Background
Social impairment is one of the two core symptom domains in autism spectrum disorder (ASD) [
1], but we have limited understanding of the mechanisms underlying this impairment. One way to probe the molecular underpinnings of social difficulties in ASD is to study single gene syndromes that frequently include prominent ASD symptoms, such as fragile X syndrome (FXS), which is found in about 1% of children with ASD [
2‐
4]. Between one quarter and two thirds of children with FXS meet criteria for ASD, depending upon the study [
5‐
15]. In some but not all studies, the average child with FXS and autism spectrum disorder (FXS-ASD) has fewer ritualistic behaviors and perhaps fewer social deficits than the average child with autism spectrum disorder without FXS [
10,
16‐
20], but these differences are not large enough to identify a behaviorally defined subgroup without genetic testing.
FXS is caused by expansion of a trinucleotide repeat in the 5′ untranslated region of the
FMR1 gene leading to gene hypermethylation and transcriptional silencing.
Fmr1 null mice have been a primary tool to investigate the cellular, synaptic, and molecular changes underlying FXS [
21]. These mice display behavioral, structural, and molecular abnormalities that mirror some of those seen in FXS [
22]. While the preponderance of evidence suggests that
Fmr1 null mice do have changes in social behavior compared to wildtype controls, results have been inconsistent across labs and appear to be dependent on the inbred strain background of the mice being studied, with abnormalities more frequently reported on an FVB/129S mixed background [
22‐
25]. The most common finding is a loss of preference for a novel mouse over a familiar mouse, but some laboratories have also reported that
Fmr1 mice fail to prefer a novel mouse over a novel object, show diminished free social interactions, or are more likely to back out of the tube test for dominance [
23‐
43]. In addition to abnormal social behavior,
Fmr1 null mice have been reported to show several other behavioral deficits, including hyperactivity, attention deficits, learning deficits, and abnormal reactions to sensory stimuli [
25,
44‐
47], although these behavioral changes have also been inconsistent across labs.
Many studies have focused on the molecular consequences of
FMR1 loss with the goal of developing an effective treatment for FXS. Considerable data support a critical role for the gene’s protein product, FMRP, downstream of group 1 metabotropic glutamate receptor (mGluR) signaling, leading to the mGluR theory of fragile X [
48]. Although improvement of behavioral and synaptic abnormalities was observed in animal models following treatment with mGluR5 negative allosteric modulators [
49,
50], clinical trials have not yet yielded benefit in adolescents or adults with FXS [
51]. Other potential treatments have also been studied, including lithium [
52], minocycline [
53], and arbaclofen [
54], but no treatment has yet shown consistent improvement in behavior in the human population.
Despite extensive characterization of the Fmr1 null mouse, the role of FMRP in mediating social behavior remains relatively unknown. The aim of the current investigation was to use the Fmr1 (FVB/129) null mouse as a tool to further investigate molecular changes that may impact social behavior. We designed a series of experiments to understand the brain areas involved in response to social stimuli in these animals and to identify changes in levels of RNA expression that may identify molecular pathways altered in this model. Beyond insight into potential mechanisms underlying altered social behavior in the Fmr1 null mouse, we hoped to use this well-studied model as a proof-of-concept of this use of social stimuli to identify genes and pathways that can be cross-validated with other data from FXS and ASD.
Discussion
Using the Fmr1 null mouse as a model of one of the most common ASD-associated syndromes, we were able to identify changes in immediate early gene activation and gene expression following the presentation of a novel mouse, in comparison to a novel object. This may provide insights into pathways and networks underlying altered social response in the Fmr1 null mice and potentially in people with FXS. Further, the strong overlap with SFARI genes suggests that a social stimulus can be used to provoke changes in gene expression that correspond with ASD-relevant brain regions and molecular pathways within a genetic mouse model of ASD.
Our initial immunohistochemistry in wildtype C57BL/6J mice revealed the LA and MA as having the greatest degree of differential activation, as evidenced by c-Fos levels, in response to a social stimulus compared to a non-social stimulus, while similar findings in
Fmr1 null mice also showed differential activation in the PLC. These findings are in line with recent work by Ferri and colleagues [
59], who demonstrated greater levels of immediate early gene products c-Fos and Egr-1 in the BLA of juvenile C57BL/6J mice in response to a social stimulus, as compared with either a non-social stimulus or no stimulus exposure (home cage controls). In contrast, however, they did not observe an increase in c-Fos in response to a social stimulus compared to non-social stimulus in the MA; both the social and non-social stimulus conditions were elevated above the level of home cage controls for the MA. They propose that the MA is more likely to be responsive to various social stimuli after adolescence, which could explain the difference in findings between our studies. In addition, they found only a trend toward a significant difference in c-Fos cell counts in response to stimuli in the PLC, as in our results, though we also found a significant effect of stimulus on staining intensity, and the direction of the effect was the same in both studies, with the social stimulus exposure corresponding to lower c-Fos levels than non-stimulus exposure in the PLC.
In the
Fmr1 null animals, we found hyperactivation of the amygdala of
Fmr1 null animals following the presentation of a novel mouse. These findings suggest that the activation is dysregulated by the deletion of
Fmr1 and support human imaging studies that associate the amygdala with social response in FXS. For example, Watson and colleagues [
71] noted increased activation of the left amygdala in response to successive eye gaze exposures in children with FXS compared to controls. Structural imaging suggests a reduction in amygdala volume in FXS [
72,
73], and Hoeft and colleagues [
74] found that decreased amygdala volume in FXS patients is correlated with altered activation patterns. Recent imaging results in the
Fmr1 null mouse model on the FVB background showed no significant change in the frontal cortex or amygdala volumes, with quite subtle changes overall including primarily white matter, as well as a small increase in ventricle size; although the resolution of MRI for individual brain structures is substantially reduced simply as a result of mouse brain size [
75,
76]. The hypoactivation of the PLC in response to the social stimulus was significant for c-Fos intensity but not cell count, but the direction of the effects observed in both measures corresponds with human imaging findings. An fMRI study in children with FXS revealed decreased PFC activation in response to direct eye gaze in comparison to age-matched, typically developing children [
71]. Decreased gray matter in the mPFC has also been reported in children with FXS [
74].
Our immediate early gene findings also overlap with the broader literature on neural mechanisms of social behavior. For instance, a wide array of data support a role for the medial amygdala in social recognition [
77], aggression [
78,
79], and mating behavior [
80] in mice. The positive correlation between lateral and medial amygdala c-Fos levels in response to a social stimulus align with recent work demonstrating the importance of the intra-amygdalar pathway in processing social behavior and formulating an appropriate response in the context of social fear learning [
81]. The high, inverse correlation between amygdala and PLC immediate early gene activation aligns with previous literature suggesting that these two regions communicate during social behavior. The involvement of prefrontal-limbic circuitry, which includes both PFC and the amygdala, has been suggested in several social behaviors in humans including social anxiety [
82], social phobia [
83], deception [
84], social cooperation [
85], and empathy [
86]. Swartz and colleagues [
87] also found a decrease in connectivity between the PFC and amygdala while viewing sad faces as measured by functional magnetic resonance imaging (fMRI). Together with the connectivity and correlational findings between the PLC and amygdala, the findings of lower activation of PLC and higher activation of the amygdala activity in response to a social stimulus suggests dysregulation of at least one node within this circuit in
Fmr1 null mice. This dysregulation in response to a social stimulus may be involved in the differences in social behavior observed in this mouse model of FXS.
The patterns of differentially expressed genes in the LA, MA, and PFC also show convergence with previous data. It is reassuring that this convergence includes highly significant overlap for genes bound by FMRP. Further, it is interesting that this overlap is primarily in the mutant versus wildtype comparison in LA after exposure to the non-social stimulus and in PFC after exposure to either the social or non-social stimulus. The absence of significant overlap with FMRP-bound genes in the genotype comparison in the MA is surprising and could indicate that this is a brain region where FMRP plays less of a direct role in regulating gene expression. Previous work in
Fmr1 null mice has identified changes in electrophysiology and dendritic spine morphology in the basolateral amygdala and the medial prefrontal cortex [
88‐
92]; whereas previous studies have not examined the MA in these animals. These data suggest that the MA could be a brain region that lacks substantial baseline changes due to the direct effects of
Fmr1 deletion but may still show dysregulated activity within a broader circuit, but further work is necessary to examine this possibility. Interestingly, in addition to genotype-driven differences, we also saw a significant enrichment of FMRP-bound genes in the social versus non-social comparison in wildtype animals in both the MA and the PFC, although we did not observe a significant interaction between stimulus and genotype. This suggests that FMRP-bound genes may play an important role in response to a social stimulus, aligning with previous data implicating FMRP as a central hub that regulates genes implicated in ASD [
93].
In addition to FMRP-bound genes, which could be expected to show differences in
Fmr1 null mice, we also observed substantial overlap of DEGs with genes implicated in ASD. Our primary analysis for this comparison was the genotype × condition interaction, which showed a significant enrichment for SFARI ASD risk genes for the LA and MA, but only a trend for the PFC. This suggests that a social stimulus may be useful to focus further study on DEGs that are particularly relevant to ASD. Interestingly,
Th, encoding tyrosine hydroxylase, was one of three genes that was differentially expressed in the genotype × stimulus interaction in all three brain regions. Tyrosine hydroxylase is the rate-limiting enzyme in dopamine synthesis, and the dopamine system has been implicated in FXS, in ASD, and in social response in general. For example, loss of FMRP alters signaling downstream of the dopamine D
1 receptor [
94]. Recent work also shows increased dopamine levels in the striatum of
Fmr1 null mice [
27]. De novo and inherited single nucleotide variants have been detected in a number of genes in the dopamine system in ASD. Elegant work using optogenetic and pharmacogenetic approaches has demonstrated that dopaminergic signaling is critical for a number of social functions, including social approach, social reward, and response to social deprivation [
95‐
99]. Beyond the significant interactions, the social versus non-social contrast showed highly significant overlap with the SFARI gene set for
Fmr1 null animals in all three brain regions, further supporting the use of a social stimulus to reveal ASD-relevant DEGs in a genetic mouse model.
Network analysis also pointed toward convergence with previous data, with DEGs clustering in networks that are associated with nervous system development. In addition to further validating the model and social stimulus procedure as producing gene expression differences in pathways relevant for ASD, the analyses may provide focus for future research and potential treatments. By identifying networks in specific brain regions that are differentially regulated in
Fmr1 null mice, we can consider different ways to probe or manipulate the pathway to potentially identify a time during development or adulthood in which perturbation of the dysregulated pathway may rescue specific phenotypes. Additionally, the networks identified may provide specific molecules to consider targeting for treatment, including non-DEGs such as NFκB that occupy a position as molecular hubs interacting with multiple DEGs (Figs.
7 and
9). Finally, the upstream regulator analyses may highlight potential treatments that could reverse the observed pattern of differential gene expression, such as beta-estradiol, which was recently shown to rescue behavioral phenotypes in zebrafish lacking
Cntnap2, an ASD-risk gene [
100]. Most importantly, these pathway and network findings direct us to specific aspects of development, for example axonal guidance signaling or immune response pathways that will be beneficial to examine in
Fmr1 null mice, particularly in relation to their social behavior.
These findings also have some limitations, including the moderately high degree of variability within the RNA sequencing data. Of the three brain areas, biological replicates had a high degree of convergence with the exception of one sample within the medial amygdala that we excluded from all analyses. Additionally, our RNA sequencing approach was focused primarily on pathway and network analyses, with few individual genes reaching significance after correction for multiple testing. We set an α level of 0.05 as a threshold for these analyses to balance a desire for specificity with a desire to eliminate false negatives, which could obscure convergence across multiple DEGs. Adjusting the threshold in either direction could yield different findings. Similarly, application of the Benjamini-Hochberg correction at the level of 0.05 for canonical pathway analysis resulted in no enriched pathways in the LA and PFC passing the threshold for significance. These individual DEGs and pathways are therefore exploratory and should be replicated before targeting specific, individual genes.
Another factor to consider is the introduction of the stimulus into the homecage. We minimized the potential for aggression by only housing the test animal in the new cage for 3 days, and containing the stimulus in the pencil cup, but it is possible that the reaction to the stimulus was influenced by territoriality in a way that might be lessened (or increased) by other circumstances.
Finally, the statistical differences between measurements of c-Fos intensity and cell counts should be considered. The effect of the interaction between genotype and stimulus was only significant in c-Fos intensity, although virtually all findings were in the same relative directions between analyses. The significant interaction effect on intensity was due to a larger effect size, and it is possible that this measure is more sensitive to biological variation than cell count. For example, the level of c-Fos expression may vary between cells, leading to cell-specific transcriptional responses. Intensity measurement has the added advantage of being more objective, not requiring an experimenter to apply a threshold as in the cell count. The meaning of each measurement and resulting findings should be considered, but one is not necessarily preferable to the other.