Social behavior was assessed by means of a modified three-chambered social task apparatus (Moy et al., 2004; Nadler et al., 2004). The chamber was an opaque glass rectangular box (75 cm long × 30 cm wide × 35 cm tall) divided into three equal compartments, connected through small rectangular doors (7 cm × 7 cm) allowing free access into each chamber. Different illumination conditions were used for pharmacological (infrared light; Lux < 5) and non-pharmacological experiments (Lux < 30). The procedure involved two phases: habituation and sociability. The test mouse was first placed in the middle chamber and allowed to explore all three chambers for 10 min. After this habituation period, a novel unfamiliar mouse (sex, strain and age-matched) was placed into a mesh cylinder (15 cm tall, 7 cm diameter) in the least explored side chamber, whereas an identical empty mesh cylinder was placed in the opposite chamber. The mesh cylinder allowed for air exchange, visual, olfactory and auditory interaction, but prevented fighting. The test mouse was then allowed to explore the chambers for 10 min (sociability). Measurements during the test phase included: distance traveled, latency to explore side chambers at the beginning of the test, time spent in each chamber and in close proximity to the mesh cylinders (<5 cm). Tracking and scoring was performed using Ethovision XT 10 software (Noldus; RRID:SCR_000441). The social preference index for conspecific chamber time was calculated as follows:

Tc = Time spent in conspecific chamber

To = Time spent in object chamber

whereas the social preference index for time in close interaction was calculated as follows:

Tnc = Time spent in close proximity to the conspecific

Tno = Time spent in close proximity to the object

Object interaction was assessed in an identical apparatus as described above. Under infrared light conditions, mice were first allowed to explore the three chambers of the empty apparatus for 10 min. After this habituation, an empty mesh cylinder was placed in the most preferred side chamber and mice were allowed to freely explore the apparatus for 30 min. Tracking and scoring was performed using Ethovision XT 10 software (Noldus; RRID:SCR_000441). Time spent in the object chamber was scored and expressed as the percentage of total test time.

Anxiety-like behavior was tested using the light-dark test. The apparatus (TSE Systems, Bad Homburg, Germany) consisted of a dark (<10 Lux) “safe” compartment and an illuminated aversive compartment (400 Lux). The compartments were connected by a small opening (7 cm × 7 cm wide) located in the center of the partition at floor level. Animals were individually placed in the apparatus facing the opening to the dark compartment and allowed to freely explore the apparatus for 10 min. Time spent in the light compartment was measured using Ethovision (Noldus). In pharmacological experiments, mice tested 5 min after injection in the three-chambered social task were re-used after 48 h in the light-dark test using a counterbalanced design. Conversely, independent mouse cohorts were used for the 1 h post-injection experiments.

Adult male WT and Grm5-/- mice were single-housed for 72 h prior to the test. Mice from each genotype were divided into three groups (n = 7/group): A group exposed to an empty mesh cylinder (object) in the home cage for 10 min, a second group exposed in the home cage to a mesh cylinder enclosing an unfamiliar sex and age-matched mouse (conspecific) for 10 min and a control group maintained undisturbed in the home cage (HC). After the test, the cylinder was removed and the mice left in their home cages undisturbed. Mice were then perfused 2 h after the end of the experimental manipulation. Tracking and scoring of the time spent in close proximity (<3 cm) with the object or conspecific was performed using the Ethovision XT 10 software (Noldus). Measurement of exploration time of the novel object or novel conspecific was obtained from an independent batch of mice from those used for c-Fos quantification.

Mice were deeply anesthetized by i.p. injection of Thiopental (150 mg/kg) and were perfused with a fixative made of 4% w/v paraformaldehyde and 15% v/v of a saturated solution of picric acid in phosphate buffer (PB) 0.1 M pH 7.4, for 12 min. Brains were removed from the skull, washed in 0.1 M PB and sliced coronally in 50 μm-thick sections on a vibratome (VT1000S, Leica Microsystems, Vienna, Austria).

Immunocytochemistry was performed as previously described (Sreepathi and Ferraguti, 2012). Briefly, free-floating sections were first washed with Tris-buffered saline (TBS; 0.9% NaCl, pH 7.4) and then incubated in 20% normal horse serum in TBS and 0.3% Triton X100 (TX) for 1 h at room temperature (RT, 21–23°C). After blocking, sections were incubated with a polyclonal goat primary antibody against c-Fos (1:300; Santa Cruz Biotechnology, Santa Cruz, CA, United States, catalog #sc-52, lot #F1112) for ∼72 h at 6°C. After three washes in TBS, the biotinylated secondary antibody (horse anti-goat IgG 1:500, Vector Laboratories, Burlingame, CA, United States, catalog #BA-9500,) was applied overnight at 6°C at a dilution of 1:500 in a buffer with the same composition as for the primary antibody. The sections were then washed and incubated in Vectastain elite ABC complex (diluted 1:100; Vector Laboratories) in TBS at RT for 1 h. Subsequently, the sections were washed in TB several times, pre-incubated with 3,3′-diaminobenzidine (DAB; 0.5 mg/ml) for 10 min and then H₂O₂ was added to the solution at a final dilution of 0.003% for 2–5 min. Sections were then washed with TBS, mounted in gelatin onto glass slides, air-dried, and then treated with graded ethanol (50, 70, 90, 95, and 100%) and butyl acetate. Finally, slides were coverslipped with Eukitt (Agar Scientific Ltd., Stansted, United Kingdom).

The following brain structures, relevant for social behavior (Kim et al., 2015), were selected for c-Fos quantification and identified based on the mouse brain atlas of Franklin and Paxinos (2007): medial orbital cortex (MO; bregma between +2,8 and +2,22 mm), prelimbic cortex (PrL; bregma between +2,34 and +1,54 mm), infralimbic cortex (IL; bregma between +1,94 and +1,54 mm), accumbens nucleus shell (AcbSh; bregma between +1,42 and +1,18 mm), accumbens nucleus core (AcbC; bregma between +1,42 and +1,18 mm), lateral septal nucleus intermedial part (LSI; bregma between +0,62 and +0,14 mm), lateral septal nucleus dorsal part (LSD; bregma between +0,62 and +0,14 mm), piriform cortex (Pir; bregma between +0,98 and +0,50 mm), medial septal nucleus (MS; bregma between +0,98 and +0,50 mm), medial preoptic nucleus medial part (MPOM; bregma between +0,0,02 and -0,22 mm), paraventricular thalamic nucleus (PV; bregma between -0,22 and -0,58 mm), paratenial thalamic nucleus (PT; bregma between -0,22 and -0,58 mm), reuniens thalamic nucleus (RE; bregma between -0,46 and -0,70 mm), basolateral amygdala (BLA; bregma between -0,94 and -1,46 mm), dorsal hippocampus (CA1, CA2, CA3, DG, bregma between -1,58 and -1,94 mm), lateral hypothalamic area (LH; bregma between -2,18 and -2,46 mm), posteromedial cortical amygdaloid area (PMCo; bregma between -2,18 and -2,46 mm) and periaqueductal gray (PAG; bregma between -2,92 and -3,16 mm). The number of c-Fos positive cells/area was semi-automatically counted with the Neurolucida software (Version 11, MBF Bioscience, RRID:SCR_001775) coupled to an Olympus BX51 Microscope by an experimenter blinded to the treatment condition and genotype. Each brain area was analyzed bilaterally across at least three sections using a sampling window (200 μm × 200 μm) placed always in the same position within the selected area.

Network analyses were performed as previously described (Tanimizu et al., 2017a). Briefly, Pearson r-values from interregional c-Fos expression data from home cage, object-exposed and conspecific-exposed groups from both genotypes were obtained and used to generate correlation matrices. In order to compare average correlations between groups/genotypes, r-values were transformed to Fischer Z-values, statistics were calculated, and values were retransformed to r values for graph representation. To characterize the generated social and object interaction networks in both genotypes, positive (r > 0.60) interregional c-Fos correlations with a significance level of p < 0.05 were used to generate unweighted network graphs. Community clustering to generate weighted network graphs was performed based on modularity optimization, according to Newman and Girvan (2004). Finally, participation coefficient and within-community z scores were calculated as defined in Guimerà and Amaral, 2005 and plotted as described by Tanimizu et al. (2017a) in order to visualize the main hubs in the generated networks. Interregional correlation matrices were obtained with Prism 7 software (GraphPad Software Inc., RRID:SCR_002798). Network construction and visualization were performed in R (version 3.3.3) using the igraph (version 1.1.2; Csardi and Nepusz, 2006) and brainGraph (version 1.0.0) packages.

Data were analyzed with the Prism 7 software (GraphPad Software Inc.) using two-tailed Student’s t-tests or analysis of variance (one-way or two-way ANOVA, factorial or repeated measures). Whenever an ANOVA resulted significant, Holm–Sidak post hoc comparisons were applied to analyze the effects of group, genotype, treatment, time and chamber in the behavioral experiments. Two-way ANOVA followed by a post hoc Newman–Keuls comparison was used to analyze the effects of genotype and groups in the c-Fos mapping experiment. Two-way ANOVA followed by a post hoc Bonferroni comparison was used to analyze differences in network density. Data were considered significant when p < 0.05.

We investigated the role of mGlu5 receptors in sociability using the classical three-chambered social task apparatus, where sociability is measured as the preference for interacting with an enclosed conspecific placed in one of the side chambers as compared to a novel object (an empty cage) placed in the opposite side chamber. At first, we examined germ-line Grm5-/- mice and compared them to age-matched WT C57BL/6j mice. During the sociability test, both Grm5-/- and WT control mice displayed sociability, spending significantly more time in the social chamber than in the novel object chamber [2-way ANOVA: chamber F(2,90) = 133, p < 0.001; chamber × genotype F(2,90) = 8.06, p < 0.001; novel object vs. novel mouse chamber: WT: p < 0.001; Grm5-/-: p < 0.001] (Figure 1A,B). However, Grm5-/- mice spent less time than WT mice investigating the novel object chamber (p < 0.05) and spent significantly more time in the middle chamber during the test (p < 0.05). Similarly, time in close proximity to the novel mouse was higher than for the novel object for both genotypes [2-way ANOVA: close interaction F(1,60) = 49.7 p < 0.001, and genotype F(1,60) = 6.97, p < 0.05] (Figure 1C), whereas the overall time in close interaction with the conspecific or object did not differ between genotypes [2-way ANOVA: close interaction × genotype F(1,60) = 2.52 p = 0.11]. Grm5-/- mice showed a higher social preference index for both conspecific chamber time [t-test: t(1,30) = 2.18, p < 0.05] (Figure 1D) and time in close interaction with the conspecific [t-test: t(1,30) = 3.178, p < 0.01] (Figure 1E) when compared to control mice. During the test we observed that the total distance traveled by Grm5-/- mice was significantly lower than the control animals [t-test: t(1,30) = 2.32, p < 0.05] (Figure 1F). Since Grm5-/- mice are known to display normal locomotion (Chiamulera et al., 2001), this effect could be attributed to a reduced exploratory activity. Unlike WT control, Grm5-/- mice explored more actively the novel conspecific during the last 5 min of the test [0–5 vs. 5–10 min: WT, paired t-test: t(1,15) = 0.25, p = 0.79; Grm5-/-: t(1,15) = 2,68, p < 0.05] (Figure 1G,H). Grm5-/- mice also exhibited a longer latency to explore the side chambers of the apparatus at the beginning of the test [t-test: t(1,30) = 4.39, p < 0.001] (Figure 1I). These findings suggest that gene-targeted deletion of Grm5 leads to enhanced social interaction, as measured by the higher social preference index, but also to a reduced exploratory activity or enhanced anxiety as suggested by the reduced distance traveled and high latency to explore the side chambers. To assess whether the lower time spent in the object chamber was due to the putative anxiogenic phenotype or an exploration deficit, we performed an additional experiment in which mice were presented only with the novel object (i.e., the empty enclosure), while the opposite chamber of the apparatus was left empty. In the first 5 min of a 30 min session, Grm5-/- indeed explored significantly less the object chamber than WT mice [2-way repeated measures ANOVA: time F(5,85) = 0.36; genotype F(1,17) = 3.66; time × genotype F(5,85) = 7.38 p < 0.001; 5 min: WT vs. Grm5-/-: p < 0.001] (Figure 2A). Conversely, in the remaining time of the session the two genotypes showed no difference in time spent in the object chamber (Figure 2A), therefore, showing no generalized deficit in exploration. These findings strongly suggest that Grm5-/- mice have an axiogenic phenotype.

Thus, we specifically tested these mice for measures of anxiety-like behavior using the light-dark-box test. Grm5-/- mice spent indeed significantly less time in the bright side of the box as compared to WT mice [t-test: t(1,22) = 2.42, p < 0.05] (Figure 2B).

Taken together, these results indicate that Grm5-/- mice show an anxiogenic phenotype although with a paradoxically enhanced sociability as measured with social ratios.

We next assessed the effects of mGlu5 receptor negative allosteric modulation in WT C57BL/6J mice in sociability and anxiety. We assessed the effects of MTEP (10 mg/kg) at two different time points, when MTEP receptor occupancy should be at its peak (5–15 min post-i.p. injection) and when it should have returned at baseline levels (1 h post-i.p. injection) (Anderson et al., 2003). Vehicle- and MTEP-treated animals displayed sociability, spending more time in the conspecific chamber than in the novel object chamber both at 5 min [2-way ANOVA: chamber F(2,72) = 233, p < 0.001; treatment F(1,72) = 0.002, p > 0.05; and chamber × treatment F(2,72) = 4.58, p < 0.05; object chamber vs. conspecific chamber: WT p < 0.001, Grm5-/-p < 0.001] and 1 h post-injection [2-way ANOVA: chamber F(2,66) = 196.8, p < 0.001; treatment F(1,66) = 0.001, p > 0.05; chamber × treatment F(2,66) = 0.59, p > 0.05] (Figure 3A,G). MTEP had no effects on the overall time spent in the novel conspecific or novel object chamber either at 5 min (WT vs. Grm5-/-: p > 0.05) or at 1 h post-injection (WT vs. Grm5-/-: p > 0.05) (Figure 3A,G). However, MTEP reduced the amount of time spent interacting closely with the conspecific and object in a non-specific manner at 5 min [2-way ANOVA: treatment F(1,48) = 13.6, p < 0.05; close interaction F(1,48) = 84.9, p < 0.05; treatment × close interaction F(1,48) = 1,93, p = 0.17] (Figure 3B) and at 1 h post-injection [2-way ANOVA: treatment F(1,44) = 6.14, p < 0.05; close interaction F(1,44) = 54.8, p < 0.05; treatment × close interaction F(1,44) = 0.53, p = 0.47] (Figure 3H). Social preference indexes for both conspecific chamber time and time in close interaction with the conspecific did not differ between vehicle and MTEP-treated animals both at 5 min [chamber: t-test: t(1,24) = 0.102, p > 0.05; close interaction: t-test: t(1,24) = 0.312, p > 0.05] (Figure 3C,D) and 1 h post-injection [chamber: t-test: t(1,22) = 0.95, p > 0.05; close interaction: t-test: t(1,22) = 0.36, p > 0.05] (Figure 3I,J). These findings indicate that the interaction with a conspecific is not altered by acute MTEP treatment. The lowered active exploration of the conspecific in mice treated with MTEP was accompanied by a marked decrease in object exploration and an increase in locomotor activity during the test, as shown by the distance traveled during the test at 5 min post-injection [t-test: t(1,24) = 4.56, p < 0.001] (Figure 3E) as well as by a statistical trend toward significance at 1 h post injection [t-test: t(1,22) = 1.77, p = 0.09] (Figure 3K).

We then tested whether the reduced active exploration of the object and conspecific during the test was due to the proposed anxiolytic activity of MTEP (Klodzinska et al., 2004; Stachowicz et al., 2007; Lee et al., 2017, 2018; but see Ade et al., 2016). Similar to sociability, we tested the effects of MTEP at both 5 and 1 h post-injection in the light-dark test. MTEP-treated animals spent significantly more time in the bright side of the box, both at 5 min [t(1,24) = 2.55, p < 0.05] (Figure 3F) and at 1 h post-injection [t(1,26) = 2.08, p < 0.05] (Figure 3L). Taken together, these results confirm the anxiolytic action of MTEP and suggest that mGlu5 receptor NAM does not influence sociability in WT mice when administered acutely.

In order to investigate whether the increased sociability observed in Grm5-/- mice could result from different patterns of brain activation upon exposure to social and non-social cues, we exposed WT and Grm5-/- mice in their home cage to either a novel conspecific or novel object and quantified the expression of the immediate early gene c-Fos. We used three independent experimental groups for each genotype: a group exposed to an age-matched conspecific enclosed into a wire mesh cage (conspecific group), a group exposed only to the wire mesh cage (object group), and a third control group left undisturbed in the home cage (HC). Similar to the three-chambered social task, Grm5-/- mice displayed a reduced exploration of the novel object in comparison to WT [2-way ANOVA: cue F(1,22) = 6.05; genotype F(1,22) = 2.45; cue × genotype F(1,22) = 4.80, p < 0.05; WT vs. Grm5-/- object: p < 0.05] and a similar exploration of the novel conspecific (Figure 2C). Two hours after the exposure, mice were perfused and processed for immunocytochemistry. Twenty-one brain regions, previously reported to be activated after social interaction (Kim et al., 2015), were preselected for c-Fos analysis. In mice kept undisturbed in their home cage, no significant differences between WT and Grm5-/- mice were detected in the number of c-Fos+ neurons in any of the areas analyzed (see Table 1 for statistical significance, Figure 4 and Supplementary Figure S1). This suggests that under resting conditions basal activity in the set of brain areas that we have analyzed is not altered by the lack of mGlu5 receptors. Conversely, compared to the HC group a significant increase in the number of c-Fos+ cells was observed in most of the areas analyzed with the exception of AcbC, LSI, CA1, CA2, and PMCo in WT mice after exposure to a novel object; AcbC, LSI, CA2, and PMCo in WT mice after exposure to a conspecific and AcbC, LSI, CA1, and PMCo in Grm5-/-mice after exposure to a novel object (see Table 1).

A significant group × genotype interaction in the 2-way ANOVAs was observed only in 4 brain areas, namely IL, MO, PrL, and LSD (see Table 1 and Figure 4). The interaction with a conspecific triggered higher activation as compared with the object in Grm5-/- mice, whereas no statistical significant differences were observed between object and conspecific exposed-groups in WT mice. These four brain areas had also a higher number of c-Fos+ cells in Grm5-/- mice as compared to WT mice when exposed to the novel conspecific, but not to the novel object (see Table 1 and Figure 4). The arousal produced by the exposure to the novel object or conspecific in our paradigm may have masked in WT mice distinct pattern of c-Fos activation. On the other hand, the lack of mGlu5 receptors was able to induce region-specific changes in the number of c-Fos+ cells specifically related to social interaction. We, thus, reasoned that social interaction, contrary to non-social, rather than producing a higher degree of activation, namely number of c-Fos+ cells per area, it enhances functional connectivity among a set of brain regions. In addition, the high activation of prefrontal areas and LSD observed in Grm5-/- mice, could underlie a disrupted activity coordination. To explore this possibility, we analyzed the functional connectivity generated during social or object investigation in WT and Grm5-/- mice.

In order to infer interactions between neural elements, we computed correlation coefficients across subjects using our c-Fos expression data set (Horwitz et al., 1995; Tanimizu et al., 2017a,b; Rogers-Carter et al., 2018). This allowed us to obtain an approximation of the strength of the coordinated activity changes among brain regions following social and non-social interactions in Grm5-/- and WT mice. We first computed inter-regional correlations for each experimental group (Figure 5A). As shown in Figure 5B, changes in network density upon social investigation were observed both in WT and Grm5-/- mice [2-way ANOVA: genotype F(1,836) = 6.50, p < 0.05; group F(1,836) = 6.66, p < 0.01; group × genotype F(1,836) = 47.51, p < 0.001]. A higher functional connectivity (mean r) was observed upon conspecific as compared to object interaction in WT mice (p < 0.05). Conversely, in Grm5-/- mice functional connectivity was higher upon non-social interaction (p < 0.001). Interaction with a conspecific led to higher functional connectivity in WT mice as compared to Grm5-/- mice (p < 0.05).

These results suggest that in WT mice social investigation leads to a higher functional connectivity than upon interaction with an object. On the contrary, the lack of mGlu5 receptors inverts the strength of the functional connectivity toward non-social interaction.

Based on interregional matrices (Figure 5A), we generated network graphs (Figure 6A), where nodes represent brain regions and edges represent above-threshold (Pearson’s r > 0.6) significant (p < 0.05) correlations (Figure 6A). We further applied graph theoretical analysis to the network graphs to explore if social investigation induces changes to the relative weight of any of the identified nodes and whether they are affected by the lack of mGlu5 receptors (Figure 6B). We then computed the within-community z score and participation coefficient for each node (Figure 6C). In correlation-based networks, the within-community z score measures how well connected a region is to its own community and the participation coefficient of a node constitutes a measure of the degree to which a node is linked with nodes in other communities (Guimerà and Amaral, 2005). The participation coefficient, therefore, denotes “hubness” (Power et al., 2013; Rogers-Carter et al., 2018). Nodes displaying a high value of within-community z score and participation coefficient are thought to be key hubs, crucial for coordinating other nodes and, thus, the overall activity of the network (Tanimizu et al., 2017a). Our analysis identified the RE, AcbSh, PAG, and CA3 as key hubs in the social interaction functional network and the PT in the object functional network in WT mice. In addition, the hippocampal regions CA1, CA2, and DG as well as the IL appear as key regions in coordinating the activity between communities in the social interaction network given their high participation coefficient. Of note is the differential participation of the PrL and IL in the object and conspecific functional networks, respectively. In Grm5-/- mice, we observed a complete derangement in the role of nodes in both the non-social and social functional networks. In these animals, while the hippocampus and mPFC lost their coordinating role in conspecific network activity, the lateral septum increased it. Moreover, the PT transferred its role as a key hub from the non-social to the social functional network.