- Top of page
- Materials and methods
Deficits in social interaction are important early markers for autism and related neurodevelopmental disorders with strong genetic components. Standardized behavioral assays that measure the preference of mice for initiating social interactions with novel conspecifics would be of great value for mutant mouse models of autism. We developed a new procedure to assess sociability and the preference for social novelty in mice. To quantitate sociability, each mouse was scored on measures of exploration in a central habituated area, a side chamber containing an unfamiliar conspecific (stranger 1) in a wire cage, or an empty side chamber. In a secondary test, preference for social novelty was quantitated by presenting the test mouse with a choice between the first, now-familiar, conspecific (stranger 1) in one side chamber, and a second unfamiliar mouse (stranger 2) in the other side chamber. Parameters scored included time spent in each chamber and number of entries into the chambers. Five inbred strains of mice were tested, C57BL/6J, DBA/2J, FVB/NJ, A/J and B6129PF2/J hybrids. Four strains showed significant levels of sociability (spend- ing more time in the chamber containing stranger 1 than in the empty chamber) and a preference for social novelty (spending more time in the chamber containing stranger 2 than in the chamber containing the now-familiar stranger 1). These social preferences were observed in both male and female mice, and in juveniles and adults. The exception was A/J, a strain that demonstrated a preference for the central chamber. Results are discussed in terms of potential applications of the new methods, and the proper controls for the interpretation of social behavior data, including assays for health, relevant sensory abilities and motor functions. This new standardized procedure to quantitate sociability and preference for social novelty in mice provides a method to assess tendencies for social avoidance in mouse models of autism.
Autism is a severe neurodevelopmental disorder defined by social and communication deficits and ritualistic-repetitive behaviors that are typically detectable in early childhood and continue throughout life (American Psychiatric Association 1994; see also Folstein & Rosen-Sheidley 2001; Kanner 1973; Lord et al. 2000a; Paul 2003; Piven et al. 1997; Piven 2001; Schloper & Mesibov 1987). The prevalence of autism is estimated at 1/1000 (with rates of 2.5/1000 when individuals with Asperger's syndrome and pervasive developmental disorder are included) (Fombonne 2003). A large body of evidence from twin and family studies supports the important role of genetic factors in the etiology of autism (Bailey et al. 1995; Ritvo et al. 1989; Szatmari et al. 1998). Linkage and candidate gene studies, aimed at identifying autism susceptibility genes, are currently the focus of research in numerous laboratories (Barrett et al. 1999; Folstein & Rosen-Sheidley 2001; Philippe et al. 2002; Shao et al. 2002). Neuroimaging and post-mortem studies have implicated a number of neural structures and circuits in autism (Bauman & Kemper 1985; Happéet al. 1996; Schultz et al. 2000). A non-human primate model, resulting from neonatal lesions to the medial temporal lobe, has been described (Bachevalier 1991). Given the complexity of this syndrome, multiple, complementary approaches will be necessary to identify the underlying neural and etiologic mechanisms, and gene-brain-behavior relationships, in this disorder. The study of genes, brain and behavior relationships in mice potentially provides an important addition to current approaches to studying the pathogenesis of autism. However, to date, behavioral models of autistic-like responses in mice have received little attention.
Social behavior is a complex construct. The particular set of behaviors that define the syndrome of autism are wide ranging in severity and type, and change in form with chronological development. So, for example, in the original descriptions of autism by Kanner (1943), autistic children were characterized by their dramatic lack of interest in others, whereas current DSM-IV criteria identify substantially more subtle deficits in reciprocal social interaction as sufficient for meeting criteria for this disorder (American Psychiatric Association 1994). Behavioral abnormalities in this domain in autism are varied and include such characteristics as deficits in non-verbal expression (e.g. deficits in eye-to-eye gaze and diminished expression of emotion as measured by lack of gesturing or facial expression), abnormalities in the social use or understanding of language, deficits in social approach (such as the absence of coming for comfort upon injury), or markedly diminished peer relationships.
To quantitate sociability tendencies in mice, we developed a set of tasks that measure (a) time spent with a novel conspecific and (b) preference for a novel vs. a familiar conspecific. Conceptually similar to previous social interaction tests (Bales & Carter 2003; File & Seth 2003; Gheusi et al. 1994; Insel & Young 2001; Tang et al. 2003; Winslow 2003), the present behaviors focus on a more narrowly defined set of parameters. We reasoned that most autistic individuals demonstrate reduced or unusual social approach. Therefore, subject mice were given the choice between exploring an empty, novel chamber vs. exploring an identical novel chamber containing a mouse with which the subject had no previous interaction (stranger 1). Autistic individuals may further avoid unfamiliar social partners and display diminished interest in novelty (American Psychiatric Association 1994). Subject mice were given another choice between exploring the chamber containing the now-familiar stranger 1, vs. exploring an identical chamber containing a new mouse (stranger 2). Time spent in each chamber and number of entries into each chamber were scored by observation. These measures were subsequently automated, as described in a companion paper (Nadler et al. 2004).
Previous methods for testing social behavior in mice employed tethered novel conspecifics and other approaches to maximize the contribution of the subject mouse to the social interactions, and minimize the contributions of the target stranger mouse (Brodkin et al. 2004; Carter et al. 1995; Winslow 2003). The present method contains the target stranger within an inverted round wire cage that permits visual, tactile, auditory and olfactory communication. This modification focuses the task on social approach tendencies, relevant to the social approach deficit in autism, while avoiding potential confounds resulting from aggressive or sexual interactions.
To ensure that differences in social behaviors were not artifacts of other aspects of mouse behaviors, a series of independent control experiments was conducted to evaluate potential confounding factors. Measures of general health, home cage behavior and neurological reflexes were conducted to rule out gross physical abnormalities (Crawley 2000). Olfaction abilities are considered essential for mouse social interactions (Cheal & Sprott 1971; Liebenaur & Slotnick 1996; Wrenn et al. 2003). Retrieval of food buried in cage bedding was quantitated in a 15-minute olfactory test, to ensure that the subject mice were not grossly impaired in olfaction abilities. Further controls were designed to detect additional symptoms relevant to autism, such as abnormalities in locomotion or motor coordination (Hallett et al. 1993; Piven et al. 1997), that could directly affect performance of the social task.
Parameters scored within the test sessions in the social apparatus provided additional information on procedural abilities necessary for social behaviors. Before the start of the choice tests, subject mice were given a habituation session in a neutral central chamber. The number of times that the mouse exited from the habituated empty central chamber into the novel empty chamber provided a general measure of locomotor function. In addition, time spent in the habituated center chamber provided a preliminary measure of anxiety-like behaviors, analogous to the open field emergence test (Holmes et al. 2003; Smith et al. 1998).
Five inbred strains of mouse were chosen for the present evaluation of autistic-like social behaviors in mice: C57BL/6J, DBA/2J, FVB/NJ, A/J and the hybrid B6129PF2/J. Many approaches to studying mouse models of human disease have focused on mice with specific targeted mutations in single genes. Available evidence from family and twin studies, and the results to date from linkage studies, suggest that, for most cases, the simplest model of autism involves the interaction of multiple genes (Risch et al. 1999). Given data from family genetic studies showing a broad range of the behavioral expression of the underlying genetic liability for autism in nonautistic relatives (Piven et al. 1997), the possibility must also be considered that variation in the social and cognitive behaviors manifested by autistic individuals represents one end of the extreme of normal genetically-controlled variation. As an initial effort to characterize social behaviors in mice that might be relevant to the study of autism, we therefore focused on comparing and contrasting the continuum of social behaviors that could be observed across several ‘normal’ strains of inbred mice.
The present study presents the first results using the three-chambered apparatus to score social approach behaviors relevant to autism. For the purposes of this study, ‘sociability’ is defined as a significant propensity to spend time with another mouse, as compared to time spent in an identical but empty chamber. ‘Preference for social novelty’ is defined as a significant propensity to spend more time with a new mouse than with a familiar mouse. Most experimental groups consisted of male mice, for consistency with the 4:1 male:female difference in the diagnosis of autism (Folstein & Rosen-Sheidley 2001). Mice were tested at both juvenile and adult ages, consistent with the neurodevelopmental aspects of autism. The choice of inbred strains was designed to represent those commonly used in behavioral genetics, and to provide baseline scores that could later be compared to lines of targeted gene mutations bred into these background strains.
- Top of page
- Materials and methods
Deficits in social interaction and communication are primary diagnostic indicators of autism (e.g. American Psychiatric Association 1994; Lord et al. 2000b). From an early age, autistic children show low levels of eye contact and affective responses, less orientation toward others and language deficits, with a concomitant paucity of speech and social communication (for review, see Tager-Flusberg et al. 2001). Given the primacy of social impairment in autism, we considered this symptom to be fundamental to a mouse model of autism (Insel 2001). Therefore, the present methods were designed to quantitate levels of sociability, i.e. tendency to initiate social contact, and preference for social novelty, i.e. tendency to initiate social contacts with a new individual as compared to someone familiar from past experience, using a simple scoring approach for laboratory mice.
In the present studies, sociability was defined as the tendency to approach and remain proximal to an unfamiliar conspecific, vs. avoidance of the stranger mouse by remaining in the center chamber or exploring an equally novel chamber devoid of another mouse. Overall, four of the five inbred and hybrid strains tested spent more time in the side of the social test box containing the unfamiliar stranger, vs. time in the empty side. Juvenile male mice from three different inbred strains (C57BL/6J, DBA/2J and FVB/NJ) and one hybrid cross (B6129PF2/J), as well as older adult male mice (C57BL/6J and DBA/2), all showed significant levels of sociability, as evidenced by longer durations for time in the stranger side. Similar levels of sociability were clearly evident in a group of young female C57BL/6J mice.
When a novel unfamiliar conspecific was placed in the formerly empty side, a reversal of chamber preference was noted in all groups of juvenile mice from the three inbred strains and the hybrid cross. The higher duration of time spent with the novel stranger (stranger 2) was interpreted as a demonstration that the mice preferred to approach a new stranger, rather than restrict contact to a more familiar individual. It was also an indication of the ability of the mice to discriminate between the two strangers, and to recognize the one that had not been encountered before. Previous work in social recognition has shown that mice show both habituation of olfactory responses directed toward an unfamiliar conspecific, and dishabituation, with a reinstatement of social sniffing, when a novel stimulus animal is presented (Bluthe et al. 1993; Dluzen & Kreutzberg 1993; Ferguson et al. 2000; Kogan et al. 2000; Winslow & Camacho 1995). The choice procedure used in the present study allows the investigator to determine whether the test mice prefer proximity to a novel stranger mouse, vs. proximity to a previously encountered mouse, which may be more relevant to the core symptoms of autism than simple recognition tests.
The social preference pattern observed for the duration measures was not evident for the number of entries into each side for the initial set of juvenile male mice (C57BL/6J, DBA/2J and FVB/NJ). In fact, within each strain, the number of entries into either side of the social test box was almost equal for both the sociability and social novelty tests. However, differences between the three inbred strains were found, with the DBA/2J mice having significantly lower entries into each side, in comparison to the C57BL/6J and FVB/NJ mice. While the DBA/2J group also had significantly lower levels of locomotor activity in the open field test, the significantly higher locomotor scores observed in the FVB/NJ strain, in comparison to the other two groups, were not reflected in an overall higher number of entries. Therefore, for the juvenile male mice, number of entries did not reflect the trends for social preference seen in the duration measures, and only partially correlated with their locomotor activity independently quantitated in the open field.
In contrast, the hybrid B6129PF2/J mice, when tested for sociability, had more entries into the side of the social test box containing the unfamiliar stranger, in comparison to the empty side. There was a reversal of this pattern during the social novelty test, when a higher number of entries were made into the side with the second unfamiliar stranger (formerly the empty side). The pattern for the entries measured in the hybrid mice reflected the social preference pattern observed for the duration measure. In the second group of juvenile C57BL/6J mice, both males and females made equal numbers of entries into each side of the social test box during the assay for sociability, but showed an increased number of entries into the side of the social box containing the second, ‘novel’ unfamiliar stranger during the test for social novelty. Finally, the adult C57BL/6J and DBA/2 mice had higher numbers of entries into the side of the chamber containing the unfamiliar conspecific in the sociability test. Therefore, number of entries appears to vary in terms of its correlation to other parameters. Future studies with these and other inbred strains will be designed to understand the usefulness of the number of entries parameter as a measure of sociability and preference for social novelty.
Based on the present results, time spent in each chamber appears to be the most useful parameter. However, it will be important to differentiate time spent in the chamber from true social investigation of the stranger target mouse in its wire cage. One approach to confirming the social nature of the chamber exploration is to score sniffs directed towards the stranger. Number of sniffs of the wire cage containing the stranger vs. number of sniffs of an empty wire cage were compared in a companion study with an automated version of the present social test chambers (Nadler et al. 2004). Higher numbers of sniffs were directed toward stranger 1 than towards the empty wire cage, and towards the novel stranger 2 vs. the now-familiar stranger 1, in C57BL/6J, DBA/2J and FVB/NJ male mice (Nadler et al. 2004).
A particularly interesting finding from the present strain distribution analysis is the low level of sociability in the A/J inbred strain. Adult A/J males failed to show a significant preference for the side containing stranger 1, as compared to the empty side. The A/J mice spent significantly more time in the center chamber, indicating a strong preference for the start box that was not seen in the other strains. These findings may indicate a specific social deficit in the A/J strain. Alternatively, these data may reflect the low levels of open field exploratory behaviors and high levels of anxiety-like behaviors in the light/dark transitions test reported for A/J mice in other studies (Bouwknecht & Paylor 2002; Mathis et al. 1994, 1995; Thifault et al. 2002; Wahlsten et al. 2003) and reflected in the low number of entries in the present study. Arguing against an artifactual interpretation that hypolocomotion caused the social deficit in the A/J group are the data of Fig. 8. Repeated testing of A/J in the modified apparatus increased the number of entries, but did not yield significant sociability scores. In addition, low locomotor scores in DBA/2J and high locomotor scores in FVB/NJ appeared to be unrelated to amount of time spent in the chamber with stranger 1 and stranger 2. It will be important to address the potential confounding variable of locomotor activity in future investigations.
The present procedures for the sociability and social novelty tests are designed to be simple, quantitative and easily mastered by investigators. Methods are based on procedures examining pair-bonding in voles (Dean, S.M. & Vandenbergh, J.G. as above), as well as many other elegant tests for social behavior in rodents (Bluthe et al. 1993; Carter et al. 1995; Dluzen & Kreutzberg 1993; Gheusi et al. 1994; Insel and Young 2001; Kogan et al. 2000; Winslow & Camacho 1995; Winslow 2003). In most cases, these procedures have involved placing two animals together and recording behavioral responses, or using a stimulus animal that is tethered or otherwise confined to one side of a multichambered testing apparatus. Observations of mice in these settings have shown that exposure to an unfamiliar stranger typically evokes a period of intense social sniffing, which may be followed by escalating aggressive responses or sexual behavior, dependent upon such factors as age and sex of the unfamiliar conspecific (Bluthe et al. 1993; Kogan et al. 2000; Winslow & Camacho 1995). Use of the three-chambered apparatus for demonstrating social preference (Insel 2001) was adapted by Brodkin et al. (2004) to investigate social approach and avoidance in female mice. The center chamber was smaller than the side chambers, and Plexiglas cylinders with small air holes were used to contain the stranger. Juvenile Balb/cJ females demonstrated social avoidance, rather than social approach, in this study (Brodkin et al. 2004). The present protocol used a small wire cage to enclose the stimulus mouse, which allowed visual, auditory and olfactory contact between the mice, but prevented the occurrence of aggressive and other responses. This method also simplified the quantification of data. As noted by Winslow (2003), the analysis of the full behavioral series that can occur in social tests demands familiarity with the mouse ethogram, extensive training and the ability to discern the sometimes rapid and fleeting responses that mice can emit. In our experience to date with the present procedures, human observers attain a high level of proficiency after only one or two practice sessions, with high rates (greater than 95%) of interrater reliability between trained observers.
Several additional behavioral assays were performed in the initial set of juvenile male mice (C57BL/6J, DBA/2J and FVB/NJ). These assays were designed as controls for sensory and motor abilities, etc., to ensure that the test mice were physically able to perform the procedures necessary for the social tasks described above. Further, we included home cage observational periods to identify other behavioral parameters that could be useful in a mouse model of autism. Nesting and sleep patterns, and any incidents of stereotypy, hyperactivity and seizures were noted, relevant to common symptoms of autism. Observational findings in some of these behavioral domains have been reported in some targeted gene mutations, e.g. deficits in social interaction within the home cage have been reported for mice deficient in the dishevelled-1 gene (Lijam et al. 1997) or the NMDA NR1 subunit (Mohn et al. 1999). In the present study, the juvenile male mice made nests and huddled together while sleeping. General health, neurological reflexes, activity levels, motor functions and sensory abilities were normal in most regards for the strains tested. All of the inbred strains were able to locate buried food, supporting an interpretation that none were anosmic, which could have severely impacted social behaviors by blocking detection of olfactory cues emitted by the stranger mice. The FVB/NJ mice demonstrated a significantly shorter latency for locating the buried food, but as they did not have markedly higher levels of sociability or preference for social novelty, it seems likely that the shorter latency reflected higher levels of exploratory activity in these mice, as shown in Table 1.
These control parameters are not only important for the interpretation of the results from the social tests, but suggest additional components for putative mouse models of autism. For example, some autistic patients demonstrate aberrant motor responses, including repeated hand flapping, jumping and other repetitive and stereotyped behaviors (Bodfish et al. 2000; Tager-Flusberg et al. 2001). Altered sensory responses have been reported, including both hypo- and hyper-reactivity to environmental stimuli (Baranek 2002). Children with fragile X syndrome often show autistic behavior (Hagerman et al. 1986), and the mouse model of fragile X syndrome has been characterized by changes in activity (Bakker et al. 1994; Peier et al. 2000) and reactivity to acoustic stimuli (Chen & Toth 2001; Nielson et al. 2002). Further work has suggested that the changes found in the mouse models of fragile X are dependent upon the particular background strain for the mutation (Dobkin et al. 2000; unpublished observations from our laboratory).
Development of a new animal model of a human disease is a long and iterative process. The present mouse tasks were designed to maximize face validity to some of the core symptoms of autism. Issues of construct validity and predictive value remain to be investigated. For example, the low sociability scores of the A/J strain require further consideration in terms of sex differences, age of onset and lifetime expression, previous life experiences and potentially confounding artifacts. Following detection of a social deficit in a line of mice, the model can be expanded to encompass a wider set of parameters, including other complex and sensitive measures of social behaviors, quantitated by videotracking software. More in-depth analyses of other behaviors potentially relevant to autism will be useful, including measures of motor stereotypies, sleep dysfunction, sensory hypersensitivity and anxiety-like behavior. Potential applications of this social task include the testing of hypotheses about the etiology of autism, investigating transgenic and knockout mice with mutations in candidate genes for autism, screening large numbers of offspring obtained from chemical mutagenesis for mutations relevant to autism and comparing strains of mice with varying levels of social skills using DNA microarray and proteonomics analyses. Most importantly, the utility of these social behavior tasks in mice will require comprehensive testing to determine their predictive value for evaluating potential therapeutic approaches to treat or prevent the social deficits in autistic patients.