Autism spectrum disorder causes, mechanisms, and treatments: focus on neuronal synapses

Autism spectrum disorder (ASD) is a group of developmental disabilities characterized by impairments in social interaction and communication and restricted and repetitive interests/behaviors. Advances in human genomics have identified a large number of genetic variations associated with ASD. These associations are being rapidly verified by a growing number of studies using a variety of approaches, including mouse genetics. These studies have also identified key mechanisms underlying the pathogenesis of ASD, many of which involve synaptic dysfunctions, and have investigated novel, mechanism-based therapeutic strategies. This review will try to integrate these three key aspects of ASD research: human genetics, animal models, and potential treatments. Continued efforts in this direction should ultimately reveal core mechanisms that account for a larger fraction of ASD cases and identify neural mechanisms associated with specific ASD symptoms, providing important clues to efficient ASD treatment.


INTRODUCTION TO AUTISM SPECTRUM DISORDER
Autism spectrum disorder (ASD) is a group of developmental disabilities characterized by abnormal social interaction and communication, and stereotyped behaviors with restricted interest. Autism was first reported by Kanner (1943) with a clinical description of 11 children showing "extreme aloneness from the very beginning of life, not responding to anything that comes to them from the outside world." He proposed the behavioral combination of autism, obsessiveness, stereotypy, and echolalia as childhood schizophrenia. However, until the 1980s, ASD was not accepted as an individual developmental disorder with a biological origin. In the early 1980s, studies demonstrated the high heritability of ASD and its association with other genetic syndromes (Gillberg and Wahlstrom, 1985;Wahlstrom et al., 1986), providing compelling evidence for a genetic etiology of ASD and fueling the conceptualization of autism as a distinct neurodevelopmental disorder. From the definition of "childhood or early-onset schizophrenia" put forward by Kanner, autism was renamed "infantile autism" in 1980Kanner, autism was renamed "infantile autism" in , "autism disorder" in 1987 and, more recently, "autism" or the umbrella term "ASD".

DIAGNOSIS
Currently, ASD is included in the diagnostic category of a neurodevelopmental disorders in the Diagnostic and Statistical Manual of Mental Disorders V (Grzadzinski et al., 2013). The diagnosis of autism is mainly based on the presence of two major aforementioned symptoms: social-communication deficits, and restricted and repetitive interests/behaviors (Grzadzinski et al., 2013). These symptoms must be shown from early childhood of individuals with ASD. But autism is also associated with various comorbidities, including sensory and motor abnormalities, sleep disturbance, epilepsy, attention deficit/hyperactivity disorder (ADHD)-like hyperactivity, intellectual disability, and mood disorders such as anxiety and aggression (Goldstein and Schwebach, 2004;Simonoff et al., 2008;Geschwind, 2009). Some monogenic syndromes including fragile X syndrome and Rett syndrome also have autistic features, while we should be cautious to directly interpret the disorders as autism since the major symptoms for these syndromes are intellectual disabilities.

PREVALENCE
An early study conducted in the UK in 1966 reported a prevalence rate of autism of 4.5 in 10,000 children (Lotter, 1966). The estimated prevalence increased to 19 in 10,000 American children in 1992 and rose steeply to 1 in 150 in 2002(Autism et al., 2007 and 1 in 110 in (Autism et al., 2009) (see also data from the US Centers for Disease Control and Prevention [CDC]). The currently accepted prevalence of ASD, based on consistent reports of ASD prevalence by multiple sources in different populations, is ∼1% worldwide, placing this disorder as one of the most common pervasive developmental disorders and elevating public concerns.

GENETICS
On the basis of numerous studies that have been undertaken to elucidate the pathogenic mechanisms underlying ASD, it is widely accepted that ASD is a disorder with strong genetic components. In support of this notion, the concordance rates for autism reach up to 90% in monozygotic twins and 10% in dizygotic twins (Rutter, 2000;Folstein and Rosen-Sheidley, 2001;Veenstra-Vanderweele et al., 2003).

ANIMAL MODELS FOR ASD
Animal models of human diseases need to satisfy three major criteria; face validity, construct validity, and predictive validity. Animal models for ASD should display behavioral abnormalities, including impaired sociability, impaired social communication, and repetitive and restricted behaviors (face validity). These models should share analogous genetic or anatomical impairments with humans (construct validity), and show similar responses to the medications used to treat ASD in humans (predictive validity).
Dedicated efforts of many behavioral neuroscientists including Jacqueline Crawley led to the establishment of several well-known assays for rat/mouse models of ASD (Silverman et al., 2010b). Examples include 3-chambered test to assess sociability and social novelty recognition of rodents, ultrasonic vocalization (USV) test to measure the communication patterns of rodents, T-maze test for restricted interests, and home cage behavior or marble burying assay for repetitive behaviors. Through these assays, many genetic and non-genetic animal models of ASD have been characterized and used to identify the etiology of ASD and develop novel treatments (see Tables 3-6 for four different groups of ASD models).
Although animal models are useful for exploring ASD mechanisms and testing novel interventions, we should be cautious in interpreting the results from animal models of ASD because what we are observing in animals are behavioral features that look similar to some of the ASD symptoms in humans. This notion partly stems from the fact that the brains of humans and rodents are fundamentally different. For instance, there are small but significant differences in gene expression patterns in the cerebral cortex in different species (Zeng et al., 2012), suggesting that the   same cell types in different species may have different functions. Moreover, the size, structural complexity, and neural connectivity of the human brain are much greater than those in rodent brains. These functional and anatomical differences between species may create difficulties in translating the ASD-related mechanisms identified in model organisms into human applications. However, some fundamental aspects of the neural mechanisms identified in animal models such as alterations in synaptic transmission, excitation-inhibition balance, and neuronal excitability might be conserved across species and translatable. In addition, given that stem cell technologies are rapidly improving, it is becoming easier for the changes observed in rodent neurons to be compared with those in human neurons derived from individuals with neuropsychiatric disorders (Brennand et al., 2011).

POTENTIAL MECHANISMS UNDERLYING ASD
Mechanisms underlying autism have been extensively studied using various approaches. Neuroanatomical studies have reported macrocephaly and abnormal neuronal connectivity in autistic individuals, while genetics studies using mouse models have implicated a variety of neuronal proteins in the development of ASD. More recently, defects in a number of synaptic proteins have been suggested to cause ASD via alterations in synaptic structure/function and neural circuits, suggesting that "synaptopathy" is an important component of ASD.

NEUROANATOMICAL ABNORMALITIES
A change frequently observed in the brains of individuals with ASD is the overgrowth of the brain termed macrocephaly, which is observed in ∼20% of autistic children (Bolton et al., 2001;Courchesne, 2002;Courchesne et al., 2003Courchesne et al., , 2007Fombonne et al., 1999;Hazlett et al., 2005). Aberrations in cytoarchitectural organization in autistic brains are observed during early brain development in regions including the frontal lobe, parietotemporal lobe, cerebellum, and subcortical limbic structures (Fombonne et al., 1999;Bolton et al., 2001;Courchesne, 2002;Courchesne et al., 2003Courchesne et al., , 2007Hazlett et al., 2005). The cerebellum is a strong candidate for anatomic abnormalities in autism (Courchesne, 1997(Courchesne, , 2002. Magnetic resonance imaging (MRI) studies have found hypoplasia of the cerebellar vermis and hemispheres, and autopsy studies have reported a reduction in the number of cerebellar Purkinje cells. In line with these anatomical changes, cerebellar activation is significantly reduced during selective attention tasks (Allen and Courchesne, 2003), whereas it is enhanced during a simple motor task (Allen et al., 2004). Although the putative role of the cerebellum in ASD has been restricted to sensory and motor dysfunctions, it is becoming increasingly clear that the cerebellum is associated with the core symptoms of autism.
In support of this notion, selective deletion of Tsc1 (tuberous sclerosis 1) in cerebellar Purkinje cells is sufficient to cause all core autism-like behaviors in mice in association with reduced excitability in Purkinje cells (see also Table 4 for summary of syndromic ASD models) (Tsai et al., 2012). In addition, mice lacking the neuroligin-3 gene (Nlgn3 −/− mice), another autism model with an Nlgn3 deletion identified in autistic patients, show occluded metabotropic glutamatergic receptor (mGluR)dependent long-term depression (LTD) at synapses between parallel fibers and Purkinje cells in association with motor coordination deficits (see also Table 5 for summary of synaptopathy ASD models) (Baudouin et al., 2012). Both synaptic and behavioral perturbations are rescued by Purkinje cell-specific re-expression of Nlgn-3 in juvenile mice, suggesting the interesting possibility that altered neural circuits can be corrected after completion of development.
The cerebral cortex is another brain region frequently affected in ASD. Abnormal enlargement or hyperplasia of the cerebral cortex has been reported in MRI studies on young children with ASD (Sparks et al., 2002;Herbert et al., 2003). Because frontal and temporal lobes are important for higher brain functions including social functioning and language development, these anatomical anomalies are likely to underlie the pathophysiology of autism.
The amygdala and hippocampus are subcortical brain regions associated with ASD (Aylward et al., 1999;Schumann et al., 2004;Schumann and Amaral, 2006). Some studies have reported that the autistic amygdala exhibits early enlargement, whereas others have reported a reduction in neuron numbers and amygdala volume. Increases and decreases in the volume of hippocampus are also associated with ASD.
Aberrant connectivity is an emerging theory to account for anatomical abnormalities in autism. Neuroimaging techniques, such as diffusion tensor imaging (DTI) and functional MRI (fMRI), have suggested that ASD involves abrogation of white matter tracts in brain regions associated with social cognition, such as the prefrontal cortex, anterior cingulate cortex, and superior temporal regions (Barnea-Goraly et al., 2004;Minshew and Williams, 2007). Alterations in connectivity across diverse brain regions associated with language, working memory, and social cognition have also been linked to autism. In general, it appears that autism subjects display local over-connectivity and longrange or inter-regional under-connectivity (Herbert et al., 2003(Herbert et al., , 2004Baron-Cohen and Belmonte, 2005;Herbert, 2005;Just et al., 2007). Potential ASD-related neural circuitries have also been proposed based on animal studies. Shank3b −/− mice, which exhibit autistic-like behaviors, have striatal dysfunctions (Table 5) (Peca et al., 2011). In addition, a shift in the balance between excitation and inhibition (E-I balance) toward excitation in the mouse medial prefrontal cortex (mPFC) induced by optogenetic stimulation causes sociability impairments (Yizhar et al., 2011). These results suggest that the striatum and mPFC are components of ASD-related neural circuits.
Although various neuroanatomical defects are observed in autistic brains, a direct linkage between neuroanatomical anomalies and behavioral symptoms of ASD remains to be elucidated. Uncovering the detailed circuitries underlying autistic behaviors would help us understand higher cognitive functions, such as language and sociability.

EXTRACELLULAR FACTORS
It has been found that growth factors and neurotrophic factors are associated with ASD. Genetic and protein expression studies have shown that MET, a transmembrane receptor for hepatocyte growth factor (HGF) with tyrosine kinase activity, is associated with ASD. Genetic variations including rs1858830 in the promoter region that abrogate MET transcription are associated with ASD in Italian and American families and case/control studies, and the levels of MET mRNA and protein are reduced in the cortex of autistic patients (Campbell et al., 2007(Campbell et al., , 2006. However, this association between rs1858830 and ASD failed to replicate in another study (Sousa et al., 2009). By binding to MET, HGF acts as a neurotrophic factor for neurons to influence neurite outgrowth and dendritic morphology (Figure 1) (Powell et al., 2001(Powell et al., , 2003Sun et al., 2002;Gutierrez et al., 2004), implicating abnormal neuronal structures in ASD pathology.
WNT2 is a secreted growth factor that has been linked to ASD. Acting through the canonical Wnt pathway, WNT2 triggers a signal transduction cascade mediated by Dishevelled (Dvl1). WNT2 is a critical regulator of multiple biological functions, including embryonic development, cellular differentiation, and cell-polarity generation. It also regulates neuronal migration, axon guidance, and dendrite branching (Figure 1) (Logan and Nusse, 2004). Multiple lines of evidence have implicated the WNT2 locus in ASD: the WNT2 gene is located at the autismsusceptibility chromosomal locus 7q31 (Vincent et al., 2000;Warburton et al., 2000), and single nucleotide polymorphisms (SNP; rs3779547, rs4727847, and rs3729629, in a case/control study in a Japanese population) and several WNT2 locus variants (R299W and L5R, in autism-affected sibling pair [ASP] and trio families) are associated with autism (Wassink et al., 2001;Marui et al., 2010), although a subsequent study in Han Chinese trios failed to replicate the SNP association with ASD (Chien et al., 2011). While the majority of Wnt2 −/− mice are lethal (Goss et al., 2009), null mutants of Dvl1 show deficits in nest building and home-cage huddling (see also  of non-synaptopathy ASD models) (Lijam et al., 1997;Long et al., 2004). Moreover, the Wnt signaling pathway is associated with and is regulated by chromodomain-helicase-DNA-binding protein 8 (CHD8; Figure 1), de novo mutations of which are repeatedly detected in autistic patients (Neale et al., 2012;O'Roak et al., 2012b;Sanders et al., 2012). Brain-derived neurotrophic factor (BDNF) is associated with ASD. BDNF is a member of the neurotrophin family of growth factors that supports neurogenesis, axodendritic growth, neuronal/synaptic differentiation, and brain dysfunctions (Figure 1) (Huang and Reichardt, 2001;Martinowich et al., 2007). Elevated levels of BDNF were reproducibly found in the sera of Japanese and American autistic individuals (Connolly et al., 2006;Miyazaki et al., 2004). Another clue comes from calcium-dependent secretion activator 2 (CADPS2), a calcium binding protein in the presynaptic nerve terminal that interacts with and regulates exocytosis of BDNF-containing dense-core vesicles (Figure 1) (Cisternas et al., 2003). CADPS2, located at the autismsusceptibility locus on chromosome 7q31, is abnormally spliced in autism patients, and Cadps2 −/− mice exhibit social interaction deficits, including maternal neglect ( Table 5) (Sadakata et al., 2007). Hence, although it is unclear how BDNF contributes to autism pathogenesis, evidence for its role in ASD is becoming clear.
Reelin is also involved in autism. Reelin is a large secreted extracellular matrix glycoprotein that acts as a serine protease for the extracellular matrix, a function that is essential for neuronal migration, cortical patterning, and brain development (Figure 1) (Forster et al., 2006). The RELN gene is located in an autism susceptibility locus on chromosome 7q22, and triplet GGC repeats in 5 untranslated regions (5 UTR) in the RELN gene have been associated with autism in a Caucasian population (Persico et al., 2001;Skaar et al., 2005) (Table 2). Expression levels of Reelin are decreased in postmortem autism brains (Fatemi et al., 2005). Reelin has also been implicated in pathogenesis of various neuropsychiatric disorders, including schizophrenia, bipolar disorder, lissencephaly, and epilepsy (Fatemi, 2001). Reeler mice, with a 150-kb deletion of the Reln gene, exhibit deficits in motor coordination, increased social dominance, and learning and memory impairments ( Table 6) (Salinger et al., 2003;Lalonde et al., 2004).

TRANSCRIPTION FACTORS
Syndromic forms of ASD frequently involve transcription factors. This is likely because defective transcription factors have Frontiers in Molecular Neuroscience www.frontiersin.org August 2013 | Volume 6 | Article 19 | 9 significant influences on many genes and their downstream molecules, affecting diverse neuronal functions. MeCP2 (X-linked gene methyl CpG binding protein 2) is one of the best examples. It is a member of a large family of methyl-CpG binding domain (MBD) proteins that selectively binds to methylated DNA and represses gene transcription (Figure 1) (Bienvenu and Chelly, 2006). Its downstream targets encompass ASD-related genes such as BDNF and CDKL5. Mutations in MeCP2 are the major cause of Rett syndrome, a progressive neurodevelopmental disorder with autistic features (Amir et al., 1999;Bienvenu and Chelly, 2006;Chahrour and Zoghbi, 2007). Mecp2-null mice, an animal model for Rett syndrome, recapitulate most symptomatic traits of Rett syndrome such as respiratory dysfunction, forelimb and hindlimb clasping stereotypy, motor dysfunction, tremor, hypoactivity, anxiety, cognitive impairments, and altered sociability ( Table 4) (Shahbazian et al., 2002;Moretti et al., 2005).
Engrailed-2 is a homeodomain transcription factor associated with ASD. Engrailed-2 is involved in a diverse range of biological processes from embryological development and segmental polarity to brain development and axon guidance (Figure 1) (Brunet et al., 2005;Joyner, 1996). The Engrailed-2 gene on human chromosome 7q36 is in the autism susceptibility locus, and an association between two intronic SNPs rs1861972 and rs1861973 at Engrailed-2 locus and ASD has been repeatedly identified in 518 ASD families (Gharani et al., 2004;Benayed et al., 2005) ( Table 2). However, these SNPs were not found to be associated with ASD in Han Chinese trios (Wang et al., 2008). This association between Engrailed-2 and ASD was further confirmed by animal model studies, which showed Engrailed-2 null mice display social dysfunction and cognitive impairments ( Table 6) (Brielmaier et al., 2012). Because Engrailed-2 is expressed upon activation of WNT2-Dvl1 signaling, it appears that the WNT2-Dvl1-Engrailed-2 pathway, which regulates neuronal migration and axonal guidance, may significantly contribute to ASD pathogenesis via neuroanatomical abnormalities. In addition, a base substitution (A218G) mutant of HOXA1, another homeobox gene, was reported in autistic individuals (Ingram et al., 2000b), indicating the importance of homeobox genes in normal brain function and ASD.

EXCITATORY AND INHIBITORY IMBALANCE
Mutations identified in important synaptic molecules including neuroligins (Jamain et al., 2003), neurexin (Autism Genome Project et al., 2007;Kim et al., 2008) and Shank (Durand et al., 2007;Berkel et al., 2010;Sato et al., 2012) in autistic subjects have prompted investigations into exploring the roles of synaptic dysfunctions in ASD pathogenesis. This "synaptopathy" model of autism has provided much insight into the field ( Table 5).
Abnormal GABAergic system is also proposed as a potential mechanism for ASD. Reduced expression levels in a ratelimiting enzyme for GABA synthesis, glutamic acid decarboxylase (GAD), and GABA receptors with altered subunit composition were observed in autistic brains (Fatemi et al., 2002(Fatemi et al., , 2010. Furthermore, linkage disequilibrium and transmission disequilibrium between GABRB3, a gene encoding the β3 subunit of GABAα receptors, with Angelman syndrome and autism has been reported (Cook et al., 1998;Bass et al., 2000;Buxbaum et al., 2002) ( Table 2).
The serotonergic system would also play a role in ASD pathogenesis by modulating the E-I balance. Serotonin levels in blood or urine are increased in subjects with autism (Cook and Leventhal, 1996;Burgess et al., 2006), and various genes in the serotonin system are linked to autism. Among them are genes encoding the serotonin transporter 5-HTT (transmission disequilibrium at the 5-HTT locus in 86 autism trios) (Cook et al., 1997), and a rate-limiting enzyme for serotonin synthesis TPH2 (two intronic SNPs rs4341581 and rs11179000 at introns 1 and 4, respectively, have been associated with autism) (Coon et al., 2005) ( Table 2).
Neurexins and neuroligins are synaptic cell adhesion molecules enriched at pre-and post-synaptic membranes, respectively (Figure 1) (Craig and Kang, 2007;Sudhof, 2008). Specific interactions between neurexins and neuroligins regulate various aspects of both excitatory and inhibitory synaptic development and function, affecting the E-I balance in postsynaptic neurons. Many mutations in genes encoding neurexins (including hemizygous CNV deletions and missense mutations) and neuroligins (e.g., R451C for NLGN3 and a frameshift insertion mutation for NLGN4) have been associated with ASD, intellectual disability, and schizophrenia (Jamain et al., 2003;Laumonnier et al., 2004;Autism Genome Project et al., 2007;Kim et al., 2008;Walsh et al., 2008) (Table 2). Neuroligin3 knockin mice with the R451C mutation found in autistic patients recapitulate autistic features including moderately impaired sociability ( Table 5) (Tabuchi et al., 2007). Notably, inhibitory transmission was enhanced in the cortical regions of the mutant brains of these mutant mice, suggesting that disrupted E-I balance may contribute to ASD.
SHANK family genes encode scaffolding proteins enriched in the postsynaptic density (PSD), a postsynaptic membrane specialization composed of multi-synaptic protein complexes (Figure 1) (Sheng and Kim, 2000). The Shank family contains three known members, Shank1, Shank2 and Shank3, also known as ProSAP3, ProSAP1, and ProSAP2, respectively. The idea that Shanks are involved in the etiology of ASD firstly emerged from Phelan-McDermid syndrome (PMS) or 22q13 deletion syndrome, a neurodevelopmental disorder caused by a microdeletion on chromosome 22 (Boeckers et al., 2002;Wilson et al., 2003;Phelan and McDermid, 2012). The association between SHANK and ASD became evident by identifying numerous mutations including de novo frameshift, truncating, and missense mutations on SHANK3 locus in autistic individuals (Durand et al., 2007) (Table 2). Mutations in SHANK2 and SHANK1 including de novo CNV deletions and missense mutations in Canadian and European populations were also identified in individuals with ASD and intellectual disability (Berkel et al., 2010;Leblond et al., 2012;Sato et al., 2012). Multiple lines of transgenic mice with Shank mutations found in human patients have been reported. Shank3 heterozygous mice show sociability deficits and reductions in miniature excitatory postsynaptic currents (mEPSC) amplitude and basal synaptic transmission (Bozdagi et al., 2010); mice with deletion of exon 4-9 of Shank3 are socially impaired and exhibit alterations in dendritic spine morphology and activity-dependent surface expression of AMPARs ; Shank1 −/− mice display reduced basal synaptic transmission in the hippocampal CA1 region and reduced motor function and anxiety-like behavior, although they show normal sociability (Hung et al., 2008;Silverman et al., 2011); mice expressing Shank2-R462X in hippocampal CA1 neurons exhibit cognitive dysfunction accompanied by reduced mEPSC amplitude and changes in neuronal morphologies ( Table 5) (Berkel et al., 2012).
CNTNAP2, a neuronal transmembrane protein, is a member of the neurexin family localized at juxtaparanodes of myelinated axons. Here, CNTNAP2 regulates neuron-glia interactions and potassium channel clustering in myelinated axons (Figure 1) (Poliak et al., 1999). Several SNPs (e.g., rs2710102, rs7794745, rs17236239) and nonsynonymous variants (e.g., I867T) in CNTNAP2 locus were found to be associated with ASD, language impairment, and cortical dysplasia-focal epilepsy syndrome in humans (Alarcon et al., 2008;Arking et al., 2008;Bakkaloglu et al., 2008;Vernes et al., 2008) (Table 2). In a case-control association study in Spanish autistic patients and controls, however, CNTNAP2 SNPs rs2710102 and rs7794745 did not associate with ASD (Toma et al., 2013). Cntnap2 −/− mice recapitulate all three core symptoms of autism, and display abnormal neuronal migration, reduced number of GABAergic interneurons, and abnormal neuronal synchronization (Table 4) (Penagarikano et al., 2011). Excessive grooming and hyperactivity in these mice were restored by the treatment of the antipsychotic risperidone (Table 4), suggesting the possibility of therapeutic intervention for certain symptoms of autism.
SynGAP is a GTPase-activating protein for the Ras small GTPase. SynGAP directly interacts with PSD-95, and negatively regulates the Ras-MAPK signaling pathway, excitatory synapse development, and synaptic transmission and plasticity (Figure 1) (Chen et al., 1998;Kim et al., 1998). In humans, de novo mutations of SYNGAP1 have been associated with intellectual disability and autism . In addition, a genetic case/control study in European populations associates a rare de novo copy number variation in SYNGAP1 with ASD . Syngap1 heterozygous mice show schizophrenia-like phenotypes including hyperactivity, impaired sensory-motor gating, impaired social memory and fear conditioning, and preference to social isolation (Guo et al., 2009) (Table 5). In a more recent study, Syngap1 heterozygous mice showed premature dendritic spine development together with enhanced hippocampal excitability and abnormal behaviors, suggesting that over-paced excitatory synaptic development during a critical time window of postnatal brain development causes intellectual disability and ASD (Clement et al., 2012).
Several genes associated with X chromosome-linked intellectual disability (XLID) and synaptic regulations have been associated with ASD. One of them is interleukin 1 receptor accessory protein-like 1 (IL1RAPL1) that encodes a synaptic transmembrane protein (Carrie et al., 1999). Recently, a systematic sequencing screen of X chromosomes of ASD-affected individuals has identified a de novo frameshift mutation in IL1RAPL1 (Piton et al., 2008). IL1RAPL1 plays an important role in the formation and stabilization of excitatory synapses by recruiting the scaffolding protein PSD-95 to excitatory postsynaptic sites through the JNK signaling pathway (Figure 1) (Pavlowsky et al., 2010). In addition, IL1RAPL1 induces the presynaptic differentiation through its trans-synaptic interaction with protein tyrosine phosphatase δ (PTPδ) (Figure 1) (Valnegri et al., 2011b;Yoshida et al., 2011). This interaction between IL1RAPL1 and PTPδ recruits RhoGAP2 to the excitatory synapses and induces dendritic spine formation (Valnegri et al., 2011b). Interestingly, IL1RAPL1 regulates the development of inhibitory circuits in the cerebellum, an ASD-related brain region, and disrupts the excitatory and inhibitory balance, as determined by a study using Il1rapl1 −/− mice (Gambino et al., 2009). These results suggest that IL1RAPL1 is involved in the regulation of excitatory synaptic development and the balance between excitatory and inhibitory synaptic inputs.
TM4SF2 or tetraspanin 7 (TSPAN7), another X-linked gene which encodes a membrane protein which belongs to transmembrane 4 superfamily (TM4SF), plays important roles in the cell proliferation, activation, growth, adhesion, and migration (Maecker et al., 1997). TM4SF proteins form a complex with integrin, which regulates cell motility and migration by modulating the actin cytoskeleton (Berditchevski and Odintsova, 1999). A balanced translocation and mutations (a nonsense mutation and a P172H missense mutation) of TM4SF2 was firstly discovered in the individuals with XLID (Zemni et al., 2000). In subsequent studies, the P172H missense mutation was found in individuals with XLID (Maranduba et al., 2004) and ASD . A microduplication in the locus of TM4SF2 was revealed, but this duplication was also present in unaffected controls, suggesting that it may be a neutral polymorphism (Cai et al., 2008). In neurons, TM4SF2 regulates excitatory synaptic development and AMPA receptor trafficking by binding to the synaptic PDZ protein PICK1 (Figure 1) (Bassani et al., 2012).

SYNAPTIC SIGNALING
Disrupted synaptic signaling may be a key determinant of ASD. Components in mGluR-or NMDAR-dependent signaling cascades have recently been implicated in ASD. Neurofibromin 1 (NF1), tuberous sclerosis complex (TSC1/TSC2), and phosphatase and tensin homolog (PTEN) are genes associated with neurological diseases with common autistic symptoms including neurofibromatosis (Rasmussen and Friedman, 2000), tuberous sclerosis (van Slegtenhorst et al., 1997), and Cowden/Lhermitte-Duclos syndrome (Pilarski and Eng, 2004). They are tumor suppressors sharing a common function; they negatively regulate the mammalian target of rapamycin (mTOR) signaling pathway. Although Tsc1 null mice are embryonically lethal (Wilson et al., 2005), mutant mice with loss of Tsc1 in cerebellar Purkinje cells display autistic-like behaviors (Tsai et al., 2012), and Tsc2 heterozygote mice exhibit abnormal social communication (Young et al., 2010); Nf1 mutant mice show aberrant social transmission of food preference and deficits in hippocampus-dependent learning (Costa et al., 2001(Costa et al., , 2002; Pten deficient mice show altered social interaction and macrocephaly with hyperactivation of mTOR pathway (Table 4) .
Signaling molecules in the downstream of mTOR in the mTOR pathway play crucial roles in ASD pathogenesis. Upon phosphorylation by mTORC1, 4E-BP proteins are detached from eIF4E to promote eIF4E-dependent protein translation (Figure 1) (Richter and Sonenberg, 2005). A SNP at eIF4E promoter region which increases its promotor activity was found in autism patients (Neves-Pereira et al., 2009). Implications of mTOR downstream signaling in ASD were demonstrated as 4E-BP2 knockout mice and eIF4E overexpression mice display autistic-like behaviors. 4E-BP2 knockout mice show enhanced translational control of neuroligins and increased excitatory transmission in the hippocampus ( Table 6) (Gkogkas et al., 2013), while eIF4E overexpressing transgenic mice show impaired excitatory/inhibitory balance in the mPFC and increased LTD in the hippocampus and striatum ( Table 6) (Santini et al., 2013). Autistic features of these mutant mice were ameliorated by 4EGI-1 infusion, which inhibits the eIF4E-eIF4G interaction.
Fragile X syndrome is the most common cause of intellectual disability and autism. It is mostly caused by the expansion of CGG trinucleotide repeats in the promoter region of the FMR1 gene, which enhances the methylation of the promoter and represses generation of FMR1-encoded protein (FMRP), which binds to target mRNAs and regulates their translation and transport of mRNA into dendrites and synapses (Figure 1) (Bassell and Warren, 2008). In the absence of FMRP, target mRNA translation becomes excessive and uncontrolled, leading to an aberrant activity-dependent protein synthesis. Fmr1 mutant mice show enhanced protein synthesis-dependent mGluR-mediated LTD and dendritic spine elongation, together with cognitive deficits, social anxiety and impaired social interaction (Table 4) (Bernardet and Crusio, 2006). Interestingly, target molecules of FMRP include Shank3, GluN2A, mTOR, TSC2, NF1, neuroligin2, and neurexin1 (Darnell et al., 2011), which are associated with ASD pathogenesis.
It should be noted that the ASD-related signaling molecules mentioned above are also associated with NMDAR and mGluR signaling pathways. NMDARs and mGluRs play critical roles in the regulation of synaptic function and plasticity at excitatory synapses. NF1 interacts with the NMDAR complex and regulates GluN2A phosphorylation (Figure 1) (Husi et al., 2000). FMRP and TSC have profound effects on mGluR-dependent LTD and protein synthesis, which are upregulated in Fmr1 −/y mice, while downregulated in Tsc2 +/− mice (Auerbach et al., 2011). FMRP is also in the downstream of mGluR signaling (Figure 1) (Bassell and Warren, 2008).
Defects in NMDAR function and associated signaling are also observed in nonsyndromic ASD models with Shank mutations. Shank proteins are physically connected to both NMDARs and mGluRs, suggesting that Shank may regulate signaling pathways downstream of NMDAR or mGluR activation, and the functional interaction between the two receptors (Figure 1). Shank2 −/− mice with the deletion of exons 6 and 7 display autistic-like behaviors and reductions in NMDAR function and associated signaling, without affecting mGluR-dependent LTD (Won et al., 2012), while Shank2 −/− mice with exon 7 deletion show similar behavioral abnormalities with NMDAR hyperfunction ( Table 5) (Schmeisser et al., 2012). Although how similar exon deletions in Shank2 lead to comparable behavioral abnormalities but different changes in NMDAR function remains to be determined, these results point to that Shank2 is an important regulator of NMDAR function, and that NMDAR function and NMDAR-associated signaling are associated with ASD.

NEUROIMMUNE RESPONSE
The implication of the immune system on autism was initially proposed in 1976 based on that some autistic children do not have detectable Rubella titers in spite of previous vaccination (Stubbs, 1976). Levels of serum IgG and autoantibodies to neuronal and glial molecules were elevated in autistic patients (Singh et al., 1997;Croonenberghs et al., 2002), proposing involvement of autoimmune responses in autism. In addition, plasma or cerebrospinal fluid (CSF) levels of pro-inflammatory cytokines and chemokines including MCP-1, IL-6, IL-12, IFN-γ and TGFβ1 were increased in autistic individuals (Ashwood and Van de Water, 2004;Ashwood et al., 2006).
Astrocytes and microglia are two glial cell types important for immune responses in the brain as well as regulation of neuronal functions and homeostasis (Fields and Stevens-Graham, 2002 neuroinflammatory responses in autistic brains (Vargas et al., 2005). Transcriptome analysis of autistic postmortem brain tissues has also revealed upregulation in the expression of genes belonging to immune and inflammatory networks (Voineagu et al., 2011). Reactive astrocytes were also detected in Cntnap2 −/− brains, a well-established autism model (Penagarikano et al., 2011). These results clearly suggest the association between neuroimmune defects with ASD, although further details remain to be determined.

NON-GENETIC MODELS OF ASD
Although we have thus far described genetic factors underlying ASD, environmental factors also have strong influences on ASD.
Epidemiologic studies suggest that maternal exposure to stress, viral or bacterial infection, thalidomide, and valproic acid can increase the risk for ASD in offspring (Grabrucker, 2012). Maternal immune activation (MIA) induced by poly(I:C), the synthetic doublestrand RNA polyriboinosinic-polyribocytidilic acid, in pregnant mice leads to the development of core ASDlike phenotypes in the offspring, including impaired sociability, decreased USV, and increased repetitive behaviors (Malkova et al., 2012). MIA by lipopolysaccharide (LPS) treatment during pregnancy can also induces ASD-like phenotypes in rodent offspring, including impaired social interaction (Hava et al., 2006;Kirsten et al., 2010) and reduced USV (Baharnoori et al., 2012). IL-6 is thought to play a critical role in this process, as IL-6 knockout mice do not show poly(I:C) induced social deficits (Smith et al., 2007), and IL-6 levels are significantly elevated in the cerebellum of autistic subjects (Wei et al., 2011). Although further details remain to be determined, the underlying mechanisms may include IL6-dependent regulation of excitatory and inhibitory synaptic transmission and neuroprotection (Sallmann et al., 2000;Biber et al., 2008;Dugan et al., 2009).
Prenatal exposure to teratogens can increase the risk for ASD in animals, as in humans. Thalidomide (THAL) and valproic acid (VPA) cause rat offspring to display brain morphological abnormalities observed in ASD, including altered cerebellar structures and reduced number of cranial motor neurons (Rodier et al., 1997;Ingram et al., 2000a). Behaviorally, VPA-exposed rats show decreases in prepulse inhibition, stereotypy, and social behaviors (Schneider and Przewlocki, 2005). VPA-exposed rats display elevated serotonin levels and abnormal serotonergic neurons (Anderson et al., 1990;Narita et al., 2002;Miyazaki et al., 2005), decreased parvalbumin-positive interneurons in the neocortex (Gogolla et al., 2009), and elevated NMDA receptor levels and enhanced LTP (Rinaldi et al., 2007), suggesting that these mechanisms may contribute to the development of ASD-like phenotypes.
Positive allosteric modulators of mGluR5 receptors were first developed to alleviate symptoms of schizophrenia (Gregory et al., 2011). Although antipsychotics are available for positive symptoms of schizophrenia, such as hallucinations, no medications are currently available for negative symptoms or cognitive impairments. Two main hypotheses have been proposed for schizophrenia: dopaminergic hyperactivity and NMDA hypofunction. Dopaminergic hyperactivity can be treated by dopamine receptor-antagonistic antipsychotics such as risperidone, but NMDA hypofunction is difficult to modulate given the expected side effects of enhancing NMDAR functions.
Therefore, the concept of augmenting NMDAR signaling via mGluR potentiation was proposed to improve negative symptoms of schizophrenia (Uslaner et al., 2009;Stefani and Moghaddam, 2010). mGluR positive allosteric modulators increase the function of NMDAR only when they are occupied by the endogenous ligand glutamate (Figure 1). mGluR positive allosteric modulators have significant advantages over the conventional mGluR agonist, (RS)-3,5-dihydroxyphenylglycine (DHPG). While DHPG has poor specificity toward particular mGluR subtypes, mGluR positive allosteric modulators offer high subtype specificity. Some positive allosteric modulators have high brain blood barrier penetrance, which enables the systemic administration of the drugs. Furthermore, whereas direct mGluR agonists cause rapid receptor desensitization, mGluR positive allosteric modulators potentiate mGluR function with minimal desensitization, because they bind to an allosteric site on the receptor distinct from the orthosteric glutamate binding site. These properties of positive allosteric modulators are predicted to minimize their excitotoxicity and enable high-dose administrations.
A large number of mGluR5 allosteric modulators have been developed (Williams and Lindsley, 2005;Gregory et al., 2011). Of these, CDPPB, ADX47273, MPPA, and VU0092273 readily cross the blood-brain barrier, and CDPPB, particular, has been examined in various behavioral assays and model animals. In CHO (Chinese hamster ovary) cells expressing human mGluR5, CDPPB treatment was shown to enhance mGluR5 activity in a concentration-dependent manner (Kinney et al.,  2005). Behaviorally, CDPPB alleviates prepulse inhibition and hyperactivity produced by amphetamine, suggesting that CDPPB could be a potential antipsychotic agent. Because NMDARs play an essential role in learning and memory, indirect potentiation of NMDARs by mGluR5 positive allosteric modulators may facilitate synaptic plasticity and learning and memory. Indeed, CDPPB and ADX47273 enhance the performance of wild-type mice in the Morris water maze test, a hippocampus-dependent learning and memory paradigm (Ayala et al., 2009). In addition, VU-29 and ADX47273 potentiate two forms of NMDAR-dependent synaptic plasticity-LTP and LTD-in the CA1 region of the hippocampus (Ayala et al., 2009). DFB-treated rats make fewer errors in the Y-maze spatial alternation task (Balschun et al., 2006), and CDPPB and ADX47273 enhance performance in novel object recognition and five-choice serial reaction time tasks Uslaner et al., 2009). CDPPB not only improves learning and memory performance of wild-type mice, but also reverses cognitive dysfunction and behavioral inflexibility induced by the NMDAR antagonist MK-801 (Uslaner et al., 2009;Stefani and Moghaddam, 2010). These results suggest that mGluR5 positive allosteric modulators have the potential to improve cognitive impairments associated with brain disorders including schizophrenia and autism.
Indeed, CDPPB has recently shown promise as a potential treatment for ASD. In a study using Tsc2 +/− mice, a mouse model of tuberous sclerosis characterized by intellectual disability and autism, Mark Bear and colleagues showed that cognitive impairments observed in these mice could be alleviated by CDPPB administration (Table 4) (Auerbach et al., 2011). In addition, social deficits of Shank2 −/− mice are rescued by CDPPB treatment (Won et al., 2012), implicating hypofunction of mGluRs and NMDARs in social impairment, and suggesting mGluR positive allosteric modulators as novel therapeutics for the treatment of social deficits ( Table 5). More recently, CDPPB was shown to reverse defects in social novelty recognition induced by neonatal phencyclidine treatment (Clifton et al., 2012).

D-CYCLOSERINE
Although it is well established that NMDARs critically regulate normal brain functions, the excitotoxicity and poor bioavailability of direct NMDAR agonists have hampered attempts to control brain activity by modulating NMDARs (Quartermain et al., 1994). D-cycloserine is a high-affinity partial agonist of NMDA-coupled, strychnine-insensitive glycine receptors (Figure 1) . Similar to glycine, D-cycloserine also binds to the glycine site of NMDARs as a partial agonist, potentiating NMDARs by increasing the frequency of channel opening. In addition, because NMDARs are not maximally potentiated by endogenous glycine, there is room for D-cycloserine to further potentiate NMDARs. These properties enable Dcycloserine to act as a positive modulator of NMDARs.
D-cycloserine is a viable drug candidate because it is a partial agonist, displaying efficacy of 40-50% relative to glycine, and has low toxicity and decent bioavailability . Although the brain penetrance of D-cycloserine is not high, it can nonetheless infiltrate the blood-brain barrier, exerting a peak effect 1 h after intraperitoneal administration (Peterson, 1992). D-cycloserine shows dose-dependent elimination (higher elimination rates with lower doses) and a half-life of 7-15 h in humans and 23 min in mice (lwainsky, 1988;Wlaz et al., 1994).
Putative effects of D-cycloserine on ASD have been suggested by previous studies. Mice with a neuroligin1 (Nlgn1) deficiency exhibit abnormally increased grooming behavior, and this behavioral anomaly is reversed by D-cycloserine treatment ( Table 5) (Blundell et al., 2010). Low-dose D-cycloserine alleviates negative symptoms of schizophrenia-affected individuals (Goff et al., 1999), and reduces social withdrawal and increases social responsiveness in autistic patients (Posey et al., 2004). Moreover, Dcycloserine partially rescues social deficits of Shank2 −/− mice, supporting the role of NMDAR functionality in autism (Table 5) (Won et al., 2012).

BENZODIAZEPINES
Recently, benzodiazepines were suggested as putative therapeutic agents for Dravet's syndrome, which is a developmental disorder with myoclonic infantile seizure, ADHD-like inattention and hyperactivity, motor impairment, sleep disorder, anxiety-like behaviors, cognitive defects, autism-like social dysfunction, and restricted interests. Mice heterozygous for a deletion of the αsubunit of the type 1 voltage-gated sodium channel (Scn1α +/− mice), an animal model for Dravet's syndrome, recapitulate most features of the disorder, including epilepsy, ataxia, sleep disorder, anxiety-like behaviors, hippocampus-dependent learning impairments, sociability deficits, and excessive repetitive grooming behaviors ( Table 4) Kalume et al., 2007;Han et al., 2012). In Scn1α +/− mouse brains, expression of the voltagegated sodium channel type-1 (Na v 1.1) is decreased in GABAergic interneurons, and GABAergic transmission onto postsynaptic neurons was reduced. This would cause a shift in the balance between excitation and inhibition in postsynaptic neurons toward excitation, which may be corrected by stimulating GABA receptors in these neurons. Indeed, it was shown that both behavioral abnormalities and aberrant GABAergic transmission are rescued by low-dose administration of clonazepam (Table 4) (Han et al., 2012).
Clonazepam, a type of benzodiazepine, is a positive allosteric modulator of GABA A receptors that exerts sedative, hypnotic, anxiolytic, anticonvulsant, and muscle relaxing effects (Figure 1) (Rudolph and Knoflach, 2011). Similar to the action of mGluR positive allosteric modulators, clonazepam potentiates GABA signaling only when GABA A receptors are bound by their endogenous ligand, GABA. Therefore, these results indicate that normalization of disrupted E-I balance may be a novel and promising strategy for treating symptoms of ASD.

mGLuR NEGATIVE ALLOSTERIC MODULATORS
The potential of mGluR of negative allosteric modulation as a therapeutic strategy in ASD was first proposed based on studies in Fmr1 −/y mice, an animal model for fragile X syndrome (Bakker et al., 1994). The enhanced mGluR5-dependent LTD and protein synthesis observed in Fmr1 −/y mice provided a conceptual framework for the mGluR theory of fragile X pathogenesis (Bear et al., 2004;Bassell and Warren, 2008). Synaptic protein synthesis is stimulated by local mRNA translation, a process that depends on group I mGluR activation. FMRP, encoded by the Fmr1 gene, is a repressor of mRNA translation; thus, mGluR-mediated protein synthesis could be enhanced in the absence of FMRP. Therefore, attempts have been made to correct fragile X syndrome by suppressing abnormally enhanced mGluR5-dependent synaptic plasticity and protein synthesis. Two approaches have been used to normalize behavioral and neuronal deficits of Fmr1 −/y mice: genetic crossbreeding with Tsc2 +/− mice, which exhibit suppressed mGluR activity, and acute administration of an mGluR antagonist (Auerbach et al., 2011). Administering the mGluR5 antagonist 2-methyl-6-(phenylethynyl) pyridine hydrochloride (MPEP) to Fmr1 −/y mice normalizes defective phenotypes, including cognitive deficits, perturbed mGluR-dependent LTD and protein synthesis, and excessive filopodia-like long and thin spines (Figure 1, Table 4) (Yan et al., 2005;de Vrij et al., 2008). In line with this, mGluR negative allosteric modulators are now in clinical trials for fragile X syndrome patients (Krueger and Bear, 2011).
The therapeutic potential of mGluR antagonists in ASD has also been suggested. Repetitive grooming behaviors in BTBR and valproic acid (VPA) mouse models of autism are significantly alleviated by MPEP treatment (Silverman et al., 2010a;Mehta et al., 2011). Impairments in social interaction of BTBR mice are also ameliorated by MPEP administration (Silverman et al., 2010a). GRN-529, a selective negative allosteric modulator of mGluR5 developed by Pfizer, was shown to fully rescue excessive repetitive grooming behavior and social dysfunction in BTBR mice and jumping stereotypy in C58/J mice . These findings suggest that mGluR negative allosteric modulators have novel therapeutic potential in autism, in addition to fragile X syndrome.

NMDAR ANTAGONISTS
NMDAR antagonists including amantadine and its close analogue memantine are now in clinical trials for autistic patients (Nightingale, 2012;Spooren et al., 2012). Amantadine and memantine are non-competitive antagonists for NMDARs with multiple clinical uses (Chen et al., 1992;Blanpied et al., 2005). Memantine is currently being used for Alzheimer's disease, while it is also useful for viral infection and Parkinson's disease. Because both drugs are weak NMDAR antagonists with moderate affinity, prolonged receptor blockade during treatment is unlikely to cause significant side effects.
In a double-blind, placebo controlled study, amantadinetreated group show significant improvements in hyperactivity and inappropriate speech (King et al., 2001). Memantine is also effective in improving language and social behavior and clinical global impressions (CGI) scale in autistic patients (Chez et al., 2007;Erickson et al., 2007;Niederhofer, 2007).
With regard to mechanisms of memantine and amantadine underlying the treatment of ASD remains, both medications are highly likely to exert their therapeutic effects by suppressing NMDAR function and modulating excitotoxicity in autistic subjects. However, care should be taken because other possibilities exist. For instance, memantine treatment promotes excitatory synapse formation and maturation and cell adhesion properties of cerebellar granule cells (CGCs) of Fmr1 knockout mice (Wei et al., 2012). In addition, memantine exerts neuroprotective activities by promoting glia-derived neurotrophic factor (GDNF) release and preventing migroglial inflammatory responses . Memantine can also act as a non-competitive antagonist for 5-HT receptors (Rammes et al., 2001) and nicotinic acetylcholine receptors (Aracava et al., 2005), while it functions as an agonist for D2 dopamine receptors (Seeman et al., 2008).

IGF-1
A new approach for alleviating phenotypic traits of ASD in animal models has come from research on Rett syndrome. Rett syndrome is an X-linked neurological disorder caused by mutations in the MeCP2 gene. MeCP2 is a transcriptional repressor and activator, which binds widely across the genome and influence a large number of genes (Chahrour et al., 2008). One of the best characterized targets of MeCP2 is BDNF, a neurotrophic factor that regulates neuronal development and synaptic plasticity (Figure 1) (Greenberg et al., 2009). Bdnf conditional knockout mice show features analogous to Rett syndrome, including smaller brain size and hindlimb-clasping behavior (Chang et al., 2006). Mice with double knockout of Bdnf and MeCP2 show earlier onset of Rett-like symptoms, whereas overexpression of Bdnf in MeCP2 knockout mice delays the onset and relieves the electrophysiological defects of MeCP2 mutants. Moreover, restoring Bdnf expression through ampakine administration alleviates respiratory problems of MeCP2 mutant mice (Ogier et al., 2007). Although BDNF appears to have significant effects in Rett syndrome model animals, it poorly penetrates the blood-brain barrier, limiting its therapeutic application.
Because IGF-1 crosses the blood-brain barrier, it may be a viable alternative to BDNF as a therapeutic agent for Rett syndrome. Indeed, IGF-1 and IGFBP have been implicated in Rett syndrome and autism: IGFBP3 levels are abnormally elevated in MeCP2 mutant mice and Rett syndrome patients (Itoh et al., 2007), and the concentration of IGF-1 in CSF is reduced in autistic individuals (Riikonen et al., 2006). The therapeutic utility of IGF-1 in Rett syndrome was originally suggested by Mriganka Sur and coworkers, who reported that lethality, hypoactivity, and respiratory problems of MeCP2-null mice are partially rescued by IGF-1 treatment in association with normalization of impaired spine density, synaptic transmission, and cortical plasticity ( Table 4) (Tropea et al., 2009). IGF-1 also reverses the reduction in excitatory synapse number and density of neurons derived from Rett patients (Marchetto et al., 2010).
The therapeutic utility of rapamycin in ASD was suggested in 2008 based on studies in Tsc2 +/− mice (Ehninger et al., 2008). The mTOR pathway is associated with TSC because TSC1 and TSC2 are upstream inhibitory regulators of mTOR activity (Han and Sahin, 2011). In this study, the learning and memory deficits, lethality, aberrant brain overgrowth, and altered synaptic plasticity of Tsc2 +/− mice were ameliorated by acute treatment with rapamycin ( Table 4). The social dysfunction and behavioral inflexibility of Purkinje cell-specific Tsc1 mutant mice were also improved by rapamycin (Tsai et al., 2012), further suggesting that rapamycin may be useful in reversing core symptoms of autism.

OXYTOCIN
Oxytocin is a nine amino acid neuropeptide hormone synthesized by magnocellular neurons in paraventricular and supraoptic nuclei of the hypothalamus and secreted from the posterior pituitary gland into the circulation (Figure 1). Oxytocin acts through oxytocin receptors (OXTRs), which are abundantly expressed in the amygdala, hippocampus, and hypothalamus (Gould and Zingg, 2003). Oxytocin is associated with various social behaviors including affiliation, maternity, aggression, and pair bonding (Lee et al., 2009;Caldwell, 2012;Feldman, 2012). Given the prominence of oxytocin in the regulation of social behavior, the association of oxytocin with autism pathogenesis has been extensively examined.

PERSPECTIVES HOMEOSTATIC MECHANISMS UNDERLYING ASD
Tuberous sclerosis and fragile X syndrome are disorders with common symptoms including intellectual disabilities, seizures, and autism. While their genetic determinants are different (TSC1/TSC2 for tuberous sclerosis and FMR1 for fragile X syndrome), their gene products both regulate protein synthesis in neurons (Bassell and Warren, 2008;Ehninger et al., 2009). Interestingly, animal models of tuberous sclerosis (Tsc2 +/− mice) and fragile X syndrome (Fmr1 −/y mice) display abnormal protein synthesis in opposite directions (Auerbach et al., 2011).
Tsc2 +/− mice exhibit diminished mGluR-dependent LTD and protein synthesis in the hippocampus, whereas Fmr1 −/y mice show excessive mGluR-dependent LTD and protein synthesis. Consistent with this, cognitive impairments of the two animal models are corrected by drugs that modulate mGluR5 in the opposite manner (CDPPB for Tsc2 +/− mice and MPEP for Fmr1 −/y mice). In addition, crossbreeding of these two mouse lines rescues behavioral impairments and synaptic dysfunctions. These results strongly suggest that mGluR5-mediated synaptic plasticity and protein synthesis in the normal range is important and that deviation in either direction from a normal range can cause brain dysfunctions that yield similar behavioral manifestations.
Another such example comes from two mouse models with different mutations in the same gene. Shank2 −/− mouse lines lacking exons 6 and 7 (Won et al., 2012) or exon 7 only (Schmeisser et al., 2012), both of which mimic mutations found in humans (Berkel et al., 2010), display similar autistic-like behaviors, but NMDAR function in their brains shows opposite changes: NMDAR hypofunction with exons 6 and 7 deletion and NMDAR hyperfunction with exon 7 deletion. Although further details remain to be explored, this is another example suggesting that NMDAR function in a normal range is important, and that deviations in either directions can lead to similar behavioral abnormalities. Therefore, individuals with mutations in the same gene may have to be carefully diagnosed, for example by high-through sequencing, in order to receive proper treatment.

CORE MECHANISMS UNDERLYING ASD
Given the diverse genetic variations underlying the development of ASD, one obvious challenge in understanding how ASD develops is the wide range of mechanisms associated with it. This diversity poses a serious additional problem in treating ASD: a single medication is likely to cover only a small fraction of individuals with ASD, or a limited spectrum of ASD symptoms. A related well-known example is the selective effect of risperidone. Risperidone, a dopamine antagonist, is an antipsychotic mainly used to treat schizophrenia and bipolar disorder, and it is currently one of the few FDA-approved medications for autism. The drug mainly ameliorates irritability, hyperactivity, and repetitive and restricted behaviors, but is largely ineffective against social withdrawal and language deficits of autistic individuals (McPheeters et al., 2011). Similarly, risperidone rescues repetitive grooming and hyperactivity, but not social deficits, in Cntnap2 −/− mice (Penagarikano et al., 2011). Another example is the demonstration that CDPPB rescues only social interaction in Shank2-deficient mice but fails to rescue impaired pup retrieval, repetitive jumping, hyperactivity, and anxiety-like behavior (Won et al., 2012). The fact that some medications reverse only selective symptoms/phenotypes of ASD, however, may provide an opportunity to further explore detailed mechanisms underlying particular aspects of ASD etiology. This would, in principle, allow us to dissect and study synaptic or circuit mechanisms that are specifically associated with certain aspects of ASD, such as impaired social interaction, impaired social communication, repetitive behavior, restricted interests, intellectual disability, anxiety, and hyperactivity.
A possible solution to the apparent diversity of ASD-related mechanisms is to identify "core" mechanisms that cover a large fraction of genetic variations, or a broader spectrum of ASD symptoms. The concept of core mechanisms is based on the assumption that a fraction of ASD-related proteins may act together and converge on a common pathway. A possible core mechanism could be excitatory synaptic transmission. Excitatory synaptic development can be regulated by a number of factors including synaptic adhesion molecules, synaptic scaffolding proteins, and actin-regulatory proteins. In addition, excitatory synaptic transmission, which is mainly mediated by AMPAR receptors, can be determined by the regulators of the synaptic trafficking and stabilization of AMPA receptors, and regulated by the signaling pathways in the downstream of NMDA receptors, mGluRs, and monoamine receptors. Another core mechanism could be the E-I balance, which is determined by the relative amounts of excitatory and inhibitory synaptic transmissions, and, together with the excitability of postsynaptic neurons, determines firing patterns of postsynaptic neurons and, subsequently, network activities across the brain. Establishing these core mechanisms, if any, would require rigorous and timeconsuming verifications using a range of approaches, including mouse genetics, electrophysiology, and behavior.

INTEGRATING THREE ASPECTS OF ASD RESEARCH: HUMAN GENETICS, MOUSE MODELS, AND POTENTIAL TREATMENTS
An important starting point for ASD research using mouse models would be to select best possible genetic variations that can provide us decent insights into the underlying mechanisms and potential treatments. Luckily, a large number of ASD-related papers are being published each year (i.e., ∼2500 papers in 2012 when "autism" was used as a search key word in PubMed). These publications, which use diverse genetic and genomic approaches and often large size samples, have identified overlapping genes and mutations, which are likely to have greater influences on the development of ASD. Characterization of transgenic mouse lines that carry these frequent genetic variations would help us efficiently obtain ASD mechanisms with a greater impact.
The synaptic and circuit mechanisms derived from ASD mouse model researches would provide clues to the ways to rescue synaptic/circuit phenotypes and ASD-like behaviors in mice. Given that there is no FDA-approved treatment for social deficits in ASD as of now, these rescue results will only be useful in supporting that the candidate mechanisms are indeed causing the ASD-like phenotypes in mice. Importantly, however, some of these mechanism-based rescues may serve as the basis for clinical trials. Eventually, some of the clinically verified medications may return to basic ASD research and be used to identify additional ASD mouse models with similar or novel underlying mechanisms, which will help us understand a bigger picture, where many synaptic and circuit mechanisms act together and converge into more comprehensive mechanisms.  Gharani, N., Rossman, I., Mancuso, V., Lazar, G., Kamdar, S., et al. (2005). Support for the homeobox transcription factor gene ENGRAILED 2 as an autism spectrum disorder susceptibility locus.  Zemni, R., et al. (1998a). Oligophrenin 1 encodes a rho-GAP protein involved in X-linked mental retardation. Pathol. Biol. 46, 678. Billuart, P., Bienvenu, T., Ronce, N., des Portes, V., Vinet, M. C., Zemni, R., et al. (1998b). Oligophrenin-1 encodes a rhoGAP protein involved