Abstract
Autism spectrum disorder (ASD) is a highly heritable complex neurodevelopmental disorder. While the core symptoms of ASD are defects of social interaction and repetitive behaviors, over 50% of ASD patients have comorbidity of intellectual disabilities (ID) or developmental delay (DD), raising the question whether there are genetic components and neural circuits specific for core symptoms of ASD. Here, by focusing on ASD patients who do not show compound ID or DD, we identified a de novo heterozygous gene-truncating mutation of the Sentrin-specific peptidase1 (SENP1) gene, coding the small ubiquitin-like modifiers (SUMO) deconjugating enzyme, as a potentially new candidate gene for ASD. We found that Senp1 haploinsufficient mice exhibited core symptoms of autism such as deficits in social interaction and repetitive behaviors, but normal learning and memory ability. Moreover, we found that the inhibitory and excitatory synaptic functions were severely affected in the retrosplenial agranular (RSA) cortex of Senp1 haploinsufficient mice. Lack of Senp1 led to over SUMOylation and degradation of fragile X mental retardation protein (FMRP) proteins, which is coded by the FMR1 gene, also implicated in syndromic autism. Importantly, re-introducing SENP1 or FMRP specifically in RSA fully rescued the defects of synaptic functions and core autistic-like symptoms of Senp1 haploinsufficient mice. Taken together, these results elucidate that disruption of the SENP1-FMRP regulatory axis in the RSA may cause core autistic symptoms, which further provide a candidate brain region for therapeutic intervene of ASD by neural modulation approaches.
Introduction
Autism spectrum disorders (ASD) is a group of neurodevelopment disorders, characterized by defects in social communication and stereotypic behaviors (Lord et al., 2020). Although ASD is the highest heritable mental disorder among neuropsychiatric disorders (Sandin et al., 2017; Wang et al., 2017), the genetic landscape of ASD was not revealed until the application of genome-wide sequencing technologies such as whole-exome sequencing and whole-genome sequencing (De Rubeis et al., 2014; Iossifov et al., 2014). The major contributory genetic components of ASD include de novo variations and rare inherited variations, as well as structural variations (Iakoucheva et al., 2019; Searles Quick et al., 2021). Among various genetic variants, the de novo variants of large effect, for example protein-truncating variants, would account for 10-15% of total contributing genetic causes of ASD, which may provide the first line of clue for illustrating the abnormal neural circuits disrupted in autism patients (Searles Quick et al., 2021).
In the recent ASD genetic study using the largest cohort so far, researchers propose that there are a group of ASD candidate genes primarily affecting core symptoms of autism such as social communication and repetitive behavior (referred to as ASD predominant genes), whereas another group of ASD candidate genes also lead to defects in cognition caused by neurodevelopmental delay (referred to as ASD & NDD genes) (Satterstrom et al., 2020). Although whether ASD predominant and ASD & NDD genes could be clearly separated is still arguable (Myers et al., 2020), focusing on the ASD candidate genes mainly affecting symptoms in social domains, instead of cognitive functions, may provide insights for understanding the genomics and neural mechanisms underlying social behaviors in mammals.
In a whole-exome sequencing study on over 600 ASD trios in Shanghai Xinhua Hospitals, we identified a de novo heterozygous protein-truncating mutation in the Senp1 gene of an autism patient with impaired social functions, but without developmental delay (normal Intelligence Quotient (IQ)/ developmental Quotient (DQ)). SENP1 (Sentrin-specific peptidase1) plays a decisive role in post-translational SUMOylation modifications by releasing SUMO groups from proteins (Flotho and Melchior, 2013). Homozygous mutation of SENP1 has been implicated in severe neurometabolic diseases (Tarailo-Graovac et al., 2016). Although SENP1 has been found to play critical roles in regulating various physiological functions including metabolism, as well as neural injury (Sun et al., 2020b; Wang et al., 2019), the role for SENP1 in the central nervous system remain largely unclear (Choi et al., 2016; Sun et al., 2014).
In this study, we showed that Senp1 haploinsufficient mice (Senp1+/−) exhibited defects of social behaviors and increased stereotypic behaviors, without deficiency in learning and memory tasks. Interestingly, we found that the inhibitory neurons were specifically affected in the retrosplenial agranular (RSA) cortex of Senp1 +/− mice. The inhibitory and excitatory synaptic transmission of layer II/III pyramidal neurons in the RSA region were altered in the Senp1 haploinsufficient mice comparing to wild-type littermates. The orchestrating alteration of inhibitory and excitatory synaptic transmission was also seen in Cntnap3 knockout mice, another autism mouse model reported by our lab (Tong et al., 2019), suggesting that the co-regulation of inhibitory and excitatory synapse may be one of the key signatures of autism pathophysiology. Remarkably, we could rescue the defects of the social behaviors of Senp1 haploinsufficiency mice by introducing wild-type Senp1 specifically into the RSA region. These data suggest that Senp1 is a candidate gene for “ASD predominant gene” affecting primarily social functions and the RSA region may serve as a circuitry node implicated in regulating social behaviors.
Results
Identification of a de novo mutation in the SENP1 gene in a ASD proband and autistic-like behaviors in Senp1 haploinsufficient mice
In order to study ASD candidate genes in Chinese cohorts, we performed whole-exome sequencing to over 600 ASD probands with their parents to search for de novo variants which may contribute to ASD symptoms. All ASD probands were further examined with Childhood Autism Rating Scale (CARS), Autism Diagnostic Observation Schedule (ADOS) and Intelligence Quotient (IQ)/developmental Quotient (DQ) tests to acquire the comprehensive appearances of social and cognitive functions. We identified a de novo protein-truncating mutation in the Senp1 gene (NM001267594:c.151C>T:p.Q51X), validated by Sanger sequencing (Figure 1A). The mutation takes place at a conserved region of the Senp1 gene across mammals, which lead to premature termination of the SENP1 protein during translation (Figure 1B, C).
The male patient carrying the SENP1 mutation exhibited typical deficits in social behaviors including communication and interaction, measured by ADOS and CARS scores (Table S1). Interestingly, the patient exhibited largely normal Developmental Quotient measured by the Gesell development scale (Table S1), suggesting that mutation of SENP1 may specifically affect the behaviors in social domain, instead of cognition domain, in human.
To investigate roles of the SENP1 gene in social behavior, we use laboratory mouse as an animal model to examine whether genetic deletion of the Senp1 gene may affect social behaviors in mouse (Wang et al., 2019). Since homozygous deletion of the Senp1 gene causes lethality of mouse, we examine a battery of behavioral tests in the Senp1 haploinsufficient mice which carry heterozygous deletion of the Senp1 gene (Figure S1A). We first examined the anxiety level using the open field test and found that Senp1 haploinsufficient mice exhibited the same level of anxiety as wild-type (WT) littermates (Figures S1B, S2A-S2E). We next examine the social behaviors of Senp1 haploinsufficient mice along with WT littermates in the classic three-chamber test (Figure S1C). We found that although Senp1 haploinsufficient mice showed the same social approach feature comparing to WT littermates (Figure S2F-S2I), Senp1 haploinsufficient mice displayed significant defects in the social novelty test, by not showing increased preference with novel partners in comparison with WT littermates (Figure 1D, E). This suggests that lack of Senp1 may lead to abnormal social behaviors. To further validate whether Senp1 haploinsufficient mice are able to recognize novel objects, we used the object cognition test (Figure S1D) and found no significant difference in object cognition preference between Senp1 haploinsufficient and WT mice (Figure S2J-S2M). Similarly, no difference in the novel object cognition test was found (Figure 1F, G). This data suggests that Senp1 haploinsufficient mice have specific defects in social cognition behaviors.
Stereotypic and repetitive behaviors are also core symptoms of ASD. Thus, we examined whether Senp1 haploinsufficient mice have stereotypic and repetitive behaviors via the restrictive and repetitive behaviors test (Figure S1E). Interestingly, Senp1 haploinsufficient mice showed more stereotypic behaviors by measuring the marbles buried and times for self-grooming (Figure 1H-J). Lastly, we investigated the spatial learning ability of mutant and WT mice by using the Barnes maze test (Figure S1F). We found that Senp1 haploinsufficient mice exhibited the normal spatial learning ability as WT mice (Figure 1K, 1L; Figure S2N). Together, these results indicate that Senp1 haploinsufficient mice display abnormal social behaviors but normal object cognition and spatial learning ability, which is in consistent with clinical symptoms of the ASD patient carrying Senp1 mutant and suggests that SENP1 in the brain may specifically contribute to the social interaction behaviors.
Compromised brain development in Senp1 haploinsufficient mice
To examine whether Senp1 haploinsufficiency may lead to abnormal neural development, we first analyzed whether brain structures were altered due to loss of Senp1. Surprisingly, we found that the brain width and thickness of cerebral cortex significantly increased in both male and female Senp1 haploinsufficient mice at 2 months age, comparing to WT littermates (Figure 2A–2D). Brain overgrowth is an intriguing symptom often seen in ASD patients and ASD mouse models such as PTEN mutant mice (Kwon et al., 2006).
To further investigate how the brain structure is altered in Senp1 haploinsufficient mice, we examine expression levels of various neuronal markers in the brain. First, we found that the expression of TBR1, an important cortical marker gene and ASD risk gene, was dramatically decreased in Senp1 haploinsufficient mice across various cortices at 2 months of age, in comparison with WT mice (Fazel Darbandi et al., 2018; Huang et al., 2014; Huang et al., 2019) (Figure 2E, 2F), suggesting that the cortical development program was impaired in the Senp1 haploinsufficient mice.
We next determined the expression pattern of SENP1 in the mouse brain. With immunostaining using antibody against SENP1 in Senp1 haploinsufficient and WT mice, we found that SENP1 was highly expressed in various cortical region of mouse brain and decreased expression in Senp1 haploinsufficient mice (Figure 2G, 2H). Notably, we found that the downregulation of SENP1 proteins was most significant in the retrosplenial agranular (RSA) cortex, a subregion of retrosplenial cortex, comparing to other cortices in the Senp1 haploinsufficient mice (Figure 2G, 2H).
In RSA cortex of Senp1 haploinsufficient mice, two kinds of GABAergic neurons, parvalbumin (PV) and somatostatin (SST) positive neurons, were specifically decreased in the retrosplenial agranular (RSA) cortex of Senp1 haploinsufficient mice, comparing to WT mice (Figure 2I, 2J, Figure S3A-S3E). In particular, the number of PV and SST positive neurons was increased in the layer II/III and layer V of the Senp1 haploinsufficient mice, in comparison with WT mice (Figure S3A, S3B, S3F, S3G), suggesting that the inhibitory synaptic connections in RSA may be altered in Senp1 haploinsufficient mice.
The retrosplenial cortex (RSC), composing of two parts — RSA (retrosplenial agranular) and RSG (retrosplenial granular), is implicated in the top-down modulation of sensorimotor information from primary sensory cortices (Bicanski and Burgess, 2020). Recently, researchers reported that ketamine treatment in mouse specifically activated the slow wave oscillation in the retrosplenial cortex (van der Meer et al., 2020). Furthermore, the rhythmic activation of deep layer neurons (layer V) in the retrosplenial cortex causally led to a series of dissociation-like phenotype in mice, including withdraw of social interaction (van der Meer et al., 2020). Since defects in sensorimotor functions are prominent in ASD patients, it is plausible that the RSC may regulate social behavior via modulating sensorimotor information integration. We hypothesized that abnormal synaptic connections in the RSA region may cause deficits of social interaction behaviors, due to incapable of proper processing sensory information to higher centers.
Disrupted inhibitory and excitatory synaptic functions in the RSA neurons of Senp1 haploinsufficient mice
Next, we used immunostaining and electrophysiology to examine whether Senp1 haploinsufficiency alters the structure and function of GABAergic synapses in the RSA cortex (Figure 3A–3D). To examine the GABAergic synapses formed on RSA neurons in WT and Senp1 haploinsufficient mice, we injected two adeno-associated viruses (AAV-CAG-FLEX-EGFP, AAV-hSyn-Cre) to sparsely label layer II/III neurons in the RSA region and imaged the synapses formed on the soma of labelled neurons by con-focal microscopy (Figure 3A, 3B). We found that numbers of inhibitory synapses, measured by punctum co-labelled with vesicular GABA transporter (VGAT) and Gepyrin, were significantly increased in soma of RSA layer II/III neurons of Senp1 haploinsufficient mice, comparing to WT mice (Figure 3E–3G). Consistently, the punctum co-labelled with VGAT and the α1 subunit of GABAA receptors were also increased in RSA layer II/III neurons of Senp1 haploinsufficient mice, in comparison with WT mice, suggesting that Senp1 haploinsufficiency leads to more inhibitory synapses formation on RSA layer II/III neurons (Figure 3H–3J).
To further address whether the inhibitory synaptic transmission of RSA layer II/III neurons of Senp1 haploinsufficient mice was altered, we performed the whole-cell patch clamp recording on the layer II/III pyramidal neurons in RSA brain slices from mice of both genotypes. We found that the frequency of miniature inhibitory post-synaptic currents (mIPSCs) in RSA layer II/III neurons of Senp1 haploinsufficient mice was significantly higher than that in WT mice (Figure 3K–3M). Together, this data indicates that inhibitory synaptic transmission of RSA layer II/III pyramidal neurons in Senp1 haploinsufficiency mice was increased likely due to more inhibitory synapses formed on these neurons.
We then wonder whether Senp1 haploinsufficiency may alter the structure and functions of excitatory synapses in RSA neurons. We performed sparse labeling by AAV injections followed by immunostaining experiments and electrophysiology (Figure 4A–4D). Interestingly, we found that the numbers of excitatory synapses, represented by punctum labeled by co-staining of PSD95 and vesicular glutamate transporter 1 (VGlut1), as well as co-staining of glutamate ionotropic receptor AMAP type subunit 2 (GRIA2) and VGlut1, were dramatically decreased in soma of RSA layer II/III neurons of Senp1 haploinsufficient mice, comparing to WT mice (Figure 4E–4J). In consistent, we found that the frequency, but not amplitude, of miniature excitatory postsynaptic current (mEPSCs) significantly decreased in Senp1 haploinsufficient mice, in comparison with WT mice, suggesting that the function of excitatory synapses was compromised in RSA neurons of Senp1 haploinsufficient mice (Figure 4K–4M).
It is worthy to note that the orchestrating alteration of inhibitory and excitatory synaptic transmission found in this study echoed a previous finding, in which there were also increased inhibitory synaptic transmission and decreased synaptic transmission in Cntnap3 knockout mice, another autism mouse model that was previously reported by our lab (Tong et al., 2019). These shared phenotypes in different autism mouse models strongly suggest that the co-regulation of inhibitory and excitatory synapse may be one of the key features of autism pathophysiology in the brain.
Impaired dendritic growth and immature spine of the RSA neurons of Senp1 haploinsufficient mice
To investigate the role of SENP1 in the neuronal development, we constructed short hairpin RNA (shRNA) against the mouse Senp1 gene (Figure S4A, S4B). By transfecting shRNA against Senp1 and control shRNA in mouse cortical neurons, we found that the dendritic complexity was dramatically decreased when the endogenous SENP1 is knocked down by shRNA, which could be fully rescued by re-introducing human SENP1 open reading frame (SENP1 ORF), but not human SENP1 ORF carrying the stop-gain mutation found in the ASD patient (SENP1 stop) (Figure S4C, S4D). When SENP1 ORF was overexpressed in mouse cortical neurons, the overall dendritic complexity was increased, suggesting that SENP1 indeed plays a critical role in regulating dendritic growth (Figure. S4C, S4D). Interestingly, we also found that there are much more filopodia appeared in the SENP1 knocked down neurons comparing to GFP-expressing neurons which could be rescued by SENP1 ORF expression, suggesting that deletion of SENP1 may lead to impairment of spine maturation (Figure S4C, S4E).
We next examined the dendritic morphology of the layer II/III neurons in the RSA cortex of Senp1 haploinsufficient mice by sparse labeling methods described above (Figures 3A). We found that the basal dendrites of layer II/III neurons exhibited less complexity in Senp1 haploinsufficient mice comparing to WT mice, whereas apical dendrites appeared to be more complex in Senp1 haploinsufficient mice suggesting that basal and apical dendrites were regulated differentially in vivo (Figures 5A, 5B). When we examined the spine morphology of layer II/III neurons in the RSA cortex, we surprisingly found that although the total spine numbers increased in the basal and apical dendrites of layer II/III neurons in Senp1 haploinsufficient mice (Figure 5C, 5D), there were much less spines in the mature form and more immature spines in layer II/III neurons of Senp1 haploinsufficient mice (Figure 5E–5H), indicating that spine development in the RSA cortex was hindered by the haploinsufficiency of Senp1.
We then examine the fine structure of excitatory synapse of the layer II/III neurons in RSA of Senp1 haploinsufficient and WT mice with electron microscopy. Consistently, we found that the synaptic cleft appeared to be wider and the length of postsynaptic density (PSD) decreased in Senp1 haploinsufficient mice (Figure 5I–5L), suggesting that SENP1 plays a critical role in promoting excitatory synapse maturation in vivo.
Downregulation of FMRP in the RSA cortex of Senp1 haploinsufficient mice
We would like to investigate the molecular mechanism by which SENP1 regulates synaptic development. Since SENP1 is one of the major SUMOylation deconjugating enzymes, we reasoned that SENP1 may regulate synapse development by conjugating small uniqitin-like modifier (SUMO) to candidate proteins, which performed critical functions in syanpse development. In a previous study, researchers identified that SENP1 proteins existed in the post-synaptic complex and implicated in brain disorders using mass spectrometry (Li et al., 2017). Among many synaptic proteins which are associated with SUMOylation modification, fragile X syndrome mental retardation protein (FMRP) was an intriguing candidate of SENP1-regulated deSUMOylation. The mutation of FMR1, the coding gene of FMRP, caused Fragile X syndrome (FXS), which shares autistic symptoms with ASD (Khayachi et al., 2018; Tang et al., 2018). It was reported that SUMOylation of FMRP was critical for its transporting mRNAs critical for synaptic functions in to synapses and release of mRNA from FMRP proteins (Khayachi et al., 2018). The SUMOylation and deSUMOylation is a dynamic process, dysregulation either of which will sure lead to compromise of proper functions of FMRP. Thus we set out to examine whether FMRP is dysregulated in the brain of Senp1 haploinsufficient mice.
We first tested whether FMRP is expressed in the RSA of mouse brain. By immunostaining experiment using antibodies against SENP1 and FMRP, we found that FMRP closely colocalized with SENP1 in the RSA neurons of the mouse brain and was downregulated in Senp1 haploinsufficient mice (Figure 6A, 6B). Moreover, we examined the expression of SENP1 and FMRP in the RSA cortex during mouse development and found that the expression levels of SENP1 and FMRP exhibit similar curve during mouse pre and post-natal brain development (Figure 6C–6E). Next we examined whether the expression of FMRP was regulated by SENP1 in vitro and in vivo. In the cultured mouse cortical neurons, we found that the level of FMRP protein significantly decreased after SENP1 knockdown (Figure 6F, 6G). We then performed Western blot on the tissue lysate collected from RSA region of Senp1 haploinsufficient and WT mice (Figure 6B). We found that the protein level of FMRP in the RSA region was clearly downregulated in Senp1 haploinsufficient mice comparing to its of WT mice (Figure 6H, 6I), strongly suggesting that FMRP was regulated by SENP1 in the RSA of the brain.
If SENP1 indeed plays a critical role in removing the SUMO groups from FMRP protein, the SUMOylation level of FMRP in the brain of Senp1 haploinsufficient mice may elevate. To determine the level of FMRP SUMOylation in the RSA region, we collected the RSA tissue lysate from Senp1 haploinsufficient and WT mice, then performed immunoprecipitation assay using antibody against FMRP (Figure 6B). We found that in the RSA lysate from Senp1 haploinsufficient mice, there is clearly an up-shifting band with FMRP positive signals, but not in WT mice (Figure 6J), suggesting that FMRP-SUMO signals are enriched in the RSA region of Senp1 haploinsufficient mice. Together, these results suggest that SENP1 regulates the SUMOylation of FMRP in the brain, hereby over SUMOylation of FMRP protein in the brain of Senp1 haploinsufficient mice may lead to its degradation (Khayachi et al., 2018).
Restoring the expression of SENP1 or FMRP in RSA rescue autistic-like behaviors of Senp1 haploinsufficient mice
Finally, we wonder whether restoration of SENP1 or FMRP in RSA could rescue the autistic-like behaviors of Senp1 haploinsufficient mice. We injected AAVs harboring control vector, SENP1 or FMRP cDNA bi-laterally in RSA of Senp1 haploinsufficient mice at 2 months of age and examined behaviors one month later (Figure 7A).
We found that injection of control AAV (AAV-vector) did not change the deficits in social novelty test of Senp1 haploinsufficient mice (Figure 7B), whereas injection of AAV-SENP1 into the RSA of Senp1 haploinsufficient mice significantly increased the time Senp1 haploinsufficient mice spent in sniffing novel partners (Figure 7C), suggesting that the SENP1 in the RSA is crucial for social behaviors in mouse. Similarly, injection of AAV-FMRP in the RSA region of Senp1 haploinsufficient mice also alleviated the social deficits (Figure 7D).
To further examine the specificity of RSA in social behaviors, we asked whether injection of AAV-SENP1 into the anterior cingulate cortex (ACC), a well-known region directly connected to RSA and also implicated in autistic behaviors (Guo et al., 2019). Surprisingly, we found that injection of AAV-SENP1 into the ACC had no effects on social behaviors of Senp1 haploinsufficient mice (Figure 7E, 7F). Consistently, we found that the stereotypic and repetitive behaviors of Senp1 haploinsufficient mice were fully rescued by injection of AAV-SENP or AAV-FMRP into RSA, but not ACC (Figure 7G–7I).
Next, we wondered whether synaptic abnormalities in the RSA cortex of Senp1 haploinsufficient mice may be rescued by genetic manipulations. First, we examined levels of various synaptic protein with biochemical methods. We collected tissue lysates from RSA regions of WT and Senp1 haploinsufficient mice with or without AAV injection (Figure 7J). Using Western blots with antibodies against various proteins, we found that re-introducing SENP1 in the RSA of Senp1 haploinsufficient mice restored the increased level of the GABAAR α1 subunit and gephyrin to the normal level as WT mice (Figure 7K–7N), as well as the decreased level of GRIA2 and PSD95 back to levels in the WT mice (Figure 7O, 7P). Consistently, injection of AAV-FMRP was also able to rescue the biochemical defects in the Senp1 haploinsufficient mice (Figure S5A-S5G).
Synaptic defects in the RSA of Senp1 haploinsufficient mice were rescued by restoration of SENP1 or FMRP
We then examined whether the structure and function of RSA neurons in Senp1 haploinsufficient mice could be rescued by re-introducing SENP1 or FMRP. By the sparse labeling methods used above, we first measured the inhibitory and excitatory synapses in RSA layer II/III neurons of Senp1 haploinsufficient mice injected with AAVs expressing control vector, SENP1 or FMRP, perspectively (Figure 8A, 8B).
We found that re-introducing SENP1 or FMRP into the RSA of Senp1 haploinsufficient mice significantly decreased the abnormal high inhibitory synapse number, represented by punctum co-labelled with Gephyrin and VGAT (Figure 8C). By performing whole-cell patch clamp experiments, we found that mIPSC frequency also accordingly decreased in the layer II/III RSA neurons in Senp1 haploinsufficient mice injected with AAV-SENP1 or AAV-FMRP, comparing with AAV-vector injecting mice (Figure 8D, 8E).
Interestingly, re-introducing SENP1 or FMRP increased the number of excitatory synapse in the layer II/III RSA neurons of Senp1 haploinsufficient mice, measured by punctum with PSD95 and VGlut1 signals (Figure 8B, 8F). Consistently, mEPSCs frequency of RSA layer II/III neuron in Senp1 haploinsufficient mice injected with AAV-SENP1 or AAV-FMRP significantly increased, comparing to mice injected with AAV-vector (Figure 8G, 8H). Together, these data indicate that the SENP1-FMRP axis regulates the balance of inhibitory and excitatory synapse in the RSA region of the mouse brain.
Discussion
Genomic studies using whole-exome sequencing technology for large ASD cohorts since 2011 have proven to be very fruitful for identification of ASD-causing genetic mutations (Iossifov et al., 2012; Neale et al., 2012; O’Roak et al., 2011; O’Roak et al., 2012; Sanders et al., 2012). Besides the classic dominant and recessive mutations, the de novo mutations taking place usually at male germ cells are found to play a critical role in contributing to the etiology of ASD. The gene-disrupting mutations (usually referred as likely gene disrupting—LGD) are more often to be discovered in the ASD probands, comparing to their unaffected siblings (Iossifov et al., 2014). Although the overall contribution of de novo LGD mutations may account for around 10% of genetic causes of ASD, they open a venue for us to understand the mechanism by which the brain receives and processes social signals and perform social interaction behaviors accordingly. 50-70% of ASD cases would have comorbidity of intellectual disability or developmental delay (Lord et al., 2020). Searching for genetic causes in ASD patients without comorbidity of developmental delay may lead to discovery of genes specific for social behaviors in mammals (Satterstrom et al., 2020). There certainly is a possibility that there are not any genes specific for social behaviors, since different penetrance of genetic mutations on individual may lead to various extent of symptoms (Myers et al., 2020). However, it is still the consensus that focusing on the genetic mutations primarily affecting core symptoms of autism may shed new light on the understanding of neural circuits underlying social interaction behaviors (Myers et al., 2020).
In this work, we found that in the Senp1 haploinsufficient mice, mimicking the human ASD patient, both inhibitory and excitatory synaptic transmission in the RSA region of retrosplenial cortex was specifically affected, which causally lead to deficits in social novelty tests and could be rescued by re-introducing SENP1 or its downstream molecule FMRP during adulthood (Figure 9A, 9B). This data suggests that the RSA region plays a critical role in regulating social behaviors in the mouse brain. From previous works, the retrosplenial cortex is implicated in the top-down control of sensorimotor information integration, as well as playing a role in coordinated oscillation with hippocampal neurons (Mao et al., 2018). Together with the recent finding that ketamine treatment in mice specifically activated the RSC neurons in mice and lead to dissociative-like symptoms, it is intriguing to hypothesize that the RSC region may play a critical role in orchestrating sensory information from primary cortices to higher centers, in which abnormal neural activity caused either by acute ketamine treatment or developmental defects due to lack of SENP1, would lead to various behavioral deficits, including social interactions (van der Meer et al., 2020).
Although ASD is a developmental disorder, in which genetic defects usually affect the brain development from early stages, it is widely recognized that autistic-like phenotypes in autism mouse models, such as Mecp2 mutant or duplication mice, could be rescued with genetic methods (Guy et al., 2007; Sztainberg et al., 2015) or by gene editing tools (Qiu, 2018; Sun et al., 2020a; Yu et al., 2020) during adulthood. These works indicate that the neural circuits governing social behaviors still remain plastic after brain is fully grown, thus identification of which will provide the possibility of manipulation impaired social behaviors in adolescent or adult ASD patients.
The neural mechanisms by which RSA regulates social behaviors is intriguing. The role of retrosplenial cortex is implicated in integration of sensory information from primary cortices. The sensorimotor problems are prevalent in ASD patients which also become one of major targets for recovery therapies (sensory integrative training). One of primary hypothesis for etiology of ASD is that abnormal development may disrupt the sensory integration and in turn compromised social interaction behaviors. Although the brain processes many kinds of sensory inputs during everyday life, whether there is a specific neural circuit responsible for process and relay social-related sensory inputs is unknown. Thus, our current work leads to the hypothesis framework that RSA may be able to extract the social information received from primary sensory cortices and present them to higher centers. Further works mapping the neural circuits upstream and downstream of RSA neurons would illustrate the comprehensive picture of social information processing in the brain. Furthermore, the availability of non-human primate models for autism may facilitate the translational efforts in which non-invasive or invasive neural modulation methods targeting RSA would be applied and yield valuable insights for intervention of ASD, prior to clinical practices in patients (Liu et al., 2016; Zhou et al., 2019)
Star methods
Detailed methods are provided in the online version of this paper and include the following:
Key resources table
Resource availability
➢ Lead contact
➢ Materials availability
➢ Data and code availability
Experiment model and subject details
➢ Animals
METHOD DETAILS
➢ Diagnosis for high-functioning autism
➢ Behavioral test
➢ Immunofluorescence assays
➢ Viruses
➢ Stereotactic injection of AAV virus
➢ Slice electrophysiological recording
➢ Western blot and co-IP assays
➢ Primary neuron culture and transfection
➢ Transmission electron microscopy assay
QUANTIFICATION AND STATISTICAL ANALYSIS
➢ Statistical Analysis
Author contributions
K.Y. and Z.Q. designed the experiments, analyzed the data, and wrote the manuscript. K.Y. performed biochemistry, immunohistochemistry, and animal behavioral analysis experiments. Y.-H.S. performed stereotactic injections for immunohistochemistry experiments and electrophysiological recordings with the assistance of L.F. and H.-T.X.. X.-J.D. and J.-H.Y performed autism diagnosis under the supervision of F.L.. Y.-F.Z. and Y.-T.Y. performed behavioral testing. S.-F.S. performed cell culture and tissue dissection. R.-Q.W., C.-H.Z., Y.-T.L., Z.-L.C. and Y.-Z.W performed Western blot and immunofluorescence. J.-K.C. provided Senp1-haploinsufficient mice.
Declaration of interests
The authors declare no competing interests.
Acknowledgements
We thank Drs. Xiaohong Xu, Ji Hu, Xiaoke Chen, Yu Fu and the Neuroscience Pioneering Club for valuable comments and Qian Hu for excellent technical assistance. This work was supported by grants from the NSFC Grants (#31625013, #81941015, #82001211, #82021001, #81761128035, #81930095), Strategic Priority Research Program of the Chinese Academy of Sciences (XDBS01060200), Program of Shanghai Academic Research Leader, the Open Large Infrastructure Research of Chinese Academy of Sciences, and the Shanghai Municipal Science and Technology Major Project (#2018SHZDZX05), CPSF-CAS Joint Foundation for Excellent Postdoctoral Fellows (2017LH036) and China Postdoctoral Science Foundation (2017M620173). Shanghai Municipal Commission of Health and Family Planning (#GWV-10.1-XK07, #2020CXJQ01), Shanghai Committee of Science and Technology (#2018SHZDZX01, #19410713500, #16JC1420501), Guangdong Key Scientific and Technological Project (#2018B030335001)
Footnotes
Leading contact: zqiu{at}ion.ac.cn (Z. Qiu)