Background
Unraveling the biological mechanisms underlying pathological conditions characterized by atypical social behavior, such as autism spectrum disorder (ASD), is currently one of the main challenges in the field of social neuroscience [
1]. Addressing this challenge requires the use of standard behavioral paradigms that typify the behavior of animal models in an unbiased way. In a seminal work published more than a decade ago, Moy and colleagues [
2] presented the three-chamber test, which has become a standard way to evaluate social behavior in animal models of ASD.
This paradigm is mostly used to assess how much a rodent subject prefers a social stimulus over an object one (social preference (SP), also termed sociability), as well as how much it prefers a novel social stimulus over a familiar one (social novelty preference (SNP)), which is the innate tendency of mice and rats [
3,
4]. The test, based on measuring the time spent by the subject in either a central chamber or each of two lateral chambers where distinct stimuli are located, suffers from several caveats. First, it largely depends on the preference of the subject to locate itself in one of the chambers, hence should be sensitive to parameters that influence spatial navigation, memory, and preference. Under certain conditions, such parameters may vary independently of the motivation of the subject for a direct social interaction and thereby interfere with the efficiency of the test to directly measure social motivation and preference. Second, this test is mostly used to measure the total time spent by the subject in each of the chambers, while neglecting the behavioral dynamics.
Here, we present a novel experimental apparatus and automated analysis system that offer an alternative to the three-chamber test and enable performing the same behavioral examinations while solving the aforementioned caveats. The custom-made apparatus is simple for production, and the analysis system is publically available as an open-source program, thereby allowing any lab to easily employ it. We demonstrate the ability of this system to measure novel parameters of murine social behavior, thus to detect previously unidentified sex- and strain-specific differences in the dynamics of social preference and social novelty preference.
Discussion
A main goal of social neuroscience, one of the most rapidly developing fields in neuroscience, is to reveal and characterize the biological mechanisms underlying deficits in social behavior displayed by humans in pathological conditions such as ASD [
1]. The use of animal models of such conditions seems to be crucial for the achievement of this aim [
6‐
8]. Animal models enable examination and manipulation of biological mechanisms and serve to explore the effects of various interventions designed to correct their impaired social behaviors. Yet, efficient use of animal models requires standardization and automation of behavioral tests that will enable assessing social behavior in an unbiased manner, independently of the observer or the specific laboratory performing the experiments [
9]. While such methods are widely used in other fields of neuroscience, they are hard to implement for exploration of the highly complex mammalian social behavior [
10‐
12].
A breakthrough in this field was reported a dozen years ago by Moy et al. [
2], who introduced the three-chamber test for the assessment of SP (sociability) and SNP. This test was shown to be efficient in revealing differences in social behavior between various strains of mice [
13]. Since then, this test has become a standard procedure for assessing social behavior in mouse models of ASD [
9,
14,
15].
Yet, despite several attempts to automate the three-chamber test [
16,
17], it still suffers from several caveats: first, it measures the time spent by the subjects in each of the three compartments of the apparatus, rather than directly measuring social investigation behavior. While a good correlation was found between the two parameters in general [
16], one cannot exclude the involvement of other behavioral parameters, such as spatial navigation and place preference, which may be independent of the motivation to investigate a specific social stimulus at certain conditions. Second, the test is mostly used in a manner which does not take in account the dynamics of the social behavior during the test (but see [
18]). Third, the location of the stimulus in a round wire cage within its compartment creates difficulties to precisely relate specific investigation events with other measured parameters, such as vocalization or electrophysiological activity. Importantly, significant differences in both SP and SNP of several mouse strains were observed when either the time spent in the chamber or the time spent sniffing the stimulus were measured in the three-chamber test [
13,
19]. It should be noted that several automated systems aiming to analyze social behavior in a higher resolution than the three-chamber test were recently published, each with its own advantages [
20,
21].
Here, we presented a novel apparatus and analysis system that enable the same type of experiments for which the three-chamber apparatus is used, while solving the aforementioned caveats. The use of triangular chambers located in two corners of the arena restricts the area of interaction with the stimulus to an easily defined plane, which allows precise automated detection of investigation behavior events. This advantage is used by our open-source software to measure the dynamics of animal behavior in each of the employed tests. It also allows the random relocation of the chambers in opposite corners of the non-compartmentalized arena in each stage of the test, thus neutralizing any effect of spatial navigation, preference, or memory. These characteristics of our system enable direct assessment of the motivation for social investigation of each of the stimuli by the experimental subject. Using these advantages of the system, in combination with its ability to measure social behavior of subjects connected to cables, would allow recordings of physiological parameters during specific events of social investigation. Finally, the apparatus is compact, simple, and affordable, and the analysis software is publically available, making this system ready to use in any laboratory.
Using this system, we sought to characterize sex- and strain-specific differences in the SP and SNP tasks, for which the three-chamber test is widely used. We demonstrated, for the first time to our knowledge, that a main difference between male and female C57BL/6J mice in these tests is the dynamics of their behavior. In both tests, male mice were more persistent in their preference than females, which completely lost their preference towards one of the stimuli after 3 min of each test. It should be noted that no difference in SP or SNP between male and female C57BL/6J mice was observed by Moy et al. [
2] using the three-chamber test. Nevertheless, they reported a significant reduction in the total time spent by females, as compared to males, in the social chambers during the SNP test. This observation is in agreement with our results, showing that females spent less time sniffing both social stimuli, as compared to males. Thus, by tracking the dynamics of social investigation using the system presented here, we were able to show previously identified and unidentified differences in social behavior between male and female mice.
Using our system, we compared two types of chambers, differing only in the level of separation between the subject and stimuli. While the grooved chambers enable rather limited social interactions, the meshed chambers allow better exposure of the social stimulus to the subject, with only a metal mesh separating between them. Whereas we found that both chambers can be efficiently used in the SP and SNP tests, the dynamics of social behavior is significantly different between them. Specifically, the meshed chambers seem to enhance longer bouts of social investigation towards the social stimulus, while reducing the number of short events, as compared to the grooved chambers. Since we found that most of the difference in investigation time between the stimuli was in longer bouts, we expect that the ability of our system to categorize investigation bouts according to their length will enhance the sensitivity of the preference analysis in the SP and SNP tests.
Using the meshed chambers, we analyzed the behavioral dynamics of male mice in both the SP and SNP tests and defined several new useful parameters, such as the duration of investigation bouts and number of transitions between stimuli. By analyzing these parameters, we found that both tests can be divided into initial exploration phase characterized by multiple transitions between stimuli and short investigation bouts, which is followed by interaction phase characterized by lower level of transitions and higher level of longer investigation bouts, mainly with the preferred stimulus. Interestingly, this dynamics was found to be altered in BTBR mice that showed high level of transitions and low level of long bouts throughout the SP test, suggesting a difficulty to shift from exploration to interaction. Notably, a similar reduction in the length of social interactions of BTBR mice as compared to C57BL/6J was recently reported using a different automated behavioral system [
20]. Thus, the analysis of the behavioral dynamics in our system allowed us to define a new type of impaired social behavior in a mouse model of ASD. Further experiments with other types of ASD animal models will reveal if such impairment is a hallmark of their social behavior.
Conclusions
To summarize, here, we present a novel design of a simple and affordable behavioral system that enables automated and precise measurements of social investigation behavior. Using the ability of this system to measure new parameters of social investigation, we demonstrated sex- and strain-specific differences in the dynamics of social behavior during the SP and SNP tasks. Thus, unlike the three-chamber test, our system provides multiple automatically measured parameters that enable a thorough analysis of the dynamics of social behavior. Such analysis should allow a detailed classification of mouse strains and genetically modified lines according to their specific social behavioral deficits. Moreover, the possibility of precise temporal detection of investigation bouts towards social stimuli demonstrated by us here, combined with the ability of the system to monitor the behavior of subjects connected to electrical cables or optical fibers, should enable using this system to record physiological parameters, such as brain neural activity, while the animals are investigating specific social stimuli. We believe that the system presented here would enhance the efforts to reveal the role of various brain networks and molecular mechanisms in mammalian social behavior and to examine their function in animal models of pathological conditions characterized by atypical social behavior, such as ASD.