Background
Autism spectrum disorder (ASD) is a behaviorally defined condition involving a broad range and degree of symptom expression for which the etiology is unknown [
1]. The study of highly penetrant ASD susceptibility genes have identified convergent downstream pathways involved in ASD [
2,
3]. Furthermore, systems biology approaches have combined disease risk, protein-protein interaction, gene ontology, conserved domains, and transcriptomics databases to create functional networks of genes implicated in ASD. The results of these studies group genes in functional networks enriched for processes such as neuronal migration, cytoskeleton dynamics, axon growth, and importantly, trans-synaptic signaling as a major network hub [
4,
5].
Neuroligins are synaptic cell-adhesion molecules that mediate trans-synaptic signaling through their binding partner neurexins and shape neural network properties by specifying synaptic functions [
6]. Importantly, mutations in genes encoding neuroligins [
7] and neurexins [
8] have been identified in ASD patients. To study the mechanisms by which these mutations and subsequent disruptions to cellular processes contribute to ASD, animal models have been generated that recapitulate important aspects of the human disorder. These animal models have become a major
in vivo tool to investigate the neurobiological mechanisms leading to the expression of behavioral traits in ASD.
Mice expressing the arginine to cysteine residue 451 substitution (R451C) mutation in the synaptic adhesion gene neuroligin-3 (NL3) have been proposed as a model of ASD. The ASD-causing point mutation in neuroligin-3 causing R451C substitution [
7] impairs intracellular trafficking and reduces neuroligin-3 protein levels at the cell membrane by 90 % [
9,
10]. NL3
R451C mice show a diverse range of behavioral abnormalities and synaptic dysfunction as shown by increased cortical inhibition together with enhanced hippocampal excitation in brain slices [
9,
11‐
14]. However, conflicting reports have arisen from studies of social behavior in NL3
R451C mice [
9,
15].
While the core diagnostic symptoms include social communication deficits alongside the presence of repetitive behaviors, patients also exhibit a variety of comorbid traits such as aggression, which adversely impact on patient quality of life [
16]. Aggression is observed in up to 70 % of individuals with ASD [
17] and significantly limits patient access to education, health care, and employment [
18]. The atypical antipsychotic risperidone is frequently prescribed for treatment of aggressive behaviors in children with ASD [
19]; however, side effects significantly limit its use [
20‐
22]. The serious negative outcomes associated with behavioral impairments in ASD highlight the inherent importance of understanding the underlying neurobiological mechanisms with the aim of developing more effective treatments.
Aggression in mice is a robust, innate, social behavior and serves to assist the acquisition of social ranking and resources from the environment. Dominance hierarchies are established and maintained through confrontations between rival male mice. Territorial aggression between males can be measured utilising the resident-intruder test whereby a juvenile intruder mouse is introduced to the home cage of a test mouse [
23]. This paradigm mimics many elements of the behavior displayed by resident males to exclude other breeding males from their home territory and their mates [
23]. Escalated, pathological, and abnormal forms of aggression are characterised by prolonged and frequent attacks and brief latencies to attack following the introduction of an intruder mouse [
24,
25].
In order to determine whether the NL3 R451C mutation impacts on social and repetitive behavior in mice, we assessed aggressive interactions in this model. Juvenile social interaction and adult sociability were also investigated in an attempt to resolve previously conflicting reports in the literature [
9,
11,
13,
15]. Repetitive and restrictive behavior was also investigated using an adaptation of a clinical task applied to children with ASD.
Methods
Animals
B6;129-Nlgn3tm1Sud/J mice were obtained from Jackson Laboratories (Bar Harbor, Maine USA) and maintained to generation F9 on a Sv129/C57BL6 background. NL3
R451C and wild-type (WT) animals were derived by mating heterozygous females with NL3
R451C males, which produced 50:50 WT and NL3
R451C male offspring (Y/+ and Y/R451C) that were genotyped as previously described [
9]. Experimental animals were weaned at 4 weeks of age and housed in groups of four per cage with food and water available
ad libitum. The holding room was maintained on a 12:12-h light/dark cycle with lights on at 7.00 a.m. and at an ambient temperature of 20 ± 1 °C. All procedures were approved by the Florey Institute of Neuroscience and Mental Health Animal Ethics Committee. All behavioral testing was conducted in the light cycle, the animals were allowed at least 1 h to habituate to testing rooms, and the experimenter was blinded to genotype and exact sample sizes for each test are reported in figure legends. Four separate cohorts of mice were exposed to behavioral testing: cohort 1: repetitive object novel contact test (6 weeks of age) and olfactory discrimination (7 weeks of age); cohort 2: reciprocal juvenile social interaction test (3–4 weeks of age), social approach (9 weeks of age); cohort 3: light/dark arena test (7 weeks of age), elevated plus test (8 weeks of age); and cohort 4: resident-intruder test (10 weeks of age).
Repetitive object novel contact test
This test was designed to measure second-order repetitive behaviors (e.g., ritualistic patterns of behavior and insistence on sameness) and was based on a clinical test [
26] and originally adapted for use on a BTBR mouse model of ASD [
27]. On the day of testing, the WT and NL3
R451C mice were habituated to the experimental room for 30 min and to the testing arena for 20 min immediately prior to testing. The mice were placed in the 30 cm
2 testing arena (lined with 1 cm of sawdust) with four distinct objects. One object occupied each corner of the arena during a 10-min trial. The four objects were as follows: a standard 6-sided dice (1.5 cm in length), a 20-sided dice (1.5 cm in diameter), a casino chip (2 cm in length), and a lego piece (2 cm in length) and placement was randomised to avoid any effect of circling or thigmotaxis behavior. Recorded video files were scored for object investigation by an observer blinded to genotype. Investigation was defined as clear facial or vibrissae contact and/or sniffing of the object. To assess for sequential patterns of investigation of the four objects, each object was allocated a number from one to four. The pattern of visitation of each individual mouse was then scored, and a record of investigation for each mouse was generated in a Microsoft Word document. For example, if the objects (designated 1–4) were visited in the designated order, the resulting sequence would be 1-2-3-4 in the word document. Subsequent visits were added contiguously to the visitation data string in the word document. The investigation data were subsequently searched using the ‘Find and Replace’ function in Microsoft Word for all possible four object visitation sequences. According to our criteria, a visitation sequence could include multiple visits to the same object, provided they were interrupted by a visit to another object. To determine if mice showed recurring visitation patterns, the total number of repetitions of identical four object visitation sequences was calculated for each mouse and then averaged by genotype.
Grooming
The mice were placed in a clear Plexiglas cage (32 x 18 x 10 cm) lined with 1 cm of sawdust and allowed to habituate for 30 mins. They were then videoed for 10 mins and scored by a rater, blinded to genotype.
Reciprocal juvenile social interaction
The mice aged 3–4 weeks were placed in a 30 cm2 cage with an unfamiliar mouse for 30 min. Equal representations of all possible pairs was achieved (WT/WT, WT/NL3, NL3/NL3). Testing sessions were recorded by a video camera mounted above the testing arena and multiple body point tracking software Cleversys LTD (Virginia, USA) was used to automatically measure time in social interaction.
Social approach (three-chamber) test
The social approach test was performed as described [
28]. In brief, after 10 min of habituation, a mouse was placed in the central chamber of a clear Plexiglas box (60 x 40 x 25 cm) divided into three interconnected chambers and was given the choice to interact with either an empty wire cup (located in one side of the chamber) or a similar wire cup with an unfamiliar 7–8-week-old C57BL/6J male mouse inside (located in the opposite chamber). Time in each chamber and interacting with each cup was measured by an observer blinded to genotype. The chambers were cleaned, and fresh bedding was added between trials. The position of the stranger mouse was randomised between the left and right side chambers to control for any chamber preference. The animals serving as strangers were habituated to placement under the wire cage for 5 min prior to the commencement of the test. Testing sessions were recorded by a video camera mounted above the testing arena, Cleversys LTD (Virginia, US) was used to automatically measure time in chambers, and time in direct social interaction was manually scored by a rater blinded to genotype.
Olfactory discrimination test
This test was conducted in order to determine the ability of the mice to discriminate between and habituate to different odor types, e.g., social vs. non-social odors. The experimental room was cleared of all strong odors, and the mice were habituated to the room for a minimum of 1 h prior to testing. Concurrently, they were habituated to a clean cotton tip applicator placed in a weigh boat and inserted through the wire bars of the home cage. The testing arena consisted of a clear Plexiglas cage (32 × 18 × 10 cm) lined with 1 cm of sawdust. The odors were presented in sequential order: water, vanilla, almond, female urine, and male urine in accordance with a protocol previously described [
29]. Vanilla and almond cooking essence were used, and urine was obtained from adult C57Bl/6J male and C57Bl/6J female mice in estrous. Each odor was presented for a period of 2 min, repeated over three trials with an inter-trial interval of 1 min. All odors were diluted in distilled water at a 1:4 dilutions factor and the investigator pipetted 10 μL of each odor onto a fresh cotton tip applicator, at the beginning of every trial. Recorded video files were scored for the amount of time the mouse spent sniffing the cotton tip applicator during each odor presentation by an observer blinded to genotype.
Light-dark arena
Photo beam arenas (E63-10, TruScan, Coulbourn Instruments, Allentown, PA, USA) with a light-dark box insert placed over half of the arena were used to monitor the exploratory activity of the mice. The animals were placed in the light area facing into the dark area and allowed to enter. The light (750 lx) was then switched on, and the 10 min trial started.
Elevated plus maze
The elevated plus maze consisted of two open (25 cm x 8 cm x 0.5 cm) and two closed (25 cm x 8 cm x 20 cm) arms emanating from a common central platform (8 cm x 8 cm) to form a plus shape and was elevated to 80 cm above floor level. The mice were habituated to the testing room for 1 h and then placed onto the central platform facing an open arm for a 6 min trial. The maze was thoroughly cleaned using 70 % ethanol between subjects, and scoring was performed using the Noldus Ethovision automated tracking system (version 3).
Resident-intruder test
The male resident mice were isolated for 1 week, during which their home cages were not changed. Aggressive behaviors in 3-month-old mice were monitored during four 5 min test exposures to 8-week-old C57BL/6 male intruder mice conducted over 4 days as aggression is a highly variable behavior in mice [
30,
31]. Mice were injected intraperitoneally with a non-sedative [
32‐
37] dose of risperidone (0.05 mg/kg; Sigma) or saline 15 min prior to testing (Additional file
1: Figure S1). Testing was conducted by an experimenter blinded to genotype, and drug treatment and treatment group assignment was randomised ensuring sufficient spread of animals from different litters, housing boxes, and genotypes in each group. The trials were aborted if the experimenter observed tufts of hair being removed from either animal. Two NL3
R451C animals were excluded from testing on the first day due to extreme aggression towards the intruder. Latency to first attack, attack incidence, attack duration, tail rattles, and non-aggressive social interactions (sniffing, climbing, and grooming) were scored by a rater, blinded to genotype from video recordings of each test session.
Statistical analyses
Where data distribution were normal and variance comparable (juvenile social interaction, adult social approach test, olfactory discrimination test, light-dark arena, elevated plus maze, repetitive object novel contact test, grooming), unpaired Student’s
t test or ANOVAs with repeated measures where appropriate followed by Bonferroni post hoc tests were used as indicated. Aggression data was not normally distributed, and regression models were applied to estimate the effect size (Additional file
2: Figure S3). Two independent analyses were conducted in accordance with our hypotheses; (a) estimating the effect of the NL3R451C mutation on aggression and (b) estimating the effect of risperidone treatment on aggression in NL3
R451C mice. Due to absence of prior estimates of effect sizes, no power analysis was conducted. In both analyses, the animals were repeatedly tested over 4 days, thus these observations are correlated within a given animal. Hierarchical regression models were used to estimate the effect of the NL3R451C gene mutation on aggression and the effect of risperidone treatment on NL3
R451C mice. Both Bonferroni adjusted for multiple comparisons and unadjusted two-sided
p values were reported together with appropriate effect size estimates and 95 % confidence intervals (95 % CI) to indicate the precision. The latency data describes the time to an attack and may be censored (e.g., when an animal does not attack during the 300-s observation period). Survival analysis, in particular the Cox regression model, is appropriate for analysis of these data [
38]. A hierarchical Cox regression model that accounts for the effect of multiple test days was used to estimate the effect size, measured as the hazard ratio of the first attack occurring at any time over the 300-s observation period. Attack incidence is a count variable, and a hierarchical Poisson regression model was used to estimate the effect size, measured as the ratio of expected number of attacks. Although the duration of social interaction data violated the normality assumption, a hierarchical random effects generalised least squares regression model did not violate the assumptions of model fit. Both Bonferroni adjusted for multiple comparisons and unadjusted two-sided
p values were reported together with appropriate effect size estimates and (95 % CI) to indicate the precision. Results for aggression data are graphically presented as a box and whiskers plot showing the 25th to the 75th percentile and the minimum to maximum of the data range and the median was shown by a line. Statistical analyses were performed with STATA and SPSS software.
Discussion
The development of mouse models of ASD is crucial to study the disorder at molecular and cellular levels, to gain insight into disease mechanisms, and test potential pharmacological interventions. Here, we have shown that risperidone efficiently reduces aggression in the NL3R451C mouse, while having no effect on the WT behavior or on locomotor activity levels. This is the first demonstration of risperidone rescuing aggression in mice with an ASD-associated genetic mutation and demonstrates predictive validity in this model. The NL3R451C mice exhibit normal anxiety levels and olfactory discrimination indicating that abnormal aggressive behavior is not due to altered anxiety or olfaction. Furthermore, while we uncover subtle differences in social behavior in the juvenile NL3R451C mice, as adults, mutants show similar sociability to the WT littermates. Using a novel assay of repetitive and restrictive behavior, we show that the NL3R451C mice exhibit stereotypical object exploration, further confirming the face validity of the NL3R451C model in the context of ASD.
Previous characterisations of NL3
R451C mouse behavior have focused on potential deficits in social interactions [
9,
13‐
15], but less attention has been given to the analysis of repetitive and restrictive movements and routines. In order to assess this domain in the NL3
R451C mice, we employed a behavioral assay based on findings from a clinical trial demonstrating that children with ASD tended to play with toys in a particular order [
26]. Upon exposure to four objects, we observed that the NL3
R451C mice repeated particular patterns of visitation more frequently than the WT littermates. Compared to the initial study by Pearson et al., utilising this task, we report a higher frequency of four sequence patterns, despite time and total visitations being similar. This discrepancy is most likely due to the fact that both studies defined a visitation sequence as one that could include multiple visits to the same object, provided they were interrupted by a visit to another object. It is possible therefore that the mice in the study by Pearson et al. made more consecutive visits to the same object, which would lower the frequency of four-object sequences recorded. The emphasis of this task is primarily on the repetition of these sequences of investigation rather than the specific order of objects. If mice were retested, it is likely that each NL3
R451C mouse would generate a different pattern of investigation, but that the object exploration would be repeated to a higher degree than WT mice. The exact order of object exploration would depend upon the time between the initial and subsequent tests and the ability of mice to recall their previous visitation pattern. Given the novelty of this test, this point has not been addressed in the behavioral literature and is one that we may explore in the future.
It has also been recently shown that NL3
R451C mutants show enhanced acquisition of repetitive motor routines on the rotarod [
43]. This repetitive phenotype was recapitulated by the conditional deletion of NL3 in adult mice and was shown to be exclusively mediated by a reduction of synaptic inhibition onto ventral striatum dopamine D1 medium spiny neurons. While the neural mechanisms underlying object exploration are likely to differ from those governing motor habit formation, it is likely that the NL3
R451C mutation could disrupt striatal circuits to shape a broad range of repetitive and stereotypic behaviors associated with ASD. This has been demonstrated by volumetric MRI studies that have shown smaller striatal volume in the NL3
R451C mice [
44,
45]. Furthermore, the disruption of striatal circuitry demonstrated in the NL3
R451C mouse model is in keeping with striatal structural and functional alterations in ASD patients [
46‐
48].
NL3
R451C mice have been assessed for sociability in a number of studies using the three-chamber social interaction arena assay; a task that tests for preference for an empty cage vs. caged novel mouse simultaneously. We show that the NL3
R451C mice show a strong preference for the novel mouse over the empty cage, indicating no difference in sociability compared to the WT animals. While the first investigation of adult social behavior in NL3
R451C mice on the C57Bl6-SV129 hybrid background showed decreased time in the chamber containing the novel mouse compared to WT mice, mutant mice still displayed a preference for the novel mouse (more time with novel mouse than with novel object) [
9]. There was no significant genotype difference when reciprocal social interactions were measured. Other behavioral assays conducted in that study [
9] do not represent standard measures of social choice. Using the three-chamber arena, the same group later characterised social interaction in the NL3
R451C mice on a pure C57Bl6 background and reported a strong preference for the novel mouse over the novel object in both the WT and mutant mice, indicating no decrease in sociability [
11]. When the object was replaced with another mouse, however, the NL3
R451C mice were shown to have a deficit in social recognition. Our results are in agreement with Crawley et al. assessing sociability in NL3
R451C mice on the C57Bl6 background [
15]. This study also conducted the three-chamber social interaction assay on a cohort of mice naïve of testing, removing any potential confound of prior experience. Two research groups have shown deficits in social interaction in this mutant on the same C57Bl6-SV129 hybrid background and on a pure 129S2/SvPasCrl background [
13,
45]. Discrepancies in social behavior assessed in the three-chamber arena could be attributable to differences in genetic background strain, differences in methodology (e.g., lighting conditions and previous behavioral testing experience) or even as a result of testing in different laboratories. We assessed juvenile social interaction in the WT and NL3
R451C mice at 4 weeks of age and found no difference in total time spent interacting with the novel mouse between genotypes, in line with another study investigating juvenile interaction in NL3
R451C mice [
13]. Although the increased frequency of interaction seen in the mutants in the current study contrasts with one report that juvenile NL3
R451C mice show no difference in this parameter [
15], our present finding is likely a result of hyperactivity shown by the NL3
R451C mice during the 30-min testing period. No aggression was noted during juvenile social interaction testing, suggesting that the heightened aggression expressed in the adult NL3
R451C mouse could be territorial in nature.
We have demonstrated that heightened aggression is a robust phenotype in the NL3
R451C mouse model of ASD. While we assessed aggression specifically towards a novel juvenile mouse following a period of isolation, a lower incidence of aggression was noted in the home cages of the animals. This observation has implications for interpretation of adult behavior in studies utilising this mutant, as repeated exposure to aggression can lead to stress, alter baseline behavior in the WT mice and impact on health [
49]. Furthermore, this is the first demonstration that risperidone can ameliorate aggression in a mouse model of ASD. How risperidone decreases aggression in NL3
R451C mice is not known. Risperidone has a broad antagonist receptor profile with greatest affinity for serotonin (5-hydroxy-tryptamine (5-HT)) 2A receptors, followed by 5-HT1B, 5-HT7, and dopamine D2 receptors [
50]. Both antagonists acting at 5-HT2A [
51,
52] and D2 receptors [
53,
54] reduce aggressive behaviors in mice, suggesting that these specific receptor candidates may mediate the effects of risperidone in NL3
R451C mice; however, this requires further examination.
Competing interests
The authors declare that they have no conflicts of interest.
Authors’ contributions
ELB, ELH, and AJH conceived of the study and participated in its design and coordination. ELB carried out and supervised all behavioral tests, collated the data, and performed the statistical analysis. LL carried out the juvenile social interaction test, olfactory, grooming, repetitive and restrictive behavior studies and the data analysis and contributed to the drafting of the manuscript. LK carried out the anxiety tests and the data analysis and revised the manuscript. LK and ELH carried out the initial aggression study that identified the phenotype. LC performed the statistical analysis for aggression data and revised the manuscript. ELB, ELH, AJH and JCB drafted and edited the manuscript. All authors read and approved the final manuscript.