During development, the human central nervous system (CNS) carries out an extraordinary undertaking: it forms on the order of 1015 connections between 1012 neurons. Considering this huge bioengineering challenge, it may be surprising that for the majority of humans the end result is immensely successful at performing a wide range of tasks. These tasks range from monitoring of sensory inputs to motor control output, and to higher cognitive tasks such as memory storage and retrieval, and social interaction. Unfortunately, though, things do go wrong. This is underlined by the appearance of symptoms of neurodevelopmental disorders (NDDs) such as autism spectrum disorders (ASD) and intellectual disability that coincides with the period of massive synaptogenesis in the developing cortex during the first 2 years of an infant’s development (Huttenlocher and Dabholkar, 1997).

Here, we will use NDDs to collectively designate all disorders that impair the development of the CNS. These commonly include autism and ASDs, attention deficit-hyperactivity disorder (ADHD), intellectual disability (ID), Down’s syndrome, and Tourette’s syndrome (TS). In addition, a large body of evidence suggests that schizophrenia (SCZ) and some forms of bipolar disorder may have developmental roots. Some studies indicate a correlation between individuals destined to manifest SCZ as adults and delays in motor and cognitive milestones during childhood, reaching back to as early as 6 months of life (reviewed by Weinberger and Levitt, 2011). It is possible that some forms of bipolar disorder manifest themselves in children as pediatric mania; however, this is difficult to diagnose and is often misdiagnosed with ADHD and conduct disorder (Biederman et al, 2000). Thus, we consider any possible CNS disorder that may have roots during childhood as a NDD.

The temporal correlation between diagnosis of a subset of NDDs and the period of synapse assembly led to the hypothesis that the deficit underlying many cases of these disorders may reside at the synapse (Bourgeron, 2009; Zoghbi, 2003). Since formulation of the idea of a synaptic pathology (or ‘synaptopathy’) lying at the heart and being a common denominator for NDDs (Brose et al, 2010), many of the genes that have been implicated in these disorders by genetic studies have been examined either in the test tube, in tissue cultures, or in genetically modified organisms (primarily mice) with a possible synaptic etiology in mind. Astonishingly, a large proportion of the genes that have been identified in cases of NDDs have identified roles at the synapse either during development or in the adult (Aldinger et al, 2011; Peca et al, 2011b). Although not all of the studies argue for strictly developmental deficits, a few recent observations argue for shortfalls during synapse assembly and maturation in at least some forms of ID, ASD, and SCZ. Here we will examine the current understanding of the assembly of synapses and how mutations in some genes identified in NDDs affect this process.


Synapse formation, as we currently understand it at the molecular and cellular level, can be dissected into three major steps: (1) contact formation, (2) transport of synaptic components within those neurites, and (3) immobilization of the synaptic components at the nascent synaptic contact sites (McAllister, 2007) (Figure 1). Genes implicated in NDDs have been found to be involved with all three aspects of synaptogenesis. In general, deficits in NDD susceptibility genes are uncommon, up to 2% of NDD cases for the more common variants (Toro et al, 2010), and collectively only account for between 10 and 20% of NDDs. However, identifying mutations that genuinely cause dysfunction of a gene, and thus implicating it in the etiology of NDDs, tells us much about the underlying biology of CNS and synapse development. Below, we present examples of genes with mutations and copy number variations (CNVs) that are both highly likely to result in protein function deficits or protein expression level differences, and for which studies clearly indicate an involvement in the three steps of synapse assembly (Table 1). This should not be understood as a comprehensive list, but serves as a demonstration of how genetic studies of the NDD population and basic research, primarily in mouse models, have helped further our understanding of synapse assembly and some aspects of NDDs. We then examine genes identified in syndromic forms of ASDs, and how the encoded proteins might function during synaptic development.

Figure 1
figure 1

The three basic steps of synapse assembly. (1) First neuronal contact triggers a cascade of events through cell adhesion molecules such as Nlgns, Nrxns, and SynCAMs. Pre- and postsynaptic precursors need to be efficiently transported in axons and dendrites (2), by associating with motor proteins, such as Kinesin17, that ‘walk’ along microtubules. Neuronal contact must result in stopping, perhaps by means of kinases such as Cdk5, and maintenance of the synaptic material at the nascent synaptic site because of scaffolding molecules, such as SHANKs and CASK (3). This process eventually leads to the formation of a functional synaptic contact (4).

PowerPoint slide

Table 1 Summary of Genes Associated with NDDs That May Be Involved in Steps of Synapse Assembly

Neuronal Contact

Genes involved in neuronal cell adhesion and synaptic scaffolding are collectively the most prevalent genes found to date in genome-wide association studies (GWASs) of NDDs (Corvin, 2010; Kumar et al, 2011; O’Dushlaine et al, 2011). This suggests that cell adhesion events are critical for the correct assembly of synapses (Figure 1, top left). There are hundreds of genes encoding potential cell adhesion molecules (CAMs) expressed in the CNS; however, to date only a handful of them have been found to be able to induce synapse assembly (Bukalo and Dityatev, 2012; Washbourne et al, 2004). This has largely been defined by the expression of candidate CAMs in nonneuronal cells and their ability to induce either pre- or postsynaptic assembly in cocultured neurons (Biederer et al, 2002; Biederer and Scheiffele, 2007; Fu et al, 2003; Scheiffele et al, 2000; Siddiqui et al, 2010).


The first family of CAMs that were shown to be sufficient to drive synapse assembly in vitro was the neuroligins (Scheiffele et al, 2000). These had been identified as ligands of the presynaptically localized α-latrotoxin receptors, the neurexins (Song et al, 1999). Indeed, subsequent studies demonstrated that binding of neuroligins (Nlgns) to the neurexins (Nrxns) resulted in assembly of presynaptic terminals in axons (Dean et al, 2003), and that binding of Nrxns to Nlgns drove postsynaptic differentiation in dendrites (Barrow et al, 2009; Nam and Chen, 2005). The in vitro studies suggest that Nlgn1 is key in triggering the recruitment of NMDA-type glutamate receptors (Barrow et al, 2009; Hoy et al, 2013), but this may be a late maturational step of synapse formation.

The neuroligins are encoded by five genes in humans (1, 2, 3, 4X, and 4Y), and mutations in NLGN1, 3, and 4X genes have been associated with autism (Sudhof, 2008) (Table 1). Whereas loss-of-function mutations have been found in patients with ASDs for NLGN3 and NLGN4X (Jamain et al, 2003; Laumonnier et al, 2004), an increase in copy number of the NLGN1 gene has been associated with ASD (Glessner et al, 2009; Szatmari et al, 2007) (Table 1). Analyses in mice have confirmed that loss-of-function mutations in Nlgn3 and Nlgn4 result in abnormal synaptic function and behavioral deficits (Ellegood et al, 2011; Jamain et al, 2008; Radyushkin et al, 2009; Tabuchi et al, 2007). It is important to note that studies performed with the same Nlgn3 and Nlgn4 mutations as characterized previously (Jamain et al, 2008; Tabuchi et al, 2007), but carried out in different labs, showed no significant effects on social behaviors or ultrasonic vocalizations (Chadman et al, 2008; Ey et al, 2012). These results open the possibility that genetic background and environmental effects may contribute to the subtle behavioral phenotypes. Since then, social deficits have been confirmed for the Nlgn3 knock-in mice, expressing a mutant form of Nlgn3 (R451C) identified in patients with autism, using multiple background strains of mice (Jaramillo et al, 2014). This result adds weight to the argument that mutations in these genes may be the cause of some cases of ASD in humans. Similarly, mice in which Nlgn1 was overexpressed, to mimic the copy number increase, also displayed anomalous synaptic characteristics (Dahlhaus et al, 2010; Hoy et al, 2013). Although overexpression of Nlgn1 resulted in deficits in learning and memory behavior, social behaviors tested were not affected (Dahlhaus et al, 2010; Hoy et al, 2013). In conclusion, the Nlgn family of genes has been clearly implicated in ASD etiology through genetic association and mouse studies.


The Nrxns are encoded by three genes in mammals; and all of them generate two forms of the protein from different start codons: the long α-Nrxns and the short β-Nrxns (Tabuchi and Sudhof, 2002). Approximately 0.5% of ASD cases harbor mutations in Nrxn 1α (NRXN1), underlining the importance of this gene in the etiology of NDDs (Reichelt et al, 2012). Analyses with loss-of-function mutations of the Nrxn genes have confirmed the role of α-Nrxns in shaping synapse function (Missler et al, 2003; Sons et al, 2006) and complex behaviors (Etherton et al, 2009; Laarakker et al, 2012), including social behaviors (Grayton et al, 2013). Deletion of all three α-Nrxn genes results in death of newly born mouse pups because of reduced synaptic transmission (Missler et al, 2003).

Further study of the extracellular domains of Nrxns demonstrated that they are relatively promiscuous, participating in molecular interactions with a suite of other synaptic proteins in addition to the Nlgns. These include LRRTMs, cerebellin, and the GABA A receptor. LRRTMs appear to act in a manner similar to Nlgns in postsynaptic assembly (de Wit et al, 2009; Ko et al, 2009; Linhoff et al, 2009; Wright and Washbourne, 2011). One important difference is that Nlgns primarily affect NMDA-type glutamate receptor recruitment and maintenance at synapses (Barrow et al, 2009; Hoy et al, 2013), whereas LRRTMs influence AMPA-type glutamate receptor presence at synapses (Soler-Llavina et al, 2011). Binding of Nrxns to cerebellin (Uemura et al, 2010), which in turn binds to glutamate receptor subunit δ2 (Matsuda et al, 2010), may be a specialization in the cerebellum to control synapse number between parallel fibers and Purkinje cells (Hirai et al, 2005). It is unclear how the cis interaction between Nrxns and the GABA A receptor contributes to synapse formation (Zhang et al, 2010). Presumably this interaction occurs in dendrites at an early stage during development, before Nrxns redistribute to axons (Barker et al, 2008). Thus, these studies suggest that Nrxns are key organizers of the pre- and postsynaptic specializations.


Another family of molecules, identified by the ability to induce the formation of presynaptic terminals in coculture experiments, is the SynCAMs (Biederer et al, 2002). These proteins resemble nectins, in that they contain three extracellular immunoglobulin domains for homo- and heterophilic interaction. There are four SynCAM genes in humans (CADM1–4) (Biederer, 2006), one of which (CADM1) has been linked to ASD by virtue of two missense mutations that render SynCAM1 unable to be transported out of the endoplasmic reticulum to the plasma membrane (Zhiling et al, 2008). Knockout mice for SynCAM1 present a subtle reduction in synapse number (<10%), perhaps because of compensation by the other family members (Robbins et al, 2010). Examination of synapses by electron microscopy suggests an impact of SynCAM1 ablation on the size of both the presynaptic active zone and the postsynaptic density (Robbins et al, 2010). SynCAM1 suppresses long-term depression (LTD), as knockout mice show enhanced LTD, with no effect on long-term potentiation (Robbins et al, 2010). As a result, SynCAM1 KO mice exhibit enhanced spatial reference memory. The in vitro studies of SynCAM1 have suggested that the majority of this CAM function resides in its ability to modulate cytoskeletal dynamics during assembly of the presynaptic terminal (Stagi et al, 2010) and during formation of the postsynaptic spine (Cheadle and Biederer, 2012). Thus, SynCAM1 is clearly implicated in the development of fully functional synapses.

Mechanisms of CAM-triggered synapse assembly

To date, there appears to be a discrepancy between the sufficiency for CAMs to induce the formation of synaptic terminals in coculture systems and their necessity in the formation of synapses, especially in vivo. Expression of SynCAMs and Nlgns in nonneuronal cells and culturing of these cells with neurons suggests a potent ability of presynaptic SynCAMs and Nrxns in coordinating the recruitment of presynaptic components (Biederer et al, 2002; Scheiffele et al, 2000). But in vitro and in vivo studies of loss-of-function mutations of these genes have not shed light on how these molecules might perform this function (Robbins et al, 2010; Varoqueaux et al, 2006). This discrepancy may be because of compensation by other genes within the family, or by other CAM families. Alternatively, the effects of the loss of synaptic CAM function may only be discerned when there is competition between individual neurons, such that complete knockout does not show a difference, whereas if some neurons have reduced CAM expression they will exhibit reduced synapse numbers (Kwon et al, 2012).

Despite the lack of a strong effect on synapse formation by loss-of-function mutations, it was hypothesized that protein–protein interactions of the short intracellular tails of both Nrxns and SynCAMs might be the trigger for the accumulation of presynaptic components. Specifically, it was assumed that the presynaptic MAGUK protein CASK, which can bind to the terminal PDZ domain of both Nrxns and SynCAMs, might be the presynaptic organizer. This view was recently challenged by experiments in which the PDZ domain was switched for one that cannot bind to CASK, yet still resulted in functional presynaptic assembly (Gokce and Sudhof, 2013). Indeed, expression of a mutant form of Nrxn in which the entire intracellular tail was eliminated was still sufficient for presynaptic assembly to occur, including CASK recruitment (Gokce and Sudhof, 2013). These experiments suggest that Nrxns, and based on the similarity of the intracellular regions of perhaps also SynCAMs, function through an as yet unidentified transmembrane receptor to induce presynaptic assembly.

In contrast, the intracellular regions of Nlgns appear to play a central role in coordinating postsynaptic assembly and maturation. Protein–protein interactions of the intracellular tails of Nlgn1 help traffic NMDA receptors to synapses (Barrow et al, 2009) and mediate the switch in NMDA receptor type (from NR2B to NR2A containing) that occurs at synapses during development (Hoy et al, 2013). In addition, novel interaction domains within the C-terminus of Nlgns have been identified that are critical for both NMDA and AMPA receptor recruitment to synapses (Etherton et al, 2011; Shipman et al, 2012). Thus, whereas recruitment and maintenance of presynaptic components is not mediated by intracellular interactions of Nrxns and SynCAMs, it appears that protein–protein interactions within the C-terminal tail of Nlgns has a profound effect on postsynaptic assembly.

Axonal and Dendritic Transport of Synaptic Components

The vast majority of synaptic components are synthesized within the cell body. Thus, for the assembly of synapses to occur, synaptic precursor material must be transported within dendrites and axons to nascent synaptic sites (see Figure 1, top right). Transport of synaptic precursors to synapses involves motor proteins, dynamins, and kinesins that walk along microtubule tracks in both dendrites and axons (Goldstein et al, 2008; Hirokawa et al, 2010). Synaptic vesicle precursors and active zone precursors are transported in a microtubule-dependent manner in both antero- and retrograde directions (Ahmari et al, 2000; Bury and Sabo, 2011; Sabo and McAllister, 2003). Similarly, NMDA and AMPA receptors are transported along microtubules in both directions in dendrites (Washbourne et al, 2002). Although it is possible that some transmembrane synaptic components diffuse within the plasma membrane to synaptic sites (Groc et al, 2009), it is unlikely that this mechanism would deliver enough receptors, especially during peak synaptogenesis. Supporting a role for microtubule-based transport in synapse assembly, genes that encode axonal or dendritic motor proteins, microtubule-organizing proteins or regulators of microtubule transport have been uncovered as candidate genes in NDDs.


As microtubules are the rails along which long-distance axonal and dendritic transport is achieved, one would expect mutations that affect microtubule stability or organization to result in NDDs. Indeed, rearrangements of human chromosome 17 that affect the microtubule-associated protein Tau (MAPT) have been identified in patients with learning disability and developmental delay (Koolen et al, 2006; Rovelet-Lecrux et al, 2012; Shaw-Smith et al, 2006). Although MAPT has been more commonly associated with neurodegenerative disorders, because of its presence in neurofibrillary tangles (Goedert et al, 2006), this gene plays an active role during development. MAPT promotes microtubule assembly and stability and may act as a linker protein between the microtubules and the plasma membrane. Thus, a reduction in MAPT can affect neuronal migration and deficits in dendritic arbor elaboration and synapse number (Sapir et al, 2012).


A nonsense mutation in Kinesin17 gene (KIF17) was found in a case of schizophrenia (Awadalla et al, 2010; Tarabeux et al, 2010). The mutation resulted in truncation of the protein before the tail domain, the region important for interaction with the transport cargo (Goldstein et al, 2008). Similar truncations result in dominant negative motor proteins (Setou et al, 2002), explaining how such a mutation could result in deficits even when present in the heterozygous state (Tarabeux et al, 2010). Kinesin17 is an interesting candidate with regard to NDDs, as it was originally identified as a motor protein for NMDA receptor transport (Kayadjanian et al, 2007; Setou et al, 2000). Thus, overexpression of Kinesin17 results in enhanced spatial learning and memory in transgenic rodents (Wong et al, 2002). The ability to release the NMDA receptor-containing cargo is critical, as mutations that block phosphorylation of serines in the tail domain result in reduced synaptic NMDA receptor localization and deficient spatial learning and memory (Yin et al, 2012).


Mutations affecting splicing and expression levels of the glycogen synthase kinase 3β gene (GSK3B) are associated with schizophrenia (Blasi et al, 2013). Many of the roles of this signaling enzyme in neural function remain unclear. However, with regard to synapse formation, Gsk-3β has been implicated in the transport of amyloid precursor protein (APP) in axons, and in the wnt signaling pathway. Although the role of APP outside of the pathway to Alzheimer’s disease remains elusive, interactome studies suggest that APP can bind synaptotagmin-1, the calcium sensor for synaptic vesicle release (Kohli et al, 2012) and the axonal motor protein Kinesin-1 (Kamal et al, 2001). However, there has been some controversy as to whether APP is important for the transport of these components in axons (Lazarov et al, 2005). Gsk-3β phosphorylates APP, reducing its axonal transport (Takashima et al, 1995), suggesting that Gsk-3β may play a generalized role in regulating transport of synaptic components to synapses. In another potential synaptic function, Gsk-3β regulates wnt signaling at vertebrate synapses and the invertebrate neuromuscular junction to modulate presynaptic microtubules during synapse formation (Salinas, 2005). Gsk-3β plays a role in this pathway by phosphorylating microtubule-associated protein 1B (MAP1B), resulting in more dynamic microtubules (Goold et al, 1999). In addition, Gsk-3β plays a role in the mobilization of postsynaptic density protein-95 (PSD-95) (Nelson et al, 2013), suggesting a wide variety of roles for this kinase in shaping the synapse.

Recruitment and Maintenance of Synaptic Components at Nascent Synaptic Sites

Once synaptic precursors reach the nascent synaptic site, they need to stop, reorganize, and be maintained at the developing synapse (see Figure 1, bottom left). Kinases may regulate synaptic precursor stopping at the synapses, whereas scaffolding molecules will lock synaptic components into a network of interactions to hold them at the mature synapse (Figure 1, bottom right).


Mutations and microdeletions in the SHANK genes have been found among individuals with ASDs (Awadalla et al, 2010; Berkel et al, 2010; Guilmatre et al, 2014; Leblond et al, 2012; Lennertz et al, 2012; Sato et al, 2012). These genes encode scaffolding molecules, molecules whose sole job is to bind to other proteins inside the cell exclusively at synapses (Grabrucker et al, 2011). Studies of knockout mice for SHANKs 1, 2, and 3 suggest that reduced levels of these SHANK proteins result in synaptic deficits (Bozdagi et al, 2010; Sala et al, 2001; Wang et al, 2011) with consequences for cognitive and social behavior (Bozdagi et al, 2010; Peca et al, 2011a; Silverman et al, 2011; Wang et al, 2011; Wohr et al, 2011; Won et al, 2012). SHANKs are interacting partners of Nlgns and form a link between the transsynaptic CAMs and important synaptic signaling molecules inside the postsynaptic terminal including NMDA receptors (Petralia et al, 2005). Indeed, SHANK2 deficits can be reverted in mice by restoring NMDA receptor function (Won et al, 2012). The wealth of mutational information from the ASD population and the strong indications from mouse experiments point toward a significant contribution of SHANK gene deficits underlying ASD etiology.


Although scaffolding molecules such as SHANKs can explain the maintenance of a variety of components at synapses, it is less clear how these components might initially be recruited to a nascent synaptic site. An inkling into mechanisms of recruitment can be gained by examining cyclin-dependent kinase 5 (CDK5). The gene for the regulatory subunit of CDK5 (CDK5R1) has been linked to nonsyndromic intellectual disability (Venturin et al, 2006) and dysregulation of CDK5 expression has been found in the brains of patients with schizophrenia (Engmann et al, 2011; Ramos-Miguel et al, 2012). CDK5 has been ascribed many neuronal functions ranging from neuronal migration to the regulation of synaptic strength (Cheung and Ip, 2007; Dhariwala and Rajadhyaksha, 2008; Smith and Tsai, 2002). However, a number of studies have specifically examined the role of CDK5 in transport of synaptic components to synapses. In the assembly of presynaptic terminals in C. elegans, CDK5 regulates the deposition of all presynaptic components (Park et al, 2011). In contrast, in vertebrates CDK5 specifically regulates the appearance of CASK (Samuels et al, 2007) and synapsin at synapses (Easley-Neal et al, 2013). In fact, CDK5 activity must be local to allow synapsin transport packets to stop at sites of nascent synapse formation (Easley-Neal et al, 2013). These studies suggest that kinases may play critical roles in regulating the recruitment of presynaptic components by generating a local STOP signal at sites of synapse formation. What exactly constitutes the STOP signal is unclear, but could include inhibition of the motor proteins (Yin et al, 2012) or uncoupling of the cargo from the motor proteins by phosphorylation (Samuels et al, 2007).


Mutations in the calcium/calmodulin-dependent serine kinase (CASK) gene have been identified in cases of X-linked optic atrophy, X-linked mental retardation, and microcephaly with disproportionate pontine cerebellar hypoplasia with associated intellectual disability (Dimitratos et al, 1998; Froyen et al, 2007; Hackett et al., 2009; Hayashi et al, 2008; Najm et al, 2008). CASK protein appears to have two very distinct functions (Hsueh, 2006). During embryonic development, 20% of the protein is present in the nucleus where it can interact with the transcription factor TBR1 (Hsueh et al, 2000). However, the protein also contains PDZ, SH3, and a guanylate kinase domain that are characteristic of membrane-associated guanylate kinase (MAGUK) proteins. MAGUK proteins, the most famous of which is perhaps PSD-95, are classic synaptic scaffolding proteins (Craven and Bredt, 1998). Indeed, CASK localizes at synapses and can interact with the intracellular tail of SynCAMs and Nrxns (Biederer et al, 2002; Hata et al, 1996). CASK associates with the α1B subunit of the voltage-gated calcium channel (VGCC) and this interaction determines VGCC localization at synapses (Maximov and Bezprozvanny, 2002; Maximov et al, 1999). As mentioned above, the synaptic localization is regulated by phosphorylation by Cdk5 (Samuels et al, 2007). Interestingly, from the perspective of synaptogenesis, CASK assembles with Mint1 and Veli to form a motor protein/cargo adaptor (Setou et al, 2000). This suggests that CASK may not only help maintain synaptic proteins, such as calcium channels at synapses, but also be involved in the transport of synaptic cargo to developing synapses (Figure 1, top right). Thus, mutations in CASK may contribute to neuronal and synaptic development at multiple levels; as a transcription factor during neuronal differentiation and migration, and during synaptic assembly and maturation.


There are several syndromes that manifest with social deficits, bringing them into the autism spectrum. Three of these strictly mendelian syndromes, tuberous sclerosis, fragile X mental retardation, and Rett syndrome, are because of mutations in genes that have now been shown to have roles in synaptic function and potentially also synapse formation. Here we discuss how the gene products of the TSC1 and 2, FMR1, and MECP2 genes may be involved in synapse formation.

Tsc1 and Tsc2

Tuberous sclerosis (TSC) is an inherited disease manifested with nonmalignant growths in the brain, skin, and other tissues. These growths, often called tubers in the brain, arise because of imbalances in cell growth signals. Importantly, 60% of individuals with TSC also manifest deficits in intellectual ability and social behavior, placing TSC within the autism spectrum (Napolioni and Curatolo, 2008). The genes affected are the tumor suppressor genes Tsc1 and Tsc2. The encoded proteins, hamartin and tuberin, interact with Rheb GTPase thereby inactivating the mammalian target of rapamycin (mTOR) complex (Napolioni and Curatolo, 2008). The mTOR is a kinase that regulates not only cell growth, but also cell motility, transcription, and protein synthesis. Importantly, recent studies have uncovered the importance for mTOR signaling at the synapse, potentially explaining the association with neurodevelopmental disorders: mTOR acts as a regulator of activity-dependent transcription of synaptic plasticity genes (Gipson and Johnston, 2012; Troca-Marin et al, 2012). More recently, Tsc1 and Tsc2 have been shown to function antagonistically with another mendelian NDD gene, Fmr1 (Auerbach et al, 2011). Thus, Tsc1 and Tsc2 may be involved in the activity-dependent consolidation of synapses during development.


Fragile X mental retardation is the most common form of inherited intellectual disability and the most common single gene cause of autism. The disease arises because of chromosomal instabilities of the X chromosome (Xq27.3). The affected gene, FMR1, encodes for an mRNA binding protein that has been shown to regulate the transcription of synaptic mRNAs in an activity-dependent manner (McLennan et al, 2011). Examples of the genes that are regulated include those encoding glutamate receptor subunits and calcium-calmodulin kinase II (CaMKII). The chromosomal instabilities result in increased numbers of CGG repeats in the promoter region that in turn cause decreased expression levels of Mecp2 protein. This decrease in expression level results in synaptic deficits, including an increase in synaptic spine density and an increase in immature spine shapes in humans (Irwin et al, 2001) and mice (Greenough et al, 2001). In addition, knockout mice exhibit an increase in metabotropic glutamate receptor-dependent LTD (Huber et al, 2002; Martin and Huntsman, 2012). Interestingly, as the LTD effect was opposite to the LTD deficit observed in Tsc2 mice, crossing of the two mutations into the same mouse resulted in the effects canceling each other out (Auerbach et al, 2011). Thus, it has been hypothesized that Tsc2 and Fmr1 lie in parallel, antagonistic pathways that both control the synthesis of LTD-specific proteins at synapses (Auerbach et al, 2011). In another comparative study of knockout mice, it was shown that both Nlgn3 knockout mice and Fmr1-null mice share a similar phenotype, including perturbed metabotropic glutamate receptor-dependent LTD (Baudouin et al, 2012). These studies suggest that both syndromic and nonsyndromic forms of autism may share synaptic mechanisms. In addition, one can conclude that, in the case of FMR1 and Tsc2, mutations affect common pathways, although in different directions. In conclusion, with regard to synapse development, FMR1 appears to be critical in the pruning back of inappropriate synapses.


Rett syndrome is caused by mutations in the methyl-CpG binding protein 2 (MECP2) gene. This gene is located on the X chromosome in humans, leading to more severe defects in affected males than females. In fact, the majority of males die within the first 2 years after birth. The encoded protein has a DNA-binding domain that preferentially interacts with methylated DNA in chromatin. Based on this observation, it was assumed that the protein is a transcriptional repressor (Boyes and Bird, 1991); however, it has been also seen to drive the expression of some genes (Della Ragione et al, 2012). Importantly, it has been shown to affect the development of dendrites and synapses, regardless of when the gene is inactivated using a conditional mouse knockout model (Nguyen et al, 2012). Furthermore, it appears that the deficit may lie within glial cells. Coculturing of neurons with Mecp2-deficient astrocytes is sufficient to elicit the neuronal phenotype (Ballas et al, 2009). This suggests that Mecp2 may regulate the expression of secreted glial factors that contribute to neuronal development (see below).


During early studies of the development of synapse formation in primary neuronal cultures, it was apparent that ‘feeder layers’ of glial cells, in particular astrocytes, were critical to drive synapse assembly between various populations of neurons (Banker, 1980; Kaech and Banker, 2006; Pfrieger and Barres, 1997). Since then, a number of factors released by astrocytes have been identified, including thrombospondins, glypicans, even cholesterol, and their mechanisms for driving synapse assembly and elimination have been examined (comprehensively reviewed by Clarke and Barres, 2013). As mentioned previously, mutations in Mecp2 (Ballas et al, 2009), and also FMR1 (Jacobs and Doering, 2010), result in astrocytes that do not maintain correct synaptic development, clearly implicating some of these factors in the etiology of ASD-linked disorders.

For this review, we will consider the glypican 6 (GPC6) gene in more detail, as it has been linked to ADHD and neuroticism (Calboli et al, 2010; Lesch et al, 2008), a character trait characterized by anxiety and depressed mood. Biochemical fractionation studies of astrocyte-conditioned media led to the identification of Gpc4 and Gpc6 as factors that could significantly increase the formation of synapses (Allen et al, 2012). Whereas astrocyte-secreted thrombospondins induce the formation of structural, but nonfunctional, synapses (Christopherson et al, 2005), Gpc4 and Gpc6 administration to neurons in culture results in the addition of functional synapses. Conversely, knockout mice for Gpc4 exhibit deficits in synapse formation (Allen et al, 2012). Together, these results argue for a significant role for this astrocyte-secreted factor in synaptic development. It will be interesting to discover the neuronal ligand that mediates the effects of Gpcs, thereby uncovering the molecular mechanism for astrocyte-induced synaptogenesis.


In conclusion, genetic studies combined with studies primarily in mouse models of the candidate gene defects have contributed to our understanding of the underlying mechanisms of NDDs. Although much work is still needed, it is apparent that, at least for a subset of NDDs, deficits during development have a significant impact on synapse assembly and function, and consequently on behavior. Much work still needs to be carried out. The number of mouse models of candidate genes needs to be expanded to gain a more complete picture of the underlying molecular mechanisms of NDDs. Existing mouse models, or perhaps even zebrafish with high-throughput capacity (Tropepe and Sive, 2003), can be systematically tested for pharmacological rescue of behavioral deficits, thereby generating lead drug candidates for treatment studies. In addition, it will be important to study whether candidate gene function is absolutely necessary during development of the synapse or whether some aspects of synaptic function can be regained in the adult.

Although determining whether a genetic defect primarily affects development of synapses rather than mature synapse function may seem, at first glance, an esoteric endeavor, this difference has important implications for therapy. With the advent of viral-based gene therapy applications almost a realistic possibility in humans (Witt and Marks, 2011), it will be important to know where and also when an intervention is critical for a successful outcome. For example, if a synaptic deficit can be overcome in the adult, then this provides hope for many adults with this form of ASD. In contrast, when it is known that a developmental synaptic deficit is hardwired into the mature CNS, it advocates for an intervention as early as possible. Although the genetic critical periods are not the only deciding factors in the ability to treat a deficit, especially by pharmacological means, information on when and where a gene product’s function is necessary will certainly give insight into windows of opportunity.

To date, only a few of the NDD candidate genes have been studied with temporal requirements in mind. These kinds of experiments are difficult, requiring either the ‘rescue’ of deficits with viruses in a known location during development and in the adult, or the generation of conditional mutations, allowing for genetic rescue at given times. Such studies have been performed for Mecp2 and Nlgn3, and for Nlgn1 overexpression, but are necessary for more models of NDDs.

The most striking temporal rescue was demonstrated using a floxed STOP allele of the Mecp2 gene in mice. Re-expression of Mecp2 protein starting at 10 weeks after birth ameliorated neurological symptoms and promoted survival of male null mice many weeks beyond their null littermates that did not re-express Mecp2 (Guy et al, 2007). These experiments conclusively demonstrated that Rett-like defects in mice can be rectified in adult mice. Similarly, late expression (P30) of Nlgn3 in a null background rescued an ectopic synapse formation phenotype in the cerebellum (Baudouin et al, 2012). Nlgn3 knockout mice demonstrate a motor coordination phenotype that is rescued by re-expression during development. Amelioration of the motor coordination phenotype was not tested in the juvenile rescue experiment (Baudouin et al, 2012). It is also important to note that pharmacological intervention has been successful in improving synaptic and social deficits in adult mice. Shank2-deficient mice, which have the same microdeletion as seen in cases of ASD, manifest decreased NMDA receptor function, reduced social interaction, and reduced ultrasonic vocalization (Won et al, 2012). Treatment of adult mice with D-cycloserine, a partial NMDA receptor agonist, or CDPPB, a positive allosteric modulator of the metabotropic glutamate receptor 5 (mGluR5, which modulates NMDA receptor function), reversed the synaptic deficits and enhanced social interaction (Won et al, 2012).

To contrast these examples of juvenile/adult rescue of neuronal and neurological deficits, adult reduction of Neuroligin1 overexpression did not rescue synaptic or behavioral symptoms (Hoy et al, 2013). Neuroligin1 overexpression in the hippocampus results in increased spine head size and reduced learning and memory behavior. When overexpression was turned off at 2 months of age, spine heads remained larger in size and repeated testing with the novel object recognition test showed no improvement of memory for up to 4 weeks (Hoy et al, 2013). In contrast, 4 weeks of overexpression that began in the adult did lead to decreased novel object recognition. This study suggests that overexpression of Neuroligin1 is sufficient to hardwire synapses into a pathological state during development, and that this state cannot be reversed by reducing expression levels in the adult. It is, however, possible that a pharmacological intervention might be successful in improving the behavioral deficits.

Given the heterogeneity of results from these models, that is, that Neuroligin1 overexpression deficits cannot be rescued after development (Hoy et al, 2013) whereas MeCP2, Neuroligin3, and SHANK2 deletions can (Baudouin et al, 2012; Guy et al., 2007; Won et al, 2012), it is imperative that we widen these types of studies to other genetic models. It is tempting to hypothesize that overexpression can generally not be rescued, whereas lack of expression can. However, this general rule does not seem to be the case as the increased number of synapses driven by SynCAM1 overexpression during development can only be maintained by continuous SynCAM1 overexpression in the adult (Robbins et al, 2010). As rescue is not equally possible in developing and adult mice, it seems that each individual mutation may have a defined critical period and will have to be experimentally determined in each case.

With the advent of personalized medicine, we can expect to see prenatal genomic profiling to become increasingly routine. Genomic data will need to be supplemented with a basic understanding of (1) the genes involved in NDDs, (2) the parts of these genes that are essential for synaptic function, and (3) the critical periods for the function of each gene. It is difficult to extrapolate genetic critical periods from mouse studies to humans, and much work is necessary to confirm these trends in human patients. However, an integration of these studies with an individual’s genome sequence may eventually allow personalized interventions, such as gene therapy and/or pharmaceutical treatments for the most severe cases of NDDs.


The author declares no conflict of interest.