SYN2 is an autism predisposing gene: loss-of-function mutations alter synaptic vesicle cycling and axon outgrowth

An increasing number of genes predisposing to autism spectrum disorders (ASDs) has been identified, many of which are implicated in synaptic function. This ‘synaptic autism pathway’ notably includes disruption of SYN1 that is associated with epilepsy, autism and abnormal behavior in both human and mice models. Synapsins constitute a multigene family of neuron-specific phosphoproteins (SYN1-3) present in the majority of synapses where they are implicated in the regulation of neurotransmitter release and synaptogenesis. Synapsins I and II, the major Syn isoforms in the adult brain, display partially overlapping functions and defects in both isoforms are associated with epilepsy and autistic-like behavior in mice. In this study, we show that nonsense (A94fs199X) and missense (Y236S and G464R) mutations in SYN2 are associated with ASD in humans. The phenotype is apparent in males. Female carriers of SYN2 mutations are unaffected, suggesting that SYN2 is another example of autosomal sex-limited expression in ASD. When expressed in SYN2  knockout neurons, wild-type human Syn II fully rescues the SYN2 knockout phenotype, whereas the nonsense mutant is not expressed and the missense mutants are virtually unable to modify the SYN2 knockout phenotype. These results identify for the first time SYN2  as a novel predisposing gene for ASD and strengthen the hypothesis that a disturbance of synaptic homeostasis underlies ASD.


INTRODUCTION
Autism spectrum disorders (ASDs) are a heterogeneous group of disorders characterized by impaired social relationships, rigid and repetitive behavior, restricted interests and abnormal language development (1). Genetic factors are playing an important role in ASD (2 -4), and an increasing number of genes predisposing to the disease have been identified over the past 10 years. Notably, the majority of the ASD-predisposing genes thus far identified encode for synaptic proteins, such as the postsynaptic proteins neuroligins 3 and 4 (NLGN3, NLGN4), their cytoplasmic interactors SHANK2 and SHANK3 and their presynaptic partner neurexin-1 (NRXN1) (5)(6)(7)(8)(9)(10)(11)(12)(13). Genetic variants in another member of the neurexin superfamily, contactin-associated protein-like 2 (CNTNAP2) have also been associated with ASD (14)(15)(16)(17), as well as mutations in two additional synaptic genes, namely IL1RAPL1 and RIMS3/NIM3 (18,19). The fact that the majority of the identified ASD-predisposing genes encode for synaptic proteins led to the 'synaptic autism pathway' hypothesis, holding that ASD is due to abnormal synaptic function and neural connectivity in the time window in which neuronal circuits are remodeled by experience (20,21).
Epileptic seizures are observed in up to one-third of ASD individuals (1,22) and autistic features are commonly observed in severe forms of epilepsy (23). Recently, mutations in human SYN1 have been reported to represent a common basis for both ASD and epilepsy (24,25). Although most of the known epilepsy predisposing genes implicate voltage-gated or ligand-gated ion channels (26), defects in synaptic proteins implicated in neurotransmitter release and synaptic vesicle (SV) trafficking have been frequently associated with an epileptic phenotype in mouse models (26 -30).
Synapsins (Syns) are a family of neuron-specific SV phosphoproteins implicated in synaptic transmission and plasticity (31). In mammals, Syns are encoded by three distinct genes (SYN1, SYN2 and SYN3) located on chromosomes Xp11.23, 3p25.2 and 22q12.3, respectively. Alternative splicing generates distinct isoforms, termed a, b and b-like composed of a mosaic of individual and shared domains. Synapsins I and II are selectively expressed at nerve terminals in mature neurons, whereas the expression of Syn III is downregulated in mature neurons and the protein is not strictly confined to synaptic terminals. Synapsins contribute to the clustering of SVs and regulate their trafficking between the recycling pool (RP) and the readily releasable pool (RRP), thus defining SV availability for release in a phosphorylation-dependent fashion. Perturbation of Syn function in a variety of models leads to disruption of the organization of SV pools in the presynaptic compartment and to an increase in synaptic depression, underlining the role of Syns in sustaining neurotransmitter release in response to high-frequency activity (31)(32)(33). Moreover, recent studies have shown that Syns also play a role in the post-docking stages of release and their perturbation leads to an imbalance between the activities of excitatory and inhibitory neurons (34 -37). Besides the well-documented role at mature synapses, a plethora of data also implicates Syns in neuronal development, from the early stages of neurite sprouting to the regulation of synapse formation and refinement (38).
Each SYN gene has been inactivated in mice (27,28,35,39 -41). Despite the absence of gross defects in brain anatomy, SYN1 and SYN2 knocked out (KO) mice exhibit spontaneous seizures, whereas SYN3 KO mice are not epileptic. Interestingly, both SYN2 and SYN3 KO mice also display cognitive impairments, suggesting that Syns could be involved in the regulation of higher brain functions (42)(43)(44)(45). Consistent with these observations, we recently showed that deletion of SYN1, SYN2 or SYN3 widely impairs social behaviors, resulting in an ASDrelated phenotype that is more pronounced in SYN2 KO mice. Interestingly, social impairments in both SYN1 and SYN2 KO mice were observed before the onset of seizures, suggesting that the behavioral impairments are not merely a consequence of the epileptic phenotype (46).
Based on these observations, we hypothesized that mutations in SYN2, similarly to mutations in its functionally related gene SYN1, can predispose to ASD or epilepsy. SYN2 was found among the five neural genes whose single nucleotide polymorphisms (SNPs) contribute to epilepsy predisposition in a screening of 279 prime candidate genes in 2717 cases of epilepsy (47). An intronic polymorphism in SYN2 was also found to be associated with idiopathic epilepsy (48).
In this study, we report three mutations in SYN2 (one nonsense and two missense) identified by direct sequencing of the entire gene that are associated with ASD. While the nonsense variant was not expressed in either cell lines or neurons, the translated proteins of the missense variants were normally targeted to nerve terminals, although they were unable to rescue the SYN2 KO phenotype in terms of SV pool dynamics and/or axon elongation. Taken together, these results support the view that rare genetic variations in SYN2 predispose to ASD and emphasize the key role of the Syn family of presynaptic proteins in the pathogenesis of the disease.

Identification of nonsense and missense SYN2 mutations in ASD patients
We sequenced the coding and splicing junction regions of SYN2 in 190 cases with ASD and 143 cases with partial epilepsy. We found a total of four novel genetic variations, including one nonsense mutation in one ASD case (p.A94fs199X), two missense mutations in ASD cases (p.Y236S and p.G464R) and one missense mutation in an epilepsy case (p.P253A) ( Fig. 1A -E). These genetic variants were not reported in the public databases (1000 Genomes, dbSNPs). The three missense mutations were located in exons 6 and 13 of SYN2. Sequencing of these exons in 90 additional ASD cases did not reveal any variations. Sequencing the entire SYN2 coding regions in 335 control individuals did not show any rare variant, suggesting that the novel genetic variants found in ASD and epilepsy are indeed mutations or at least very rare variants. Statistical analysis showed that the excess of rare missense variants is significant in the ASD, but not in the epilepsy cohort (P ¼ 0.0471 and P ¼ 0.299, respectively). All SYN2 mutation carriers with ASD were male and the mutation was transmitted by the unaffected mother (Tables 1  and 2). We are not aware of any affected sibling with ASD in SYN2 mutation carriers. Considering the positive association of SYN2 mutations with ASD, we further analyzed the impact of mutant SYN2 in vitro.
In silico predictions of the impact of SYN2 mutations Synapsins are composed of a mosaic of shared, evolutionarily conserved, domains (A, B, C, E) and of individual domains (D, F, G, H, J) (31). The nonsense mutation p.A94fs199X is caused by a deletion (del 282_288fs) that leads to an early frameshift, followed by a premature stop codon (pos.598). If translation occurred, the frameshift would cause the expression of a protein truncated at position 199, and aberrant in the sequence from amino acid 95 to amino acid 198, thus removing the majority of the structure of both the a and b isoforms of SYN2 ( Fig. 1A and D; Supplementary Material, Fig. S1A). The missense mutation Y236S involves the highly conserved central C-domain of Syn II and the G464R mutation hits the proline-rich H-domain that is specific for the Syn IIa isoform (Fig. 1A). Thus, while the A94fs199X SYN2 mutation is likely to have a dramatic impact on the translation and/or functional properties of the gene product, the effects of the missense mutations (Y236S and G464R) were hardly predictable by in silico analysis. Indeed, while the Y236 residue can show a conservative substitution with F in mammals, the G464 residue is conserved between man, rat and mouse, but not in lower vertebrates (Supplementary Material, Fig. S1B). Analysis of the missense substitutions by POLYPHEN-2 software (http://genetics.bwh.harvard.edu/pph/) predicted a significant impact on the human Syn II ortholog (hSyn II) protein function for the G464R mutation (HumDiv, 0.97 score; HumVar, 0.78 score), but not for the Y236S mutation.

Expression pattern of hSyn II mutants in HeLa cells and primary neurons
We first investigated the expression of the nonsense and the Y236S/G464R missense mutations in a human cell line (HeLa cells), as well as in murine hippocampal neurons prepared from SYN2 KO mice (Fig. 2). While transfection of HeLa cells or nucleofection of primary hippocampal neurons with the wild-type (WT) hSyn II promoted a reliable expression of the protein at the expected molecular mass, the expression of the A94fs199X-hSyn II mutant was totally undetectable in either cell system and no immunoreactive band in the expected molecular mass range of 20 kDa was seen, even under conditions of gel overloading ( Fig. 2A, left). In contrast, the expression of the missense mutants Y236S-Syn II and G464R-hSyn II in neurons nucleofected at early stages of development with ires-Tomato constructs ( Fig. 2A, right) or in HeLa cells transfected with constructs coding for fusion proteins with mCherry ( Fig. 2B) elicited expression levels that were fully comparable with those of WT-hSyn II.
Confocal imaging of Hela cells transfected with mCherrytagged variants of either WT-hSyn II, A94fs199X-hSyn II, Y236S-hSyn II or G464R-hSyn II fully confirmed these results ( Fig. 2C): WT-hSyn II and its missense variants were expressed in the cytoplasm of transfected cells, whereas A94fs199X-hSyn II was virtually absent. The practically complete lack of expression of the A94fs199X-hSyn II mutant after cDNA transfection could be attributable to poor translation efficiency and/or fast degradation of the aberrant protein. Hsyn II missense mutants are unable to rescue the SYN2 KO presynaptic phenotype We next evaluated the impact of the two ASD-linked missense mutations on presynaptic function. To this aim, we performed dynamic imaging of exo-endocytosis by assessing the effects of the expression of the WT or the missense SYN2 mutants at the level of single synaptic boutons. Based on ultrastructural investigations and immunoblotting with SV-specific markers (27,41), it has previously been reported that SYN2 KO neurons display a reduced population of SVs, consistent with the intense depression observed in response to high-frequency stimulation (27). However, no data are available on the dynamics of the exo-endocytotic cycle of SVs and on SV trafficking in these neurons. Thus, before testing the biological activity of the Syn II mutants, it was necessary to study the SYN2 KO phenotype and whether the potential impairments in SV dynamics could be rescued by expression of the WT-hSyn II. WT hippocampal neurons, SYN2 KO neurons and SYN2 KO neurons expressing the mCherry variant of hSyn II were transfected with synaptophysin-pHluorin (SypHy), a chimeric SV probe whose fluorescence is low in the acidic intravesicular environment and strongly increases when exposed to the extracellular medium during exocytosis (49)(50)(51). To evaluate the effect of SYN2 deletion and hSyn II reconstitution on SV trafficking evoked by electrical activity, stimulation protocols triggering action potentials (APs) were applied that allowed to estimate the size of either the RRP (40 APs at 20 Hz) or the RP (1600 APs at 20 Hz in the presence of bafilomycin), followed by exposure to 50 mM of the alkaline agent NH 4 Cl to reveal the total SV pool that also includes release-reluctant SVs [(52); Fig. 3A -C].
The size of the RRP and the kinetics of endocytosis were not significantly affected in SYN2 KO neurons ( Fig. 3D-F), which is consistent with a substantially preserved number of morphologically docked SVs in SYN2 KO terminals (41). In contrast, the size of the RP was markedly decreased in SYN2 KO neurons with only a slight and non-significant increase of its depletion rate ( Fig. 3G-I). Interestingly, the expression of hSyn II was able to fully rescue the SYN2 KO phenotype and returned the RP size and time constant of depletion to the levels observed in WT neurons ( Fig. 3G-I). In agreement with previous observations obtained with different methodologies, the total content of SVs, evaluated as DF MAX /F 0 was significantly impaired in SYN2 KO terminals with respect to WT terminals, and was virtually normalized upon reconstitution of WT-hSyn II (Fig. 3L).
We next used the same experimental system with the missense mutants to determine their ability to rescue the phenotype when compared with WT-hSyn II. However, before a presynaptic effect of the mutants could be studied, it was necessary to demonstrate that, albeit normally expressed (Fig. 2), the hSyn II mutants were also correctly targeted to nerve terminals to exert their physiological actions. Thus, we expressed mCherry chimeras of WT-hSyn II, Y236S-hSyn II or G464R-hSyn II in SYN2 KO hippocampal neurons that were co-transfected with SypHy and measured the mCherry fluorescence intensity at SypHy-positive puncta in which the SypHy fluorescence was dynamically regulated by activity. As shown in Figure 4A, both hSyn II mutants were correctly targeted to nerve terminals and their presynaptic levels were indistinguishable from those of WT-hSyn II.
Finally, we analyzed whether the expression of either hSyn II missense mutant was able to rescue the impairment in the RP size observed after constitutive deletion of endogenous . Hela cells were transfected with either WT-hSyn II or A94fs199X-hSyn II in a pCAGGS-ires-Tomato vector and harvested 48 h later. SYN2 KO hippocampal neurons were nucleofected at 0 DIV with pCAGSS-ires-Tomato vector encoding for either WT-hSyn II, A94fs199X-hSyn II, Y236S-hSyn II or G464R-hSyn II and lysed at 3 DIV. The expression of the A94fs199X-hSyn II isoform was undetectable in both Hela cells and neurons. (B) Expression of mCherry-tagged missense hSyn II mutants in HeLa cells. mCherry-tagged variants of either WT-hSyn II, Y236S-hSyn II or G464R-hSyn II were transiently transfected in HeLa cells, with the empty vector (2) used as control. Cells were harvested 48 h after transfection. The expression of recombinant hSyn II variants was analyzed by SDS page and immunoblotting (A and B) using a monoclonal antibody (Mab19.21) directed to the hSyn IIa/IIb sequence S 12 SFIAN located NH 2 terminal to the frameshift (72). In B, the expression of the mCherry fusion proteins was double checked using anti-Cherry antibodies. Mutant hSyn II isoforms were expressed at levels similar to those of the WT protein in both cell systems. In all immunoblots, actin staining was used as a control of equal loading. (C) Confocal images of Hela cells transfected with mCherry-tagged variants of WT-hSyn II, A94fs199X-hSyn II, Y236S-hSyn II or G464R-hSyn II and counterstained with AlexaFluor-488 phalloidin to visualize the F-actin cytoskeleton. WT-hSyn II and missense variants were expressed in the cytoplasm of transfected cells, whereas the expression of A94fs199X-hSyn II was totally absent. Scale bar, 30 mm.
Syn II as effectively as the WT-hSyn II did. As shown in Figure 4B and C, either mutant was unable to correct the SYN2 KO phenotype and the RP size remained at the same SYN2 KO level, markedly and significantly lower than the level reached after rescue with WT-hSyn II. No detectable effects were observed on the kinetics of RP mobilization that was not significantly altered in the SYN2 KO neurons (Figs 3I and 4D). The Y236S and G464R mutants were also less effective than WT-hSyn II in rescuing the total SV content of SYN2 KO terminals, with a clear-cut and significant loss-of-function for the G464R mutant (Fig. 4E).
The G464R-hSyn II mutant impairs axon outgrowth and dendritic branching in SYN2 KO hippocampal neurons at early stages of in vitro development Syn II expression peaks with synaptogenesis and remains elevated in mature neurons. However, the protein has been shown to be involved also in the early stages of neuronal development. Indeed, SYN2 KO neurons display delayed and abnormal neurite outgrowth (38). While the A94fs199X-hSyn II mutant, due to the lack of expression, implies a full lossof-function, the ability of the missense SYN2 mutants to rescue the developmental phenotype of SYN2 KO neurons is testable. To this aim, we nucleofected primary hippocampal neurons at 0 days in vitro (DIV) and followed the emergence and elongation of the axon as well as the development of the dendritic tree over the early developmental stages in vitro (Figs 5 and 6). As expected, SYN2 KO neurons displayed a significant impairment in axon outgrowth at 3 DIV. Interestingly, this impairment was totally rescued by the expression of WT-hSyn II, demonstrating that mouse hippocampal neurons are a reliable model to assess the impact of mutations in neural development. When the two ASD-linked missense hSyn II mutants Y236S and G464R were expressed in place of the WT isoform, we found that the former promoted an increase in axon elongation that was indistinguishable from that of WT neurons, while the latter was virtually ineffective in rescuing the SYN2 KO phenotype (Fig. 5).
A similar picture was observed when the development and complexity of the dendritic arborization was analyzed at 7 DIV by Sholl analysis (Fig. 6). Also in this case, the number of dendritic branchings was significantly decreased in SYN2 KO neurons over a large range of radii from the cell body, and expression of either WT-hSyn II or its Y236S mutant was able to fully rescue the phenotype, bringing the number of branches to the levels observed in WT neurons. However, the expression of G464R-hSyn II was unable to correct the KO phenotype, indicating that this mutant, albeit normally expressed, exhibits a developmental loss-of-function.

DISCUSSION
In addition to the several 'synaptic' ASD candidate genes identified in the last few years, including NLGN3, NLGN4, SHANK2/ 3 and IL1RAPL1 coding for postsynaptic proteins, as well as NRXN1 and its homolog CNTNAP2 and RIMS3/NIM3, coding for presynaptic proteins, we have recently reported that nonsense and missense mutations in SYN1 (p.Q555X; p.A51G, p.A550 T and p.T567A) are associated with epilepsy and/or ASD (24,53), adding another presynaptic candidate gene as a common basis for these related diseases. Moreover, another nonsense mutation in SYN1 (p.W356X), causing mRNA decay, was identified as the cause of epilepsy in a family with the history of syndromic epilepsy associated with behavioral disturbances, learning disabilities and one case of autism (25,54).
By screening a cohort of ASD individuals affected by ASD and/or epilepsy for SYN genes other than SYN1, we have identified several rare variants in SYN2, namely one nonsense (n ¼ 1) and two missense (n ¼ 2) mutations in ASD patients and one missense (n ¼ 1) mutation in an epileptic patient. While the association did not reach the significance level for epilepsy, the rare variants were significantly associated with ASD. Both the ASD-linked nonsense and missense mutations in SYN2 led to a loss of function being either non-expressed (frameshift/nonsense) or expressed and targeted to nerve terminals, but inactive in rescuing the SYN2 KO phenotype (missense). The identification of SYN2 as a predisposing gene for non-syndromic ASD further reinforces the notion that the SYN gene family is an important component of the 'synaptic ASD pathway'. Interestingly, all the ASD individuals with SYN2 mutations described here are males (n ¼ 3) and the mutant alleles have been transmitted by unaffected females. This observation, although on a limited number of individuals, is consistent with the recent report on the autosomal SHANK1 gene deletions that have been associated with ASD in males, but not in females (55). The occurrence of ASD only in males with SHANK1 and SYN2 would thus represent an example of autosomal sex-limited expression. The mechanism underlying this higher penetrance in males remains to be determined (56), although these observations are consistent with the 'protective factor' model, suggesting that there is a higher threshold in females in order to develop ASD (57). Autosomal sex-limited expression, in addition to rare mutations in X-linked genes (e.g. SYN1, NLGN3 and NRLGN4, IL1RAPL1, PTCHD1), may thus contribute to the increased prevalence of ASD in males when compared with females (5,18,24,58). Previous studies have suggested an association of SYN2 rs3773364 A.G polymorphism with febrile seizures in the UK, Irish and Finnish cohorts [EPIGEN Epilepsy Genetic Consortium; (47)] and in Indian patients with idiopathic epilepsy (48), but not in the Australian cohort (47) or in Malaysian epileptic patients (59). Moreover, a positive association between schizophrenia and variants of the SYN2 gene, such as SNPs and insertion/deletion polymorphisms, has been observed in Chinese subjects (60,61), in the Korean population (62) and in northern European families (63). Interestingly, these genetic studies, together with the rare variants described here, are reflected in the phenotype of SYN2 KO mice. These mice are epileptic from 2 to 3 months of age and display behavioral abnormalities that are reminiscent of deficits observed in other known animal models of intellectual disability, schizophrenia and ASD (27,34,(42)(43)(44)46,64). We have recently shown that the deletion of Syn isoforms widely affects social and repetitive behaviors, resulting in ASD-related phenotypes. In particular, SYN2 deletion causes the most severe phenotype with respect to the deletion of other Syn isoforms, with extensively impaired social behavior and memory, altered exploration of a novel environment and increased self-grooming (46).
Syn II shares with Syn I a large NH 2 -terminal domain and the property of being phosphorylated by several kinases (PKA, CaMKI/IV, MAPK/Erk and Src), but diverges in its C-terminal domain (31). Similarly to Syn I, Syn II is specifically associated with the cytoplasmic membrane of SVs and interacts with actin filaments in a phosphorylation-dependent manner, although its interactions appear stronger than those of Syn I (65). While both isoforms are thought to control the density of SVs at the nerve terminal and regulate their availability for release, they display non-overlapping roles in nerve terminal function. We recently found that, although the total nerve terminal SV complement is markedly decreased in both SYN1 and SYN2 KO mice (28,41,53), the resulting effects on synaptic transmission are isoform specific. The lack of either isoform has distinct effects on SV pool size and dynamics. While SYN1 deletion decreases both the RRP and the RP (24,36,66), deletion of SYN2 only impairs the RP, leaving the RRP virtually unaffected (41). Moreover, synaptic depression is more intense in SYN2 KO mice (27). In addition, we recently showed that Syn I and Syn II play opposite roles in regulating the release dynamics of inhibitory neurons, with Syn II facilitating the delayed asynchronous release of GABA and contributing to the extent of tonic inhibition, and Syn I enhancing the synchronous release response versus the asynchronous one (37,41).
The dynamic analysis of SV exo-endocytosis with SypHy fluorescence revealed that SYN2 deletion does not affect the size of the RRP or the kinetics of endocytosis, but specifically decreases the size of the RP and the total SV population without major effects on the kinetics of SV depletion. These observations are consistent with: (i) the decreased expression of SV proteins (but not of synapse density or pre-and postsynaptic proteins) observed in SYN2 KO brains (27); (ii) the decreased overall SV density together with a substantially preserved number of SVs docked to the active zone observed by ultrastructural studies of SYN2 KO central and peripheral synapses (41,67); (iii) the pivotal role of the Syn IIa isoform in controlling the size of the RP of SVs, being the only Syn isoform capable of rescuing synaptic depression in SYN1/2/3 triple KO mice (68); (iv) the unaffected RRP size upon expression of SynIIa in SYN1/ 2/3 triple KO central synapses under physiological Ca 2+ conditions (68). The latter finding is not in agreement with the observation that microinjection of a truncated form of recombinant SynII into SynII KO neuromuscular junctions reduced the number of docked SVs under low Ca 2+ conditions (69). However, the substantial differences in synapse type, extracellular Ca 2+ levels and type of expressed protein do not allow a direct comparison. Overall, these data underline the primary effects of Syn II in the formation and maintenance of SV pools in nerve terminals.
Before studying the impact of SYN2 mutations on SV trafficking and neural development, we checked whether the expression of the homologous hSyn II was able to efficiently rescue the synaptic and developmental phenotypes of mouse SYN2 KO neurons (27,34,38). The full rescue of SV pool size, axon outgrowth and neuropil complexity by the human ortholog confirmed the relevance of our experimental system.
Except for the A94fs-hSyn II mutant that was not translated in either HeLa cells or primary neurons, we found that the missense hSyn II mutants are expressed and targeted to nerve terminals in SYN2 KO neurons. However, both missense hSyn II mutants displayed a clear loss-of-function when challenged with the rescue of the SYN2 KO presynaptic phenotype and were both unable to correct the impaired RP size and total SV content that were virtually normalized by the expression of WT-hSyn II.
In neuronal development, Syn II plays an important role in regulating the early steps of neurite outgrowth and synapse formation, without affecting the number of synapses at the end of synaptogenesis (38). In contrast to Syn I, the Syn II-linked developmental phenotype also affects dendrite arborization, an effect that is likely to be secondary to the more marked delay in development displayed by SYN2 KO neurons. This indicates that Syn II has a more profound structural role compared with Syn I, likely attributable to the stronger interactions of the former isoform with the cytoskeleton (65). With respect to WT-hSyn II, the G464R mutation predicted to be deleterious was totally unable to rescue the developmental KO phenotype, while the Y236S mutation, predicted to have a milder impact on Syn II function, was similar to WT-hSyn II.
The SYN2 gene encodes for both hSynIIa and hSynIIb. While the A94fs199X and Y236S mutations affect both SynIIa and SynIIb isoforms, the G464R mutation affects SynIIa only. Although SynIIb was reported to have a role in synaptogenesis in neuroblastoma × glioma cells (32), we only concentrated on the hSynIIa isoform for the following reasons: (i) the hSynIIa isoform is involved in all mutations and therefore represented the only means to directly compare the effects of the three mutations; (ii) SynIIa was the only Syn isoform able to rescue the synaptic depression phenotype of SYN1/2/3 triple KO neurons, while SynIIb had a much weaker effect (68) and (iii) hSynIIa was able to fully rescue both the developmental and the synaptic SynII KO phenotype under our experimental conditions. While the hSyn II mutants described here displayed a predominant loss-of-function character in primary SYN2 KO neurons, their pathogenic role in patients in whom only one allele is mutated could result from either a haploinsufficiency with impaired Syn II expression or a dominant-negative effect towards the endogenous Syn II. The fact that the G464R mutation affecting only SynIIa, but not SynIIb, had a stronger impact on neuronal development compared with the Y236S-hSyn II mutant further indicates that SynIIa may be sufficient to perform all functions of the SYN2 gene, as previously reported (68). The stronger effect the G464R mutant could also be ascribed to the appearance of a new canonical Class II SH3 binding motif (xPxxPxR/K) by the R substitution in position 464 of SynIIa. The uncovering of a novel SH3-binding site could activate anomalous interactions with Src, PI3 K or proteins involved in SV cycling, with potential impact on the molecular interactions and adhesive properties of the mutant protein.
The strong association between epilepsy and ASD suggest that common mechanisms of synaptic dysfunction may underlie both diseases. Syn II is not essential for exocytosis, but is important for synapse maturation and remodeling. ASD manifestations begin in the second to third year of life, a period of intense refinement, remodeling and experience-dependent plasticity of synapses. This period fully overlaps with the developmental expression pattern of Syn IIa/IIb that peaks at 1 -3 months after birth (70). In the healthy brain, a balance of excitation and inhibition is essential for all functions, from physiological network activity and oscillations to cognitive processes. Based on mouse KO studies, disruption of Syn isoforms is associated with an altered excitatory/inhibitory balance with serious consequences in postsynaptic integration, network computation, excitability and activity-dependent plasticity. Interestingly, dysfunctions of GABAergic signaling are strongly associated with the behavioral deficits observed in ASD patients (71). Hence, in addition to SYN1, also SYN2 can participate in the 'synaptic autism pathway', implicating the whole Syn family of synaptic genes essential for activity-dependent changes in neuronal function in the pathogenesis of ASD.

Characterization of epilepsy and ASD phenotypes
Affected individuals gave informed consent, and the study was approved by the ethics committee of the CHUM. Clinical evaluations for the epilepsy phenotype included interview with patients and relatives, standard EEG and brain MRI. Seizure and syndrome classification were made according to the International League Against Epilepsy (1989). In the affected individual bearing the P253A mutation, the clinical phenotype consisted of complex partial seizures. In this individual, EEG did not reveal interictal epileptic activity, and brain MRI was unremarkable. Otherwise, all the affected individuals from our cohort exhibited clinical features compatible with idiopathic partial epilepsy. Standardized Assessment for Diagnosis of Autism (ADI-R and/or ADOS-G Module-3) was performed in our ASD cohort, as previously described (11). Neuropsychological assessment included the WAIS III and WISC Q5 IV in some individuals. Psychological and psychiatric assessments were performed by neuropsychologists and psychiatrists blinded to the genetic status (Table 2; see also Supplementary Material, Information).

Sequencing of SYN2 gene
The genomic organization of the human SYN2 gene was determined by aligning sequences from SYN2 mRNA (Genbank NM_133625) to the corresponding genomic sequence on chromosome 3 (Genbank NC_000003.11). Primers were designed to amplify 400-600 bp fragments from genomic DNA to screen all coding portions of the gene. Portions of the SYN2 gene were amplified by polymerase chain reaction (PCR) and analyzed by direct sequencing on an automatic sequencer (ABI3730; Applied Biosystems). Due to technical difficulties, the first exon was not sequenced. Primer sequences are available upon request.

Cloning of SYN2 and mutagenesis
Primers were designed to amplify by PCR the complete open reading frame of hSyn IIa from a human brain cDNA library (Marathon-ready; Clontech). The PCR products were subcloned into either pCAGGS-ires-Tomato or pmCherryC1 (Clontech) vector. The SYN2 mutants were generated from WT-hSyn II cDNA, using site-directed mutagenesis (Quikchange II kit; Stratagene). Primer sequences are available upon request.

Expression of hSyn II in HeLa cells
HeLa cells were cultured in Dulbecco's MEM (DMEM; Gibco) supplemented with 10% fetal bovine serum (Gibco), 1% penicillin -streptomycin (Gibco) and maintained at 378C in a 5% CO 2 humidified atmosphere. The day before transfection 3.5 × 10 5 cells were plated on 60 mm plates. The medium was replaced with OPTIMEM medium (Gibco) 1 h before the transfection. Cells were co-transfected with 1.5 mg of pCAGGS-ires-Tomato or mCherry-tagged WT or mutated hSyn II cDNA and 2.5 ml of Lipofectamine 2000 reagent (Invitrogen) in OPTIMEM medium. After 4 h of incubation with the transfection mix, the medium was replaced with fresh DMEM and the cells were incubated under standard growth conditions for 48 h and then processed for western blotting or immunocytochemistry.

Expression of hSyn II in primary hippocampal neurons
Homozygous SYN2 KO mice and C57BL/6J WT littermates were used in accordance with the guidelines established by the European Communities Council (Directive 2010/63/EU of 22 September 2010) and were approved by the Italian Ministry of Health. Hippocampal neurons were prepared from WT and SYN2 KO E17-18 mouse embryos, plated onto poly-L-lysine coated coverslips (Sigma) at 6 × 10 4 cells/coverslip and maintained in a culture medium consisting of Neurobasal (Gibco) plus B-27 (Gibco, 1:50 v/v), Glutamax (Gibco, 1% w/v), penicillin -streptomycin (Gibco, 1%) at 378C in a 5% CO 2 humidified atmosphere. Before plating, neurons (1.5-4 × 10 6 ) were nucleofected with Amaxa basal nucleofector kit for primary neurons (Lonza) and 4 mg of plasmid DNA (pCAGGS-ires-Tomato hSyn II) in the Amaxa nucleofector device according to manufacturer's protocol. Cells were incubated under standard growth conditions until 3 DIV and then processed for analysis of axon elongation or western blotting. For Sholl analysis, neurons were transfected at 4 DIV and fixed at 7 DIV (0.6 mg of cDNA and 6 ml of lipofectamine 2000 reagent; Invitrogen). For functional experiments, neurons were double transfected at 14 DIV with superecliptic SypHy and mCherry-tagged hSyn II, using 0.6 mg of cDNA and 6 ml of lipofectamine 2000 reagent (Invitrogen), diluted in Neurobasal medium without antibiotics, for each coverslip. Live imaging experiments were performed between 16 and 19 DIV.

Morphometric analysis
Hippocampal neurons were either nucleofected at 0 DIV and fixed at 3 DIV to analyze axon outgrowth, or transfected at 4 DIV and fixed at 7 DIV to analyze dendritic arborization. A retrospective b-tubulin staining was performed to discriminate between glia and neuron cells. Axons at 3 DIV, identified as the longest process of each neuron with non-tapering morphology, were measured using the ImageJ software based on b-tubulin staining. The identity of the axons was retrospectively confirmed using SMI-312 staining. Dendritic arborization at 7 DIV was analyzed by Sholl analysis, counting the number of intersections of dendrites with concentric circles centered on the cell body whose radius was increasing at regular steps of 10 mm.

Live-cell imaging assay
Optical recordings were performed at 16-19 DIV (i.e. 48 -72 h after transfection) when a proper degree of maturation of the network was reached. Transfected hippocampal neurons, grown onto poly-L-lysine-treated glass coverslips, were transferred to a stimulation chamber (volume 100 ml; Warner Instruments, Hamden, CT, USA), immersed in Tyrode Solution (140 mM NaCl, 3 mM KCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 10 mM HEPES-buffered to pH 7.4, 10 mM glucose) supplemented with glutamate receptor inhibitors (10 mM CNQX/50 mM APV) and positioned on the stage of an IX-81 motorized inverted epifluorescence microscope (Olympus). An MT20 Hg-Xe lamp (Olympus) was used as light source with 480 + 20 nm excitation, 495 nm dichroic and 525 + 50 nm emission filters to detect the SypHy signal and 560 + 40 nm excitation, 585 nm dichroic and 630 + 75 nm emission filters to detect the mCherry signal. One image of mCherry-hSyn II was acquired at the beginning of each experiment to verify the presence of the protein in the presynaptic boutons to be analyzed and used for the region-of-interest (ROI) selection. Time-lapse images were acquired at 1 Hz for 100 s with a Hamamatsu Orca-ER IEEE1394 CCD camera (Hamamatsu Photonics, Hamamatsu City, Japan) using a UplanSapo 60X1.35 NA oil-immersion objective (Olympus). Cells were maintained in a saline solution through a laminar-flow perfusion and the chosen field was stimulated after 10 s of baseline acquisition. APs were evoked by passing 1 ms current pulses, yielding fields of 10 V/cm, through platinum iridium electrodes using an AM 2100 stimulator (AM Systems, Carlsborg, WA). Images were analyzed using Excellence RT software (Olympus). Circular ROIs of 1.7 mm diameter were positioned at the center of manually selected individual puncta that were selected on the basis of: (i) a clear response to stimulation at 20 Hz (2 s) in a pretrial round and (ii) the concomitant expression of hSyn II in the red channel. Fluorescence time courses were analyzed with SigmaPlot 10.0 (Systat Software, Chicago, IL).

Statistical analysis
The excess of rare variants found in ASD (n ¼ 380) versus control (n ¼ 670) chromosomes was assessed by using the Fisher's exact test. The two-sided P-value was considered. The normal distribution of experimental data was assessed using the Shapiro -Wilk test. Data with normal distribution were analyzed by one-way analysis of variance (ANOVA) followed by the post hoc Bonferroni's multiple comparison test. Nonnormally distributed data were analyzed by ANOVA on ranks test (Kruskal-Wallis test) followed by Dunn's test. The statistical analysis was carried out using the Sigmaplot 12.0 software (Systat Software, Chicago, IL, USA). Significance level was preset to P , 0.05. Data are expressed as means + SEM throughout for the number of cells (n).