Social behavior in a genetic model of dopamine dysfunction at different neurodevelopmental time points

Abstract

Impairments in social behavior characterize many neurodevelopmental psychiatric disorders. In fact, the temporal emergence and trajectory of these deficits can define the disorder, specify their treatment and signal their prognosis. The sophistication of mouse models with neurobiological endophenotypes of many aspects of psychiatric diseases has increased in recent years, with the necessity to evaluate social behavior in these models. We adapted an assay for the multimodal characterization of social behavior at different development time points (juvenile, adolescent and adult) in control mice in different social contexts (specifically, different sex pairings). Although social context did not affect social behavior in juvenile mice, it did have an effect on the quantity and type of social interaction as well as ultrasonic vocalizations in both adolescence and adulthood. We compared social development in control mice to a transgenic mouse model of the increase in postsynaptic striatal D2R activity observed in patients with schizophrenia (D2R-OE mice). Genotypic differences in social interactions emerged in adolescence and appeared to become more pronounced in adulthood. That vocalizations emitted from dyads with a D2R-OE subject were negatively correlated with active social behavior while vocalizations from control dyads were positively correlated with both active and passive social behavior also suggest social deficits. These data show that striatal dopamine dysfunction plays an important role in the development of social behavior and mouse models such as the one studied here provide an opportunity for screening potential therapeutics at different developmental time points.

Several psychiatric disorders, most notably autism spectrum disorders (ASD) and schizophrenia, include deficits in social behavior (Abdi & Sharma 2004). Because these are developmental disorders, mouse models can be used productively to investigate the etiology of these deficits, to identify potential early markers of disease and to develop behavioral and pharmacological interventions that can then be tested in patients. Historically, studies of social behavior in rodents have focused on rats mostly in the context of dam/pup interactions, reproductive contexts focusing on male sexual behavior and male agonistic encounters (Agmo 1997; Blanchard & Blanchard 1988; Hofer 1994). With the recent emergence of new mouse models of neuropsychiatric disease, other social contexts such as social play, sibling reciprocal social interactions and adult female–female interactions tests of sociality have been examined (e.g. Scattoni et al. 2011 and for review see: Ricceri et al. 2007). The majority of this work has been carried out in the context of autism research, where neurobiological studies have dramatically increased in recent years. Identifying robust tests of sociality that capture the developmental trajectory of schizophrenia is also important given the number of mouse models that have recently been generated. Therefore, we set out to develop methods that could be adapted to assess social behaviors in genetic mouse models that are sensitive to developmental changes in species-typical social behavior.

The form and function of social behavior changes across the lifespan in mice, as in all mammals. For example, at the time of weaning, as peer relationships become more critical for survival, juvenile C57BL/6J mice display increased investigative behaviors toward conspecifics (Panksepp & Lahvis 2007). In adolescence, high rates of investigatory behavior are still observed in C57BL/6J male resident/female and female resident/female pairs, but decline in female resident/male and male resident/male pairs (Panksepp et al. 2007). Around puberty, male/male mouse interactions shift from being submissive-philopatric to being agonistic-dispersive (Ricceri et al. 2007) and male agonistic behaviors are highest around P35, coincidental with increased sex hormone (androgen) levels (Ricceri et al. 2007).

These developmental trajectories of normal behaviors are especially important to consider when modeling psychiatric disease because of the developmental nature of social deficits observed in these disorders. For example, externalizing behavior, including traits such as aggression, bullying, disruptiveness and defiance, is a fairly sensitive and specific predictor of schizophrenia that can emerge in preschool-age children (Tarbox & Pogue-Geile 2008). By adolescence, however, social withdrawal becomes a more predictive marker (Tarbox & Pogue-Geile 2008). Understanding the onset and development of prodromal symptomatology is of critical importance because early intervention may delay the occurrence of the first psychotic episode and lessen the severity of the disease (Hafner et al. 2004; Insel 2010; Mcgorry et al. 2002; Tarbox & Pogue-Geile 2008).

Here, we first characterize the development of social interactions in control mice to provide a reference framework of normal social development. We then characterize a transgenic mouse model of the increase in striatal dopamine D2 receptor activity that has been observed in schizophrenia patients (Abi-Dargham et al. 2000). Like patients, mice overexpressing dopamine D2 receptors selectively in the striatum (D2R-OE mice) display impairments in cognitive function, including impairments in working memory, sustained attention, behavioral flexibility, conditioned associative learning and interval timing (Bach et al. 2008; Drew et al. 2007; Kellendonk et al. 2006; Ward et al. 2009). In addition to these cognitive phenotypes, D2R-OE mice also display behavioral deficits that closely resemble the negative symptoms of schizophrenia, including decreased motivation, a reduction in willingness to expend effort to earn rewards and a diminished capacity to represent the value of potential future outcomes accurately (Simpson et al. 2011; Ward et al. 2012). Given the similarity of these negative symptom phenotypes in patients and the D2R-OE mice (Simpson et al. 2012), we next wanted to know whether striatal D2R overexpression also results in a deficit in other negative symptom domains, namely social behaviors.

In control mice, we describe interactions between all social pairings (male or female subject with male or female stimulus) at three ages: juvenile, adolescent and adult. In addition to describing these interactions, we also analyzed ultrasonic vocalizations (USVs). Ultrasonic vocalizations are closely associated with social investigation and have been used as an assay of deficits in communication in other social paradigms (Fischer & Hammerschmidt 2011; Lahvis et al. 2011).

Next, we examined the developmental trajectory of social behavior in D2R-OE mice and found that the increased striatal D2 receptors resulted in social deficits in behavior and vocalizations that emerge between adolescence and adulthood that share qualitative similarities with social deficits in patients with schizophrenia. These results show that striatal dopamine function influences social development and suggest that social deficits in patients with schizophrenia may be linked to striatal dopamine dysfunction during development.

Materials and methods

Animals

Mice were housed under standard laboratory conditions in a colony room with a 12-h light–dark cycle. Cages were monitored daily for the presence of newborn pups and the date of birth was designated as day 0. Mice were weaned on postnatal day (P) 21 with same sex littermates in groups of 2–4 mixed genotype animals and moved to a housing room adjacent to the testing room. Animals were tested longitudinally at three developmental time points [P26 (juvenile), P45 (adolescent) and P65 (adult)] but not in the same experimental pairing. Thus, although similar numbers of mice were tested across all time points, the numbers of mice in different experimental pairings at different ages vary. The number of animals used at different ages and social pairings are detailed in the Supporting Tables. No mice were lost to sickness or death.

The generation of D2R-OE has previously been described (Kellendonk et al. 2006). Briefly, mice expressing the human D2 receptor under control of the tet-operator (tet-O) were generated on a C57BL/6J: CBA F2 background. These tet-O_hD2R mice (which have since been backcrossed more than 10 generations to the C57BL/6J background) were crossed with mice expressing the tetracycline transactivator (tTA) transgene under the calcium/calmodulin-dependent kinase IIα (CaMKIIα) promoter (Mayford et al. 1996). CaMKII–tTA mice were on 129SveV(Tac)N17 background. The F1 offspring from an intercross between these two strains were used for all experiments. Mice that inherited both transgenes are D2R-OE mice. Several previous studies did not reveal a single transgene effect (Kellendonk et al. 2006), and the a posteriori analyses, grouping data across pairings, did not show single-transgenic effects. Thus, littermate mice carrying only one of the transgenes, or neither transgene, were combined and collectively called ‘control’. Mice were genotyped by triplex polymerase chain reaction (PCR) using primers specific for tTA, tet-O and a fragment of an unrelated endogenous gene fragment (to provide a positive control for DNA quality and the PCR).

Both male and female control and female D2R-OE subject mice were tested with unfamiliar age-matched control stimulus mice of the same F1 hybrid genetic background [C57BL6(J)/129SveV(Tac) F1 hybrid]. Subject and stimulus mice came from several litters and litters were not exclusively comprised of subject or stimulus mice. We focused on social pairings with female D2R-OE subjects because pilot work showed no genotypic differences with male D2R-OE subject mice. All subject and stimulus animals were gently handled and tail-marked without restraint every 2 days to reduce experiment-induced stress (Hurst & West 2010). Estrus state was noted for all females after P35 and balanced across conditions (Byers et al. 2012). All tests and treatment procedures were approved by and were in accordance with the Institutional Animal Care and Use Committees of New York State Psychiatric Institute and Columbia University.

Testing chamber and behavioral recordings

The testing apparatus consisted of a sound-attenuated chamber made from Plexiglas, lined on the bottom and sides with foam (Sound-foam M, soundcoat.com), and the edges of the lid lined with LED lights on a dimmer with illumination output of approximately 14 lux (HitLights Cool White Flexible Ribbon LED Strip Light, 12VDC Input; Zitrades Waterproof 30 Watt LED Power Supply Driver Transformer, 120–12 Volt DC Output; PWM Dimming Controller for LED Lights or Ribbon, 12 Volt 8 Amp; Adjustable Brightness Light Switch Dimmer Controller, DC12V 8A 96 W for Led Strip Light) (Fig. 1a,b). For every subject, a clean standard rat cage (44 cm L × 23.5 cm W × 20.5 cm H) containing 2 cm of clean, familiar betachip bedding was placed within the chamber. A camera (Sony DCR-SR45; Sony: Tokyo, Japan) was affixed above the chamber and an ultrasonic microphone with a frequency range of 10–200 kHz and approximate input-referred self-noise level of 18 dB SPL (Avisoft UltraSoundGate condenser microphone capsule CM16, Avisoft Bioacoustics, Berlin, Germany) was placed overhead in one upper corner of the chamber above the bedding at a 45° angle, facing to the center of the chamber.

Figure 1. Sound-attenuated testing chamber from the (a) side and (b) top. Animals are tested on neutral bedding and stimulus animals are marked for identification.

Camera and ultrasonic microphone are affixed overhead. (c) Spectral visualization of ultrasonic vocalizations used for counting and measurement recorded during P45 female–female control pair interaction.

Behavior for all test sessions was video-recorded and video files were scored with Observer XT 7.0 software (Noldus Information Technologies, Wageningen, The Netherlands) by an investigator blind to the sex or genotypes of the animals. Behavioral interactions were scored and included ‘active’ social behavior, defined as the nose of the subject animal within 1.5 cm of the stimulus animal while sniffing/investigating or closely following/chasing, ‘passive’ social behavior, defined as the nose of the stimulus animal within 1.5 cm of the subject animal, and ‘reciprocal’ social behavior, defined as the nose of both the subject and stimulus animal within 1.5 cm of the other. Behaviors are expressed as a percentage of the total test time. Grooming was observed but not scored and allogrooming was not observed.

Ultrasonic vocalizations were captured using Avisoft RECORDER software version 4.2, and spectrograms were created and analyzed using Raven Pro (Fig. 1c; Bioacoustics Research Program, Cornell University, Ithaca, NY 14850, USA). The band limited energy detector was used as a first pass to automatically detect calls. The target signal parameters were a minimum frequency of 30 kHz and a maximum frequency of 125 kHz, selecting sounds with a minimum duration of 0.008 seconds and a maximum duration of 0.200 seconds with a minimum separation of 0.021 seconds. The signal-to-noise parameters required a minimum occupancy of 40.0%, with signal noise ratio (SNR) threshold set above 5.0 dB. The noise power estimate parameters were defined as 19.999 seconds block size, with a 19.999 seconds hop size, and 20.0 percentile. These parameters were chosen based on software developer recommendations with minor adjustments made to capture as many vocalizations and fewest noise readings as possible. No exclusion band or bandwidth filter was used. After the band limited energy detector was run on the call, each selection was reviewed by a trained human observer and those selections that contained background noise were deleted. That is, selections that captured bedding noise were excluded while calls lower in amplitude were retained. This scoring method was validated by comparing a subset of automatic detection (M = 161.8, SD = 231.7) to hand-selected calls (M = 202.6, SD = 283.1); t(61) = 0.62, P = 0.53.

In addition to call number, measurements include begin time (time at which the first call began; seconds), aggregate entropy (measurement of overall disorder in the sound by analysis of the energy distribution within a selection; bits), average power (the value of spectrogram’s power spectral density summed and divided by the number of time-frequency bins in the selection; dB), delta time (the difference between begin and end time for the selection; seconds), average entropy (measurement of disorder for a typical spectrum within the selection calculated by finding the entropy for each frame in the selection and then taking the average of these values; bits), energy (the total energy within the selection bounds; dB), peak frequency (the frequency at which the peak power occurred within the selection; Hz), interquartile bandwidth (the difference between the first and third quartile frequencies; Hz) and peak power (the power at the darkest point in the selection; dB) (Charif et al. 2010). Our preliminary analysis revealed that the measures of average and aggregate entropy were highly positively correlated measures, r = 0.77, P < 0.0001, with a R² = 0.597, as were the measures of average and peak power, r = 0.96, P < 0.0001, with an R² = 0.912. Therefore, average entropy and peak power measures were excluded from the results presentation to reduce redundancy.

An additional examination of the D2R-OE data included additional measures that have been reported previously (Panksepp et al. 2007) and that are described by Lahvis et al. (2011) as measures of prosody: intercall interval (the duration of time between USVs), frequency modulation (the absolute value of the degree of the angle of the USV relative to the horizontal) and average frequency.

Testing procedure

The day before testing, subject mice were isolated in standard mouse cages (28 cm L × 17 cm W × 13 cm H) with clean bedding. Stimulus animals remained group-housed. The day of testing, subject and stimulus animals were moved to the testing room and allowed to habituate for 1 h. Subject animals were placed singly in the testing apparatus and allowed to explore for 5 min. Habituation and testing times vary widely in the literature. We have observed that this amount of time is sufficient for mice to overcome the novelty of a new environment enough to direct behavior toward an incoming conspecific but not long enough to establish territoriality. This is supported by data presented in subsequent sections that show an effect of the first occupant of the chamber and that no territorial fighting was observed, including in adult male dyads. At 5 min, a stimulus mouse was marked in the center of the back (BIC^® – Wite-Out Water-Based low VOC Correction Fluid) and placed in the testing apparatus opposite of the subject mouse. Behavior and USVs were recorded for 5 min, when the peak of social interactions and vocalizations typically occur (D’Amato & Moles 2001; Scattoni et al. 2011). After the 5-min test, both mice were removed from the testing apparatus, weighed and returned to pre-test group-housing configurations. All testing took place during light phase.

Statistical analysis

Statistical tests for behavioral and USV measures were performed using GraphPad Prism (version 5.0, San Diego, CA: GraphPad Software).

Physical interaction measures

For the control pair analysis, two-way repeated measures analyses of variance (ANOVAs) were conducted to evaluate the effect of pairing type on social behavior at each age point. No data were excluded and total sample sizes are shown in Table S1. The within-subject factor was type of social behavior with three levels (active, passive and reciprocal) and the between-subject factor was pairing type with four levels [female subject with female stimulus (female–female), female subject with male stimulus (female–male), male subject with male stimulus (male–male) and male subject with female stimulus (male–female)]. In a second analysis (the D2R-OE analysis), two-way repeated measures ANOVAs were conducted to evaluate the effect of genotype on social behavior for each age point and each pairing type. The within-subject factor was type of social behavior with three levels (active, passive and reciprocal) and the between-subject factor was genotype with two levels (control and D2R-OE). Post hoc analyses of all significant main effects and interactions were assessed with Tukey’s multiple comparisons tests.

We did not statistically evaluate a main effect of age because the study design required that, at each testing point, the subject was paired with an unfamiliar stimulus animal. This often resulted in different social pairings for an animal across the testing time points, e.g. an individual female subject may have been tested with a female stimulus at P26 and a male stimulus at P45 and P65, and that at times a subject mouse may act as a stimulus mouse and visa versa.

Ultrasonic vocalizations

Because of the highly variable nature of USVs, the data set was trimmed for analyses by removing values two standard deviations from the group means for each measure. By age, this resulted in 3.5% of data eliminated at P26 and P45 and 3.7% of data eliminated at P65 in the control pair analyses and 3.6% of data eliminated at P26 and P45 and 2.7% of data eliminated at P65 in the D2R-OE analyses (see Table S2 and Table S4). For the control pair analysis, one-way ANOVAs were conducted to evaluate the effect of pairing type on USV measures at each testing time point. When significant, post hoc multiple comparisons were evaluated using Tukey’s test. For the D2R-OE analysis, independent-samples t-tests were conducted to evaluate the effect of genotype on USV measures at each age and pairing using the Holm–Šidák correction. Histograms were analyzed with the Kolmogorov–Smirnov test and also corrected at each age and pairing using the Holm–Šidák correction. We did not statistically evaluate the USV data across age for reasons described above (Section Physical interaction measures).

Results

Normal development of mouse social behaviors in different social contexts

Physical interaction

We found that total social behavior did not differ across social pairings until the mice reached adolescence. At adolescence (P45), differences emerged in females interacting with male stimulus mice compared to pairings with male subject mice. By adulthood (P65), female–male pairings interacted less than all other pair types. To characterize the developmental trajectory in more detail, we measured the active, passive and reciprocal interactions of male or female subjects with male or female stimulus animals at the three developmental time points tested: juvenile (P26), adolescent (P45) and adult (P65). We found that although there were no differences in total social behavior, juvenile animals engaged in significantly more active social behavior than passive or reciprocal behavior (P < 0.001), regardless of pairing type (Fig. 2a; Table S1). In adolescence, a significant interaction between pairing and type of social behavior emerged, F_6,80 = 5.42, P < 0.001 (Fig. 2b; Table S1). At this age, total social behavior was lower in pairs with a female subject and a male stimulus compared to pairings with a male subject. Further analysis of the pairing × behavior interaction revealed that this reduction in total social behavior was driven by lower active behavior (female investigation of the male) (P < 0.001). No differences in passive or reciprocal social behaviors were observed.

Figure 2. Percent time (a) juvenile, (b) adolescent and (c) adult control mice pairs spent in social contact.

(female–female: female subject, female stimulus; female–male: female subject, male stimulus; male–male: male subject, male stimulus; male–female: male subject, female stimulus). Main effect of behavior type: ^###P < 0.001; post hoc comparisons: *P < 0.05, **P < 0.01, ***P < 0.001).

We also observed a significant pairing × behavior type interaction in adulthood, F_6,86 = 22.27, P < 0.001 (Fig. 2c; Table S1). Total social behavior was lower in pairings with an adult female subject and an adult male stimulus compared to all other social pairings at this age [female–male vs. female–female (P < 0.01); female–male vs. male–female (P < 0.01); female–male vs. male–male (P < 0.05)]. As in adolescence, active social behavior was lower in female–male compared to all other pairings (P < 0.001). It was also lower in male–male compared to female–female and male–female pairs (P < 0.001). Consistent with the transition in the nature of the adult social interaction, passive social behavior was higher in female–male pairings compared to female–female and male–female pairings (P < 0.05) reflective of male stimulus investigation in the opposite-sex pairings. No significant differences were observed in reciprocal social interactions. In sum, females’ active social behavior was lowest and passive social interactions were highest in adult female–male pairs while male–male pairs engaged in less active social behavior than male–female pairs.

These observations reveal that there are different developmental trajectories in different social contexts with the result that both the quantity and type (e.g. active, passive and reciprocal) of social interactions develop in ways unique to social pairing types. Therefore, the standard simple analysis of total social behavior focusing on any one sex pairing at any one developmental time point may miss important changes in components of total social behavior.

Ultrasonic vocalizations

To investigate further the nature of the social interactions in control mice at different developmental time points, we analyzed the USVs emitted during the test sessions. Several attributes of USVs were analyzed, details are provided in the methods section. As observed with physical interactions, no differences in any USV measure were observed at the juvenile time point (Fig. 3; Table S2).

Figure 3. Ultrasonic vocalizations emitted by control pairs across age and social pairing, measuring (a) count, (b) energy, (c) time, (d) length, (e) frequency, (f) interquartile bandwidth, (g) aggregate entropy and (h) power of the calls (female–female: female subject, female stimulus; female–male: female subject, male stimulus; male–male: male subject, male stimulus; male–female: male subject, female stimulus).

Post hoc comparisons: *P < 0.05, **P < 0.01, ***P < 0.001).

In adolescence, female–female pairs emitted fewer USVs compared to male–female pairs (P < 0.01) (Fig. 3a; Table S2). Overall aggregate entropy in the calls of female–female pairs was also reduced compared to the male–male pairs (Fig. 3g; Table S2; P < 0.05), suggesting more concentrated power within the frequency domain.

We found no differences in the number of USVs across groups in adulthood (Fig. 3a; Table S2), but other USV measures were lower in female–female pairs. Female–female pairs emitted shorter calls than pairs with a male subject mouse (Fig. 3d; Table S2; P < 0.05), and the power of their USVs was more broadly distributed within the frequency domain compared to all other pair types as reflected in the aggregated entropy measure (Fig. 3g; Table S2; female–male: P < 0.001; male–male, male–female: P < 0.01).

Across all control data, the number of USVs emitted was sometimes correlated with active and/or passive social behavior (Fig. 4). In female–female pairs, USV count was correlated with active social behavior at P26 (P < 0.05) and P65 (P < 0.01) but not at P45 (Fig. 5). In female–male pairs, USV count was correlated with active and passive social behavior only at P65 (P < 0.001). In male–male pairs, USV count was correlated with passive social behavior at P26 and active social behavior at P65 (P < 0.05). In male–female pairs, USV count was only correlated with active social behavior at only at P26 (P < 0.01).

Figure 4.

Vocalizations emitted during active, passive and reciprocal social investigation of control pairs (top: open circle) and pairs with a D2R-OE subject mouse (bottom: closed circles), pooled across age and social pairings.

Figure 5.

Vocalizations emitted during active, passive and reciprocal social investigation of pairs with a female control subject and female control stimulus, separated by age.

Summary of developmental changes

In sum, female–male pairs displayed an overall reduction in social behaviors in adolescence and adulthood compared to the other types of dyads. Female–female pairs produced vocalizations that were briefer, and more clearly defined contours in the spectrogram at those same ages.

Development of social behaviors in D2R-OE mice

Physical interaction

Given our demonstration that social behaviors in mice changed throughout development, we performed a longitudinal study with a mouse model of dopamine dysfunction that is relevant to schizophrenia (see introduction). Because we determined that the most dynamic changes in social interaction at the different ages were in female–female and female–male pairs, we focused on these two pair types for this study. Although there were no differences in total social behavior in both social pairings of juvenile mice, we replicated the finding of significant main effects of behavior type, female–female: F₂₃₀ = 107.80, P < 0.001; female–male: F₂₃₄ = 40.07, P < 0.001 (Fig. 6a,b; Table S3). Juvenile mice engaged in significantly more active social behavior than passive or reciprocal behavior (P < 0.001). At this age, there were no genotypic differences in total social behavior or in any specific type of social contact.

Figure 6. Percent time (a) juvenile, (b) adolescent and (c) adult female D2R-OE mice spend in social contact with controls (female–female pairs: female subject, female stimulus; female–male pairs: female subject, male stimulus; control: control subject, control stimulus; D2R-OE: D2R-OE subject, control stimulus).

Main effect of behavior type: ^###P < 0.001; post hoc comparisons: *P < 0.05, ***P < 0.001.

In adolescence, again female–female pairs exhibited more active social behavior than passive or reciprocal behavior when the stimulus mouse was female but not when the stimulus animal was male (Fig. 6c,d; Table S3, P < 0.001).

Adult female–female but not female–male pairs showed a genotypic difference in total social behavior. In adult female–female pairs, there was a significant genotype × behavior interaction, F₂₃₄ = 6.17, P < 0.01, such that adult female D2R-OE subjects engaged in significantly less total social behavior than control female–female pairs (P < 0.05). This was driven by less active social behavior from the adult female D2R-OE mouse (Fig. 6e; Table S3; P < 0.001). In the female–male adult pairings, there was no effect of genotype (Fig. 6f; Table S3).

Overall, in adult mice, there was a significant reduction in total social behavior in female D2R-OE mice paired with a control female stimulus mouse. These overall genotypic differences in adulthood were based on differences in active social behavior.

USVs in D2R-OE mice

We analyzed the USVs made by mice during the testing sessions described above. For the averaged measures, we observed no effect of genotype on any of the USV measures during the juvenile or adolescent time points (Table S4). Differences emerged in adulthood, where vocalizations emitted by female–female pairs with a D2R-OE subject occurred earlier, P < 0.05 (Fig. 7c), were significantly shorter, P < 0.05 (Fig. 7d), and with a lower interquartile bandwidth frequency, P < 0.05 (Fig. 7f), compared to control pairs. The complete set of results for this experiment is presented in Table S4.

Figure 7. Ultrasonic vocalizations emitted from female D2R-OE subject pairs with female stimulus mice across age, measuring (a) count, (b) energy, (c) time, (d) length, (e) frequency, (f) interquartile bandwidth, (g) aggregate entropy and (h) power of the calls (control: control subject, control stimulus; D2R-OE: D2R-OE subject, control stimulus.

Holm–Šidák correction, *P < 0.05).

The additional histogram analysis revealed genotypic differences in intercall interval in female–female pairs at P26 and P65 and in female–male pairs at P45 (Figure S1, Supporting Information; P < 0.0001). Frequency modulation histograms differed by genotype in female–female pairs at P26 and P45 (P < 0.001) and in female–male pairs at all ages (Figure S2; P26 and P45, P < 0.001; P65, P < 0.01). Mean frequency histograms differed by genotype in both pairings at all ages (Figure S3; all P < 0.001 except female–male P26, P < 0.01). For all D2R-OE data, USV count was negatively correlated with active social behavior (Fig. 4; P < 0.001). In female–female pairs with a D2R-OE subject, USV count was positively correlated with passive social behavior at P26 (P < 0.01) and with active social behavior at P65 (Fig. 8; P < 0.05). In female–male pairs with a D2R-OE subject, USVs did not correlate with social behavior.

Figure 8.

Vocalizations emitted during active, passive and reciprocal social investigation of pairs with a female D2R-OE subject and control female stimulus, separated by age.

Summary of genotypic differences in social behavior

Adult female–female pairs with D2R-OE subject mice displayed differences in both the quantity and quality of USVs in the same social context and developmental time points that they display reductions in social interaction behavior.

Discussion

We examined the development of social interactions in mice and found that different behaviors have unique developmental trajectories in different social contexts. Juveniles expressed few differences in social behavior based on pairing type whereas adolescents and adults show reliable differences. Coincident with the emergence of sexual maturity, pairs with a female subject and male stimulus showed lower total social behavior compared to pairs with a male subject in adolescence and compared to all other pairings in adulthood. Thus it appeared that, from adolescence on, females avoided males but were able to do so successfully only when they were placed in the social space first. Because our subjects were resident in the test chamber for only 5 min before the introduction of the stimulus mouse, it is apparent that even very brief occupancy of an environment affects social behavior. This is further supported by data showing that vocalizations positively correlate to active and reciprocal investigation, which depend on the behavior of the subject mouse, and are not correlated with passive social investigation. Thus, the method is socially non-symmetrical, with the subject mouse not ‘equal’ to the stimulus mouse.

Our study also revealed that the type of the interaction (active, passive, reciprocal) is sometimes more informative than is measurement of total physical interaction, which may miss key changes in age-appropriate, sex pair-specific behaviors. For example, although total social behavior in male subject pairs appears similar in adolescence and adulthood, underlying differences in active social behavior in adulthood show qualitative differences between pairs with a male or female stimulus. Indeed, active social behavior in adulthood was highest in pairs with a female stimulus, interpreted as affiliative when the subject was female and reproductive when the subject was male.

The purpose of our initial study was to provide a reference framework of normal social development with which genetic mouse models can be compared. The transgenic mouse model we tested was created to model the increase in striatal D2 receptor activity observed in patients with schizophrenia (Abi-Dargham et al. 2000). We have previously found that D2R-OE mice exhibit behavioral deficits highly relevant to both the cognitive and negative symptoms of schizophrenia (Bach et al. 2008; Drew et al. 2007; Kellendonk et al. 2006; Simpson et al. 2011; Ward et al. 2009, 2012). The negative symptoms of schizophrenia include a flattened affect, deficits in incentive motivation, poor social function and poverty of speech. Studying social behaviors allowed us to determine that increased striatal D2R activity leads to deficits in social functions in addition to the previously documented motivational deficits, which would suggests that these two distinct types of negative symptoms may share an underlying etiologic cause.

For this analysis, we focused on female–female and female–male pairs because these are the social pairs in which we saw the largest changes in social interaction over development. We found that female subjects showed more investigation of female conspecifics than male conspecifics progressively with age. Female subjects were also investigated more by male conspecifics than female conspecifics in adulthood. With D2R-OE and control littermate mice, we saw genotypic differences in social behaviors that were selective for social context and developmental stage. In female–female pairs with a D2R-OE subject, we observed significantly less social interaction in adulthood, which is remarkably similar to the developmental stages at which patients start to show social deficits during a schizophrenia prodromal period and become fully expressed in early adulthood (Insel 2010). Active social behavior (driven by the subject) increases with age in female–female pairings in control subjects, but decreases in female–female pairings with D2R-OE subjects. Therefore, the overall reduction in social behavior is driven by a lack of active behavior on the part of D2R-OE females. These observations illustrate the importance of surveying specific social behaviors in different social contexts and at different developmental stages because we observed that social deficits in the D2R-OE mouse model changed across these factors.

Impoverished speech, inability to understand social cues and deficits in prosody are all features of patients with schizophrenia. Receptive and expressive language deficits may emerge early in children during the schizophrenia prodromal period and impoverished speech production or alogia, is a negative symptom of patients diagnosed with schizophrenia (Cannon et al. 2002; Ziauddeen et al. 2011). Because language deficits in patients likely have a significant impact on their functioning, it is important to also investigate language related behaviors in biological models. USV studies often focus on the number of USVs, however other elements of vocalizations (i.e. pitch, loudness, duration and timing) are extremely important for social interaction and may differ based on age and social context (Lahvis et al. 2011). Consequently, these elemental properties of USVs are likely aspects of vocalization related to function. For example, female–female mouse vocalizations, which we found were briefer with more clearly defined contours than vocalizations in other pairings, are often used affiliatively rather than to determine rank (Moles et al. 2007). These parameters may be related to this functional role.

How USVs are captured, selected and analyzed remains methodologically challenging. In this study, it was not possible to determine the source of the calls. In other studies, the individual source of calls was resolved by playing back vocalizations behind a devocalized mouse (e.g. Hammerschmidt et al. 2009). This procedure was appropriate to observe social approach but could not be adapted for freely interacting pairs of mice. Devocalizing the stimulus mouse or simply playing vocalizations back to the subject mouse each remove important modalities of the social landscape that were of interest. We chose to use automated scoring because hand-selecting calls in the sonogram is both time-consuming and increased the probability that calls containing background noise were selected. However, automated scoring may result in ‘misses’ of calls that are faint or in some way anomalous to the set parameters. Faint calls may occur because of differences in body weight, for example, a common feature of many mutant strains. In this study, it should be noted that female control and D2R-OE mice do not show a genotypic difference in body weight (Simpson et al. 2011), so this factor should not have made mutant calls less likely to be detected. The ability to measure several USV features also presents opportunities and challenges. New critical call features may be discovered but, until these are completely understood and standardized, there exists the danger of alpha error accumulation when multiple aspects of the vocalizations are statistically analyzed. Failure to correct across multiple comparisons could result in type I error but increasing the alpha may result in type II error. It is therefore necessary to increase sample sizes and to test in a way that can be analyzed by repeated measures, although repeating the same social experience at different age points would reduce the novelty of the experience.

Female–female pairs in which a D2R-OE was the subject emitted vocalizations earlier and that were briefer compared to adult female–female control pairs. They also emitted markedly fewer vocalizations in adulthood, but this did not reach significance after the correction for multiple comparisons. Genotypic differences in measures related to prosody (Lahvis et al. 2011) were also observed at most ages in both social pairings. However, as the identity of the mouse in a pair producing USVs cannot be determined using the current methodology, interpretation of USV data is preliminary and suggested for future study. The data clearly demonstrate a change in vocal interaction but do not unambiguously specify the individual contributions to those differences. Across all data with a D2R-OE female subject, the correlation between vocalizations and active social investigation was not lost, but inverted compared to data with a control female subject. This suggests that vocalizations are related to the behavior of the subject, not the stimulus mouse. That USV counts were negatively correlated with active social behavior in female–female pairs with a D2R-OE subject is interesting because USVs usually correlate with investigation (Panksepp et al. 2007; Scattoni et al. 2011). This suggests that D2R-OE females did not vocalize while actively investigating stimulus mice and that vocalizations occurred during passive social behavior when the control stimulus mouse investigates the subject, which seems similar to the impoverished speech production of schizophrenia patients. Thus, the data from the mouse model suggest that the pathophysiology of such deficits in patients may arise from the dysfunction of striatal dopamine D2 signaling.

In sum, our study of D2R-OE mice revealed deficits in social behavior in the female D2R-OE mouse when paired with a female conspecific. The USV results may indicate a deficit in vocal communication if the calls are coming from the mutant or a reflection of a lack of interest in control mice to communicate with the mutant. However, the correlational analysis is suggestive that D2R-OE mice do not call normally during active social investigation. These deficits emerged between adolescence and adulthood, mirroring the developmental trajectory of social deficits in schizophrenia (Tarbox & Pogue-Geile 2008). Overall, our data suggest that in mice, social deficits can be driven by an increase in dopamine D2 receptor activity in the striatum. This may be an indication that the social deficits in patients are related to the well-observed changes in striatal dopamine function (Howes et al. 2012).

The data presented here highlight a critical concept for designing studies of social behaviors in rodent models of pathophysiology. If in our study of the D2R-OE mouse model we had tested only one age or one type of social pairing, the full picture of their social deficit would not have been observed. Instead, by examining same sex and different sex pairs at different developmental time points, we were able to identify social deficits that emerged at a time relevant to the disease model. It now becomes of interest to identify the causal mechanisms by which increased striatal D2 receptor activity leads to these social and communication deficits, so that potential new treatment targets may be identified.

Supplementary Material

Supplemental figure 1

Figure S1: Histograms depicting percentage of vocalizations at intercall intervals ranging from 0 to 800 milliseconds across age, genotype and social pairings.

Supplemental figure 2

Figure S2: Histograms depicting percentage of vocalizations with mean frequency ranging from 40 to 90 kHz across age, genotype and social pairings.

Supplemental figure 3

Figure S3: Histograms depicting percentage of vocalizations with frequency modulation ranging from 0° to 60° across age, genotype and social pairings.

Supplemental table 1

Table S1: Means, SEM, n, and statistics for control pairs across behavior and age.

Supplemental table 2

Table S2: Means, SEM, n, and statistics for control pairs across ultrasonic vocalization measures and age.

Supplemental table 3

Table S3: Means, SEM, n, and statistics for pairs with a D2R-OE vs. control female subject across behavior and age.

Supplemental table 4

Table S4: Means, SEM, n, and statistics for pairs with a D2R-OE vs. control female subject across ultrasonic vocalization measures and age *P < 0.05 after Holm-Sidak correction.