Exome sequencing in multiplex autism families suggests a major role for heterozygous truncating mutations

Article metrics


Autism is a severe neurodevelopmental disorder, the aetiology of which remains mainly unknown. Family and twin studies provide strong evidence that genetic factors have a major role in the aetiology of this disease. Recently, whole exome sequencing (WES) efforts have focused mainly on rare de novo variants in singleton families. Although these studies have provided pioneering insights, de novo variants probably explain only a small proportion of the autism risk variance. In this study, we performed exome sequencing of 10 autism multiplex families with the aim of investigating the role of rare variants that are coinherited in the affected sibs. The pool of variants selected in our study is enriched with genes involved in neuronal functions or previously reported in psychiatric disorders, as shown by Gene Ontology analysis and by browsing the Neurocarta database. Our data suggest that rare truncating heterozygous variants have a predominant role in the aetiology of autism. Using a multiple linear regression model, we found that the burden of truncating mutations correlates with a lower non-verbal intelligence quotient (NVIQ). Also, the number of truncating mutations that were transmitted to the affected sibs was significantly higher (twofold) than those not transmitted. Protein–protein interaction analysis performed with our list of mutated genes revealed that the postsynaptic YWHAZ is the most interconnected node of the network. Among the genes found disrupted in our study, there is evidence suggesting that YWHAZ and also the X-linked DRP2 may be considered as novel autism candidate genes.


Autism spectrum disorders (ASDs) represent a group of neurodevelopmental disorders characterized by impairments in reciprocal social interaction, verbal and non-verbal communication and stereotyped behavioural patterns.1 The prevalence of ASD is estimated to be around 0.6%, making it one of the most prevalent disorders in childhood.2 Family and twin studies have provided strong evidence that genetic factors have a major role in the aetiology of ASD: the sibling recurrence risk is 20-fold higher than in the general population,3 and the concordance rate is 80–90% in monozygotic twins versus 10–30% in dizygotic twins.4

Although in recent years several high-throughput genetic studies have identified novel ASD genes in idiopathic autism, the vast majority of the underlying genetic factors remain still unknown. In this regard, no mutation or copy number variant (CNV) found to be associated with autism accounts for more than 1% of cases, which makes the estimation of the number of independent risk loci extremely high.5 Nowadays, whole exome sequencing (WES) represents a powerful technology for identifying rare single-nucleotide variants (SNVs) that may help to depict the complex genetic architecture of autism and unravel some of the missing heritability. Recently, several studies performed WES in parent–child trios, focusing on de novo variants.6, 7, 8, 9, 10 These studies show that de novo mutations have an increased mutation rate in autism and correlate with increased paternal age.7, 9 However, even the high mutation rate observed (approximately one de novo single-nucleotide variant in the exome per autistic child) cannot fully explain the aetiology of the disorder, in which additive effects of several gene mutations are expected under a multihit model of inheritance.11, 12 The exome studies performed to date have focused mainly on de novo variants. Although this approach is useful, it is not sufficient to uncover the complete pool of aetiological variants, the majority of which would be expected to be in the inherited fraction. Several studies in singleton and multiplex families suggest a differential genetic landscape, with families with multiple affected individuals being more likely to bear inherited events than de novo variants.13, 14, 15 According to the multihit model, we would expect mutations in several genes to be at the basis of the disorder. Current data show that most aetiological variants would be heterozygous, whereas homozygous or compound heterozygous mutations would be much less frequent.16, 17, 18

The aim of this study was to uncover new ASD candidate genes by analysing the pool of inherited rare variants in multiplex families. We performed WES on 10 families, considering those rare variants not present in public or in-house databases that were predicted to be damaging and coinherited in the affected siblings.

Materials and methods

ASD patients and phenotype assessment

In our study, we considered 10 multiplex families with ASD: nine with two affected probands and one with three affected probands. Clinical data are summarized in Table 1, and more details are provided in the Supplementary Information.

Table 1 Description of the multiplex ASD families included in our study

CNVs analysis with CytoScan HD Array

Genome-wide DNA copy number analysis was performed in all ASD patients in our study (21 individuals), using the CytoScan HD Array (Affymetrix, Santa Clara, CA, USA). This study was carried out to rule out families with chromosome abnormalities previously reported in autism. Details of this study are summarized in the Supplementary Information.

Exome capture and sequencing

A total of 41 individuals including 21 probands and their parents were processed for exome enrichment using 3 μg of blood genomic DNA. The TruSeq DNA Sample Preparation kit (Illumina, San Diego, CA, USA) and the NimbleGen SeqCap EZ Human Exome Library v.1.0 kit (Roche, Basel, Switzerland), which targets approximately 20 000 genes corresponding to 34 Mb, were used according to the manufacturer’s instructions. The resulting exome libraries were applied to an Illumina flowcell for cluster generation. Paired-end sequencing was performed on a HiSeq2000 instrument (Illumina) using 76-base reads.

Exome data analysis

Reads were aligned to the reference genome (GRCh37) using Burrows–Wheeler analysis with the same option,19 and a BAM file was generated using SAMtools.20 Polymerase chain reaction duplicates were removed using SAMtools and custom scripts, and initial single-nucleotide polymorphisms calling was performed using a combination of SAMtools and Sidrón as described previously.21 Statistics for the callability (percentage of the exome covered by at least 10 reads with mapping quality 30 and base quality 20) of each sample are shown in Supplementary Table 1. On average, individuals had 83.5% of the target covered. Common variants, defined as those present in dbSNP135 with a minor allele frequency 1%, were filtered out.

Variants selection and validation

After the common single-nucleotide polymorphisms were filtered out, we selected all variants transmitted from parents to both affected probands, with the exception of family SJD_50, in which we selected variants transmitted to two out of three or to all three affected probands. All untransmitted variants or variants transmitted to only one affected sib were excluded from the study. In-house and WES false-positive databases were also used to filter the pool of variants.22 Our selection included all truncating mutations (nonsense and indels producing a frame shift), splice sites and missense variants. Missense variants were included in our selection only if they were predicted to be pathogenic by SIFT (Sorting Intolerant From Tolerant) or PolyPhen (Polymorphism Phenotyping).23, 24 Those variants that passed all filtering steps were validated by performing standard polymerase chain reactions and Sanger sequencing. On average, 90.1% of the selected variants were successfully validated. The workflow followed in this study is summarized in Supplementary Figure 1.

Statistical and bioinformatic analyses

We performed Gene Ontology enrichment analysis of our set of selected variants using the FatiGO software available at the Babelomics suite (http://babelomics.bioinfo.cipf.es).25 The database Neurocarta (http://gemma-doc.chibi.ubc.ca/neurocarta/) was used to identify potential gene–phenotype relationships in our gene list. The SFARI database (https://gene.sfari.org/autdb/Welcome.do) was used to obtain an annotated list of genes involved in autism: genes predisposing to autism in the context of syndromic disorder (category S), strong ASD candidate genes (category 2) and ASD genes with suggestive evidence (category 3).

The pool of genes selected in this study was investigated for networking and gene function relationships using Ingenuity Pathway Analysis (IPA) software (www.ingenuity.com). IPA computes a score based on the probability of finding a given pool of genes in a network from the Ingenuity’s Knowledge Database by chance (score=−log (Fisher exact test probability)). IPA was also used to assess ‘disease and disorder’ associations emerging from our pool of genes using the ‘Top Bio Functions’ category.

A multiple linear regression model was considered to assess the relation between the non-verbal intelligence quotient (NVIQ) and the different categories of genetic variants (truncating, non-synonymous damaging, non-synonymous benign, synonymous). The basic regression assumptions (errors are normally and independently distributed random variables) were checked to ensure that the model was appropriate. Following a conservative approach, to detect the regressors (the different variant categories) significantly related to the response NVIQ, the Bonferroni-adjusted procedure was considered. For all regressor variables, the sample Pearson’s correlation coefficient with the response variable and its significance was calculated. Furthermore, a simulation study was performed. By Boostrap resampling, 50 new simulated analyses with a sample size of 100 individuals each were computed and the corresponding multiple linear regression model was checked.

A multiple logistic regression model was used to establish the relation between variants status (the non-synonymous variants transmitted to both affected siblings in a family (TR) and those variants not transmitted from parents (NT)) and SIFT, PolyPhen scores.

For details about the multiple linear regression, model adequacy, Bonferroni correction, simulation study and multiple logistic regression see Supplementary Information.

Furthermore, the [0,1] × [0,1] square determined by the SIFT and PolyPhen scores of TR and NT variants was divided into four quadrants according to the pathogenicity thresholds of 0.05 and 0.5 used for SIFT and PolyPhen, respectively. We compared the proportion of TR and NT variants between the benign (values from 0.05 to 1 for SIFT and 0 to 0.5 for PolyPhen) and pathogenic (values from 0 to 0.05 for SIFT and 0.5 to 1 for PolyPhen) quadrant using the exact binomial test and Fisher’s exact test. The analyses were performed using the R package (http://CRAN.R-project.org).


In this study, we sequenced the exomes of 10 autism multiplex families and validated 220 rare variants transmitted from each parent to at least two affected siblings. All selected variants were filtered for common variants and using in-house databases and were predicted to be pathogenic. The pool of validated variants is listed in Supplementary Table 2. Most (83%) are missense changes, whereas variants predicted to cause protein truncation (nonsense mutations and frameshift indels) accounted for 16% of the total. Variants altering canonical splice sites or start codons were less represented, at 0.5% each (Supplementary Figure 2). All these variants were heterozygous, none of them was homozygous or compound heterozygous, and no gene was found mutated in more than one family. Truncating mutations represent remarkable events in the exome and may pinpoint potential candidate genes for a disease. Interestingly, in our study this kind of mutation represented a substantial proportion of the whole pool of variants, with 20 indels and 16 nonsense mutations (Table 2).

Table 2 Gene-disrupting rare variants shared by the affected sibs in each multiplex family

Chromosomal rearrangements and previously described fully penetrant deletions were ruled out in the affected probands, by karyotyping and through a CNV study. This analysis, followed by experimental validation, revealed six inherited CNVs present in five sib-pairs identifying the following genes: COL4A3-MFF, FHIT, MRPL36-NDUFS6, CTNND2, GRM1 and ASAH1 (Table 3). Structural variants spanning the genes FHIT and CTNND2 had previously been described in autism.14, 26

Table 3 Validated CNVs shared by the affected sibs in each multiplex family

To establish whether the pool of variants selected in our exome study was enriched in potential ASD susceptibility genes, we conducted a Gene Ontology study. Cellular component analysis showed ‘Cell junction’ to be a significantly enriched category after applying multiple testing corrections (P=0.043) (Supplementary Table 3). Several genes in this category such as CYFIP1, SCN1A, DNM2, FLNB, GABRA4, P2RX4, PJA2 and SV2B have been described in autism or are involved in neuronal functions. Several GO biological processes were over-represented in our study and some of them may be relevant for autism, such as ‘Neuronal tube development’ (P=0.044) and ‘Regulation of action potential’ (P=0.044) (Supplementary Table 4). In addition, we used IPA to analyze the ‘Top Bio function’ category of ‘Diseases and disorders’ and found that ‘Developmental disorders’ and ‘Neurological diseases’, among others, were the most significant groups related to the genes in our study (Supplementary Table 5). The Neurocarta database was used to build a map of psychiatric phenotypes previously linked to these genes (Figure 1). The majority of genes were associated with autism, but also with epilepsy, dyslexia, intellectual disability, attention-deficit hyperactivity disorder and schizophrenia.

Figure 1

Genes identified in our study that have already been associated with autism or with other psychiatric conditions according to Neurocarta (http://gemma-doc.chibi.ubc.ca/neurocarta/). ADHD, attention-deficit hyperactivity disorder; ASD, autism spectrum disorder; EP, epilepsy; SCZ, schizophrenia; DYS, dyslexia; ID, intellectual disability.

PowerPoint slide

We also investigated the possible correlation between the severity of the mutations and intellectual disability, by using the NVIQ. NVIQ was assessed in our sample of 21 affected siblings, with scores ranging from 35 to 139 and a mean of 91 (Supplementary Figure 3). A multiple linear regression model was applied with NVIQ as a response variable and truncating, non-synonymous (damaging and benign changes according to SIFT and PolyPhen) and synonymous variants as regressor variables. The results showed that only truncating variants contribute significantly to NVIQ (P=0.007). Furthermore, truncating variants were correlated with lower NVIQ scores (correlation coefficient r=–0.517, P=0.016; Figure 2) and explained 26% of NVIQ variance in our ASD sample (r2=0.267). A simulation study including 50 new simulated analyses of 100 individuals each obtained the same results as in the original sample (for more details see Supplementary Information, Supplementary Table 7 and Supplementary Figure 8).

Figure 2

Distribution of the average number of rare variants per proband for each interval of non-verbal intelligence quotient (NVIQ) in four variant categories: (1) truncating mutations, which include nonsense and frameshift mutations; (2) non-syn damaging, non-synonymous mutations predicted to be damaging by SIFT or PolyPhen; (3) non-syn benign, non-synonymous mutations predicted to be benign by SIFT and PolyPhen; and (4) Syn, rare synonymous variants. The pool of rare variants considered are those inherited by two or three affected sibs in a family. Multiple linear regression analysis of these data showed that the number of truncating variants contributes significantly to NVIQ (P=0.007). In contrast, the contribution of the other mutation types was not significant (P>0.385). The correlation coefficients between NVIQ and each mutation type with their significances are also presented.

PowerPoint slide

We subsequently investigated whether truncating mutations might have a major role in autism aetiology by comparing the number of those that were cotransmitted (Table 1) with those that were not transmitted (Supplementary Table 6). We found a significant difference, considering the total number of rare variants, between disrupting mutations that were transmitted to affected probands (16 nonsense variants and 20 frameshift indels) and those not transmitted (9 nonsense variants and 9 frameshift indels) (Fisher’s exact test, P=0.015; Supplementary Figure 4). Subsequently, we investigated the potential contribution to ASD of rare non-synonymous variants predicted to be pathogenic, by comparing the group of transmitted (TR) single-nucleotide variants with those that were not transmitted (NT). A multiple logistic regression analysis (fitted logit model) was performed to investigate whether the scores given by SIFT and PolyPhen were able to discriminate the single-nucleotide variants in relation to their TR or NT status, but the results were not significant (SIFT P=0.122; PolyPhen P=0.811). To further test the possible relation of the non-synonymous variants with the disease, we plotted all missense changes against the PolyPhen scores (X axis) and SIFT scores (Y axis) (Supplementary Figure 5) to determine whether there was any difference in the proportion of variants predicted to be pathogenic (PolyPhen scores>0.5 and SIFT score<0.05) between the TR and NT groups. No significant differences were detected (two-sided exact binomial test, P=0.47).

The list of 220 genes harbouring rare predicted pathogenic mutations (Supplementary Table 2) and genes found in the CNV study (Table 2) were considered together and analysed by IPA to identify canonical networks that may suggest interactions of potential ASD candidate genes. The best scoring network was obtained after considering only direct interactions (score=89) (Figure 3). It is noteworthy that this network included genes previously involved in autism or other psychiatric disorders such as SCN1A, PAX3, KDM5C, TSC1, NF1, CYFIP1, KCNQ3 and LAMA5. The YWHAZ gene was the most interconnected node in this network. Furthermore, when the protein–protein interaction analysis was performed with an additional 43 ASD genes from the SFARI Gene database (Categories S, 2 and 3), the score of this network improved (score=106). YWHAZ was again the main node and interacted directly with TSC1 and CYFIP1, two ASD genes found to be mutated in our study, and also with other ASD genes from the SFARI list such as UBE3A, DISC1, MET and TSC2 (Supplementary Figure 6).

Figure 3

Protein–protein interaction analysis (Ingenuity Pathway Analysis, IPA) including all genes found to be mutated in our study. Only direct interactions among proteins were considered. Proteins in grey or colour represent genes identified in our study, with red indicating a previous association with autism, green with epilepsy and blue with schizophrenia. Proteins depicted in white are those not present in our study. Upregulatory effects are represented by outward pointing arrows, downregulatory effects are represented by outward ticks, and circular arrows indicate homotypic interactions.

PowerPoint slide


Several exome sequencing reports of autism trios have been published in the past few years. These studies have enabled the identification of novel candidate genes for ASD by focusing on de novo variants. Despite these encouraging results, de novo variants represent probably <5% of autism risk variance,8 and hence inherited rare variants may account for a considerable proportion of the missing heritability in autism. Here we present findings from the exome sequencing of 10 multiplex families, in which only the inherited rare variants shared by the affected siblings in a family and predicted to be pathogenic were considered. The resulting list of about 220 identified genetic variants was assessed for GO enrichment analysis and networks of gene interactions (IPA). These approaches identified interesting categories related to developmental disorders or involving neuronal functions, and revealed plausible interactions with previously reported ASD genes. In our study, we identified a substantial number of genes already associated with autism or with other psychiatric conditions, suggesting that there is a common genetic background for psychiatric disorders.27, 28 The most interesting finding emerging from our study suggests a major role for truncating mutations in autism. We found that those probands with a higher number of heterozygous disrupting mutations are those with lower NVIQ scores. In addition, we found more truncating mutations that were transmitted and shared between the affected sibs (36 mutations) than those that were not transmitted (18 mutations). Interestingly, Iossifov et al.9 analysed exome data and described a twofold higher rate of disrupting de novo mutations in affected probands compared with unaffected siblings. Also, a very recent WES study on inherited homozygous or compound heterozygous loss-of-function mutations found a twofold enrichment in autism compared to a control group.18 Such data suggest a genetic model for autism based on the cumulative contribution of truncating alleles and other rare pathogenic variants, coupled with rare structural variants, in which the impact of common variants may be less important than previously thought.29 Most genes with truncating mutations found in our study have an unknown function and none have previously been described in autism. Among these, YWHAZ and DRP2 may be considered as strong novel ASD candidate genes. The NHLBI Exome Sequencing Project (ESP) database (http://evs.gs.washington.edu/EVS/) was used to verify the frequency of disrupting mutations in these genes in about 6500 individuals. No truncating mutations were listed in either YWHAZ or DRP2.

The YWHAZ gene, encoding a postsynaptic protein, is the most intriguing candidate in our study. The protein physically interacts with numerous ASD gene products such as TSC1, TSC2, DISC1, UBE3A and CYFIP1 (Supplementary Figure 6). YWHAZ belongs to the highly conserved 14-3-3 protein family that comprises seven isoforms (β, γ, ɛ, η, ζ, σ, θ) involved in signal transduction by binding specific pSer/pThr motifs. These proteins are involved in a wide range of processes including cell cycle, transcription, neuronal development, migration and neurite outgrowth. Although ubiquitously expressed, expression levels are highest in the brain.30, 31 The members of this family have been associated with several neurodevelopmental disorders, syndromes or psychiatric diseases. Deletions of the contiguous YWHAE and PAFAH1B1 genes are responsible for two distinct Mendelian disorders depending on the size of the deletion: isolated lissencephaly and Miller–Dieker syndrome.32 Duplications encompassing only the YWHAE gene are associated with a distinct phenotype involving autism and other behavioural symptoms.33, 34 This gene has also been strongly associated with schizophrenia and found to be downregulated together with other 14-3-3 isoforms in the prefrontal cortex of schizophrenic patients.35, 36 Other 14-3-3 members have also been associated with psychiatric phenotypes: heterozygous microdeletions encompassing YWHAG and HIP1 were associated with epilepsy, learning difficulties and intellectual disability,37 whereas YWHAH was associated with bipolar disorder.38, 39 Recently, Cheah et al.40 reported that YWHAZ knockout mice show neurobehavioural and cognitive deficiencies, and aberrant development of the hippocampus with migratory defects of pyramidal cells and granular neurons. Furthermore, neuroproteomic studies showed decreased expression of YWHAZ in the brain of schizophrenic patients.41 YWHAZ forms a molecular complex with DISC1 (Disrupted In Schizophrenia 1), Ndel1 and LIS1 to control the development of the hippocampus by coordinating neuronal migration, axonal pathfinding and synapse formation.40 In our study, the 1-bp insertion found in YWHAZ causes a frame shift leading to a premature stop codon after 18 amino acids (Supplementary Figure 7). If nonsense-mediated mRNA decay does not prevent the degradation of this transcript, it is possible that this abnormal protein may act in a dominant-negative manner altering the normal protein interactions or preventing homo- and heterodimerization of the protein with 14-3-3 members.

The X-linked dystrophin-related protein 2 gene (DRP2), expressed mainly in the brain and spinal cord, is also a good ASD candidate. DRP2 forms a complex with periaxin and dystroglycan, regulating the myelination of Schwann cells. Sherman et al.42 showed that loss of DRP2 affects the organization of the Schwann cell cytoplasm in the Cajal bands, although other members of the dystrophin family may partially supply the absence of DRP2. No human disease has been associated with the DRP2 gene, although deletions spanning DRP2 were described in individuals with X-linked agammaglobulinaemia and Mohr–Tranebjaerg syndrome caused by BTK and TIMM8A mutations, respectively. The four patients described with the microdeletion spanning BTK, TIMM8A, TAF7L and DRP2 presented the typical X-linked agammaglobulinaemia and Mohr–Tranebjaerg syndrome phenotypes, but also autism and language delay.43, 44, 45 Interestingly, a recent WES study reported the same mutation (E432*) that we found in DRP2 in one autistic family, which was absent from a large cohort of controls.18

This work represents one of the first comprehensive studies of multiplex families with ASD by exome sequencing. Similar to other studies using WES technologies, known limitations should be considered. First, the coverage of the exome fraction was not complete (83.5% of target sequence covered on average); thus, it is possible that we missed disease-causing mutations. Second, we used stringent criteria to filter false-positives, and as a consequence we may have missed true aetiological variants that were not considered as being pathogenic. Third, we considered only variants shared by affected probands, although aetiological variants transmitted to only one child are also likely to contribute to the disease. Fourth, repetitive elements or variants located on non-coding sequences were not explored in our study. Finally, the sample under study comprises only 10 families.

In conclusion, our data suggest that inherited disrupting mutations in multiplex families may have a major role in the aetiology of ASD. We highlight novel potential ASD candidate genes such as YWHAZ and DRP2. Further WES studies of inherited rare variants in larger ASD samples are warranted to corroborate these results and to gain more insight into the missing heritability of ASD.


  1. 1

    Lord C, Jones RM . Annual research review: re-thinking the classification of autism spectrum disorders. J Child Psychol Psychiatry 2012; 53: 490–509.

  2. 2

    Elsabbagh M, Divan G, Koh YJ, Kim YS, Kauchali S, Marcin C et al. Global prevalence of autism and other pervasive developmental disorders. Autism Res 2012; 5: 160–179.

  3. 3

    Ozonoff S, Young GS, Carter A, Messinger D, Yirmiya N, Zwaigenbaum L et al. Recurrence risk for autism spectrum disorders: a Baby Siblings Research Consortium study. Pediatrics 2011; 128: e488–e495.

  4. 4

    Ronald A, Hoekstra RA . Autism spectrum disorders and autistic traits: a decade of new twin studies. Am J Med Genet B 2011; 156B: 255–274.

  5. 5

    Sanders SJ, Ercan-Sencicek AG, Hus V, Luo R, Murtha MT, Moreno-De-Luca D et al. Multiple recurrent de novo CNVs, including duplications of the 7q11.23 Williams syndrome region, are strongly associated with autism. Neuron 2011; 70: 863–885.

  6. 6

    O'Roak BJ, Vives L, Fu W, Egertson JD, Stanaway IB, Phelps IG et al. Multiplex targeted sequencing identifies recurrently mutated genes in autism spectrum disorders. Science 2012; 338: 1619–1622.

  7. 7

    O'Roak BJ, Vives L, Girirajan S, Karakoc E, Krumm N, Coe BP et al. Sporadic autism exomes reveal a highly interconnected protein network of de novo mutations. Nature 2012; 485: 246–250.

  8. 8

    Neale BM, Kou Y, Liu L, Ma'ayan A, Samocha KE, Sabo A et al. Patterns and rates of exonic de novo mutations in autism spectrum disorders. Nature 2012; 485: 242–245.

  9. 9

    Iossifov I, Ronemus M, Levy D, Wang Z, Hakker I, Rosenbaum J et al. De novo gene disruptions in children on the autistic spectrum. Neuron 2012; 74: 285–299.

  10. 10

    Sanders SJ, Murtha MT, Gupta AR, Murdoch JD, Raubeson MJ, Willsey AJ et al. De novo mutations revealed by whole-exome sequencing are strongly associated with autism. Nature 2012; 485: 237–241.

  11. 11

    Schaaf CP, Sabo A, Sakai Y, Crosby J, Muzny D, Hawes A et al. Oligogenic heterozygosity in individuals with high-functioning autism spectrum disorders. Hum Mol Genet 2011; 20: 3366–3375.

  12. 12

    Leblond CS, Heinrich J, Delorme R, Proepper C, Betancur C, Huguet G et al. Genetic and functional analyses of SHANK2 mutations suggest a multiple hit model of autism spectrum disorders. PLoS Genet 2012; 8: e1002521.

  13. 13

    Bailey A, Le Couteur A, Gottesman I, Bolton P, Simonoff E, Yuzda E et al. Autism as a strongly genetic disorder: evidence from a British twin study. Psychol Med 1995; 25: 63–77.

  14. 14

    Sebat J, Lakshmi B, Malhotra D, Troge J, Lese-Martin C, Walsh T et al. Strong association of de novo copy number mutations with autism. Science 2007; 316: 445–449.

  15. 15

    Marshall CR, Noor A, Vincent JB, Lionel AC, Feuk L, Skaug J et al. Structural variation of chromosomes in autism spectrum disorder. Am J Hum Genet 2008; 82: 477–488.

  16. 16

    Chahrour MH, Yu TW, Lim ET, Ataman B, Coulter ME, Hill RS et al. Whole-exome sequencing and homozygosity analysis implicate depolarization-regulated neuronal genes in autism. PLoS Genet 2012; 8: e1002635.

  17. 17

    Puffenberger EG, Jinks RN, Wang H, Xin B, Fiorentini C, Sherman EA et al. A homozygous missense mutation in HERC2 associated with global developmental delay and autism spectrum disorder. Hum Mutat 2012; 33: 1639–1646.

  18. 18

    Lim ET, Raychaudhuri S, Sanders SJ, Stevens C, Sabo A, Macarthur DG et al. Rare complete knockouts in humans: population distribution and significant role in autism spectrum disorders. Neuron 2013; 77: 235–242.

  19. 19

    Li H, Durbin R . Fast and accurate short read alignment with Burrows–Wheeler transform. Bioinformatics 2009; 25: 1754–1760.

  20. 20

    Li H, Handsaker B, Wysoker A, Fennell T, Ruan J, Homer N et al. The Sequence Alignment/Map format and SAMtools. Bioinformatics 2009; 25: 2078–2079.

  21. 21

    Puente XS, Pinyol M, Quesada V, Conde L, Ordonez GR, Villamor N et al. Whole-genome sequencing identifies recurrent mutations in chronic lymphocytic leukaemia. Nature 2011; 475: 101–105.

  22. 22

    Fuentes Fajardo KV, Adams D, Mason CE, Sincan M, Tifft C, Toro C et al. Detecting false-positive signals in exome sequencing. Hum Mutat 2012; 33: 609–613.

  23. 23

    Kumar P, Henikoff S, Ng PC . Predicting the effects of coding non-synonymous variants on protein function using the SIFT algorithm. Nat Protoc 2009; 4: 1073–1081.

  24. 24

    Adzhubei IA, Schmidt S, Peshkin L, Ramensky VE, Gerasimova A, Bork P et al. A method and server for predicting damaging missense mutations. Nat Methods 2010; 7: 248–249.

  25. 25

    Medina I, Carbonell J, Pulido L, Madeira SC, Goetz S, Conesa A et al. Babelomics: an integrative platform for the analysis of transcriptomics, proteomics and genomic data with advanced functional profiling. Nucleic Acids Res 2010; 38: W210–W213.

  26. 26

    He WZ, Liu WQ, Zhong XQ, Chen XL, Li SY, Zhang HM et al. Analysis of de novo copy number variations in a family affected with autism spectrum disorders using high-resolution array-based comparative genomic hybridization. Zhonghua Yi Xue Yi Chuan Xue Za Zhi 2012; 29: 266–269.

  27. 27

    Guilmatre A, Dubourg C, Mosca AL, Legallic S, Goldenberg A, Drouin-Garraud V et al. Recurrent rearrangements in synaptic and neurodevelopmental genes and shared biologic pathways in schizophrenia, autism, and mental retardation. Arch Gen Psychiatry 2009; 66: 947–956.

  28. 28

    Brooks-Kayal A . Epilepsy and autism spectrum disorders: are there common developmental mechanisms? Brain Dev 2010; 32: 731–738.

  29. 29

    Anney R, Klei L, Pinto D, Almeida J, Bacchelli E, Baird G et al. Individual common variants exert weak effects on the risk for autism spectrum disorderspi. Hum Mol Genet 2012; 21: 4781–4792.

  30. 30

    Kleppe R, Martinez A, Doskeland SO, Haavik J . The 14-3-3 proteins in regulation of cellular metabolism. Semin Cell Dev Biol 2011; 22: 713–719.

  31. 31

    Foote M, Zhou Y . 14-3-3 Proteins in neurological disorders. Int J Biochem Mol Biol 2012; 3: 152–164.

  32. 32

    Cardoso C, Leventer RJ, Ward HL, Toyo-Oka K, Chung J, Gross A et al. Refinement of a 400-kb critical region allows genotypic differentiation between isolated lissencephaly, Miller–Dieker syndrome, and other phenotypes secondary to deletions of 17p13.3. Am J Hum Genet 2003; 72: 918–930.

  33. 33

    Bruno DL, Anderlid BM, Lindstrand A, van Ravenswaaij-Arts C, Ganesamoorthy D, Lundin J et al. Further molecular and clinical delineation of co-locating 17p13.3 microdeletions and microduplications that show distinctive phenotypes. J Med Genet 2010; 47: 299–311.

  34. 34

    Capra V, Mirabelli-Badenier M, Stagnaro M, Rossi A, Tassano E, Gimelli S et al. Identification of a rare 17p13.3 duplication including the BHLHA9 and YWHAE genes in a family with developmental delay and behavioural problems. BMC Med Genet 2012; 13: 93.

  35. 35

    Ikeda M, Hikita T, Taya S, Uraguchi-Asaki J, Toyo-oka K, Wynshaw-Boris A et al. Identification of YWHAE, a gene encoding 14-3-3epsilon, as a possible susceptibility gene for schizophrenia. Hum Mol Genet 2008; 17: 3212–3222.

  36. 36

    Middleton FA, Peng L, Lewis DA, Levitt P, Mirnics K . Altered expression of 14-3-3 genes in the prefrontal cortex of subjects with schizophrenia. Neuropsychopharmacology 2005; 30: 974–983.

  37. 37

    Ramocki MB, Bartnik M, Szafranski P, Kolodziejska KE, Xia Z, Bravo J et al. Recurrent distal 7q11.23 deletion including HIP1 and YWHAG identified in patients with intellectual disabilities, epilepsy, and neurobehavioral problems. Am J Hum Genet 2010; 87: 857–865.

  38. 38

    Grover D, Verma R, Goes FS, Mahon PL, Gershon ES, McMahon FJ et al. Family-based association of YWHAH in psychotic bipolar disorder. Am J Med Genet B 2009; 150B: 977–983.

  39. 39

    Pers TH, Hansen NT, Lage K, Koefoed P, Dworzynski P, Miller ML et al. Meta-analysis of heterogeneous data sources for genome-scale identification of risk genes in complex phenotypes. Genet Epidemiol 2011; 35: 318–332.

  40. 40

    Cheah PS, Ramshaw HS, Thomas PQ, Toyo-Oka K, Xu X, Martin S et al. Neurodevelopmental and neuropsychiatric behaviour defects arise from 14-3-3zeta deficiency. Mol Psychiatry 2012; 17: 451–466.

  41. 41

    English JA, Pennington K, Dunn MJ, Cotter DR . The neuroproteomics of schizophrenia. Biol Psychiatry 2011; 69: 163–172.

  42. 42

    Sherman DL, Wu LM, Grove M, Gillespie CS, Brophy PJ . Drp2 and periaxin form Cajal bands with dystroglycan but have distinct roles in Schwann cell growth. J Neurosci 2012; 32: 9419–9428.

  43. 43

    Sediva A, Smith CI, Asplund AC, Hadac J, Janda A, Zeman J et al. Contiguous X-chromosome deletion syndrome encompassing the BTK, TIMM8A, TAF7L, and DRP2 genes. J Clin Immunol 2007; 27: 640–646.

  44. 44

    Jyonouchi H, Geng L, Toruner GA, Vinekar K, Feng D, Fitzgerald-Bocarsly P . Monozygous twins with a microdeletion syndrome involving BTK, DDP1, and two other genes; evidence of intact dendritic cell development and TLR responses. Eur J Pediatr 2008; 167: 317–321.

  45. 45

    Arai T, Zhao M, Kanegane H, van Zelm MC, Futatani T, Yamada M et al. Genetic analysis of contiguous X-chromosome deletion syndrome encompassing the BTK and TIMM8A genes. J Hum Genet 2011; 56: 577–582.

Download references


We are grateful to all families for their participation in our study. We thank Patricia Romarís (Hospital Universitari Mútua de Terrassa) for contributing to clinical delineation of patients and Lara Nonell and Eulàlia Puigdecanet (Servei d'Anàlisi de Micorarrays, IMIM-Hospital del Mar, Parc de Recerca Biomèdica de Barcelona) for their contribution to the CNV studies. Exome sequencing services were provided by the National Centre for Genomic Analysis (CNAG). CT was supported by the European Union (Marie Curie, PIEF-GA-2009-254930) and BT by AGAUR (FI grant). Financial support was received from ‘Fundació La Marató de TV3’ (092010), ‘Fundación Alicia Koplowitz’, AGAUR (2009SGR00971) and ‘Ministerio de Economía y Competitividad, Spain’ (SAF2012-33484, SAF2010-21165).

Author information

Correspondence to B Cormand.

Ethics declarations

Competing interests

The authors declare no conflict of interest.

Additional information

Supplementary Information accompanies the paper on the Molecular Psychiatry website

Supplementary information

PowerPoint slides

Rights and permissions

Reprints and Permissions

About this article


  • autism spectrum disorder
  • exome sequencing
  • multiplex families
  • novel candidate genes
  • rare genetic variants
  • truncating mutations

Further reading