Background

Autism spectrum disorder (ASD) is a complex neurodevelopmental disorder that is characterized by two core behavioral symptoms: deficits in social communication and social interaction, and restricted, repetitive patterns of behaviors, interests, or activities [1]. Despite its high clinical and etiological heterogeneity, recent genetic studies have revealed that ASD appears to share neurobiological mechanisms and be attributable to the dysregulation of some convergent pathways [2,3,4,5]. One such pathway is the mechanistic target of rapamycin complex 1 (mTORC1) signaling cascade [6,7,8,9].

Tuberous sclerosis complex (TSC), a prototypical mTORopathy, is an autosomal dominant disorder. Up to 50% of patients with TSC have comorbid ASD. The prevalence of TSC in overall ASD was estimated to be 0.01 [10,11,12]. Thus, TSC is one of the most common diseases that is responsible for syndromic ASD. Tuberous sclerosis complex is considered to be caused by pathogenic mutations in the TSC complex subunit 1 (TSC1) or TSC2 gene and is unique in the similarity of clinical phenotypes between pathogenic mutations in the two genes [13,14,15]. TSC1 and TSC2 proteins form a complex with complemental TBC domain family member 7, which suppresses a small guanosine triphosphate-binding protein, Rheb, and negatively regulates the mTORC1 signaling cascade [16]. The excessive activation of mTORC1 signaling that is caused by pathogenic mutations in the two genes is assumed to be the primary pathophysiological mechanism of TSC [17]. According to Knudson’s two-hit theory, TSC-associated tumors develop because of biallelic inactivation of the TSC1 or TSC2 gene [18]. Indeed, the loss of heterozygosity (LOH) of either gene is often detected in most TSC-associated tumors, and more than half of brain cortical tubers lack LOH [19,20,21], suggesting that the heterozygosity of TSC1 or TSC2 is sufficient to cause neuronal abnormalities [22].

Genotype–phenotype correlation studies suggest that TSC2 pathogenic mutations rather than TSC1 pathogenic mutations are associated with severe phenotypic manifestations, including early-onset seizures, a high rate of complications that are associated with ASD, low cognitive function, and a high number of renal angiomyolipomas [15, 23,24,25,26]. Somatic TSC1 mutations were identified in lesions from patients with a pathogenic germline TSC2 mutation, suggesting that TSC1/TSC2 double heterozygous mutations may promote tumorigenesis [21, 27, 28]. Therefore, evaluations of manifestations of double heterozygous pathogenic mutations compared with mutations of either gene alone will provide insights into the pathophysiology of TSC. However, previous studies exclusively focused on pathogenic mutations of either gene alone, and few studies have described TSC1/TSC2 double mutations [29,30,31].

The homozygous loss of either Tsc1 or Tsc2 in mice causes embryonic lethality [32, 33]. Conventional heterozygous mice recapitulate some neuropsychiatric phenotypes of TSC in humans, including learning and memory deficits and social impairments without apparent brain lesions [34,35,36]. Additionally, transient treatment with an mTORC1 inhibitor successfully rescued these deficits.

The present study focused on the impact of Tsc1 and Tsc2 double heterozygous mutations in mouse models. We analyzed social behaviors and gene expression profiles in Tsc1 and Tsc2 double heterozygous mice and assessed the therapeutic effects of the mTOR inhibitor rapamycin on autistic-like behaviors.

Materials and methods

Mice

Tsc1 heterozygous knockout (Tsc1+/−) and Tsc2 heterozygous knockout (Tsc2+/−) mice were generated and maintained on a C57BL6/J background as previously described [32, 33]. A Tsc1+/− mouse and a Tsc2+/− mouse were crossed to produce Tsc1+/+/Tsc2+/+ (WT) mice, Tsc1+/−/Tsc2+/+ (Tsc1+/−) mice, Tsc1+/+/Tsc2+/− (Tsc2+/−) mice, and Tsc1+/−/Tsc2+/− (TscD+/−) mice. C57BL/6J mice were purchased from CLEA Japan (Tokyo, Japan). After weaning at 3 weeks of age, all mice were maintained separately by sex (two to five mice per cage) in ventilated racks at 22 °C ± 2 °C under a 12 h/12 h light/dark cycle (lights on 8:00 A.M. to 8:00 P.M.) with access to a standard laboratory diet and water ad libitum.

Drug

Rapamycin (LC Laboratories, Woburn, MA, USA) was dissolved to 0.5 mg ml−1 with 10% dimethylsulfoxide (vehicle) and administered intraperitoneally in a volume of 10 ml kg−1 once daily for 2 days. The behavioral tests and brain collection for gene expression analysis began 24 h after the second rapamycin administration.

Behavioral tests

The hole board test, self-grooming test, social interaction test, three-chambered sociability test, elevated plus maze test, and social transmission of food preference test were performed with male and female mice at 14.4–21.6, 44.4–46.9, 17.1–43.4, 15.6–43.4, 15.6–26.1, and 16.0–21.7 weeks of age between 10:00 A.M. and 4:00 P.M., respectively. The number of WT, Tsc1+/−, Tsc2+/−, and TscD+/− mice that were used in the behavioral tests were 13–19, 14–20, 15–20, and 14–22, respectively.

Hole board test Each mouse was placed in a hole board apparatus (500 mm × 500 mm × 400 mm; Muromachi Kikai, Tokyo, Japan) with four holes (38 mm diameter) and allowed to freely explore it for 30 min. Behavior was automatically analyzed for counts of head dipping using a video tracking system.

Self-grooming test Each mouse was individually placed in a new cage without bedding material. After an initial 10-min habituation period, the animal was video-recorded for 10 min. The number of episodes of self-grooming and rearing behavior was measured.

Social interaction test Each mouse was left alone in its home cage for 15 min. A novel mouse (C57BL/6J) of the same sex and size was then introduced to the cage. The behavior of the test mouse was video recorded for 10 min. The total duration of active interaction (anogenital sniffing, allogrooming, close following, and mounting) was measured.

Three-chambered sociability tests A three-chamber arena (500 mm × 500 mm × 400 mm) was used to evaluate social approach, preference for social novelty, and social preference. After a 10 min habituation period, an unfamiliar C57BL/6J mouse of the same sex (stranger 1) was introduced to the columnar wire cage (100 mm diameter, 100 mm height) in a side chamber, and a metal block (object) was placed in another cage on the other side chamber. The test mouse was allowed to explore the entire arena for 10 min (Session 1: social approach). The block was then replaced with another unfamiliar mouse (stranger 2), and the test mouse was examined for an additional 10 min (Session 2: social novelty). The test mouse was then presented with another unfamiliar mouse (stranger 3) and a cagemate of the test mouse for 10 min (Session 3: social preference). The amount of time spent around each cage was automatically measured using a video tracking system. The approach-avoidance score was calculated as the following: (time exploring stranger 3) − (time exploring cagemate).

Elevated plus maze test Each mouse was individually placed in the center of the elevated plus maze apparatus (Muromachi Kikai) and allowed to freely explore the apparatus for 10 min. Total distance traveled and time spent on the open arms were automatically measured using a DV-Track Video Tracking System (Muromachi Kikai).

Social transmission of food preference test Twenty-four hours before testing, the test mouse was placed in a novel cage with a specially designed food jar that was filled with powdered chow. Eighteen hours before testing, the feeding jars were removed, and the mice were deprived of food until testing. To begin testing, a “demonstrator” mouse was given a food jar that was filled with powdered chow that was flavored by mixing it with either 1% ground cinnamon or 2% powdered cocoa. After 1 h, the food jar was removed and weighed to ensure that at least 0.2 g of food had been consumed. The demonstrator mouse was then immediately placed in the cage of the “observer” test mouse, and interactions between the demonstrator mouse and observer mouse were allowed for 30 min. The demonstrator mouse was then removed, and two jars of powdered food (one of each flavor) were placed in the cage with the observer mouse for 1 h. At the end of the 1 h choice session, the food jars were removed and weighed. The demonstrator mice were C57BL/6J mice that were matched to the test observer mouse according to weight and sex.

Whole transcriptome analysis

Total RNA was isolated from the right brain hemisphere of 12-week-old male mice using Sepasol-RNA I Super G (nacalai tesque, Kyoto, Japan) and purified with RNeasy Mini (Qiagen, Tokyo, Japan). Only RNA with an RNA Integrity Number above 8.0, determined by an Agilent Bioanalyzer (Agilent Technologies, Tokyo, Japan), was used for gene expression analysis. Cy3-labeled cRNA was prepared using the Low Input Quick Amp Labeling Kit (Agilent Technologies) according to the manufacturer’s protocol, hybridized to the SurePrint G3 Mouse Gene Expression v2 8 × 60 K Microarray (Agilent Technologies), and scanned using a SureScan Microarray Scanner (Agilent Technologies).

Microarray data were measured using Feature Extraction software (Agilent Technologies) and analyzed using GeneSpring GX software (Agilent Technologies). Differentially expressed transcripts (DETs) were identified by two-way analysis of variance (ANOVA) with p < 0.05. Common transcripts between more than two DETs were selected instead of multiple testing correction to minimize false positives. Clustering analysis was performed using hierarchical clustering analysis with the default Euclidean distance metric and default Ward’s linkage rule.

Gene ontology and pathway enrichment analyses

Gene ontology (GO) and pathway enrichment analyses were performed using MetaCore version 6.31 (https://portal.genego.com/cgi/data_manager.cgi; accessed July 22, 2022). Map Folders and GO Process Networks are mainly based on biochemical and signaling cascades and cellular processes, respectively. The statistical significance value and false discovery rate (FDR) were used to rank GO. Canonical pathway modeling was used to identify pathways with statistical significance, zScore, and gScore by considering the number of objects of datasets and canonical pathways in the networks.

Statistical analysis

The statistical analyses of behavioral data were performed using SPSS Statistics 24 software (IBM Japan Ltd., Tokyo, Japan). Motor, sensory, and social behaviors between mutant and WT mice were statistically evaluated using one-way ANOVA followed by the Tukey Honestly Significant Difference (HSD) test. Unpaired t-tests were used to evaluate differences between vehicle and rapamycin treatment in the social interaction test. Paired t-tests were used to evaluate differences between familiar and stranger mice in the three-chambered sociability tests and between cued food and non-cued food in the social transmission of food preference test. Values of p < 0.05 were considered statistically significant.

Results

Autistic-like behaviors in Tsc1+/−, Tsc2+/−, and TscD+/− mice

The homozygous knockout of either the Tsc1 or Tsc2 gene causes embryonic lethality in mice, whereas both Tsc1+/− and Tsc2+/− mice are apparently healthy, do not die prematurely, and lack apparent cerebral lesions and spontaneous seizures [32, 33]. TscD+/− mice were born at the expected Mendelian frequency and exhibited no apparent spontaneous seizures, no early death, and no delay of physical development (data not shown). Adult Tsc1+/− and Tsc2+/− mice have been reported to exhibit normal motor and sensory behaviors [36]. Similar to Tsc1+/− and Tsc2+/− mice, TscD+/− mice did not display any significant differences in motor or sensory behaviors in the hole board test (Fig. 1A), elevated plus maze test, or self-grooming test compared with WT mice (Additional file 1: Fig. S1A–C). TscD+/− mice normally discriminated the scent of a food that was presented by stranger mice in the social transmission of food preference test, similar to Tsc1+/− and Tsc2+/− mice, indicating that these mutant mice can differentiate novel nonsocial stimuli that are transmitted through social interaction (Additional file 1: Fig. S1D). TscD+/− mice exhibited more frequent self-grooming behavior than WT mice, similar to Tsc1+/− and Tsc2+/− mice (p = 0.024, 0.031, and 0.042 for Tsc1+/−, Tsc2+/−, and TscD+/− mice, respectively, Tukey HSD test; Fig. 1B). TscD+/− mice also exhibited a reduction in interaction time compared with WT mice, similar to Tsc1+/− and Tsc2+/− mice, in the social interaction test (p = 0.001, 0.004, and 0.004 for Tsc1+/−, Tsc2+/−, and TscD+/− mice, respectively, Tukey HSD test; Fig. 1C). In the social approach session of the three-chambered sociability test, TscD+/− mice exhibited normal social preference, with a significant difference between the novel mouse and non-social object, similar to WT, Tsc1+/−, and Tsc2+/− mice (Fig. 1D). In the social novelty session, WT mice spent significantly more time exploring the novel mouse than the familiar mouse, whereas TscD+/− mice exhibited no significant preference for social novelty, similar to Tsc1+/− and Tsc2+/− mice (p = 0.005, 0.581, 0.099, and 0.845 for WT, Tsc1+/−, Tsc2+/−, and TscD+/− mice, respectively, paired t-test; Fig. 1E). In the social preference session, TscD+/− mice spent a similar amount of time exploring both the novel mouse and the cagemate, similar to Tsc2+/− mice, whereas WT and Tsc1+/− mice preferred the novel mouse significantly more than the cagemate (p < 0.001, p < 0.001, p = 0.725, and p = 0.422 for WT, Tsc1+/−, Tsc2+/−, and TscD+/− mice, respectively, paired t-test; Fig. 1F). The calculated approach-avoidance score revealed a severe reduction of social preference for a novel mouse in Tsc2+/− and TscD+/− mice rather than Tsc1+/− mice (p = 0.045, p < 0.001, p < 0.001 for Tsc1+/−, Tsc2+/−, and TscD+/− mice, respectively, compared with WT mice; p = 0.015, Tsc1+/− mice compared with Tsc2+/− mice, Tukey HSD test; Fig. 1G).

Fig. 1
figure 1

Social behaviors in Tsc1+/−, Tsc2+/−, and TscD+/− mice. A–C Frequencies of head-dipping (A) and self-grooming (B) and time spent in active interaction behaviors (C) in Tsc1+/−, Tsc2+/−, and TscD+/− mice in the hole board test, self-grooming test, and social interaction test, respectively. D Time spent exploring (social approach) a stranger mouse (S1) and non-social object (O) in the three-chamber test. E Time spent exploring a familiar mouse (S1) and stranger mouse (S2) in the social-novelty preference test. F Time spent exploring a cagemate (C) and stranger mouse (S3) in the social preference test. G Approach avoidance score calculated as the following: (time exploring S3) – (time exploring C). *p < 0.05; **p < 0.01; ***p < 0.001, one-way ANOVA followed by Tukey HSD test (A–C, G) and paired t-test (D–F). Each bar indicates the mean ± SEM

Full size image

Rapamycin ameliorated impairments in social behaviors in TscD+/− mice similarly to Tsc1+/− and Tsc2+/− mice

We previously reported that 5–10 mg kg−1 rapamycin ameliorated impairments in social interaction in Tsc1+/− and Tsc2+/− mice in the social interaction test [36]. The time spent engaged in active interaction was increased by 5 mg kg−1 rapamycin in TscD+/− mice in the social interaction test similarly to Tsc1+/− and Tsc2+/− mice, with no effect in WT mice (p = 0.559, 0.012, 0.016, and 0.001 for WT, Tsc1+/−, Tsc2+/−, and TscD+/− mice, respectively, unpaired t-test; Fig. 2A). Rapamycin also increased social novelty preference in Tsc1+/− mice (p = 0.600 and 0.007 for vehicle and rapamycin treatment, respectively, paired t-test; Fig. 2B), Tsc2+/− mice (p = 0.513 and 0.008 for vehicle and rapamycin treatment, respectively, paired t-test; Fig. 2B), and TscD+/− mice (p = 0.749 and 0.003 for vehicle and rapamycin treatment, respectively, paired t-test; Fig. 2B) and social preference in the three-chambered sociability tests in Tsc2+/− mice (p = 0.106 and 0.007 for vehicle and rapamycin treatment, respectively, paired t-test; Fig. 2C) and TscD+/− mice (p = 0.314 and 0.042 for vehicle and rapamycin treatment, respectively, paired t-test; Fig. 2C), which were similar to Tsc2+/− mice but not Tsc1+/− mice.

Fig. 2
figure 2

Effects of rapamycin on social behaviors in Tsc1+/−, Tsc2+/−, and TscD+/− mice. A–C Effects of rapamycin and vehicle on the time spent in active interaction in the social interaction test (A), time spent exploring a familiar mouse (S1) and stranger mouse (S2) in the social novelty session of the three-chambered sociability test (B), and time spent exploring a stranger mouse (S3) and cagemate (C) in the social preference session of the three-chambered sociability test (C). *p < 0.05; **p < 0.01; ***p < 0.001, unpaired t-test (A) and paired t-test (B, C). Each bar indicates the mean ± SEM. Veh vehicle, Rap rapamycin

Full size image

Changes in gene expression in Tsc1+/−, Tsc2+/−, and TscD+/− mice

To evaluate similarities and differences in autistic-like behaviors in mutant mice, gene expression levels were profiled in the brain in mutant mice. After excluding control and low-intensity transcripts from a total of 56,745 transcripts that were included in the microarray, 45,313 transcripts were analyzed. The standard deviations of log2 (fold-change) were 0.156, 0.165, and 0.226 for Tsc1+/−, Tsc2+/−, and TscD+/− mice, respectively, compared with WT mice, and the approximate curve showed a larger variation of log2 (fold-change) in TscD+/− mice than in Tsc1+/− and Tsc2+/− mice (Fig. 3A).

Fig. 3
figure 3

Gene expression profiles in the brain in Tsc1+/−, Tsc2+/−, and TscD+/− mice. A Histograms of the log2 (fold-change) (n = 45,313, bin width = 0.02) for each mutant mouse compared with WT (cyan: Tsc1+/− vs. WT, magenta: Tsc2+/− vs. WT, blue: TscD+/− vs. WT). Solid lines represent approximate normal distribution curves. SD standard deviation. B Venn diagram of three DETs between mutant and WT mice. Among overlapping transcripts (yellow area), transcripts that changed in different directions (▲: upregulated, ▼: downregulated) were excluded from each entity for further analysis. C Heat map of valid 1119 transcript profiles in vehicle-treated samples for four genotypes (n = 4 for each genotype) that were in common with at least two DETs and met the criteria. Hierarchical clustering was based on the relative expression value after baseline correction to the median of vehicle-treated WT samples (red: upregulated, blue: downregulated). D Top 20 rank map folders enriched for differentially expressed 1119 transcripts in mutant mice. The blue and red folders indicate cancer/neoplasm and immunity/inflammation, respectively. FDR false discovery rate

Full size image

In Tsc1+/−, Tsc2+/−, and TscD+/− mice compared with WT mice, 1523 transcripts, 1225 transcripts, and 4954 transcripts, respectively, were identified as DETs (two-way ANOVA, p < 0.05; Fig. 3B). To minimize false-positive changes in gene expression, only transcripts that were in common with at least two DETs (yellow area) were the focus of the subsequent analysis. Of the 1119 transcripts that were obtained after excluding transcripts that were regulated adversely between DETs, 289 and 830 were expressed upward and downward, respectively. Hierarchical clustering analysis showed that the expression patterns of these significant transcripts in TscD+/− mice were similar to Tsc2+/− mice but not Tsc1+/− mice (Fig. 3C).

Among the top 20 canonical processes that were identified in Map Folders, two major categories were included: cancer or neoplasms (e.g., lung cancer, prostatic neoplasm, hepatocellular carcinoma, and stomach neoplasm; nine processes) and diseases or pathways related to immunity or inflammation (e.g., systemic lupus erythematosus, allergic dermatitis, and asthma; eight processes; Fig. 3D). This result was consistent with the Tsc1 and Tsc2 gene functions of suppressing tumors and regulating immune and inflammatory responses via the mTOR pathway. Among GO process networks, the top three enrichment processes included “cellular component organization or biogenesis,” “regulation of cation channel activity,” and “nervous system development” (Additional file 3: Table S1). Furthermore, the 148 transcripts that overlapped between all three DETs were enriched in the seven significant networks in the canonical pathway modeling analysis in MetaCore (p < 0.05, zScore > 100; Additional file 4: Table S2).

Differentially expressed transcripts that were reversed by rapamycin treatment in the “STAT3, IRF1, IRF4, IL-2R alpha chain, IFN-γ” pathway initiated from PDLIM2

The effect of rapamycin on transcriptional changes was investigated, and we assessed whether the effects corresponded to the results of the behavioral tests in mutant mice. Among the DETs that were obtained to compare mutant and WT mice (1523 transcripts in Tsc1+/− vs. WT, 1225 transcripts in Tsc2+/− vs. WT, and 4954 transcripts in TscD+/− vs. WT), the expression of 180, 79, and 174 transcripts was significantly reversed by rapamycin treatment in Tsc1+/−, Tsc2+/−, and TscD+/− mice, respectively (two-way ANOVA, p < 0.05; Fig. 4A). The pathway enrichment analysis was performed with canonical pathway modeling in MetaCore to elucidate signal transduction pathways that are related to DETs between mutant and WT mice and are reversed by rapamycin. These DETs were enriched in seven networks, five networks, and six networks in Tsc1+/−, Tsc2+/−, and TscD+/− mice, respectively, and “signal transducer and activator of transcription 3 (STAT3), interferon regulatory factor 1 (IRF1), IRF4, interleukin-2R (IL-2R) α chain, and interferon-γ (IFN-γ)” was the common network in all mutant mice (Fig. 4A). Most DETs in the network were regulated by STAT3, which was downstream of IRF1, IRF4, IL-2R α chain, and IFN-γ, as well as STAT4 (Fig. 5). The network was initiated from PDZ and LIM domain protein 2 (PDLIM2), which was the only gene that overlapped among DETs in all mutant mice. The gene expression of Pdlim2 was higher in mutant mice than in WT mice and reduced by rapamycin treatment (Fig. 4B).

Fig. 4
figure 4

Rapamycin-induced changes in gene expression in the brain in Tsc1+/−, Tsc2+/−, and TscD+/− mice. A The cyan-colored circle indicates transcripts whose expression significantly changed in Tsc1+/− mice and was reversed by treatment with rapamycin ([Tsc1+/− vs. WT] ∩ [Tsc1+/− Veh vs. Rap]). Magenta- and blue-colored circles indicate gene expression changes in Tsc2+/− and TscD+/− mice, respectively. Canonical pathway modeling analysis for these three groups of transcripts identified seven, five, and six significant networks in Tsc1+/−, Tsc2+/−, and TscD+/− mice, respectively (p < 0.05, zScore > 100). B Box-and-whisker plot of normalized log2 signal intensity for the gene expression of Pdlim2 in vehicle- or rapamycin-treated mutant and WT mice. **p < 0.01; ***p < 0.001 (two-way ANOVA). Veh vehicle, Rap rapamycin

Full size image
Fig. 5
figure 5

Signaling pathway associated with DETs in Tsc1+/−, Tsc2+/−, and TscD+/− mice with rapamycin treatment. Arrows indicate directional edges, with main actions shown in green. All pathways are commonly initiated from PDLIM2 to STAT3 via STAT4, IRF1, IFNγ, and IRF in Tsc1+/−, Tsc2+/−, and TscD+/− mice (symbol definitions can be found at https://portal.genego.com/legends/MetaCoreQuickReferenceGuide.pdf; accessed July 14, 2022)

Full size image

Discussion

Autism spectrum disorder is a polygenic and genetically heterogeneous disorder for which many genes and genetic loci have been identified as causal factors in human genetic studies. The contribution of each gene or genetic locus to ASD is low (up to a few %), indicating that no genetic model covers all or most all ASD cases. The mTOR signaling pathway is a convergent pathway that is disturbed by genetic variations in the TSC1 and TSC2 (TSC), PTEN (Bannayan-Riley-Ruvalcaba syndrome, Lhermite-Duclos syndrome, and Cowden syndrome), FMRP (fragile X syndrome), and NF1 (neurofibromatosis type 1) genes. All of these disorders are associated with ASD, which together account for approximately 3–5% of overall ASD cases [12]. Thus, Tsc1/2 heterogeneous knockout mice are considered models for at least 3–5% of ASD cases.

Tuberous sclerosis complex manifests with such neurological symptoms as epilepsy, intellectual disability, and syndromic ASD, the symptoms of which do not differ from non-syndromic ASD [37]. Therefore, TSC is a disease model of the pathogenesis of ASD. Tuberous sclerosis complex is considered to be caused by pathogenic germline mutations (e.g., single nucleotide variations, deletions, insertions, and copy number variations) in either the TSC1 or TSC2 gene [38,39,40]. These mutations were reported to be mostly structural mutations that induce missense (7–33%) and frameshift (24–29%) mutations and affect splice sites (13–40%), whereas 18–28% are nonsense mutations. A total of 9–15% of TSC patients were found to have no identifiable mutation, and genotype–phenotype causal relationships have not yet been elucidated [39,40,41]. Furthermore, somatic TSC1 mutations were identified in lesions from patients with a pathogenic germline mutation of TSC2 [21, 27, 28], suggesting that TSC1 and TSC2 mutations may work in concert to elicit phenotypic differences among TSC patients. The present study was performed with TscD+/− mice as a model of TSC patients with both TSC1 and TSC2 mutations.

Tsc1+/− mice were previously reported to exhibit lower active interaction in the social interaction test [34, 36]. Tsc2+/− mice and Ekar (Tsc2+/−) rats were also reported to exhibit lower active interaction in the social interaction test [36, 42, 43] and impairments in social behavior in the social novelty session of the three-chambered test [42], but they did not exhibit abnormalities in the social approach session of the three-chambered test [42, 44]. The present results indicated that heterozygous Tsc2 mutation-induced social impairment was slightly more severe than the heterozygous Tsc1 mutation. TscD+/− mice exhibited impairments in social behaviors that were similar to Tsc2+/− mice and were associated with the mTOR pathway. Tsc1 and Tsc2 proteins form a heterodimeric complex that constitutively inactivates mTORC1 through inhibition of the small guanosine triphosphatase (GTPase) protein Rheb. Tsc2 possesses the GTPase-activating protein (GAP) domain for Rheb in the C terminus, and complex formation with Tsc1 stabilizes Tsc2 and enhances its GAP activity [16]. TscD+/− mice were expected to exhibit additive or synergistic phenotypes of both Tsc1+/− and Tsc2+/− mice. Indeed, the number of astrocytes in the hippocampus was reported to be higher in TscD+/− mice than in Tsc1+/− and Tsc2+/− mice [30]. Tsc1f/−; Tsc2f/−; hGFAP-Cre mice have a shorter lifespan than mice with mutations of any of these genes alone [31]. Additionally, glial-fibrillary acidic protein (GFAP)-positive cell-specific Tsc2 conditional knockout (Tsc2GFAP1) mice exhibited a high frequency of seizures and shorter life span than Tsc1GFAP1 conditional knockout mice [45]. Taken together with the present results, Tsc2 rather than Tsc1 appears to be dominant for causing neurological symptoms, including impairments in social behaviors that are caused by heterozygous Tsc1 and/or Tsc2 mutations, and impairments in social behaviors in TscD+/− mice would be partially mediated through different mechanisms from astrocytic abnormalities in an mTOR-dependent manner.

Unlike humans, which have marked individual differences, inbred mouse strains on homogeneous genetic backgrounds can be analyzed with much smaller sample sizes than humans. Indeed, a microarray analysis with 6–8 mice or 57 humans reported that standard deviations of the log (sample/reference) were approximately 0.3 and 0.8, respectively [46]. The Agilent microarray has superior detection compared with other microarray platforms [47]. Our transcriptome data that were based on Agilent microarrays showed that standard deviations of log (mutant/WT) in Tsc1+/−, Tsc2+/−, and TscD+/− mice were as low as 0.158, 0.165, and 0.226, respectively (Fig. 3A). Therefore, because the mice that were used in the present study were on a homogeneous genetic background and the standard deviation for each mutant group was low, the transcriptome analysis in which four mice were analyzed for each genotype and treatment group was considered to have the necessary power to detect differences between groups. Brain structures are left–right asymmetrical and have different functions that are controlled by each hemisphere [48], and these asymmetries of brain structures and functions have been reported to be broken in ASD [49,50,51]. Left hemisphere impairments and microstructural abnormalities were found in high-functioning ASD [52, 53], whereas the right hemisphere failed to respond to temporal novelty in ASD [54]. Social deficits in ASD have been reported to be associated with gamma-band electroencephalographic asymmetry in the right hemisphere [55], and social processing is right hemisphere-dominant in vertebrates [56, 57]. Unknown are the precise brain regions and cellular subtypes that are responsible for social deficits in ASD that are associated with TSC. Therefore, whole transcriptome analysis was performed using the right brain hemisphere rather than the left hemisphere and specific brain regions in the present study.

Over 1,000 genes with alterations of gene expressions in each mutant mice compared with WT mice were identified in the study. The magnitude of changes in gene expression in TscD+/− mice was higher than in Tsc1+/− and Tsc2+/− mice. The number of transcripts whose expression was significantly different from WT mice was also higher in TscD+/− mice than in Tsc1+/− and Tsc2+/− mice. Our results indicated that unlike social behaviors, gene expression in the brain in TscD+/− mice was affected by synergistic effects of heterozygous Tsc1 and Tsc2 mutations. In the present study, few genes in the mTOR1 signaling pathway had alterations of expression in addition to the Tsc1 and Tsc2 genes. Similarly, genes in the mTOR signaling pathway were not enriched in the gene expression analysis of cortical tubers from TSC patients, suggesting that transcription does not necessarily feedback from changes in translation, whereas mTOR1 signaling regulates translation via ribosomal protein S6 kinase, polypeptide 1 (S6K1), and eukaryotic translation initiation factor 4E binding protein 1 (4E-BP1) [21, 58]. Transgenic mice in which eukaryotic translation initiation factor 4E (Eif4e) gene expression was driven by the β-actin promoter exhibited autistic-like behaviors and a ~ 30% increase in cap-dependent translation, but the enhanced translation was limited to only a part of proteins [59]. Our results indicated that all mutant mice exhibited impairments in social behaviors despite large differences in the number of genes with alterations of gene expression between mutant and WT mice, suggesting that impairments in social behaviors that were caused by Tsc1 and Tsc2 mutations would be attributable to changes in specific molecular pathways and not a general increase in transcription or translation.

The hierarchical clustering analysis indicated that gene expression patterns in overlapping genes with alterations of expression between mutant and WT mice were similar in Tsc2+/− and TscD+/− mice, which corresponded to the severity of impairments in social behavior in mutant mice. Furthermore, genes whose expression was different between Tsc1+/− and Tsc2+/− mice were mostly enriched in the group of genes that were altered in Tsc2+/− and TscD+/− mice but not in Tsc1+/− mice compared with WT mice, also indicating the similarity of changes in gene expression between Tsc2+/− and TscD+/− mice (Additional file 2: Fig. S2A). The GO analysis of overlapping genes showed that they were enriched in cancer/neoplasm- and immunity/inflammation-related processes. These cancer/neoplasm- and immunity/inflammation-related processes showed the up- and downregulation of genes that were altered in Tsc2+/− and TscD+/− mice (Additional file 2: Fig. S2B), respectively, indicating that the severity of symptoms that are related to cancer/neoplasm and immunity/inflammation would be different between TSC patients with underlying TSC1 and TSC2 mutations.

STAT3 is phosphorylated by mTOR1 and activated in TSC2-deficient cells, which are inhibited by rapamycin [60,61,62]. The expression of inflammatory gene markers, including IL-1β, IL-6, and Toll-like receptors, and phosphorylated STAT3 increased in brain lesions in TSC patients [63,64,65,66]. These previous studies support our findings that the “STAT3, IRF1, IRF4, IL-2R α chain, IFN-γ” pathway network was associated with impairments in social behaviors in mutant mice and suggest that suppressing inflammatory responses in TSC patients may improve social deficits. This pathway network is initiated by Pdlim2, which is ubiquitously expressed throughout the body of mice but especially in the genitourinary, nervous, and visceral organ systems. Pdlim2 encodes an E3 ubiquitin ligase and ubiquitinates STAT3 and STAT4 to lead to their degradation [67, 68]. Pdlim2 is reported to be associated with tumor suppression [69,70,71] and tumorigenesis [72, 73], but its function in the nervous system remains unknown.

Limitations

The “STAT3, IRF1, IRF4, IL-2R α chain, IFN-γ” pathway network initiated from PDLIM2 and STAT4 was associated with impairments in social behaviors in Tsc1+/−, Tsc2+/−, and TscD+/− mice, but the regulatory mechanisms of the mTOR pathway and social behaviors by the pathway network were not elucidated in the present study. It is also unclear whether the pathway network also plays a critical role in impaired social behaviors in autism spectrum disorders not caused by Tsc1 and Tsc2 mutations.

Conclusions

In the present study, detailed analyses of social behaviors and transcriptomes in Tsc1+/−, Tsc2+/−, and TscD+/− mice revealed the genetic pathology of impairments in social behaviors. Mutations in the TSC2 gene are likely to have dominant effects on social behaviors rather than TSC1 gene if these mutations have similar potency. However, rapamycin has therapeutic effects on impairments in social behaviors caused by the mutations in both genes. It is also suggested that social behavioral abnormalities in patients with mutations in both the TSC1 and TSC2 genes are similar to those in patients with mutations in the TSC2 gene. The Pdlim2-mediated signaling pathway is suggested to be associated with altered brain function and impaired social behaviors due to mutations in the TSC1 and TSC2 genes. The present findings further our understanding of genotype–phenotype relationships and may contribute to the development of medications that can ameliorate social deficits in ASD.