difference is not an effect of the postnatal maternal environment. Realized heritability has been calculated at 0.34 for the high and 0.22 for the low line. Heritability calculated from a backcross derived from the two lines amounted to 0.54. Strain differences in other behaviors, including activity, and in physiology have also been reported. However, there may be no differences in level of testosterone between these strains.

The base population for the NC900 and NC100 consisted of outbred ICR mice. In agonistic encounters with a standard male opponent, individuals were scored for 33 variables including attack latency and number of attacks. The tests occurred in a neutral cage. An aggression score was based on both attack latency and number of attacks. For the NC900 line, high scoring males were usually mated to the sisters of high scoring males. Conversely, for the NC100 line, low scoring males were usually mated to the sisters of low scoring males. For both lines, brother by sister mating was avoided. A randomly bred control line was also maintained. There was an asymmetry in the effects of selection. Over generations, the scores of the high lines did not change, whereas for the low line, the scores decreased to essentially zero by the fourth generation. This resulted in a line difference which was maintained for the next 26 generations of selection. Cross-fostering studies indicate that this strain difference is not an effect of the postnatal maternal environment. Heritabilities have not been calculated for these lines. The strain difference in attack behavior appears to be a function of activity, reactivity, and immobility. The low line has a strong tendency to be immobile in general and to freeze on being approached by a conspecific.

There does not appear to be any strain difference in reactivity to physical stimuli. Many environmental and ontogenetic effects on attack behavior of these lines also appear to influence reactivity to social stimuli. It has been suggested that changes in receptivity to social stimuli could be involved in the rapid microevolution of offense. Alternatively, hyper- or hyporeactivity may impact on the ontogeny of offense directly or via genetic or ontogenetic effects. GABA or dopamine neurotransmitter systems have been postulated to play a significant role in the reactivity to social stimuli [30]. However, the strain differences in social reactivity, offense, and neurotransmitters are not due to difference in postnatal testosterone levels.

The base population for the SAL and LAL lines was wild mice trapped in a mansion near the town of Groningen (The Netherlands). In agonistic encounters with a standard male opponent, individuals were categorized according to their attack latency. Under the given resident-intruder testing conditions, attack latency was inversely correlated with how oftern each motor pattern of offense occurred. For the SAL line, males with low attack latencies were usually mated to the sisters of males with low attack latencies. Conversely, for the LAL line, males that did not attack were usually mated to the sisters of similar males. For both lines, these were brother by sister inbreedings in the first four generations.

A randomly bred control line was also maintained. There was an asymmetry in the effects of selection. Selection for short attack latency succeeded on the first attempt, but selection for long attack latency did not succeed until the fourth attempt. The strain difference as shown by ova transfer and cross fostering experiments is not due to maternal factors. For the high line, the realized heritability at the 11th generation of selection was 0.30.

It appears that the SAL and LAL lines represent two behavioral morphs which are also found in wild populations of mice. It has been suggested that the SAL and LAL morphs are adaptive in different phases of population growth in mice. The LAL would be adaptive when mouse demes are established and stable and the SAL morph would be adaptive when males are competing to establish new demes. Also, it has been suggested that the aggressive behavior of SAL and LAL mice reflect differences in coping style. The SAL mice actively manipulate the environment in response to changes in external stimuli and the LAL mice passively adjust to changes in external stimuli. These strains differ in perinatal and adult testosterone levels and in tissue sensitivity to testosterone.

The large literature on offensive aggression in inbred strains has been reviewed by [43], [62], [63], [65], [67], [105]. Very few of these strain differences appear to be due to effects of the postnatal maternal environment. In six of eight studies, there was no effect of cross-fostering on the strain difference in male offense [63]. In one of the eight, cross-fostering affects one but not both strains.

For several decades the strain-characteristic levels of aggressive behavior were correlated with a profile of neurochemical activity [46], [47], [120], [125]. These post-mortem studies began with whole brain assays of amines and their acidic metabolites, and advanced to the analysis of micropunched brain structures [116]. It is now possible to track several canonical neurotransmitters in in-vivo microdialysis samples in mesocorticolimbic areas of C57BL/6 and DBA/2 mice under basal and challenge conditions [44]. These methodological successes will enable a more rigorous examination of the genetic control over aminergic and peptidergic activities in hypothalamic, limbic and cortical areas that have been hypothesized to be critical in the initiation, execution and termination of aggressive encounters.

Recent research with inbred strains has also focused on correlations between neural morphology and offensive aggression. For seven inbred strains, negative genetic correlations were found for a measure of offense and the size of the IIP (intra- and infra-pyramidal) mossy fibers in the hippocampus [36], and glutamic acid decarboxylase activity (an enzyme that converts glutamic acid to GABA) in the olfactory bulb [34]. In each study, the measure of offense was proportion of males attacking a passive A/J male in the resident's cage. For 10 inbred strains, positive genetic correlations have also been obtained between the above measure of offense and a measure of fear [35]. Fear was inferred from a test that allowed avoidance of light; the measures were percentage of time spent in the light area, number of transition between dark and light areas, and number of fecal boli. Similarly, a single gene mutation in C57BL/6J with effects on size of the IIP mossy fibres of the hippocampus has pleitropic effects on percentage of attacking males in a neutral cage and resident-intruder test, on number of fecal boli in the light/dark choice test, and on size of thermal nest [108].

The research on inbred and selected strains provided much information on the behavioral and biological correlates of attack behaviors in males. The studies on selected and inbred strains described next for attack behaviors in females are a similar beginning. But they are just a beginning for the next phase of the genetic analyses of aggressive behaviors. This is to identify and map the individual genes.

Offensive aggression in female mice was first shown when female mice were paired in competitive tests [28]. The mice were food deprived and the competition was for a piece of mouse chow. In this test both males and females of the BALB/c and C57BL10 strains differ in measures of offense.

However, it initially appeared that whereas male mice displayed offense in non-competitive tests, female mice did not. This view dramatically changed when wild female mice were tested for offense against a standard male opponent in a non-competitive test. Individuals were rated on a seven-point scale, and an aggression score was based on this rating. Variation in offense was observed, and lines of female mice differing in offense were selectively bred (within family) from these wild mice, constituting replicate lines [23]. For the high lines, high scoring females were bred to the brothers of high scoring females of another litter, and for the low line, low scoring females were bred to the brothers of low scoring females from another litter. Brother by sister mating was avoided. Randomly bred control lines were also maintained. The realized heritability at the eighth generation was 0.12 for H1, 0.14 for H2, 0.34 for L1, and 0.46 for L2. More recently, similar variation for females of inbred strains has been shown [83]. Females of the AKR but not the other strains displayed offense against standard opponent ICR males.

Some genetic effects on female offense may depend on the postnatal social environment [38]. If mice are isolated after weaning, females of the NC900 and NC100 strains do not fight, whereas males of the NC900 but not NC100 strains do fight. However, if females are group housed, the females of the NC900 but not NC100 line fight. The strain difference in fighting is similar between non-isolated females and isolated males. Grouped, but not isolated, housing of the females may enhance their social reactivity and thereby fighting.

The findings with NC900 and NC100 strains suggest that there may be a genetic correlation between male and female offense, at least when males are isolated and females are grouped before testing. However, selection for female offense [23] and for male offense [123] did not have a correlated effect on offense of the other sex.

There are chromosomally distinct populations of Mus domesticus. In Italy, there are house mice with 20, 13, 12, and 11 pairs of chromosomes [15]. These latter three arise from the first by centromeric fusion of two chromosomes. The populations with 12 and 13 pairs are sympatric, and the population with 13 pairs is displacing that with 12 pairs. This may be due to differences in offense. When resident and intruder had the same karyotype, the mice with the 13 pair karyotype had a higher proportion of fighting pairs and a lower latency to attack than those with the 12 pair karyotype. When the resident and intruder had different karyotypes, the mice with the 13 pair karyotype had a shorter attack latency than those with the 12 pair karyotype. Also, in both tests, mice with the 13 pair karyotype were more likely to kill intruders than those with the 12 pair karyotype. The differences between the 12 and the 13 pair karyotypes were observed not only in laboratory cages but also in a large enclosure with a family unit of the adult male, and adult female, and her mature litter [15].

In another part of Italy, free hybridization occurs between a population with 11 pairs of chromosomes and a surrounding population with 20 pairs of chromosomes. The reproductive fitness of the hybrids is inferior to that of the parental populations. When resident and intruder had the same karyotype, mice with the 11 pair karyotype had a lower attack latency than mice with the 20 pair karyotype, but when the resident and intruder had different karyotypes, the attack latency was lower for the mice with the 20 pairs of chromosomes. This might limit the spread of the population with 11 pairs of chromosomes into that with 20 pairs.

There is evidence that one or more genes in the t region of chromosome 17 have effects on offense in males. There are many genes in this region. The array of genes is a haplotype; + and t are two of these haplotypes. When male mice of the +/+ and +/t haplotypes were paired in laboratory cages, in a large arena, and in an outdoor enclosure, more +/t males initiated fights than did +/+ males [54], [55]. Also, in the laboratory setting but not in the semi-natural situation, more of the +/t males won the fight. There is also an effect of genetic background. The difference between +/+ and +/t mice was seen on the background of wild but not laboratory mice. It has been speculated that since t/t mice are homozygous lethals, the difference in offense of +/+ and +/t mice may confer a reproductive advantage on the heterozygote thereby maintaining + and t haplotypes in the population.

The eutherian mammalian Y chromosome has two parts; one of these is male specific. The other recombines with the X chromosome; this is referred to as the pseudoautosmal region (PAR). This section will focus on the male specific part.

Several groups of investigators have suggested that there is an effect of the non-pseudoautosomal region or male specific part of the Y chromosome on offense [64]. The following Y chromosomes of the following strain pairs have differential effects on offense: DBA/1 vs. C57BL10; DBA1 vs. DBA2; CBA/FaCam vs. C57BL6; CBA/H vs. NZB; SAL vs. LAL; and PHH vs. PHL. This was shown in studies with reciprocal F1 hybrids, reciprocal backrosses, and reciprocal congenic strains. Maxson [64] has suggested that Sry (sex determining region Y) is a candidate for a Y gene with effects on offense. Regardless, the Y gene(s) with effects on offense do(es) not appear to have pleitropic effects on adult levels of serum testosterone.

The effect of the male specific part of the Y chromosome on offense often depends on genetic background. For example, when the genetic background is that of the DBA1 strain but not that of the C57BL10, the DBA1 and C57BL10 Ys differ in effect on offense. Similarly, the differential effect of SAL and LAL Y (NPAR) on offense is a function of the pesudoautosomal region. When one Y has both PAR and NPAR regions from the SAL and the other has PAR and NPAR region from the LAL, there is a differential effect of the Y on offense. A mouse with the NPAR and PAR from SAL has a shorter attack latency than those with PAR and NPAR from both strains or from only LAL.

The maternal environment can also modify the effect of Y chromosome pairs on offense [11]. A higher proportion of males from a cross of CBA/H×NZB F1 to CBA/H females attack standard opponent A/J males than those from a cross of NZB×CBA/H F1 to CBA/H females. This effect is reversed when the CBA/H ovaries for this backcross are transplanted into an F1 female. This appears to be an effect of both the prenatal and postnatal uterine environment on either the direction of the effect of the same Y gene on offense or on the effect of different sets of Y genes on offense.

The genotype of the opponent can also influence the effect of the Y chromosome (NPAR on offense). The NZB and CBA/H Ys have differential effects on offense in a homogeneous set test (tested pair has same genotype) but not in a standard opponent test (tested pair has different genotypes). Similar effect has been reported for the DBA1 and C57BL10 Ys. For this pair, the direction of the effect is different in a homogeneous set test and in a standard opponent test.

The research on the Y chromosome has clearly shown that genetic effects on male aggressive behavior are dependent on environmental influences. This interdependence of genetic and environmental effects should be considered in studies to identify and map individual genes.

Selective targeting and mutation of specific genes has identified several genes that appear to influence mouse aggression. The products of these genes range from proteins that are important during development (i.e. NCAM) to neurotransmitter receptor molecules (i.e. 5-HT1B receptor). As is shown in Table 1, altering genes can render both increases and decreases in aggression, indicating that simply ‘knocking out’ a gene does not create a uniform change in aggression. There is no obvious pattern as to whether a particular type of gene product (i.e. receptor, peptide, or enzyme) determines if a ‘knockout’ (KO) mouse is more or less aggressive, further suggesting a specific relationship between the gene knockout and the behavioral change. For example, knocking out the 5-HT1B receptor was originally reported to increase aggression, whereas knockouts of other neurotransmitter receptor molecules, the NAα2C and H1 receptors, decrease aggression.

Several mouse mutants may exhibit large changes in aggressive behavior on account of ultimate action at a common final path. The proposal of the brain serotonin system as a candidate for such a common mechanism has received support. The evidence derives primarily from clinical studies which characterized highly aggressive individuals to exhibit a serotonin-deficient trait, indicated either by the low level of the acid metabolite of serotonin in the CSF or by the blunted plasma prolactin response to a serotonin agonist challenge [8], [14], [56], [60]. A second line of support are pharmacological studies in aggressive animals that indicate inhibition of aggressive behavior by administration of the dietary serotonin precursor, l-tryptophan, the immediate precursor l-5-HTP, releasing agents such as fenfluramine, receptor agonists at various 5-HT receptor subtypes, sertonin-specific reuptake inhibitors, or metabolic enzyme inhibitors [66], [70], [71], [76], [77]. It is noteworthy that antagonists, particularly at 5-HT2C receptors, are also effective in reducing aggressive behavior, both in animals and in human patients. However, most of these pharmacological manipulations achieve their anti-aggressive effects on account of their sedative effects.

A most direct link between aggression and brain serotonin has been developed for the 5-HT1B receptor, initially via pharmacological agents with partial affinity for this receptor and significant actions at other receptors [86]. Recently, more selectively acting agonists at this receptor subtype such as CP-94,253, anpirtoline and zolmitriptan efficaciously reduced aggressive behavior particularly when this behavior occurred at very high levels [19], [27]. These observations need to be integrated with those in null mutant mice in whom the gene for the 5-HT1B receptor was deleted ([97], Table 1). These 5-HT1B KO mice exhibit aggressive behavior more rapidly and frequently than their wild-type counterparts. One major population of these receptors is located pre-synaptically and when stimulated, inhibits 5-HT transmission, as indicated by lower 5-HT release in the hippocampus and prefrontal cortex [25], [39], [49]. If in fact, a deficiency in brain 5-HT characterizes highly aggressive individuals, then the null mutation of 5-HT1B receptors should remove pre-synaptic inhibition and lead to increased 5-HT transmission. However, despite the absence of 5-HT1B receptors, there is no accompanying difference in either basal or K+-stimulated extracellular content of 5-HT [117]. A more likely explanation is that the lack of post-synaptic 5-HT1B receptors in mutant mice may prevent inhibitory action of endogenous 5-HT [19], [103].

Alterations in 5-HT neurotransmission have been found in may of the knockout mice that display unusual aggressive behavior. In general, the findings are consistent with human studies suggesting that lower levels of 5-HIAA are correlated with higher levels of aggression. Both the MAO-A and NAα2C knockout mice are more aggressive and have higher levels of whole brain 5-HIAA that their WT couterparts [12], [95]. In the MAO-A knockouts the difference between knockout and wild-type mice is largest when the mice are young. NAα2C knockout mice also have lower levels of whole brain 5-HT. Although the mice lacking NCAM and nNOS do not have altered levels of 5-HT or 5-HIAA, these mice do have some apparent abnormalities in the 5-HT system. The NCAM knockouts are more sensitive to the anxiolytic-like effects of 5-HT1A agonists and express more mRNA for the immediate early gene cFos in the dorsal raphé serotonin-containing cells after confrontation with an intruder [112], [113]. The nNOS knockouts do not show changes in the levels of extracellular 5-HT or 5-HIAA, but fail to show leptin-induces increases in 5-HT or 5-HIAA [10]. In contrast to the more aggressive knockouts, the H1 receptor knockout mice have a higher ratio of 5-HIAA to 5-HT in multiple brain regions suggesting changes in turnover, while the α-CaMKII knockout mice show diminished release of 5-HT [13], [128]. The findings on 5-HT neurotransmission in knockout mice that are either more or less aggressive than their wild-type counterparts suggest that an underlying alteration in components of the 5-HT system may mediate some of the aggressive behavioral phenotypes. While such a proposal is tempting, it is premature to accept a simple final common path without more complete comparisons, including knockout mice whose aggression is unaffected by gene deletion [4]. It is apparent that aggression is influenced by many environmental and experiential variables that affect multiple neurochemical systems [71], [76]. A more complex neural circuitry with many interacting canonical neurotransmitters such as GABA, glutamate, catecholamines, indolamines and acetylcholine as well as peptides such as vasopressin, oxytocin, and opioid peptides, and steroids such as glucocorticoids, androgens, and estrogens, each involving polygenic influences, needs to be considered when delineating the expression of aggression of genetically modified mice.

Observations of altered aggressive behavior in knockout mice have occurred via direct experimental testing and, in some cases, chance observation of excessive wounding or high mortality in group cages. The standard test for measuring aggressive behavior in knockout mice is to pair an isolated ‘resident’ male typically against a group-housed ‘intruder’ in either the home or neutral cages [16]. Some experimenters include additional measurements by housing ‘residents’ with females before confronting ‘intruders’, or to have groups of mice confront each other in a neutral arena (i.e. the homogenous set test). However, the term ‘standard’ tests of aggression has been applied to a range of conditions that differ greatly in terms of the characteristics of the opponent and the behavioral measurements recorded. As shown in Table 1, intruders range from breeding males, that are typically very aggressive, to mice from various outbred and inbred strains that differ in level of aggressive behavior, to olfactory bulbectomized mice that are non-aggressive but very active. Given the impact of the intruder on resident aggression, the use of non-standard intruders seriously limits conclusions about aggressive knockout mice. Since different genes and therefore different biological mechanisms may be involved in attack behavior on different types of intruders, the use of different types of intruders with the different knockouts may make it difficult to impossible to relate the effect of one knockout to another and to use these to dissect common pathways for gene effects on aggressive behaviors. For example, knocking out neuronal nitric oxide synthase and endothelial nitric oxide synthase produces opposite efects on aggression; nNOS knockouts are reported as more aggressive, while eNOS knockouts are reported as less aggressive. However, the isolated nNOS knockouts fought against C57BL/6×129/SvEv intruders, whereas eNOS knockout mice fought against breeding C57BL/6, suggesting differences in the stimulus properties of the intruder as a possible explanation for the reported effects of the gene deletion. Similarly, the effect of variants of the Sts (steriod sulfatase gene) on attack behavior depends on the type of opponent [33], [94]. The gene appears to effect aggressive behavior when there is no risk of the opponent retaliating and appears to have no effect when there is a risk of injury from the opponent.

The procedures evaluating aggression in knockout mice also differ in the number of tests that are performed to ascertain the behavioral effects of the knockout. Some experimenters rely on a sngle determination of aggression while others have performed duplicate and triplicate determinations; some have performed multiple tests in the same day. The importance of experience as an intervening variable can be illustrated by the results of the β estrogen receptor knock-out [84]. During the first confrontation against an olfactory bulbectomized intruder, the βER KO had a shorter attack latency, higher duration of aggressive behavior, and a greater frequency of bouts of aggression in which an attack bite occurred as compared to low levels of aggression in the wild-type mice. After repeated testing however, the aggression of the wild-type mice increased to the level of the knockout mice, prompting an alternative interpretation of the effects of knocking out of the βER gene. The authors concluded no reliable change in aggression, a conclusion that would not have been reached if based solely on a single determination of aggression. This highlights the need to perform repeated tests of aggression while examining the influence of a gene on behavior. In both isolated and breeding, outbred CFW (Swiss-Webster) mice, we have found that levels of aggression are quite variable during the initial confrontations with a CFW intruder and that the variability in aggression declines over repeated experience [127]. Moreover, repeated testing in 1 day results in lower levels of aggression due to habituation [127]. In future work, it is recommended to perform multiple determinations of aggressive behavior until a stable level has been reached.

Behavioral measurements of aggression in knockout mice range from determining the latency and/or frequency of any aggressive event, to aggressive scores based on lumping multiple aggressive or defensive behaviors. In general, across studies, measurement of the attack bite is an unambiguous measure of aggressive behavior. Composite scores of aggression are difficult to interpret because they equally weigh behaviors of different intensities, such as the attack bite and the tail rattle. There is also a heavy reliance on latency measurements although these do not always correlate with other measurements of aggression [109], [111]. In most of the knockout studies, the aggression tests last for an absolute duration regardless of latency. In these instances, mice with longer latencies for a behavior will typically have lower levels of behavior simply because there are fewer minutes remaining in the test. The practice of relying on absolute rather than relative duration of tests heavily emphasizes the latency measurement and may distort the measures of actual fighting once it is initiated.

Section snippets

Acknowledgements

Preparation of this review and the original research from our own laboratories were supported by USPHS research grants AA05122 and DA02632 and a grant from the Alcoholic Beverage Medical Research Foundation (KAM, PI). Inbred Mouse Fund of the University of Connecticut, and grants from The University of Connecticut Research Foundation (SM, PI).

References (128)

  • P.L. Huang et al.

    Targeted disruption of the neuronal nitric oxide synthase gene

    Cell

    (1993)
  • J.L. Hurst

    Behavioral variation in wild house mice Mus domesticus Rutty: a quantitative assessment of female social organization

    Anim. Behav.

    (1987)
  • A. Imperato et al.

    Strain-dependent effects of dopamine agonists on acetylcholine release in the hippocampus: an in vivo study in mice

    Neuroscience

    (1996)
  • E. Kempf et al.

    Serotonin levels and turnover in different brain areas of isolated aggressive or non-aggressive strains of mice

    Prog. Neuro-Psychopharmacol.

    (1984)
  • M. Linnoila et al.

    Low cerebrospinal fluid 5-hydroxyindoleacetic acid concentration differentiates impulsive from nonimpulsive violent behavior

    Life Sci.

    (1983)
  • R.G. Lister

    Ethologically-based animal models of anxiety disorders

    Pharmacol. Ther.

    (1990)
  • J.H. Mackintosh

    Territory formation by laboratory mice

    Anim. Behav.

    (1970)
  • J.J. Mann

    Role of the serotonergic system in the pathogenesis of major depression and suicidal behavior

    Neuropsychopharmacology

    (1999)
  • K.H. McAllister et al.

    Effects of naloxolone upon the behavioral organization of specific defeat activities in intruder mice: assessment by cluster analysis

    Gen. Pharmacol.

    (1986)
  • K.A. Miczek et al.

    Drug effects on agonistic behavior

  • K.A. Miczek et al.

    Neurobiological mechanisms controlling aggression: preclinical developments for pharmacotherapeutic interventions

    Neurosci. Biobehav. Rev.

    (1994)
  • S. Ogawa et al.

    Aggressive behavior in inbred strains of mice during pregnancy

    Behav. Neural. Biol.

    (1984)
  • T.B. Poole et al.

    Differences in aggressive behavior between male mice (Mus musculus L.) in colonies of different sizes

    Anim. Behav.

    (1973)
  • E.M. Banks

    A time and motion study of pre-fighting behavior in mice

    J. Genet. Psychol.

    (1962)
  • R.F. Benus et al.

    Routine formation and flexibility in social and non-social behaviour of aggressive and non-aggressive male mice

    Behaviour

    (1990)
  • V. Bolivar et al.

    List of transgenic and knockout mice: behavioral profiles

    Mamm. Genome.

    (2000)
  • P.F. Brain

    Differentiating types of attack and defense in rodents

  • G.L. Brown et al.

    Aggression, suicide, and serotonin-relationships to CSF amine metabolites

    Am. J. Psychiatry

    (1982)
  • R.B. Cairns et al.

    On fighting in mice: ontogenetic and experiential determinants

    J. Comp. Physiol. Psychol.

    (1971)
  • G. Calapi et al.

    Leptin increases serotonin turnover by inhibition of brain nitric oxide synthesis

    J. Clin. Invest.

    (1999)
  • M. Carlier et al.

    The Y chromosome effect on intermale aggression in mice depends on the maternal environment

    Genetics

    (1991)
  • O. Cases et al.

    Aggressive behavior and altered amounts of brain serotonin and norepinenphrine in mice lacking MAOA

    Science

    (1995)
  • C. Chen et al.

    Abnormal fear response and aggressive behavior in mutant mice deficient for alpha-calcium-calmodulin kinase II

    Science

    (1994)
  • E.F. Coccaro et al.

    Fluoxetine treatment of impulsive agression in DSM-III-R personality disorder patients

    J. Clin. Psychopharmacol.

    (1990)
  • M. Corti et al.

    The role of intermale aggression in speciation processes in chromosomal races of house mice

  • P. Crowcroft
  • P. Crowcroft et al.

    Social organization and territorial behaviour in the wild house mouse

    Proc Zool Soc Lon

    (1963)
  • DeAlmeida RMM, Nikulina E, Faccidomo S, Fish E, Miczek KA. Zolmitriptan, a 5-HT1B/D receptor agonist, alcohol, and...
  • C. De Felipe et al.

    Altered nocieption, analgesia and aggression in mice lacking the receptor for substance P

    Nature

    (1998)
  • G.E. Demas et al.

    Elimination of aggressive behavior in male mice lacking endothelial nitric oxide synthase

    J. Neurosci.

    (1999)
  • A.C. DeVries et al.

    Reduced aggressive behaviour in mice with targeted disruption of the oxytocin gene

    J. Neuroendocrinol.

    (1997)
  • P.D. Ebert

    Selection for aggression in a natural population

  • I. Eibl-Eibesfeldt

    Beiträge zur Biologic der Haus- und der Älirenmaus nebst einigen Beobachtungen an anderen Nagern

    Z. Tierpsychol.

    (1950)
  • G. Engel et al.

    Identity of inhibitory presynaptic 5-hydroxytryptamine (5-ht) autoreceptors in the rat brain cortex with 5-ht binding sites

    N-S. Arch. Ph.

    (1986)
  • E.W. Fish et al.

    Aggression heightened by alcohol or social instigation in mice: reduction by the 5-HT1B receptor agonist CP-94,253

    Psychopharmacology

    (1999)
  • E. Fredericson

    Aggressiveness in female mice

    J. Comp. Physiol. Psychol.

    (1952)
  • S.C. Gammie et al.

    Maternal aggression is reduced in neuronal nitric oxide synthase-deficient mice

    J. Neurosci.

    (1999)
  • J.L. Gariepy et al.

    Genes, neurobiology and aggression

  • B. Ginsburg et al.

    Some effects of conditioning on social dominance and subordination in inbred strains of mice

    Physiol. Zool.

    (1942)
  • Cited by (361)

    View all citing articles on Scopus
    View full text