AUTS2 Regulation of Synapses for Proper Synaptic Inputs and Social Communication

Summary Impairments in synapse development are thought to cause numerous psychiatric disorders. Autism susceptibility candidate 2 (AUTS2) gene has been associated with various psychiatric disorders, such as autism and intellectual disabilities. Although roles for AUTS2 in neuronal migration and neuritogenesis have been reported, its involvement in synapse regulation remains unclear. In this study, we found that excitatory synapses were specifically increased in the Auts2-deficient primary cultured neurons as well as Auts2 mutant forebrains. Electrophysiological recordings and immunostaining showed increases in excitatory synaptic inputs as well as c-fos expression in Auts2 mutant brains, suggesting that an altered balance of excitatory and inhibitory inputs enhances brain excitability. Auts2 mutant mice exhibited autistic-like behaviors including impairments in social interaction and altered vocal communication. Together, these findings suggest that AUTS2 regulates excitatory synapse number to coordinate E/I balance in the brain, whose impairment may underlie the pathology of psychiatric disorders in individuals with AUTS2 mutations.


HIGHLIGHTS
AUTS2 regulates excitatory synapse number in forebrain pyramidal neurons Loss of Auts2 leads to increased spine formation in development and adulthood Loss of Auts2 alters the balance of excitatory and inhibitory synaptic inputs Auts2 mutant mice exhibit cognitive and sociobehavioral deficits

INTRODUCTION
Synapses form the basis for the neuronal network and brain function. Development of synapses, synaptogenesis, is precisely regulated by genetic programs as well as synaptic activities. Even after establishment of the fundamental brain structures, synapses are dynamically formed and eliminated in response to neuroenvironmental stimuli (Holtmaat and Svoboda, 2009). However, maintenance of the number of synapses within a certain range, comprising the synapse homeostasis, assures neuronal homeostasis (Davis, 2013;Tien and Kerschensteiner, 2018;Wefelmeyer et al., 2016). It has been proposed that failure of either synapse or neuronal homeostasis results in various neuropsychiatric disorders (Bourgeron, 2009;Ramocki and Zoghbi, 2008). Consistent with this, postmortem pathological studies have revealed that aberrant regulation of dendritic spine number as well as structural abnormalities of spines were observed in patients with numerous psychiatric disorders such as autism spectrum disorders (ASDs), schizophrenia, and neurodegenerative diseases (Hutsler and Zhang, 2010;Penzes et al., 2011;Tang et al., 2014). Thus, appropriate regulation of synaptogenesis as well as synapse homeostasis is critical for normal healthy brain function; however, its molecular machinery remains elusive.
Autism susceptibility candidate 2 (AUTS2) (also termed ''activator of transcription and developmental regulator'') located on human chromosome 7q11.22 has been initially identified as a possible ASD risk gene in a study that reported a de novo balanced translocation in monozygotic twin patients with ASDs (Sultana et al., 2002). Thereafter, structural variants that disrupt the AUTS2 locus have been identified in the patients with not only autism but also other neuropathological conditions including intellectual disabilities (IDs), schizophrenia, attention deficit hyperactivity disorder (ADHD), dyslexia, and epilepsy, as well as brain malformation and craniofacial abnormalities (Amarillo et al., 2014;Bakkaloglu et al., 2008;Ben-David et al., 2011;Beunders et al., 2013;Elia et al., 2010;Hori and Hoshino, 2017;Jolley et al., 2013;Kalscheuer et al., 2007;Talkowski et al., 2012;Zhang et al., 2014). In addition, AUTS2 has been recently implicated as a potential gene in human-specific evolution .
In the developing mouse brain, Auts2 expression starts from early embryonic stages in multiple regions of the central nervous system, but particularly strong prenatal expression is observed in the regions associated with higher brain functions including neocortex, hippocampus, and cerebellum (Bedogni et al., 2010). We previously demonstrated that the AUTS2-Rac1 signaling pathway is required for neuronal migration and subsequent neurite formation in the developing cerebral cortex . However, even at postnatal and adult stages, AUTS2 expression is maintained in various types of neurons (Bedogni et al., 2010). Although this late-stage expression raised the possibility that AUTS2 may also be involved in later neurodevelopmental processes, such as synaptogenesis and synaptic homeostasis, its involvement in synapse regulation remains unknown.
In human patients, AUTS2 mutations are associated with a variety of psychiatric diseases, such as ASD, schizophrenia, depression, intellectual disabilities, and language disability. Although the underlying pathways to evoke this wide range of disorders have not been clarified, one possible mechanism is that different types of gene disruption may cause distinct types of disorders. AUTS2 is a very large gene with multiple exons and many types of gene mutations, such as deletion, duplication, single nucleotide change, and chromosomal translocation, have been reported in humans (Hori and Hoshino, 2017;.
In this study, we show that AUTS2 coordinates excitation/inhibition balance by restricting the number of excitatory synapses during development as well as at post-developmental stages. Targeted disruption of Auts2 resulted in excessive numbers of excitatory synapses without affecting inhibitory ones. Consistent with this, electrophysiological analyses showed that excitatory but not inhibitory inputs increased in the mutant hippocampal neurons where strong c-Fos signals were detected, suggesting impairment in the excitatory and inhibitory coordination in that region. Behavioral analyses on Auts2 heterozygous mutant mice revealed abnormalities in social interaction and altered vocal communication as well as the defects in recognition. Thus, our data suggest that AUTS2 regulates synapse homeostasis by restricting the number of excitatory synapses without affecting inhibitory ones and that loss of AUTS2 function leads to impaired excitatory and inhibitory coordination that may underlie the pathogenesis of some psychiatric illnesses.

Auts2 Restricts the Number of Excitatory Synapses In Vitro
To investigate the involvement of AUTS2 in synapse formation, we utilized primary cultured hippocampal neurons from homozygous Auts2-floxed (Auts2 flox/flox ) embryos. Most excitatory synapses in mammalian brain are formed on dendritic spines (Bhatt et al., 2009). We confirmed that, at 21 days in vitro (DIV21), most PSD-95 (excitatory postsynapse marker) signals were observed on the spine heads ( Figure 1A).
Deletion of Auts2 was carried out by co-introducing GFP with the Cre recombinase expression vector into the Auts2 flox/flox primary hippocampal neurons. Consistent with our previous report, loss of Auts2 resulted in the impairment of dendrite development, as shown by decreased total dendritic length (**p = 0.003, Figures S1A and S1B). Furthermore, Sholl analysis revealed that the Auts2-deficient neurons exhibited a lower dendritic arbor complexity compared with the control neurons (**p = 0.008, Figure S1C) .
Immunostaining revealed that the Auts2-deficient neurons (Auts2 del8/del8 neurons) exhibited a significant increase in the density of dendritic spines compared with the control neurons at DIV28 (***p < 0.001, Figure 1B). Consistent with the increased dendritic spines, Auts2-deficient neurons harbored a larger number of excitatory synapses defined as puncta double-positive for PSD-95 and presynaptic marker synapsin-I than the control at DIV21 (**p = 0.001, Figures 1A and 1C). The larger number of excitatory synapses were already evident at an early culture stage (DIV14) in the mutant neurons (***p < 0.001, Figure 1C). Interestingly, the number of inhibitory postsynapse marker, Gephyrin-positive puncta on the dendrites (C) The number of PSD-95 puncta colocalized with or adjacent to synapsin-I puncta in GFP-positive cells was measured at DIV14 and 21 (DIV14, n = 28 dendrites; DIV21, n = 51 dendrites of 15-22 neurons). (D) The number of Gephyrin-positive puncta colocalized with Gephyrin on the dendrites and axon initiation sites (AIS) were measured at DIV21 (n = 25 dendrites and n = 20 AIS of 20 neurons). (E) WT primary hippocampal neurons were co-electroporated with Auts2-shRNA and the indicated expression vectors and analyzed at DIV22-24. To visualize the neurons, GFP vector was co-electroporated. Graph shows the density of dendritic spines (n = 19-20 dendrites). Expression of the shRNA-resistant FL-AUTS2 (FL-AUTS2 R ) or nuclear-localized form AUTS2 (NLS-AUTS2 R ) in Auts2-knockdown neurons rescues the aberrant spine formation. (F) WT mouse hippocampal neurons at DIV16 expressed with mRFP only (WT) or mRFP plus Auts2 shRNA vector were imaged at the beginning (0 h) and 4 h after the analysis (dashed white circle, spine eliminated; white arrowheads, spines formed). (G) Gain and loss of dendritic protrusions (including spines and filopodia) in WT and Auts2 knockdown neurons were analyzed during a 6-h time window at DIV16-17 (WT, n = 7 neurons; Auts2 shRNA, n = 10 neurons).
We further observed the development of dendritic spines at different stages in culture ( Figure S2A). In control neurons, filopodia were predominantly formed during the first week of culture but gradually decreased from 2 to 4 weeks, with increasing spine formation during the same period. During the first week of culture, Auts2 mutant neurons had a similar number of protrusions including filopodia and spines as control neurons ( Figure S2A: p = 0.300 for filopodia, p = 0.321 for spine). At later stages, however, larger numbers of dendritic spines as well as filopodia were continuously formed in the Auts2 mutant neurons compared with the control neurons ( Figure S2A: DIV14, ***p < 0.001 for filopodia, ***p < 0.001 for spine; DIV21, ***p < 0.001 for filopodia, ***p < 0.001 for spine; DIV28, *p = 0.039 for filopodia, ***p < 0.001 for spine). The Auts2-deficient neurons, however, exhibited the same extent of spine maturation with that of WT neurons, as depicted by the spine maturity index ( Figure S2B: DIV7, p = 0.220; DIV14, p = 0.664; DIV21, p = 0.903; DIV28, p = 0.595) as well as the spine size ( Figure S2C: p = 0.5903 for spine length, p = 0.358 for spine). Furthermore, we observed no significant difference in the PSD-95 puncta size between the control and Auts2 mutant neurons (p = 0.794, Figure S2D). These results suggest that loss of Auts2 does not influence the maturation of dendritic spines.
Next, we introduced the expression vectors for AUTS2 isoforms or possible AUTS2 downstream factors into the Auts2-knockdown neurons . We first confirmed that knockdown of Auts2 well recapitulated aberrant spine formation as observed in Auts2 KO neurons ( Figure 1E: one-way ANOVA, p < 0.001, F (6,133) = 1.781; Dunnett's post hoc test, ***p < 0.001). This abnormality was restored by co-expression of the shRNA-resistant full-length AUTS2 (FL-AUTS2 R ), indicating that excess spine formation is the result of specific knockdown of Auts2 (p = 0.795, Figure 1E). We have previously demonstrated that a cytoplasmic AUTS2-Rac1 signaling pathway is required for neuronal migration in the developing cerebral cortex . In that study, defective cortical neuronal migration in Auts2 KO mice was shown to be rescued by introduction of either NES (nuclear export sequence)-tagged FL-AUTS2 R (NES-FL-AUTS2 R ) ( Figure S2E) or wild-type Rac1 (Rac1-WT). Overexpression of these proteins, however, failed to rescue the aberrant spine formation evoked by Auts2 knockdown ( Figure 1E: *p = 0.013 for Rac1-WT and **p < 0.001 for NES-FL-AUTS2 R ). In contrast, introduction of NLS (nuclear localization signal)-tagged FL-AUTS2 R (NLS-FL-AUTS2 R ) ( Figure S2E) was able to rescue the spine number to levels comparable with that of control neurons (p = 0.999, Figure 1E), whereas the C-terminal AUTS2 short isoforms (S-AUTS2-var.1 and 2) (Figure S3B), which are exclusively localized in nuclei , were not able to rescue the phenotype ( Figure 1E: ***p = 0.001 for S-AUTS2-var.1, **p = 0.008 for S-AUTS2-var.2). These results indicate that nuclear FL-AUTS2 is involved in the control of spine number.
In Auts2 del8/del8 brains, expression of FL-AUTS2 and S-AUTS2-var.1 proteins is eliminated, whereas another C-terminal AUTS2 short isoform variant 2 (S-AUTS2-var.2) is increased , raising a possibility that aberrant synapses in the primary Auts2 del8/del8 hippocampal culture are caused by the overexpression of S-AUTS2-var.2. However, overexpression of S-AUTS2-var.2 into wild-type primary hippocampal neurons did not affect the number and morphology of spines ( Figure S2F: one-way ANOVA, p = 0.521), suggesting that the formation of aberrant number of spines in Auts2 mutant neurons was not due to a gain-of-function effect by increased AUTS2 short isoform expression. Similarly, we also found that FL-AUTS2 or S-AUTS2var.1 did not affect the spine number ( Figure S2F).
Next, we performed live imaging to observe the dynamics of dendritic protrusions including spines and filopodia at DIV16-17. During a 6-h time window, neurons expressing the Auts2 shRNA exhibited a higher rate of protrusion gain (***p < 0.001) and a lower rate of protrusion loss (**p = 0.002) compared with WT neurons (Figures 1F and 1G). Compared with the fixed neurons, a higher number of protrusions were formed in the Auts2-knockdown living neurons during the time-lapse recording ( Figure 1G). This may be attributed to the  iScience Article difference in experimental conditions. Alternatively, the exposure to laser might have caused damage to living neurons during the time-lapse recording, which may affect the dynamics of cell protrusions.
Altogether, these in vitro experiments suggest that AUTS2 restricts the number of excitatory synapses, while not affecting inhibitory neurons.

Loss of Auts2 Results in Excessive Dendritic Spines In Vivo
To assess the involvement of AUTS2 in the regulation of dendritic spine formation in vivo, we generated forebrain-specific Auts2 conditional KO mice by crossing Auts2-floxed mice with Emx1 Cre mice (Iwasato et al., 2000) ( Figure S3A and Table S1) and examined brain tissues by Golgi staining to visualize dendrite morphology. Immunoblotting confirmed that expression of FL-AUTS2 protein in the mutant cerebral cortex was successfully eliminated (arrow in Figure S3C).
Consistent with the increase of spines, immunohistochemical analysis revealed that the excitatory presynaptic marker VGLUT1 but not inhibitory VGAT-labeled puncta at mPFC was increased in Auts2 mutant brains compared with the control mice, suggesting that loss of Auts2 leads to an imbalance of excitatory and inhibitory synapse density ( Figure S5: ***p < 0.001 for VGLUT1, p = 0.070 for VGAT).

Auts2 Deficiency Causes Aberrant Excitatory Neurotransmission
Next, we investigated the effect of Auts2 inactivation on synaptic transmission properties. To address this, we performed whole-cell patch clamp recording of spontaneous miniature excitatory and inhibitory ll OPEN ACCESS iScience 23, 101183, June 26, 2020 5 iScience Article postsynaptic currents (mEPSCs and mIPSCs, respectively) on CA1 pyramidal neurons in acute hippocampal slices from P33-44 mouse brains. In the Emx1 Cre/+ ;Auts2 flox/flox homozygous brains, the mEPSCs were increased in frequency (**p = 0.006), in agreement with increased spines (Figures 3A and 3C). Furthermore, the average paired-pulse ratio of evoked EPSCs in response to paired sets of local stimulation was iScience Article unchanged across the genotypes (p = 0.520, Figure S6), suggesting that the increase in mEPSC frequency observed in Auts2 mutant brains is probably due to an increase in the number of functional excitatory synapses rather than an increase in the probability of presynapse release. On the other hand, the mEPSC in amplitude was unaltered (p = 0.954) compared with the control (Auts2 flox/flox ) mice ( Figures 3A and 3C), suggesting that ablation of Auts2 does not further promote the maturation of excitatory synapses. We also observed no significant difference in the mIPSCs with regard to either amplitude or frequency between the control and Emx1 Cre/+ ;Auts2 flox/flox mutants (Figures 3B and 3D: p = 0.171 for amplitude, p = 0.252 for frequency). iScience Article Furthermore, we examined the expression of the immediate-early gene product, c-Fos, as a marker of neuronal activity in the brain (Sagar et al., 1988). Compared with the control (Auts2 flox/flox ) mice, a larger number of pyramidal neurons with strong c-Fos immunoreactivity were observed in the mPFC and hippocampal CA1 in Emx1 Cre/+ ;Auts2 flox/flox homozygous mutants (Figures 3E and 3F: *p = 0.023 for mPFC, **p = 0.009 for CA1). This suggests that the disturbed balance between excitatory and inhibitory synaptic inputs in local neural circuits results in increased excitability in the Auts2 mutant brains.

Auts2 Prevents Excessive Spine Formation Even after Developmental Stages
Although our ex vivo and in vivo analyses suggest that AUTS2 regulates excitatory synapse formation, it is unclear whether AUTS2 possesses such a function after establishment of brain structures. To assess this issue, we crossed Auts2-floxed mice with CaMKIIa-CreER T2 mice to generate CaMKIIa-CreER T2 ;Auts2 flox mutant mice, in which the exon 8 of Auts2 can be ablated in the forebrain projection neurons by administration of tamoxifen  ( Figure S7A and Table S1). We have previously demonstrated that Auts2 mutant mice displayed defects in neural development including neuronal migration and neurite outgrowth in a gene-dosage dependent manner (Hori et al., 2014, Cell Rep). Interestingly, however, the Emx1 Cre/+ ;Auts2 flox/+ heterozygous mutants exhibited aberrant spine formation to the same extent as the homozygotes ( Figure 2B: p = 0.394 Het (Emx1 Cre/+ ;Auts2 flox/+ ) versus Homo (Emx1 Cre/+ ;Auts2 flox/flox ) for mPFC; p = 0.305 Het versus Homo for CA1; p = 0.631 Het versus Homo for CA1, one-way ANOVA with Bonferroni post hoc test). To better understand the contribution of AUTS2 in postnatal synapse development as well as the Auts2 phenotypes on mouse behaviors as described below, we examined CaMKIIa-CreER T2 ;Auts2 flox/flox homozygotes and Auts2 flox/flox control mice (Figures 4 and S11).

Aberrant Gene Expression in Auts2 Mutant Mice
The ex vivo rescue experiments in Figure 1E showed that AUTS2 in the nucleus functions to restrict the spine number. A previous study clarified that nucleic AUTS2 works as a component of PRC1 to participate in gene transcription (Gao et al., 2014). These findings suggest that AUTS2 protein in nuclei restricts spine formation by regulating gene expression of relevant neural genes. Therefore, we examined global mRNA expression profiles for Emx1 Cre/+ ;Auts2 flox/flox homozygous brains and Auts2 flox/flox control littermate brains. In the postnatal mouse brains, the expression of AUTS2 in the cerebral cortex is downregulated to considerably lower levels and is confined to the prefrontal regions (Bedogni et al., 2010). In addition, the disturbed spine formation elicited by the ablation of Atus2 is specific to the upper-layer neurons in the cerebral cortex ( Figures 2B and S3D). In contrast, the hippocampus entirely sustains a higher level of AUTS2 expression even in mature brains. Thus, we prepared the RNA samples from the hippocampi of 2-week-old Auts2 homozygous mutants and the control littermates for RNA sequencing (RNA-seq) analysis. Through RNA-seq, we identified a total of 168 genes, whose expression levels were significantly altered (false discovery rate [FDR] < 0.05) in the mutant hippocampus, with 78 downregulated and 90 upregulated genes expressed as Log 2 FKPM (fragments per kilobase of exon per million reads mapped) (Figures 5A-5C and Data S1). Interestingly, these differentially expressed genes included the genes encoded synaptic proteins or molecules involved in synaptic functions, such as Reln, Mdga1, Camk2b, Cacna1c, and C1ql-family genes (Fink et al., 2003;Gangwar et al., 2017;Martinelli et al., 2016;Matsuda et al., 2016;Moosmang et al., 2005;Pettem et al., 2013;van Woerden et al., 2009;Wasser and Herz, 2017) (Figures 5B and 5C). Gene ontology (GO) analysis revealed that these altered genes were associated with multiple aspects of neurodevelopment including ''nervous system development,'' ''cell differentiation,'' and ''neuronal migration,'' with particular enrichment of the terms for synapse development such as ''dendritic spine morphogenesis,'' ''negative regulation of synapse assembly,'' and ''regulation of cytosolic calcium ion concentration'' ( Figure 5D and Data S2). Among the genes categorized in GO cellular components such as ''Membrane'' or ''Synapse,'' six up-regulated (e.g., Mdga1, Camk2b, and sema6b) and thirteen down-regulated genes (e.g., Dcc, Gfra1, Gpc2, Hap1) overlapped with genes categorized in the biological process ''nervous system development'' ( Figure S8). These results suggest that nucleic AUTS2 regulates the expression of genes that are related to synapse formation/function and some of which may be involved in spine number restriction. Aberrant expression of such synaptic genes may cause synaptic dysfunction in patients with AUTS2 mutations.

Loss of Auts2 Impairs Social Behaviors
In our previous studies, the heterozygotic mouse mutants for another Auts2 allele, Auts2 neo/+ , whose AUTS2 expression profile is distinct from that of Auts2 del8/+ (Table S1), displayed the behavioral abnormalities in cognition and emotional control while behaving normally in social interaction . Human genetic studies have previously reported that individuals with mutations in AUTS2 locus iScience Article exhibited common features including ID, developmental delay, microcephaly, and epilepsy but distinct psychiatric disorders such as ASDs, ADHD, and schizophrenia . One plausible hypothesis is that the heterogeneity of structural variants in the AUTS2 locus could result in the expression of phenotypic variation between the patients with AUTS2 mutations. This prompted us to examine the social behaviors of Auts2 del8/+ mice, especially focusing on mouse social communications.
We performed the reciprocal dyadic social interaction test to evaluate social behavior, in which mice were allowed to freely move and reciprocally interact with each other . Auts2 del8/+ mice displayed lower levels of active affiliative social interaction than WT mice in both session 1 and session 2 ( Figure 6A: **p = 0.001 for session 1, **p = 0.009 for session 2). Of note, the restricted ablation of Auts2 in mature excitatory neurons in the adult forebrain well recapitulated the impairment of social interaction, as depicted by tamoxifen-treated CaMKIIa-CreER T2 ;Auts2 flox/flox mutants (Figures S11A and S11D: **p = 0.001 for session 1, *p = 0.038 for session 2). Furthermore, in a three-chamber social interaction test, Auts2 del8/+ mutant mice displayed a decreased preference for a social subject (stranger mice 1 and 2) over non-social subject (empty chamber or familiar mouse) compared with WT mice in both sociability and social novelty phases ( Figure 6B). These results suggest that Auts2 mutant mice have social defects. We confirmed that sensory abilities such as olfaction and visual functioning as well as tactile response were not significantly different across the genotypes, as no phenotype was observed in the buried food finding test ( Figure S9A: p = 0.065; Figure S11C: p = 0.707), whisker twitch reflex (100% response in WT, n = 12, Auts2 del8/+ , n = 10, Auts2 flox/flox , n = 10 and CaMKIIa-CreER T2 ;Auts2 flox/flox , n = 10), and visual placing response test (p = 0.898, Figure S9B; p = 0.557, Figure S11B), respectively. To further examine the sensory function of the vibrissae, we measured thigmotactic behaviors, defined as movement along the walls so that one side of the vibrissae could contact and scan the edge of the wall (Luhmann et al., 2005;Milani et al., 1989). Auts2 del8/+ mutant and WT mice behaved similarly in this test ( Figure  In the open field test, the time that Auts2 del8/+ mice spent in the illuminated inner area was comparable with that of WT mice, although general locomotor activity was slightly reduced in Auts2 del8/+ mice as indicated by total travel distance during the test ( Figure S10B: time spent in inner sector, p = 0.697; total distance traveled, ***p < 0.001). In the elevated plus maze test, however, Auts2 del8/+ mice displayed increased exploratory behavior of the open arms compared with WT mice, suggesting that Auts2 del8/+ mice have reduced fear of height (**p = 0.008, Figure 6C).
In a novel object recognition test, Auts2 del8/+ mice exhibited impaired recognition memory performance depicted by the significant decrease of time for exploratory index to the novel object ( Figure 6D: session 3 genotype interaction, F (1,62) = 25.63, p < 0.001; genotype, F (1,62) = 25.15, p = 0.001; session, F (1,62) = 21.74, p < 0.001). Meanwhile, Auts2 del8/+ mice showed normal associative memory functions in the fear-conditioning test ( Figure S10C: context-dependent, p = 0.175; tone-dependent, p = 0.841). Interestingly, Auts2 del8/+ exhibited a higher response to nociceptive stimuli as observed in the Auts2 neo/+ mutants in our previous study (***p < 0.001, Figure S10C) . Furthermore, Auts2 del8/+ exhibited abnormal acoustic startle responses as well as sensorimotor gating deficits as indicated by decrease in the percentage of prepulse inhibition ( Figure 6E: prepulse 3 genotype interaction, F Among types of social behaviors, mouse vocal communication has recently received attention as a possible model for studying the genetic and neural mechanisms for social communication (Holy and Guo, 2005). Mice use ultrasonic vocalizations (USVs) to exchange information in a variety of social contexts (Portfors and Perkel, 2014). When interacting with females, adult WT males actively emit courtship USVs with key tone frequencies between 50 and 80 kHz, as observed in the real-time spectrograms in Figure 7A. In contrast, the USVs produced by Auts2 del8/+ males were apparently dispersive during the test ( Figure 7A). Indeed, the mean number and duration of USVs were markedly reduced in Auts2 del8/+ mice compared with WT controls ( Figure 7B: ***p < 0.001 for call number; ***p < 0.001 for duration). Similarly, CaMKIIa-CreER-T2 ;Auts2 flox/flox males also displayed the altered vocalizations ( Figure S11E: **p = 0.003 for call number; p = 0.058 for duration). The experiments of auditory playback previously showed that adult females prefer USVs with greater complexity from neonates as well as adult males (Chabout et al., 2015;Takahashi et al., 2016). We classified the acoustic structures of USVs into 12 different call patterns and grouped them into ''simple'' iScience Article and ''complicated'' syllable types ( Figure 7C). Auts2 del8/+ emitted significantly fewer numbers of the complicated syllable type, including ''harmonics,'' ''complex,'' or ''one jump + harmonics,'' whereas the simple syllable types with shorter duration such as ''downward'' or ''short'' were significantly increased (Figure 7D: **p = 0.002 for downward; *p = 0.025 for short; **p = 0.001 for complex; *p = 0.022 for harmonics; *p = 0.025 for one jump + harmonics). These findings suggest that loss of Auts2 alters mouse vocal communication, which may underlie the pathology for communication disorders in patients with ASD with AUTS2 mutations.

DISCUSSION
In this study, we found that AUTS2 restricts the number of excitatory synapses in forebrain pyramidal neurons, such as mPFC, and in the hippocampus, which are implicated as the critical regions for socio-communicative and cognitive brain functions. In Auts2 mutant forebrains, the aberrant dendritic spine formation leads to the enhancement of excitatory synaptic inputs, which results in the changes in a balance between iScience Article excitation and inhibition (E/I) that is observed in several otherwise different neuropsychiatric disorders such as ASDs and schizophrenia as well as mouse models (Lee et al., 2017;Penzes et al., 2011). These findings suggest a potential link between the behavioral abnormalities in Auts2 mutant mice and the aberrant dendritic spine development.
Interestingly, in Auts2 mutant cerebral cortex, aberrant spine formation specifically appeared in the upperlayer but not deep-layer neurons, although AUTS2 is widely expressed in both cortical layers ( Figures 2B  and S3D) (Bedogni et al., 2010). One plausible hypothesis is that AUTS2 may have distinct roles for neural development in different cerebral cortical areas, which may depend on differences of AUTS2 isoforms expressed between neurons or on co-factors that differentially interact with each AUTS2 isoform. Electrophysiological experiments revealed that excitatory but not inhibitory synaptic inputs were elevated in the Auts2 mutant hippocampal slices where strong c-Fos signals were observed, implying that the E/I balance was disturbed in that region. E/I balance in neural circuits is tightly controlled and established by contributions from a large number of factors in the normal brain. Accumulating evidence implicates a disturbed E/I balance within cortical neural circuitry in various neuropsychiatric disorders including ASD, anxiety, and ADHD (Chao et al., 2010;Edden et al., 2012;Gogolla et al., 2009;Han et al., 2012;Rubenstein and Merzenich, 2003). Although a recent report suggests that E/I imbalance is not causative for the neuropathology of the disorders but reflects a homeostatic response in some mouse models (Antoine et al., 2019), the hyperexcitability caused by an increased E/I ratio in the cerebral cortex is thought to be one potential common mechanism underlying the neurobehavioral defects of some forms of ASD via a distinct molecular pathway (Lee et al., 2017).
During the spinogenesis, a rapid increase of dendritic spine density occurs in the forebrain neurons, in which the gain of spines exceeds loss of spines, eventually causing excessive excitatory synapses for the formation of neural circuits (Chen et al., 2014;Forrest et al., 2018;Isshiki et al., 2014;Penzes et al., 2011). Thereafter, the growth of excitatory synapses is gradually downregulated and unnecessary spines are selectively pruned, after which spines are maintained during adulthood. Time-lapse imaging experiments using Auts2-knocked-down hippocampal neurons revealed that de novo formation of dendritic spines is promoted, whereas the elimination rate is decreased, resulting in the exaggerated formation of excitatory synapses. These observations suggest an important role for AUTS2 in controlling the number of spines or excitatory synapses in forebrain neurons by modulating their turnover. We found that this excess in synapses was also observed in tamoxifen-treated CaMKIIa-CreERT2;Auts2 flox/flox in which Auts2 was ablated after establishment of the brain structure. This suggests that AUTS2 is involved in regulating synaptic homeostasis at late developmental and/or adult stages.
Emerging evidence indicates that aberrant regulation of spine number and/or an increased excitatory synaptic inputs likely caused by incomplete pruning or exaggerated formation of spines is associated with numerous pathological conditions such as ASD, schizophrenia, and neurodegenerative disorders (Chen et al., 2014;Forrest et al., 2018;Lee et al., 2017;Penzes et al., 2011). Transcriptional control by epigenetic regulation including histone post-translational modification and chromatin remodeling is critical in synapse development and neurological disorders. A recent study by Korb et al. revealed that Fragile X mental retardation protein Fmr1 mutant mice exhibit widespread histone mis-modifications (Korb et al., 2017). These are associated with open chromatin caused by upregulation of epigenetic factor Brd4, resulting in alteration of the transcription levels of many critical synapse-related genes. In this study, we showed that nuclear-localizing AUTS2 functions restrict spine number. Because AUTS2 is involved in transcriptional regulation via chromatin modification as a component of PRC1 (Gao et al., 2014), and because expression of many synapse-related genes was altered in the Auts2 mutants ( Figure 5), we believe that nuclear AUTS2 restricts the excitatory synapse number via controlling the expression of relevant genes, thus maintaining the excitation/inhibition balance of the brain.
In previous and current studies, we characterized behavioral phenotypes for two lines of mutant mice with different mutations disrupting the Auts2 locus . We summarized the results from a behavioral test battery for Auts2 neo/+  and Auts2  iScience Article Figure S10D. In this study, we found that the Auts2 del8/+ heterozygous global KO as well as CaMKIIa-CreER T2 ;Auts2 flox/flox conditional KO mice exhibited autistic-like behaviors including social deficits and altered vocal communications as well as multiple other behavioral impairments. In addition, Auts2 del8/+ mice also showed altered anxiety as well as higher responses against nociceptive and auditory stimuli, both of which are often observed in patients with ASD (American Psychiatric Association, 2013). Interestingly, Auts2 del8/+ mutant mice share several behavioral phenotypes with Auts2 neo/+ mutants but also display a distinct combination of phenotypes ( Figure S10D). Although the mechanisms underlying how different mutations lead to the distinct behavioral phenotypes in mice remains unclear, it is possible that compensatory expression of an AUTS2 C-terminal short isoform (S-AUTS2 var2) in Auts2 del8/+ mutant brains negatively affects social behaviors in the social interaction tests ( Figures 6A and 6B), whereas it alleviates the cognitive dysfunctions displayed in Auts2 neo/+ mutant mice  such as the associative memory formation in fear-conditioning tests ( Figure S10C). Alternatively, structural changes of the Auts2 gene locus in these mutant mice could differentially impact on the expression of other AUTS2 isoforms, leading to the distinctive behavioral phenotypes, although we do not have a direct evidence of this. Further comparative analyses between these Auts2 mutants will help us to understand the physiological function of AUTS2 in synapse development and the pathology of the AUTS2-related psychiatric illnesses.
In humans, it has been reported that multiple types of heterozygous genomic structural variants in the AUTS2 locus including de novo balanced translocation, inversion, or intragenic deletions are associated with a wide range of psychiatric illnesses such as ASDs, ID, ADHD, schizophrenia, and dyslexia, as well as other neuropsychiatric diseases . In addition to the exonic deletions of the AUTS2 locus, some of the genomic structural variants are within non-coding regions including intronic and 5 0 upstream regions, implying that improper and disorganized expression of AUTS2 could be involved in the onset of the disorders. However, it remains largely unclear how different mutations of the same gene contribute to different diseases. Currently, eight computationally annotated AUTS2 isoforms in humans are incorporated in public databases (for example, the UCSC Genome Bioinformatics ([https://genome.ucsc.edu]). However, the study by Kondrychyn et al. revealed that auts2a, the zebrafish ortholog of Auts2, possesses 13 putative unique transcriptional start sites (TTS) and, surprisingly, more than 20 alternative transcripts are potentially produced from this gene locus by the aforementioned TSSs and/or by alternative splicing (Kondrychyn et al., 2017). These findings suggest that mammals including mouse and human could have similar or higher transcriptional complexity for Auts2/ AUTS2 than previously thought. Furthermore, Oksenberg et al. have identified several enhancer regions for the expression of auts2a/Auts2 in zebrafish and mouse brain within the intronic regions of this gene locus . Therefore, structural variants such as genomic deletions within a certain region of Auts2/AUTS2 locus could not only alter the expression of full-length AUTS2 directly but also affect the transcriptional regulation of other AUTS2 isoforms. Different mutations of the AUTS2 gene may differentially alter the temporal and spatial expression profiles of AUTS2 isoforms in various brain regions, which may distinctively affect neurobiological functions, ultimately resulting in the occurrence of multiple types of psychiatric disorders in individuals with AUTS2 syndrome. Our previous and this study, thus, highlighted that two types of Auts2 mutants with different AUTS2 protein expression profiles exhibited overlapping but distinct behavioral abnormalities. This may support the notion that different types of mutations in AUTS2 account for distinct types of neuropsychiatric illnesses. Future comprehensive studies elucidating the regulatory mechanisms for transcription/splicing of Auts2/ AUTS2 as well as neurobiological functions of the distinctive AUTS2 isoforms will help us to understand the pathogenic mechanisms underlying the occurrence of a variety of psychiatric disorders in individuals with AUTS2 mutations and could contribute to therapeutic development for AUTS2-related neurological disorders.
In conclusion, the findings presented here suggest that synaptic regulation by AUTS2 is required for proper social behaviors. Furthermore, our results from the behavioral analyses for Auts2 del/8/+ KO mice provided insight into the involvement of AUTS2 in other higher brain functions such as recognition and emotion. In addition to the AUTS2 function on synapse regulation, AUTS2 is also involved in neuronal migration and neurite formation . Therefore, the other abnormal behaviors observed in Auts2 del/8/+ or Auts2 neo/+ KO mice may partly be caused by the impairments in these developmental processes. Comparative analyses of the different forms of Auts2 mouse mutants will help us to better understand the pathological mechanisms of the psychiatric disorders caused by AUTS2 mutations. iScience Article with CaMKIIa-CreER T2 or other more restricted-expression forms of Cre will be useful for dissecting the distinct neural circuitries involved in these abnormal behaviors.

Limitations of the Study
In this study, we demonstrated that the nuclear AUTS2 controls the number of excitatory synapses in the forebrain pyramidal neurons, possibly by regulating the expression of genes for synapse development and functions. Transcriptome analysis revealed that loss of Auts2 alters the expression levels of multiple synapse-related genes as well as genes for neuronal morphogenesis. The current study, however, does not address the mechanisms underlying the regulation of AUTS2 in the expression of these synapse-related genes. Moreover, the AUTS2 downstream targets that are responsible for dendritic spine development remains to be determined. Electrophysiological experiments reveal that increased dendritic spines caused by Auts2 ablation in mice leads to the enhancement of excitatory synaptic inputs, resulting in a disturbed balance in excitatory and inhibitory synaptic inputs. We have not, however, evaluated the effects on synaptic plasticity such as long-term potentiation/depression. Further studies are required to address these issues to obtain a more complete picture of synaptic pathology caused by AUTS2 mutations.

Resource Availability Lead Contact
Further information and requests for resources should be directed to and will be fulfilled by the Lead Contact, Mikio Hoshino (hoshino@ncnp.go.jp).

Materials Availability
All unique materials generated from this study are available from the Lead Contact with a complete Materials Transfer Agreement.

Data and Code Availability
RNA-seq data have been deposited into GEO database with the accession number GSE134712.

METHODS
All methods can be found in the accompanying Transparent Methods supplemental file.

DECLARATION OF INTERESTS
The authors have declared that no conflict of interest exists. ll OPEN ACCESS Hori, K., and Hoshino, M. (2017). Neuronal migration and AUTS2 syndrome. Brain Sci. 7, 54.
Luhmann, H.J., Huston, J.P., and Hasenohrl, R.U. (2005). Contralateral increase in thigmotactic scanning following unilateral barrel-cortex lesion in mice. Behav. Brain Res. 157, 39-43.          In open field tests, Auts2 del8/+ mutant mice exhibited a decrease in total distance traveled in a test field area for 5 min (right graph) whereas there was no significant difference between genotypes in time spent in an inner area (left graph) as well as the ratio of distance traveled in an inner area scored as the percentage of total distance traveled (middle graph) (WT, n=19, Auts2 del8/+ n=15). (C) Associative memory of WT and Auts2 del8/+ mutant mice was measured by the contextual (Context-dependent) and tone cued (Tone-dependent) fear-conditioning test 24 hrs after the conditioning phase (Conditioning). Freezing responses of Auts2 del8/+ mice during contextual and cued memory test were comparable to WT mice while Auts2 mutant mice exhibited a higher response to lower nociceptive stimuli relative to WT mice (Nociceptive threshold) (WT, n=18, Auts2 del8/+ , n=15).
(D) Summary of the results from behavioral test battery for Auts2 neo/+  and Auts2 del8/+ mutant mice. Data are mean ± SEM and box-and-whisker plots (medians with interquartile range, minimum, and maximum values are represented). **P < 0.01, ***P < 0.001, (A) two-way ANOVA with repeated measures, (B) unpaired t-test, (C) two-way ANOVA with repeated measures in conditioning and Mann-Whitney U test in freezing responses.  Table S1. Summary of Auts2 mutant mouse strains, Related to Figure 2, 4, 6 and 7.

Experimental animals
Rosa26R YFP mouse line (stock no. 006148) was obtained from The Jackson Laboratory.

The buried food finding test
The buried food finding test was carried out as described below (Yang and Crawley, 2009).
Male mice were fasted for 18-24 hrs before testing. Subject mice were individually habituated in a clean cage (45 x 23 x 15 cm) for 5 min. For testing, a food pellet was buried at the end of the cage under 1 cm of wood-chip bedding. Subject mice were placed in the corner opposite to the site of the concealed food pellet. Movement of mice was recorded by video camera and time spent to explore the food pellet was measured by an examiner with stopwatch.

Visual placing response test
The function of the visual system was evaluated by the visual placing response according to the methods by Metz and Schwab (Metz and Schwab, 2004). In this test, the test mouse was suspended by its tail and lowered toward a solid object without any contact to the vibrissae.
When the head of a mouse approaches near the edge of the object, the mouse normally raises its head and extends the forelimbs to place them onto the object. The procedure was conducted by three trials and the mean response was rated with the following scoring system: 0 indicates no observable placing behavior, 1 represents a weak or delayed placing response and 2 points indicates a clear placing reaction.

Whisker twitch reflex
The whisker twitch reflex was tested by approaching from behind and lightly touching one set of vibrissae, eliciting head turning to the side on which the vibrissae was touched (Miyakawa et al., 2001).

Thigmotaxis
Mice were placed in the center of the test chamber (26 cm x 26 cm x 40 cm) under moderately bright light conditions (100 lux) and allowed to explore it. Each 20 min session was monitored by video camera and analyzed in four 5 min bins. Time spent in the marginal area defined as a 4 cm band extending from the wall was measured by examiner with a stopwatch.

Locomotor activity
Spontaneous exploratory locomotion was examined as fallows . Mice were individually placed in a transparent acrylic cage with a black frosted Plexiglas floor (25×25×20 cm) under moderate light conditions (15 lux), and locomotor activity was measured every 5 min for 60 min using digital counters with an infrared sensor (BrainScience Idea, Osaka, Japan).

Open field test
Mice were placed in the center of the test chamber (diameter, 60 cm; height, 35 cm) under moderate light conditions (60 lux) and allowed to explore it for 5 min, while their activity was automatically analyzed using the ethovision automated tracking program (Brainscience Idea Co. Ltd., Osaka, Japan) (Lee et al., 2005).

Reciprocal social interaction test
Reciprocal social interaction test was performed as described below  Behaviors were monitored with video camera and time spent in active behaviors were analyzed by examiner with a stopwatch. Active social behaviors included aggressive forms (i.e. wrestling, boxing, kicks, mounting, tail rattle, bites, sideway offense and pursuit) and affiliative forms (i.e. olfactory investigation and allogrooming).

Three-chamber social interaction test
A three-chamber arena was used to examine social approach and preference for social novelty as follows (Nadler et al., 2004). During habituation, empty cylinders were placed in each end chamber. The test subject was placed in the center chamber and its behavioral approach to the chambers was monitored for 10 min. During the sociability test, an unfamiliar male C57BL/6N mouse (stranger 1) that had no prior contact with the test mouse was put in one of the empty chambers, and the behavioral approach to the empty chamber and stranger 1 was monitored for 10 min. During the social novelty test, new unfamiliar male C57BL/6N mouse (stranger 2) was placed in another chamber, and the behavioral approach to the stranger 1 and stranger 2 was monitored for additional 10 min. The amount of time spent in each arena was measured by an ethovision automated tracking program (Noldus, Wageningen, Netherlands).

Novel object recognition test
A novel object recognition test was carried out as described below (Nagai et al., 2007). Mice The objects were a golf ball, wooden cylinder, and square pyramid, which were different in shape and color. An animal was considered to be exploring the object when its head was facing the object or it was touching or sniffing the object. The time spent exploring each object was recorded by using video camera and analyzed in a double-blind manner. During retention sessions, mice were placed back into the same box 24 h after the training session, one of the familiar objects used during training session was replaced by a novel object, and the mice were allowed to explore the two objects freely for 5 min. The exploratory index in the retention session, the ratio of the amount of time spent exploring the novel object to the total time spent exploring both objects, was used to measure cognitive function. In the training session, the preference index was calculated as the ratio of time spent exploring the object that was replaced by a novel object in the retention session to the total exploration time.

Cued and contextual fear conditioning test
Cued and contextual fear conditioning test was carried out as described below (Ibi et al., 2010). Training took place in the chamber (30 × 30 × 40 cm) equipped with a metal floor and a 15-sec white noise tone (85 dB) was delivered (conditioned stimulus). During the last 5 sec of the tone stimulus, a foot shock of 0.8 mA was delivered through a shock generator as an unconditioned stimulus (Brainscience Idea Co. Ltd., Osaka, Japan). This procedure was repeated four times at 15-sec intervals. Twenty-four hr after conditioning, the context-dependent test was performed. For the context-dependent test, each mouse was put in the training chamber, and the freezing response was monitored for 2 min in the absence of the conditioned stimulus. Tonedependent testing was performed 4 hr after the context-dependent test. For the tonedependent test, the freezing response was measured for 1 min in a standard transparent rectangular rodent cage (25 × 30 × 18 cm) in the presence of a continuous-tone stimulus identical to the conditioned stimulus using mice that had been subjected to the contextdependent test.

Prepulse inhibition (PPI) test
The PPI test was carried out as follows (Takahashi et al., 2007). The animals were placed in the chamber (San Diego Instruments, San Diego, California) and were habituated for 10 min. During the habituation time, 65 dB background white noise was delivered. Mice then received 10 startle trials, 10 no-stimulus trials and 40 PPI trials. The intertrial stimulus intervals were between 10 and 20 sec and the total session lasted 17 min. Mice were presented with a single 120 dB white noise burst lasting 40 msec during the startle trial. PPI trials consisted of a prepulse (20 msec burst of white noise at 69, 73, 77 or 81 dB intensity) followed, 100 msec later, by the startle stimulus (120 dB, 40 msec white noise). Each of the four prepulse trials (69, 73, 77 or 81 dB) was performed 10 times. Sixty different trials were pseudo-randomly delivered, ensuring that each trial was carried out 10 times and that no two consecutive trials were identical. The resulting movement of the animal in the startle chamber was measured for 100 msec after startle stimulus onset (sampling frequency 1 kHz), rectified, amplified and fed into a computer, which calculated the maximal response over the 100 msec. Basal startle amplitude was determined as the mean amplitude of the 10 startle trials. PPI was calculated according to the following formula: 100 × [1−(PPx/P120)] %, in which PPx is the mean amplitude of the 10 PPI trials (PP69, PP73, PP75 or PP80) and P120 is the basal startle amplitude.

Ultrasonic vocalizations
Ultrasonic Vocalizations were recorded using an UltraSoundGate system (Avisoft bioacoustics, Glienicke, Germany) composed of a CM16/CMPA condenser microphone, Avisoft-UltraSoundGate 116H computer interface, and Avisoft Recorder software with a sampling rate of 400 kHz. A microphone was hung 16 cm above the floor of a sound attenuating chamber. For the test, male mice were individually housed in Plexiglas cages (23 cm x 16 cm x 12 cm) for a week prior to test time to acclimate to the testing environment. Unfamiliar wild type three month old C57BL6/N female mice were placed into the test male cage and recordings begun after USV was detected and continued for 1 min period.

Syllable analysis
Vocal signals recorded in wav files were automatically detected by MATLAB-based software USVSEG with modification to mouse USVs (Tachibana et al., 2014). This software segments each syllable and exports as individual jpeg files. As Tachibana et al reported (Tachibana et al., 2020), USVSEG can detect correct vocal signals with approximately 95% accuracies compared to the information that was manually defined by a human examiner. The number of USVs and duration of each call are automatically detected. By observing jpeg files, experimenters then manually excluded the files of vocalizations that includes only click-like sounds without any tone-like signals or that could not be classified into any of the call types as noises (false positive).
The vocalizations were manually categorized into 12 types observing these jpeg files based on the previously published criteria with some modification (Kikusui et al., 2011;Yasumura et al., 2014). In the previous criteria, syllables including both jumps and harmonics were classified into One Jump or More jump. In our present methods, such syllables were classified in more detailed manner in order to demonstrate more clearly whether such Jumps include harmonics or not. In addition, these 12 call types were grouped into "simple" and "complicated" syllable types based on call duration, frequency modulation and the presence/absence of harmonics or jumps.
The call classifications we used are as follows; Upward. Syllables with upwardly modulated frequency change (> 5 kHz) Downward. Syllables with downwardly modulated frequency change (> 5 kHz).
Short. syllables which is shorter than or equal to 5 msec.
Chevron. Syllables with an upsweep (greater than 5 kHz) followed by a down-sweep (greater than half of the frequency change of the upsweep) or reversed one, formed like a U or a reversed U.
Wave. Syllables with two directional changes in frequency (> 5 kHz).
Complex. Syllables with three or more directional changes in frequency (> 5 kHz).