A Subset of Autism-Associated Genes Regulate the Structural Stability of Neurons

Autism spectrum disorder (ASD) comprises a range of neurological conditions that affect individuals’ ability to communicate and interact with others. People with ASD often exhibit marked qualitative difficulties in social interaction, communication, and behavior. Alterations in neurite arborization and dendritic spine morphology, including size, shape, and number, are hallmarks of almost all neurological conditions, including ASD. As experimental evidence emerges in recent years, it becomes clear that although there is broad heterogeneity of identified autism risk genes, many of them converge into similar cellular pathways, including those regulating neurite outgrowth, synapse formation and spine stability, and synaptic plasticity. These mechanisms together regulate the structural stability of neurons and are vulnerable targets in ASD. In this review, we discuss the current understanding of those autism risk genes that affect the structural connectivity of neurons. We sub-categorize them into (1) cytoskeletal regulators, e.g., motors and small RhoGTPase regulators; (2) adhesion molecules, e.g., cadherins, NCAM, and neurexin superfamily; (3) cell surface receptors, e.g., glutamatergic receptors and receptor tyrosine kinases; (4) signaling molecules, e.g., protein kinases and phosphatases; and (5) synaptic proteins, e.g., vesicle and scaffolding proteins. Although the roles of some of these genes in maintaining neuronal structural stability are well studied, how mutations contribute to the autism phenotype is still largely unknown. Investigating whether and how the neuronal structure and function are affected when these genes are mutated will provide insights toward developing effective interventions aimed at improving the lives of people with autism and their families.

Autism spectrum disorder (ASD) comprises a range of neurological conditions that affect individuals' ability to communicate and interact with others. People with ASD often exhibit marked qualitative difficulties in social interaction, communication, and behavior. Alterations in neurite arborization and dendritic spine morphology, including size, shape, and number, are hallmarks of almost all neurological conditions, including ASD. As experimental evidence emerges in recent years, it becomes clear that although there is broad heterogeneity of identified autism risk genes, many of them converge into similar cellular pathways, including those regulating neurite outgrowth, synapse formation and spine stability, and synaptic plasticity. These mechanisms together regulate the structural stability of neurons and are vulnerable targets in ASD. In this review, we discuss the current understanding of those autism risk genes that affect the structural connectivity of neurons. We sub-categorize them into (1) cytoskeletal regulators, e.g., motors and small RhoGTPase regulators; (2) adhesion molecules, e.g., cadherins, NCAM, and neurexin superfamily; (3) cell surface receptors, e.g., glutamatergic receptors and receptor tyrosine kinases; (4) signaling molecules, e.g., protein kinases and phosphatases; and (5) synaptic proteins, e.g., vesicle and scaffolding proteins. Although the roles of some of these genes in maintaining neuronal structural stability are well studied, how mutations contribute to the autism phenotype is still largely unknown. Investigating whether and how the neuronal structure and function are affected when these genes are mutated will provide insights toward developing effective interventions aimed at improving the lives of people with autism and their families.

INTRODUCTION
Autism spectrum disorder (ASD) is a neurodevelopmental clinical condition currently diagnosed based on the American Psychiatric Association's Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) criteria reflecting symptoms, possibly of varying severity, in social interaction, communication and behavior (American Psychiatric Association, 2013;Lord and Jones, 2013). ASD occurs in 1:68 individuals in the United States (Baio, 2014) and complex genetic interactions appear responsible for a high degree of heterogeneity of the clinical symptoms in ASD. Individuals with ASD often co-express other comorbidities including epilepsy which often complicates diagnosis and treatment. Alterations in neuronal structures in different brain regions have been reported in ASD individuals, including increased dendritic spine density in cortical pyramidal neurons (Hutsler and Zhang, 2010;Tang et al., 2014) as well as stunting of dendritic branching in the hippocampus (Raymond et al., 1996;Bauman and Kemper, 2005). In addition, subcortical band heterotopia, representing alterations in cell migration has also been found in a child with ASD (Beaudoin et al., 2007). These brain regions are often characterized with neuroanatomical irregularities in ASD (Donovan and Basson, 2016). The defective regulation for structural stability of neurons may be one of the underlying mechanisms that contribute to the anatomical changes in ASD.
Autism spectrum disorder is typically diagnosed during the first 3 years of life, a period of extensive neurite formation, synaptogenesis and refinement (Huttenlocher and Dabholkar, 1997;Zoghbi and Bear, 2012;Stamou et al., 2013;McGee et al., 2014). Indeed, brain imaging studies from individuals with ASD and anatomical measurements of neuronal structure in post-mortem tissues exhibit differences in neuronal connectivity derived from the disruption of neurite outgrowth, synapse formation and stabilization (Raymond et al., 1996;Hutsler and Zhang, 2010;Penzes et al., 2011). Studies of human induced pluripotent stem cells (iPSCs) derived from people with ASD also have identified defects of neuronal structure (Habela et al., 2015;Nestor et al., 2015). Genome-wide association studies on individuals with ASD and their families revealed several risk genes that may be the common molecular targets in autism Glessner et al., 2009;Hussman et al., 2011;O'Roak et al., 2011O'Roak et al., , 2012aBuxbaum et al., 2012Buxbaum et al., , 2014Sanders et al., 2012;Shi et al., 2013;Stamou et al., 2013;Yu et al., 2013;Brett et al., 2014;Cukier et al., 2014;De Rubeis et al., 2014;Iossifov et al., 2014;McGee et al., 2014;Pinto et al., 2014;Ronemus et al., 2014;Toma et al., 2014;Yuen et al., 2015). Animal studies of these genes further identify several specific cellular pathways during brain development that are vulnerable in ASD, including the disruption of neurite outgrowth, dendritic spine formation, and synaptic function (Figure 1) (Walsh et al., 2008;Bourgeron, 2009;Hussman et al., 2011;Penzes et al., 2011;Zoghbi and Bear, 2012;Ebert and Greenberg, 2013;Stamou et al., 2013;Bernardinelli et al., 2014;De Rubeis et al., 2014;Pinto et al., 2014;Phillips and Pozzo-Miller, 2015). Differences in environment as well as the presence of multiple gene mutations occurring in the same individual with autism complicate studies of the relationship between each gene and the phenotype observed. However, because similar cellular pathways (e.g., neurite outgrowth) are altered in different affected individuals, we can potentially develop therapeutic interventions to help mitigate the autism phenotypes.
During development, neurite outgrowth and synapse formation are dynamic processes and their maturation is FIGURE 1 | Diagram of autism-risk genes implicated in regulating the structural stability of neurons. Each circle represents a cellular pathway to regulate the structural stability of neurons, including neurite outgrowth (red), dendritic spine or synapse formation (blue), and synaptic plasticity (gold). Experimental evidence shows that many autism-risk genes regulate at least one cellular pathway to maintain the integrity of neuronal structures. Genes that regulate only one pathway are labeled in light gray. Genes that regulate two pathways are labeled in dark gray. Genes that regulate three pathways are labeled in black. The summaries of autism-risk genes that affect each cellular pathway can be found in Tables 1-3. mutually dependent on proper guidance. Neurites initially exhibit frequent branch additions and retractions. Once dendrite arbors are established, productive synapse formation later in life and the accompanying activation of post-synaptic signaling machinery promotes arbor stability (Dailey and Smith, 1996;Wu and Cline, 1998;Rajan et al., 1999;Wong et al., 2000;Cline, 2001;Niell et al., 2004). Conversely, a loss of synaptic inputs leads to dendritic loss (Jones and Thomas, 1962;Matthews and Powell, 1962;Coleman and Riesen, 1968;Sfakianos et al., 2007). This reciprocal regulation contributes to the refinement of dendrites and synapses as the neurons mature (Wu et al., 1999;Trachtenberg et al., 2002;Holtmaat et al., 2005;Koleske, 2013). Thus, maintaining the structural stability of neurons and synapses is critical for proper brain function. Alterations in these processes likely underlie the disruption of normal dendrite and dendritic spine structure in neurological disorders, including neurodevelopmental conditions, psychiatric disorders, and neurodegenerative diseases (Fiala et al., 2002;Lin and Koleske 2010;Penzes et al., 2011;Kulkarni and Firestein, 2012;Zoghbi and Bear, 2012;Koleske, 2013;Bernardinelli et al., 2014).
It is well-accepted that ASD is not a monogenetic disorder, instead, it is often a neurological condition resulted from multiple mutations of several different genes. Although knockout, knockin, or transgenic approaches of autism-risk genes in animal models have demonstrated some of the autistic-like behaviors (Kazdoba et al., 2016), the limitation of the number Adhesion molecule Esch et al., 2000;Bekirov et al., 2008;Tan et al., 2010;Friedman et al., 2015bPCDH Uemura et al., 2007Morrow et al., 2008;Keeler et al., 2015NRXN Gjorlund et al., 2012NLGN Gjorlund et al., 2012CNTNAP2 Anderson et al., 2012CNTN Ye et al., 2008NCAM2 Sheng et al., 2015 Surface receptor GRIN2B Ewald et al., 2008;Espinosa et al., 2009;Sepulveda et al., 2010;Bustos et al., 2014NTRK Joo et al., 2014 Signaling molecule Hammerle et al., 2003;Benavides-Piccione et al., 2005;Gockler et al., 2009;Lepagnol-Bestel et al., 2009CDKL5 Chen et al., 2010bAmendola et al., 2014;Fuchs et al., 2014PTEN Jaworski et al., 2005Kwon et al., 2006;Zhou et al., 2009 Synaptic protein   Galvez et al., 2003;Antar et al., 2006;Tucker et al., 2006;Berman et al., 2012;Amiri et al., 2014MECP2 Fukuda et al., 2005Jugloff et al., 2005;Zhou et al., 2006;Ballas et al., 2009;Belichenko et al., 2009;Kishi and Macklis, 2010;Cohen et al., 2011;Marshak et al., 2012;Nguyen et al., 2012;Stuss et al., 2012;Jiang et al., 2013a;Baj et al., 2014UBE3A Dindot et al., 2008Miao et al., 2013;Valluy et al., 2015TSC1/2 Floricel et al., 2007Choi et al., 2008 of genes being manipulated in animals makes it difficult to recapitulate the human condition experimentally. Furthermore, ASD is a common comorbid condition in individuals with other neurodevelopmental disorders. The similar representation of the symptoms but different contribution of genetic mutations often complicates the diagnosis and the treatment. The complex profile of gene mutations makes it difficult to call a gene "the autism gene." However, the list of autism-risk genes provides us a direction to understand the potentially vulnerable pathways in neurons that may be therapeutic targets to develop more efficient interventions for ASD. Indeed, in addition to the structural stability of neurons, several cellular pathways including transcriptional regulation (De Rubeis et al., 2014;Sanders, 2015), excitatory/inhibitory (E/I) balance (Blatt et al., 2001;Hussman, 2001;Rubenstein and Merzenich, 2003;Gao and Penzes, 2015;Nelson and Valakh, 2015), cerebellar development Hampson and Blatt, 2015), and autoregulatory feedback loops (Mullins et al., 2016) have been proposed to be vulnerable in autism. In this review, we focus on recent identified autism-risk genes that have been shown to regulate neuronal structures and circuit formation, including aspects of neurite outgrowth (Table 1), synapse formation and spine stability ( Table 2), and synaptic plasticity ( Table 3). We will discuss the known biological function of those individual autism-risk genes in neurons and how they converge into common pathways. We have categorized these genes into cytoskeletal regulators, adhesion molecules, cell surface receptors, signaling molecules, as well as synaptic proteins (Figure 2). In addition, we include genes causing syndromic disorders in the discussion to highlight the importance of maintaining the neuronal structures for proper brain function.

Actin and Microtubule Regulators Are Associated with Autism
Myosins are motors that utilize ATPase activity to provide motility of actin or cargo transport on actin filaments (Pollard and Korn, 1973;Oliver et al., 1999;Tyska and Warshaw, 2002). Several myosin isoforms play central roles in regulating neurite outgrowth, as well as dendritic spine structural plasticity (Wylie et al., 1998;Wylie and Chantler, 2003;Ryu et al., 2006;Hammer and Wagner, 2013;Kneussel and Wagner, 2013;Yoshii et al., 2013;Koskinen et al., 2014;Ultanir et al., 2014). Among all isoforms, MYO16 (Myr8 or NYAP3) was recently implicated in ASD Connolly et al., 2013;Kenny et al., 2014;Roberts et al., 2014;Liu et al., 2015b). MYO16 is expressed predominantly in the cortex and cerebellum. Levels and phosphorylation of MYO16 protein peak during early developmental stages, consistent with a role in regulating neuronal migration and neurite extension (Patel et al., 2001;Yokoyama et al., 2011). In addition to binding directly to filamentous-(F-)actin, MYO16 also physically interacts with PI3K and WAVE complex to regulate stress fiber remodeling in fibroblasts as well as the adhesiondependent neurite outgrowth in neurons (Yokoyama et al., 2011).

Small RhoGTPase Regulation Is a Key Mechanism in Controlling Neurite and Spine Stability
Small RhoGTPases including Rho, Rac, and Cdc42 are central cytoskeletal regulators that control cell motility and morphology Newey et al., 2005;Lin and Koleske, 2010;Tolias et al., 2011). Genetic mutations or dysregulation of the small RhoGTPase regulators, including guanine-exchange factors (GEFs) and GTPase-activating proteins (GAPs), have been implicated in several neurological conditions, including ASD Lin and Koleske, 2010;Antoine-Bertrand et al., 2011;Stankiewicz and Linseman, 2014). Here, we will highlight those that regulate the morphological stability of neurons.
Engulfment and cell motility 1 (ELMO1) was first identified in a complex with a RacGEF, DOCK180, to activate Rac1 activity, which is essential for cell migration and phagocytosis (Gumienny et al., 2001;Brugnera et al., 2002;Grimsley et al., 2004). In hippocampal neurons, ELMO1 and DOCK180 colocalize at synaptic sites and together are required for spine formation FIGURE 2 | Schematic illustration of how autism-risk genes regulate neuronal structure and their sites of action. An illustration of a dendritic segment containing a dendrite and a dendritic spine is enlarged from the box region on the left and shown on the right. Microtubules (green) and actin filaments (blue) are two major cytoskeletons found in dendrites and dendritic spines, respectively. Autism-risk genes (in bold font) are categorized by their main function and color coded accordingly. (1) Cytoskeletal proteins (gray rounded rectangular box): MYO16, CTTNBP2, and ADNP, directly regulate actin and microtubule function to control dendritic spine and neurite stability. ELMO1 and SYNGAP1 regulate actin dynamics to control spine stability via small RhoGTPases. (2) Adhesion molecules (colored rectangular box): Cadherins (CDHs), protocadherins (PCDHs), and neurexin (NRXN)-neuroligin (NLGN) complex, as well as surface receptors, NTRK, GRIK, and NMDAR, act at synapses to regulate synaptic function. NCAM2 and CNTNAP2, also have functions in regulating neurite outgrowth. (3) Signaling molecules (blue ellipse shape): CDKL5, DYRK1A, and PTEN regulate several signaling pathways to maintain the stability of dendritic structures. (4) Scaffolding proteins (yellow polygon): SHANK3 and DLGAP2, locate at post-synaptic density and tightly associate with PSD95 and other signaling molecules to regulate spine stability and synaptic plasticity. (5) Synaptic proteins (pink ellipse shape): STXBP5 and PRICKLE1 not only regulate synaptic vesicle release, but also play a role in regulating neurite outgrowth. (6) Syndromic molecules (clear rectangular box): FMRP and UBE3A regulate the structural stability of neurons via the regulation of protein synthesis or binding with other molecules in dendritic spines. MECP2 mainly functions in the nucleus and regulates transcription of many genes to in turn affecting the structural stability of neurons. TSC1/2 regulates the mTOR pathway and cytoskeletal machinery to maintain dendritic stability. (Kim et al., 2011a). Loss of Elmo1 shows a reduction in spine number but increased filopodia, suggesting a role in formation and/or maintenance of mature spines (Kim et al., 2011a). In addition, ELMO1 has been shown to regulate axonal and dendritic branching via Rac1 activation in response to different upstream signals (Franke et al., 2012;Lanoue et al., 2013).
Since the actin and microtubule cytoskeletons are the major components of neuronal processes, it is not surprising that manipulating the cytoskeletal machinery dramatically affects neuronal structures. A small imbalance of cytoskeletal dynamics will create a huge impact on the structural stability of neurons, which in turn alters the formation of neuronal circuitry.
Interestingly, most autism-associated cytoskeletal regulators control neurite outgrowth and synapse/spine formation thereby affecting the structural stability of neurons. These two processes are also the initial steps to establish correct neuronal connections during development. Failure to regulate these processes properly may result in significantly altered wiring of brain circuitries that is often found in ASD. The next research focus should investigate early in development to connect the dysregulatory effects of mutations in cytoskeletal genes.

TRANS-SYNAPTIC ADHESION MOLECULES PLAY IMPORTANT ROLES IN THE REGULATION OF NEURONAL STABILITY
Cell adhesion molecules (CAMs) play crucial roles in many aspects of neural circuit formation and, thus, it comes as no surprise that these molecules are found as top hits in lists of autism risk genes (Betancur et al., 2009;Pinto et al., 2010;Hussman et al., 2011;Chen et al., 2014b). Here, we discuss the current understanding of how CAMs that belong to the cadherin-, the neurexin/neuroligin-and the immunoglobulinsuperfamily regulate neuronal stability.
N-cadherin, also known as cadherin 2 (CDH2), is the best studied classical cadherin. N-cadherin functions throughout the development of the nervous system, including neurite outgrowth, axon guidance, synaptogenesis and synaptic plasticity (Takeichi and Abe, 2005;Arikkath and Reichardt, 2008;Hirano and Takeichi, 2012;Friedman et al., 2015a). N-cadherin promotes dendritic outgrowth during development and is also required for activity-dependent dendrite expansion (Esch et al., 2000;Tan et al., 2010). N-cadherin is also required for the establishment of initial contacts between axons and filopodia followed by clustering at contact points to stabilize early synapses (Benson and Tanaka, 1998;Huntley and Benson, 1999;Togashi et al., 2002). Blocking N-cadherin adhesion in hippocampal neurons perturbs synapse formation and abolishes long-term potentiation (LTP)-induced stabilization of dendritic spines (Togashi et al., 2002;Mendez et al., 2010). Neural activity increases N-cadherin protein levels and dimerization leading to increased synapse number (Bozdagi et al., 2000). In mature synapses, N-cadherin is required for the persistence of dendritic spine enlargement and LTP (Bozdagi et al., 2000(Bozdagi et al., , 2010. Together with N-cadherin, CDH8 regulates the development of the hippocampal mossy fiber pathway (Bekirov et al., 2008). CDH8 also mediates assembly and maturation of corticostriatal synapses (Bekirov et al., 2008;Friedman et al., 2015b), whereas CDH9-mediated adhesion is involved in the formation and differentiation of dentate gyrus synapses on CA3 cells where it regulates synapse density, presynaptic bouton complexity and postsynaptic morphology . In contrast to CDH8 and CDH9, an RNAi screen for molecules required for synapse development identified CDH11 and CDH13 as positive regulators of glutamatergic synapse development (Paradis et al., 2007). Interestingly, Cdh11-deficient mice revealed enhanced LTP in the CA1 region of the hippocampus and mice show reduced fear-or anxiety-related behavior suggesting that CDH11 might restrict synaptic plasticity and efficacy (Manabe et al., 2000).
Protocadherins are the largest subgroup within the cadherin superfamily and are further subtyped into clustered (α-, βand γ-PCDH) and non-clustered protocadherins (δ1and δ2-PCDH) (Frank and Kemler, 2002). They share a similar structure to classical cadherins, but with six to seven cadherin domains/EC motifs. However, the cytosolic tails of protocadherins and cadherins do not show significant homology suggesting that they likely engage distinct intracellular signaling pathways. Protocadherins are highly expressed in the nervous system and localize to synapses. Based on their spatial and temporal expression pattern in the brain and on recent reports, protocadherins have roles in dendritic development and synaptic connections (Hirano et al., 1999;Frank and Kemler, 2002;Kim et al., 2007Kim et al., , 2011bKeeler et al., 2015). For example, PCDH10 expression is regulated by neuronal activity and its function is crucial for forebrain axon outgrowth and the proper patterning of thalamocorticial projections (Uemura et al., 2007;Morrow et al., 2008). In addition, PCDH10 mediates synapse elimination by promoting proteasomal degradation of PSD-95 (Tsai et al., 2012a).
FAT atypical cadherin 1 (FAT1) belongs to the atypical cadherin family and consists of a huge extracellular domain comprising 34 cadherin domains/EC motifs (Tanoue and Takeichi, 2005;Sadeqzadeh et al., 2014). FAT1 expression is enriched during embryonic neurodevelopment and severe nervous system defects are found in FAT1-deficient mice (Ciani et al., 2003;Sadeqzadeh et al., 2014). At the cellular level, FAT1 localizes to cell-cell contacts as well as to the leading edge of lamellipodia and tips of filopodia to regulate cell polarity, cell migration, and cell-cell adhesion (Moeller et al., 2004;Tanoue and Takeichi, 2004). These functions are likely mediated through intracellular signaling via Ena/VASP proteins to regulate actin assembly and dynamics (Moeller et al., 2004;Tanoue and Takeichi, 2004). Other intracellular binding partners of FAT1 include the classical cadherin binding partner β-catenin as well as the synaptic scaffolding molecules Homer-1 and 3 (Hou et al., 2006;Schreiner et al., 2006).

Immunoglobulin Superfamily of Cell Adhesion Molecules Participate Largely in Neuronal Circuit Formation
The immunoglobulin superfamily of CAMs (IgSF-CAMs), including contactin, L1CAM, NCAM or SynCAM, make up a third large group of trans-synaptic CAMs. IgSF-CAMs have been implicated in various processes during neural circuit formation, from neurite outgrowth and axonal navigation to synapse formation and plasticity (Rougon and Hobert, 2003).
The contactin (CNTN) subfamily consists of six members (CNTN1-6), each of which contain six Ig-like and four fibronectin III-like domains that are linked to the cell membrane via a glycosylphosphatidylinositol (GPI)-anchoring domain . While CNTN1 and 2 have been extensively studied in the context of neurite outgrowth, fasciculation, and axon guidance, less is known about the function of CNTN3-6 (Karagogeos, 2003;Shimoda and Watanabe, 2009;Mohebiany et al., 2014). However, CNTN3-6 have been implicated as risk genes in ASD (Fernandez et al., 2004;Christian et al., 2008;Morrow et al., 2008;Glessner et al., 2009;Roohi et al., 2009;Cottrell et al., 2011;Hussman et al., 2011;van Daalen et al., 2011;Leblond et al., 2012;Prasad et al., 2012;Vaags et al., 2012;Cukier et al., 2014;Kashevarova et al., 2014;Nava et al., 2014;Poot, 2014;Hu et al., 2015;Liu et al., 2015a). CNTN4 is strongly expressed in a subset of olfactory sensory neurons where it guides proper targeting of axon terminals to the corresponding glomeruli for the formation of olfactory circuits (Kaneko-Goto et al., 2008). The Cntn5 knockout mice display reduced fiber density and glutamatergic synapses in the auditory brainstem Toyoshima et al., 2009). CNTN6 is highly expressed in the postnatal cerebellum and plays an important role in the formation of synapses between parallel fibers and Purkinje cells Sakurai et al., 2009). Similarly, CNTN6 regulates the formation of glutamatergic synapses in the hippocampus and the orientation of apical dendrites of layer V pyramidal neurons in the visual cortex (Ye et al., 2008;Sakurai et al., 2010).
Neural cell adhesion molecule 2 (NCAM2) belongs to the NCAM family and is a paralog of NCAM1. Similar to other members of the Ig-superfamily, NCAMs contain five Ig-and two FN3-domains in the extracellular region and are differentially spliced to produce both transmembrane and GPI-anchored variants (Winther et al., 2012). NCAM2 is predominantly expressed in the brain and required for the formation and maintenance of axonal and dendritic compartmentalization in the olfactory glomeruli (Walz et al., 2006;Borisovska et al., 2011). In addition, NCAM2 regulates filopodia formation and neurite branching of cortical neurons via a CaMKII-dependent signaling pathway (Sheng et al., 2015). SNP and chromosomal deletion including NCAM2 has been reported in individuals with autism (Haldeman-Englert et al., 2010;Hussman et al., 2011;Petit et al., 2015).
Adhesion molecules are a huge group of proteins that display many similarities in molecular structure and in signaling property. Depending on the cellular localization, functions of adhesion molecules range from neurite outgrowth and synapse/spine formation, to neuronal plasticity, further highlighting their importance in regulating the structural stability of neurons. However, whether these molecules function to compensate each other or are developmentally regulated is still not clear. The interesting question is whether the temporal and spatial expression patterns of these autism-associated adhesion molecules correlate with the affected developmental time frame and affected brain regions in ASD.

SURFACE RECEPTORS SIGNAL THROUGH INTRACELLULAR SIGNALING PATHWAYS TO REGULATE NEURONAL STABILITY
Establishment of synaptic connections and modification of their strength and stability is intimately related to the receptor populations in the plasma membranes of pre-and postsynaptic cell compartments. Thus, several ASD risk genes code for cell surface receptor proteins including the ionotropic glutamate receptors (iGluR) and the receptor tyrosine kinases (RTK).
NMDARs are composed of an obligatory GluN1 subunit and one or more GluN2 (GluN2A-GluN2D) subunits with the majority of the composition being GluN1/2A/2B (Buller et al., 1994;Petralia et al., 1994;Luo et al., 1997). The composition of GluN2 subunits are developmentally regulated and critically determine the synaptic properties (Laurie and Seeburg, 1994;Sheng et al., 1994;Li et al., 1998). The GluN2B subunit expresses early during development gradually being replaced by GluN2A indicating its role in the formation of neuronal circuitry (Sheng et al., 1994;Li et al., 1998;Bustos et al., 2014). Overexpression or knockdown of GluN2B alters dendrite arborization in neurons both in vivo and in vitro (Ewald et al., 2008;Espinosa et al., 2009;Sepulveda et al., 2010;Bustos et al., 2014). GluN2B is also required for the formation of dendritic spines, maturation of synapses, and the proper molecular compositions of several postsynaptic proteins (Akashi et al., 2009;Espinosa et al., 2009;Brigman et al., 2010;Kelsch et al., 2012). In turn, GluN2B is crucial for maintaining proper synaptic plasticity (Brigman et al., 2010;Ohno et al., 2010;Wang et al., 2011a;Yang et al., 2012a;Ryan et al., 2013;Dupuis et al., 2014). GRIN2B, an autism-risk gene, further suggests that pathways involved in early circuitry formation may be vulnerable targets in autism. Selective inhibition of GluN2B function has been shown to restore dendritic spine loss and associated behavior alterations in several experimental conditions providing insights to the potential therapeutic targets to correct some ASD phenotypes Iafrati et al., 2014;Gupta et al., 2015).
Kainate-type receptors regulate axonal filopodia motility of hippocampal mossy fibers in response to neuronal stimulation during synaptogenesis (Tashiro et al., 2003). KAR subunits, in particular GluK2, interacts with structural elements of the synapse; such as the PSD-95 and SAP-102 scaffolding molecules, as well as the N-cadherin and β-catenin adhesion molecules (Carta et al., 2014;Pahl et al., 2014), indicating that Grik genes are involved in processes that regulate synapse architecture and stability. Indeed, GluK2 regulates hippocampal synapse maturation and stability (Huettner, 2003;Lanore et al., 2012;Lerma and Marques, 2013). Animals with deficient GluK2 proteins exhibit a delay in the postnatal maturation of synaptic contacts between MF-CA3 in the hippocampus, suggesting that the expression of the GluK2 is important for the establishment of normal morphology and function of synaptic networks in the hippocampus (Contractor et al., 2001;Lanore et al., 2012). Expression of the GluK4 is mainly restricted to mossy fiber synapses in the hippocampal CA3 region where it co-assembles with GluK2 in functional pre-and postsynaptic GluK2/4 receptor complexes (Darstein et al., 2003). Mice with forebrain GluK4 overexpression exhibit altered synaptic transmission and display several autistic-like behaviors including social impairment, enhanced anxiety, and depressive states, coinciding with the finding of GRIK4 duplications in individuals with ASD (Griswold et al., 2012;Aller et al., 2015). Even though the phenotypes resulting from Grik gene dysfunction in mice are in the same general categories with symptoms of ASD, further investigation about the molecular consequences of impairments in GluK proteins in ASD is required for developing future therapeutic interventions.

Receptor Tyrosine Kinases: The NTRK Genes
Tyrosine receptor kinases (Trks) mediate neurotrophic growth factor-induced signaling via dimerization and transautophosphorylation of Tyr residues on the intracellular domains of the receptor and subsequent activation of intracellular signaling pathways (Deinhardt and Chao, 2014). This results in a number of neurogenic events, such as synaptic plasticity, maturation and stability, dendritic and axonal growth and differentiation as well as cell survival and maintenance (Martinez et al., 1998;Deinhardt and Chao, 2014). The Trk family consists of three proteins; TrkA, B and C, which are expressed by the neurotrophic tyrosine receptor kinase genes (NTRK) 1, 2 and 3, respectively. Each Trk receptor interacts selectively with a different neurotrophin resulting in preferential interaction pairs: TrkA is activated by NGF, TrkB by BDNF, and TrkC by NT-3 (Deinhardt and Chao, 2014). Considering its welldocumented function in synaptophysiology (Minichiello, 2009), it would be reasonable to suspect a correlation between genetic variations in NTRK2 and ASD. However, to date, only one study has reported a weak association between NTRK2 mutations and ASD (Correia et al., 2010), while other studies were unable to confirm that link (Chakrabarti et al., 2009). Alternatively, a growing body of evidence generated from genetic evaluation of ASD risk genes has identified NTRK3, the gene coding for TrkC, as a plausible candidate in autism (Chakrabarti et al., 2009;Hussman et al., 2011;Vardarajan et al., 2013).
In the mammalian brain, TrkC (as well as other neurotrophic receptors) is present both as full length catalytically active receptor, as well as a splice variant that lacks the Tyr kinase domain and is catalytically inactive (Ichinose and Snider, 2000). Interestingly, knockout of the non-catalytic TrkC isoform in mice yields a more severe phenotype than does the depletion of the kinase-active receptor, indicating that TrkC has important functions beyond the ability to convey classical RTK signaling (Faux et al., 2007;Deinhardt and Chao, 2014). Indeed, recent studies have begun to elucidate the function assigned to non-catalytic isoforms by demonstrating a role for TrkC in synaptic adhesion complexes (Takahashi and Craig, 2013). Postsynaptic TrkC interacts across the presynaptic cleft with protein tyrosine phosphatase (PTP) σ to form an adhesion complex crucial for development and stability of excitatory, but not inhibitory, synapses (Takahashi et al., 2011;Coles et al., 2014). Formation of this adhesion complex is enhanced by the presence of the TrkC ligand, NT-3, which facilitates glutamatergic presynaptic assembly and function (Ammendrup-Johnsen et al., 2015). NT-3 binding to kinase domain-truncated TrkC isoforms has also been shown to induce cytoskeletal changes via recruitment of the scaffold protein tamalin, leading to activation of Arf6 and induction of Rac1-GTP (Esteban et al., 2006). Interestingly, the expression of non-catalytic TrkC relative to the kinase active isoform is upregulated during the second and third postnatal weeks, the most intense period of synaptogenesis, indicating that expression of the different Ntrk3 gene products is temporally associated with synapse formation (Valenzuela et al., 1993;Menn et al., 2000). Recent studies also found that NT-3-TrkC signaling between presynaptic granule neurons and postsynaptic Purkinje cells controls dendrite morphogenesis in cerebellum (Joo et al., 2014). Although no studies have evaluated the ratio of non-catalytic to catalytic TrkC receptors in individuals with ASD, it might be speculated that certain genetic variants could cause imbalances in the expression patterns of NTRK3 isoforms.
Surface receptors respond to extracellular signals such as neurotransmitters and trophic factors to activate downstream signaling pathways to diversify the cellular responses. Each receptor may have a unique signaling pathway associated with it and therefore, the mutations on selective receptors provide us with clues about which signaling pathways may be more susceptible to perturbations in ASD. Thus, identifying the downstream effectors and signaling pathways that are affected by these autism-associated receptor mutants should be an important direction of future investigation.

Protein Kinases
The dual-specificity tyrosine-(Y)-phosphorylation-regulated kinase 1a (DYRK1A) is one of the isoforms in DYRK family and is a human homolog of the Drosophila kinase minibrain (MNB) (Shindoh et al., 1996). DYRK1A was first described as a cadidate gene for intellectual disability in Down syndrome because of its location on the "Down syndrome critical region" of chromosome 21 (van Bon et al., 1993;Shindoh et al., 1996;Hammerle et al., 2003). Interestingly, recent genetic analyses suggest that DYRK1A is also a risk gene in ASD (Iossifov et al., 2012;O'Roak et al., 2012a,b;Chen et al., 2014a;Krumm et al., 2014;Redin et al., 2014;Bronicki et al., 2015;van Bon et al., 2016). Expression of Dyrk1a in mouse brain is limited to early developmental periods and can promote neurite formation (Okui et al., 1999;Hammerle et al., 2003;Gockler et al., 2009). In addition, DYRK1A phosphorylates N-WASP, a cytoskeletal protein, to inhibit spine formation in primary hippocampal neurons . Pyramidal neurons in Dyrk1a +/− mouse cortex have reduced dendritic branches and dendritic spine density, which potentially causes the reduced brain size in these mice (Fotaki et al., 2002;Benavides-Piccione et al., 2005). On the other hand, overexpression of DYRK1A in mice causes increased spine density in cortical pyramidal neurons, and these animals show prefrontal deficits including significant impairment of spatial learning and cognitive flexbitiliy (Altafaj et al., 2001;Thomazeau et al., 2014). However, overexpressing DYRK1A in primary cortical mouse neurons significantly reduces dendrite complexity through disruption of REST/NRSF levels and REST/NRSF-SWI/SNF chromatin remodeling complex (Lepagnol-Bestel et al., 2009).
The identification of the vulnerable intracellular signaling pathways will aid us in the pursuit to find new therapeutic drug targets in patients with ASD. Interestingly, a variety of gene mutations result in disruption of the mechanistic pathways that these signaling molecules participate in. Therefore, several of the experimental pharmacological agents currently proposed as possible treatment strategies for autistic phenotypes are targeting these signaling molecules (see "Perspectives").

Scaffolding Proteins Provide Supporting Roles to Connect Structural and Signaling Molecules
Synaptic signaling processes are key to proper neural function. Some pivotal components of synapses are postsynaptic scaffolding proteins, which cluster neurotransmitter receptors, cell adhesion proteins, ion channels and cytoskeletal molecules to a confined postsynaptic region (Kim and Sheng, 2004;Sheng and Hoogenraad, 2007). Dysfunction in scaffolding proteins often has a huge impact on neuronal function, including neuronal morphology and synaptic plasticity (Ting et al., 2012). Emerging evidence has recently linked ASD with mutations of several genes encoding scaffolding proteins as described below.

Molecules Regulating Synaptic Vesicles Are Implicated in the Regulation of Neurite Outgrowth
Neurotransmitter release is regulated by the cycling of synaptic vesicles at the axonal terminal. The regulation of synaptic vesicles contains several steps and requires precise interaction of several specialized proteins, including SNARE complex for membrane fusion and syntaxin for vesicle docking. STXBP5 encodes a syntaxin-binding protein, tomosyn that negatively regulates neurotransmitter release by forming a syntaxin-SNAP25-tomosyn complex (Fujita et al., 1998;Sakisaka et al., 2004;Yizhar et al., 2004;Yamamoto et al., 2009Yamamoto et al., , 2010Bielopolski et al., 2014). Neuron-specific tomosyn deletion in mouse hippocampal dentate gyrus impairs spatial learning and memory, whereas tomosyn knockdown in dentate gyrus decreases synaptic plasticity of mossy fibers (Barak et al., 2013;Ben-Simon et al., 2015). Tomosyn also regulates SNARE complexes via ROCK phosphorylation of syntaxin-1 to control neurite outgrowth (Sakisaka et al., 2004). Recent genetic studies have identified the association of STXBP5 and ASD (Davis et al., 2009;Cukier et al., 2014;De Rubeis et al., 2014).
PRICKLE1 encodes PRICKLE1 protein, which has been traditionally thought to regulate the Wnt/beta-catenin signaling pathway to control epithelial planar cell polarity and cell migration during neural tube formation (Heitzler et al., 1993;Carreira-Barbosa et al., 2003;Veeman et al., 2003;Jenny et al., 2005). Intriguingly, the Prickle1 +/− mice exhibit autism-like behaviors, which may result from disrupted interaction with synapsin, a regulator of neurotransmitter release, suggesting that PRICKLE1 plays a critical role in synaptic vesicle regulation (Paemka et al., 2013). In addition, knockdown of PRICKLE1 in mice results in reduced axonal and dendrite formation in hippocampal neurons (Liu et al., 2013a). More recently, variants of PRICKLE1 have been found in individuals with autism (Cukier et al., 2014;Toma et al., 2014).
Synaptic scaffolds are crucial not only to maintain the structural stability of dendritic spines and synapses but also to link the signaling molecules and receptors to efficiently act in response to certain extracellular stimuli. Mutations in these molecules may disrupt several different signaling pathways and result in wide range of cellular defects, which sometimes are not limited to ASD. In addition, couple autism-associated genes that have been shown to regulate synaptic vesicles also play roles in neurite outgrowth or synaptic plasticity thereby regulating the structural stability of neurons. An interesting direction of investigation is whether the regulation of synaptic vesicles represents one of the key vulnerable cellular pathway that contributes to the alteration of neuronal structures in ASD.

SPECIFIC SYNDROMIC DISORDER RELATED GENES
Several autism-related neurodevelopmental disorders, such as Fragile X, Rett, Angelman syndromes (AS), and tuberous sclerosis are caused by a highly penetrable mutation of a single gene, e.g., FMR1 in Fragile X (Verkerk et al., 1991;Gedeon et al., 1992), MECP2 in Rett (Amir et al., 1999), UBE3A in AS (Kishino et al., 1997), and TSC1/2 in tuberous sclerosis complex (TSC; Povey et al., 1994). In recent DMS-5 criteria, however, ASD condition has been separated out from these single gene related disorders. Interestingly, these molecules all have a major function in regulating gene expression or protein synthesis, which in turn widely affects the structural stability of neurons. Because of the comorbidity between ASD and these single gene related disorders, we also review the current understanding of these genes and discuss how alterations of these genes may impair the structural integrity of neurons.

UBE3A
UBE3A gene is a paternally imprinted gene located at human chromosome 15 and encodes a member of the E3 ubiquitin ligase proteins (Huibregtse et al., 1993;Albrecht et al., 1997). Because UBE3A is selectively imprinted in mature neurons, epigenetic regulation of UBE3A has been associated with several neurodevelopmental disorders (Albrecht et al., 1997;LaSalle et al., 2015). Mutations resulting in loss-of-function in the maternally expressed copy of UBE3A causes AS, a severe developmental disorder characterized by delayed development, intellectual disability, severe speech impairment, and ataxia (Kishino et al., 1997). Maternal duplication of UBE3A results in Dup15q syndrome, a developmental disorder that has many similarities with AS but also exhibits several autistic traits (Cook et al., 1997;Wang et al., 2008;Hogart et al., 2010;Smith et al., 2011;Urraca et al., 2013;Al Ageeli et al., 2014;Germain et al., 2014). Coincidently, several genome-wide studies from individuals with autism identify UBE3A as an autism-risk gene (Nurmi et al., 2001;Glessner et al., 2009;Schaaf et al., 2011;Kelleher et al., 2012;Carvill et al., 2013;Iossifov et al., 2014;Yuen et al., 2015). In addition to its function of ubiquitin ligase to catalyze the protein degradation step, UBE3A also can act as a transcriptional coactivator for the nuclear hormone receptor superfamily of transcription factors (Nawaz et al., 1999). UBE3A localizes both in the nucleus and cytosol, including dendrite and pre-and post-synaptic compartments in neurons to regulate dendrite and dendritic spine morphology (Dindot et al., 2008;Valluy et al., 2015). Although maternal deletion of Ube3a does not affect dendrite arborization in mouse brains, knockdown of UBE3A in cultured neurons results in defects of dendrite polarization in pyramidal neurons (Dindot et al., 2008;Miao et al., 2013). Maternal-deficiency of Ube3a in mouse brain, however, shows defects in dendritic spine development in the cortex, hippocampus, and cerebellum (Dindot et al., 2008;Kim et al., 2016). Furthermore, several neuronal substrates for UBE3A have been identified, including Arc (Greer et al., 2010), the Rho-GEF Pbl/ECT2 (Reiter et al., 2006), Ephexin5 (Margolis et al., 2010), and TSC2 (Zheng et al., 2008). Their regulation by UBE3A provides molecular mechanisms to explain how synaptic integrity is maintained and how alteration of this interaction contributes in part to neuronal phenotypes in neurodevelopmental disorders. In addition, a recent study demonstrates that a PKA phosphorylation-defective mutation on UBE3A found in an individual with autism resulted in an increase of dendritic spine density (Yi et al., 2015).

PERSPECTIVES
Diagnosis of ASD cases has risen dramatically in recent years. The increased awareness of the symptoms and the broader definition of the spectrum may be major contributing factors for the rising number of ASD cases. Thus, there is an increased interest on understanding the etiologies of ASD. It is widely accepted that the genetic component plays a major role in ASD, however, except for the direct inheritance of some syndromic conditions, it is difficult to identify risk factors for autism. It is possible due to the low sample size and the high heterogeneity of genetic variances to have sufficient statistical power to make conclusive correlations (Geschwind and State, 2015). Among those autism-risk genes identified to date, some of the autism associations are due to de novo mutations, and some are familial variants (Table 4). Whether the inheritance pattern exhibits a risk factor is still not clear, however, the diverse gene mutations found in different individuals with autism suggest that instead of focusing on the genes per se, identifying the vulnerable pathways that these genes regulate may provide better clues toward understanding the contributing cellular and molecular changes that reserve in the autism phenotypes. The cellular defects resulting from different combinations of gene mutations contribute to the diverse phenotypes observed in autism. The heterogeneity of symptoms in ASD further complicates the diagnosis and treatment. However, understanding how autism-associated genes function in the regulation of key cellular pathways will provide insights to how therapeutic intervention can be more targeted and efficient to treat affected individuals.
Neuroanatomical studies of individuals with autism suggest a common disruption of neuronal structures with a decrease of dendrite arborization but an increase of dendritic spine density in select brain regions (Raymond et al., 1996;Bauman and Kemper, 2005;Hutsler and Zhang, 2010;Tang et al., 2014). This feature is distinct from other neurodevelopmental disorders, such as Rett or Fragile X syndromes, where the dendrite arbors and dendritic spine density are both downregulated (Kulkarni and Firestein, 2012). Intriguingly, dendrite arborization completes prior to dendritic spine formation during development. Although it has been proposed that the pruning mechanism of dendritic spines is defective in ASD (Frith, 2003), it is also plausible that the increase of dendritic spine density may be a compensation to re-establish the sufficient quantity of connections with fewer dendrite arbors. However, the precise spatial arborization of dendrites is critical for correct pre-and post-synaptic innervation when establishing the brain circuitry. The local increase of dendritic spine density may not be sufficient to compensate the effect from the loss of dendrite arbors, and may instead result in abnormal synaptic activity to disrupt normal neuronal function. This further emphasizes the importance of the establishment of structural integrity for neurons in order to provide proper brain function. In addition, the mechanisms of action of many current pharmacological agents for treating ASD affect normal neuronal function including the structural stability of neurons. With the early onset of ASD, the treatment often occurs at a very young age when the brain is still undergoing the period of development and maturation. As these pharmacological treatments may be beneficial to ameliorate some symptoms in ASD, the general brain development of these individuals may also be affected (Penagarikano, 2015). Thus, more precise circuitry-specific therapeutic intervention is needed to reduce the unwanted effect to the developing brain. Understanding the genetic and cellular pathways affected in ASD should provide more selective candidates for developing targeted intervention.
To date, several animal studies have tried to model the behavior phenotypes in autism, however, there remains a debate as to whether rodents can sufficiently recapitulate the complexities of the condition in human. The heterogeneity of genetic components also make it difficult to establish reliable animal models to describe the cellular and molecular mechanistic alterations in specific pathways. However, the studies on rodents can suggest which brain circuitry should be the area of interest for the corresponding behavior. iPSCs derived from ASD individuals appear to be an attractive model systems that allow researchers to directly investigate the interaction between the genetic contribution and the autism-relevant phenotypes. However, what is lacking in this system is a physiological relevant environment to correlate the behavior and the cellular phenotype. A recent emerging genetic editing technique, CRISPR/Cas9 (Jinek et al., 2012), is a powerful tool to study the mechanistic questions and identify the potential therapeutic interventions. Unlike the traditional knock-in or knock-out technique, CRISPR/Cas9 can introduce genomic editing of several genes at once. Using CRISPR/Cas9 in iPSCs can potentially determine the genetic contribution to the cellular phenotypes and provide a mechanism to correct them. However, there is still room for improvement of the efficiency and precision before this technique can be reliably used in clinical applications. Combining animal studies, iPSC models, and gene editing techniques, it is now possible to perform more comprehensive translational research in order to better understand the etiologies of ASD and design more efficient and effective therapeutic interventions.

AUTHOR CONTRIBUTIONS
All authors listed, have made substantial, direct and intellectual contribution to the work, and approved it for publication.

FUNDING
This work is supported by Hussman Foundation grant HIAS15003 to Y-CL.