Browse Subject Areas

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Defining the Contribution of CNTNAP2 to Autism Susceptibility

  • Srirangan Sampath ,

    Affiliation Center for Complex Disease Genomics, McKusick-Nathans Institute of Genetics Medicine, Johns Hopkins University School of Medicine, Baltimore, Maryland, United States of America

  • Shambu Bhat,

    Affiliation Center for Complex Disease Genomics, McKusick-Nathans Institute of Genetics Medicine, Johns Hopkins University School of Medicine, Baltimore, Maryland, United States of America

  • Simone Gupta,

    Affiliation Center for Complex Disease Genomics, McKusick-Nathans Institute of Genetics Medicine, Johns Hopkins University School of Medicine, Baltimore, Maryland, United States of America

  • Ashley O’Connor,

    Affiliation Center for Complex Disease Genomics, McKusick-Nathans Institute of Genetics Medicine, Johns Hopkins University School of Medicine, Baltimore, Maryland, United States of America

  • Andrew B. West,

    Affiliation Department of Neurology and Neurobiology, Center for Neurodegeneration and Experimental Therapeutics, University of Alabama at Birmingham, Birmingham, Alabama, United States of America

  • Dan E. Arking,

    Affiliation Center for Complex Disease Genomics, McKusick-Nathans Institute of Genetics Medicine, Johns Hopkins University School of Medicine, Baltimore, Maryland, United States of America

  • Aravinda Chakravarti

    Current address: Department of Molecular and Human Genetics, Baylor College of Medicine, Houston, Texas, United States of America

    Affiliation Center for Complex Disease Genomics, McKusick-Nathans Institute of Genetics Medicine, Johns Hopkins University School of Medicine, Baltimore, Maryland, United States of America

Defining the Contribution of CNTNAP2 to Autism Susceptibility

  • Srirangan Sampath, 
  • Shambu Bhat, 
  • Simone Gupta, 
  • Ashley O’Connor, 
  • Andrew B. West, 
  • Dan E. Arking, 
  • Aravinda Chakravarti


20 Dec 2013: Sampath S, Bhat S, Gupta S, O’Connor A, West AB, et al. (2013) Correction: Defining the Contribution of CNTNAP2 to Autism Susceptibility. PLOS ONE 8(12): 10.1371/annotation/b4552fc7-285e-42e8-b126-d498eaf9f73a. View correction


Multiple lines of genetic evidence suggest a role for CNTNAP2 in autism. To assess its population impact we studied 2148 common single nucleotide polymorphisms (SNPs) using transmission disequilibrium test (TDT) across the entire ~3.3 Mb CNTNAP2 locus in 186 (408 trios) multiplex and 323 simplex families with autistic spectrum disorder (ASD). This analysis yielded two SNPs with nominal statistical significance (rs17170073, p = 2.0 x 10-4; rs2215798, p = 1.6 x 10-4) that did not survive multiple testing. In a combined analysis of all families, two highly correlated (r2 = 0.99) SNPs in intron 14 showed significant association with autism (rs2710093, p = 9.0 x 10-6; rs2253031, p = 2.5 x 10-5). To validate these findings and associations at SNPs from previous autism studies (rs7794745, rs2710102 and rs17236239) we genotyped 2051 additional families (572 multiplex and 1479 simplex). None of these variants were significantly associated with ASD after corrections for multiple testing. The analysis of Mendelian errors within each family did not indicate any segregating deletions. Nevertheless, a study of CNTNAP2 gene expression in brains of autistic patients and of normal controls, demonstrated altered expression in a subset of patients (p = 1.9 x10-5). Consequently, this study suggests that although CNTNAP2 dysregulation plays a role in some cases, its population contribution to autism susceptibility is limited.


Autism (MIM 209850) is a highly heritable, early onset childhood neuropsychiatric developmental disorder characterized by impairments in the three core domains of social interaction and communication, language development, and repetitive restricted behavior and interests. The phenotype occurs within a wide range constituting the autistic spectrum disorders (ASD) that includes Asperger’s syndrome, childhood disintegrative disorder and pervasive developmental disorder not otherwise specified; autism is the most severe and archetype of the ASD. Among children within the ages of 4-6 years, the prevalence of autism is 20 per 10,000 live births and for ASD it is 60 per 10,000 live births [1]. Males are more frequently affected with autism than females (sex ratio ~4.2:1) [2]. The substantial increase in the prevalence of ASD in the last two decades presents an enormous public health challenge with great societal and individual consequences [3,4].

The clear evidence for an increased prevalence of autism among siblings of autistic probands, and the greater disease concordance rate among monozygous than dizygous twins, shows a significant genetic influence in autism etiology with heritability estimates of 80-90% [5-7]. Nevertheless, considering that autism represents one of the most heritable neuropsychiatric disorders, progress in identifying the molecular basis of autism has been sluggish, and its pathophysiology remains unclear. Over the last decade, numerous gene discovery studies have implicated a small number of genes with either rare highly-penetrant mutations, or low-penetrant common variants or copy-number variants (CNV) that together explain no more than 15-30% of the population prevalence [8-12]. Although these studies have implicated many genes of diverse function, the underlying genetic heterogeneity ensures that no single gene accounts for the majority of autism. Despite such heterogeneity several chromosomal loci have been shown to be linked to ASD by multiple independent linkage and association studies, including those at 2q24-2q31[13-15], 7q22-7q31 [16-23], 7q34-7q36 [24,25] and 17q11-17q21 [26-30]. Recurrent structural chromosomal changes involving deletions and duplication at 7q11, 15q11-15q13, 17p11.2, 22q11.2 and 22q13 have also been associated with syndromic forms of autism [31].

CNTNAP2, or contactin associated protein-like 2, is one of the genes with the strongest evidence of autism susceptibility with convergent evidence from several independent studies. In fact, mice lacking CNTNAP2 show striking similarity to the core deficits of behavioral and cognitive functions that are seen in patients with ASD signifying its vital role in brain development [32]. This gene was first identified in autism by linkage in a Old Order Amish family with an autosomal recessive founder null mutation leading to cortical dysplasia-focal epilepsy (CDFE; MIM 610042) characterized by the presence of neuronal migration abnormalities, seizures, intellectual disability, language regression, hyperactivity, impulsive and aggressive behavior, and ASD [33]. Subsequently, several independent studies using whole genome linkage, association and CNV analysis have demonstrated the role of rare [34], common [35-37] and deletion variants [34,38,39] at CNTNAP2 as a susceptibility factor for idiopathic ASD and autism, and related language quantitative traits. In addition, haploinsufficiency for CNTNAP2, resulting from gene disruption, has been implicated in other clinically distinct behavioral disorders such as Gilles de la Tourette syndrome [40] (GTS; MIM 137580), attention-deficit/hyperactivity disorder [41] (ADHD) and, schizophrenia and epilepsy [42]. Finally, homozygous and compound-heterozygous deletions and mutations in CNTNAP2 and NRXN1, both distant members of the evolutionarily conserved synaptic proteins of the neurexin superfamily, have been discovered in patients with severe mental retardation, autistic behavior, seizures and microcephaly, a phenotype resembling Pitt-Hopkins syndrome (PHS; MIM 610954) [43].

CNTNAP2 is one of the largest mammalian genes that spans more than 3.3 Mb and maps to the common fragile site FRA7I that is associated with genomic instability in solid tumors. CNTNAP2 encodes CASPR2 with expression restricted to neurons. CASPR2 is a transmembrane scaffolding protein belonging to the neurexin superfamily that clusters voltage-gated potassium channels at the nodes of Ranvier [44]. Consequently, CNTNAP2 is a very large mutational target, and its restricted and early expression in the nervous system, potentially deregulated by mutations in autism, fit well with current understanding of autism pathobiology.

We, and others, have identified common non-coding variants in CNTNAP2 that are associated with some features of autism: rs7794745 in intron 2 is associated with strict autism [36], whereas, the intron 13 variants rs2710102 and rs17236239 are associated with the quantitative autism endophenotypes of age-at-first word [35] and specific language impairment [37]. In addition, other variants within CNTNAP2 exons 13-15 show modest association with specific language impairment [37]. Despite some degree of overlap in the families used for analyses, these genetic signals at CNTNAP2 represent statistically significant and independent associations. Consequently, we designed a study to comprehensively search for genetic associations within the entire CNTNAP2 locus that could contribute to autism and ASD susceptibility.

In this study, we perform genetic association analyses of CNTNAP2 with four major improvements. First, we perform fine-mapping of the entire 3.3 Mb CNTNAP2 locus at high marker density by genotyping 2148 common SNPs using a custom genotyping array and at high accuracy. Second, we examine a broad segment of the underlying liability by including both multiplex and simplex families, and both autism and ASD probands. Third, we use the family-based transmission disequilibrium test (TDT) to minimize population stratification. Fourth, we examine a large sample size of families (731 probands). An independent source of evidence of CNTNAP2’s role was obtained by examining its gene expression in the occipital cortex. Using a logistic expression analysis, we observed that altered CNTNAP2 expression is highly associated with autism status (p = 1.9 x 10-5). Although we have demonstrated dysregulation of CNTNAP2 gene expression, it appears that common variants in CNTNAP2 have a limited role in autism susceptibility. It remains entirely possible that CNTNAP2 harbors a diversity of coding and non-coding mutations that will be important for understanding autism and ASD pathophysiology, but its epidemiological impact is small.

Materials and Methods

Study Samples

The multiplex families used in this study came from the Autism Genetic Resource Exchange (AGRE, [45] and the National Institute of Mental Health Autism Genetics Initiative repositories (NIMH, The majority of the families self-identify themselves as being ‘white’ (75.1%); the remainders are African-American (1.5%), Asian (2.4%), Hispanic (5.3%) and of mixed ancestry (15.7%). The simplex families were derived from the Simons Foundation Autism Research Initiative (SFARI) Simplex Collection (SSC) [46]. The self-reported race/ethnicities of these samples were: white (75.2%), African-American (4.7%), Asian (3.7%), Hispanic (3.0%) and mixed ancestry (13.4%). The ADI-R and ADOS instruments were available for only 68% and 52% of the affected subjects in this study, respectively. We excluded families with known chromosomal anomalies and four families with >10% Mendelian segregation errors.

Brain tissue

Frozen brain tissue samples from the cerebral cortex (BA 19) of 39 autism cases and 44 control subjects were obtained via the Autism Tissue Program ( that includes the Harvard Brain Tissue Resource center, the NICHD Brain and Tissue Bank at the University of Maryland and the University of California San Diego.

SNP selection and genotyping

We designed a custom, targeted genotyping array (TG-array) containing 3150 SNPs that spanned the 3.3 Mb CNTNAP2 locus (hg18; 144 751 846-148 054 557), with a median spacing between SNPs of 663 bp, using Affymetrix technology (GeneChip Custom 3K SNP Kit). The SNPs were selected from the HapMap Phase II [47] data (r22, NCBI build 36) from the European (CEU), African (YRI) and Asian (CHB-JPT) samples. This consisted of tagging SNPs selected using the Haploview Tagger software [48] with settings of r2 = 0.70 and minor allele frequency (MAF) >0.05 across all populations. We augmented this list by filling in physical gaps with tag SNPs that had MAF >0.01 and r2 = 0.95. Of the 3900 SNPs we identified, 3150 met Affymetrix manufacturing criteria and were on the array (TG-array). Of the three known common variants at CNTNAP2 associated with autism and QTL phenotypes, only rs2710102 was present on the TG-array; thus, rs7794745 and rs17236239 were separately genotyped using TaqMan assays (Applied Biosystems, Foster City, CA). The genotyping assay is described in detail in Supplementary Methods section (see Information S1).

Stage II validation studies were performed by genotyping multiplex (572 families) and simplex families (1479 trios) for six SNPs (rs7794745, rs17170073, rs2710093, rs2710102, rs2215798, rs17236239) using either TaqMan or primer extension MALDI-TOF genotyping (Autoflex HT, Sequenom) assays, except for rs2710093 which was genotyped using both TaqMan and Sequenom assays. Pre-designed assays were used for genotyping all SNPs except for rs2710093 that was custom designed. Manufacturer’s protocols were followed for all assays. Custom primer sequences for rs2710093 are provided in the Supplementary Methods section (see Information S1).

Genotyping quality control analysis

The overall call rate for the TG-array panel was 99.7% based on 1749 samples. In order to evaluate the reproducibility of TG-array genotypes, measured as the proportion of mismatch to the consensus genotype for each sample across all 3150 TG-array SNPs, we performed replicate genotyping of Affymetrix control (27 replicates) and nine simplex samples (at least 7 replicates each). The mean proportion of mismatch to the consensus genotype was 8.0 x 10-4 and 3.0 x 10-4 for the Affymetrix control and nine simplex samples. Ten percent of the TaqMan assay was also performed in replicate, with a non-missing concordance rate of 100%. Samples genotyped by TG-array in stage I were also genotyped by either TaqMan or Sequenom assays in stage II, and their concordance were as follows. In the simplex collection, for SNPs rs2215798 (TaqMan), rs2710102 (Sequenom) and rs2710093 (Sequenom) the concordance rates were 99.9%, 98.9% and 99.8%, respectively, based on 938 overlapping samples. In the NIMH collection, for SNPs rs17170073 (TaqMan), rs2710102 (TaqMan) and rs2710093 (TaqMan) the concordance rates were 100%, 99.9% and 99.1%, respectively, based on 768 overlapping samples. For rs2710093, that was genotyped by both TaqMan and Sequenom in stage II, the concordance was 100%, based on 5147 overlapping samples. The genotype concordance for rs7794745 from our previous study [36] and present study was 99.8% based on 3760 overlapping samples. These results indicate that the genotype data we generated are of excellent quality in both the simplex and multiplex families.

Association analysis

Family based association analysis using the transmission disequilibrium test (TDT) was performed using PLINK [49] software (v1.07). Multiple affected sibs in multiplex families violate the independence of genotypes and can bias transmission ratios and p-values when using PLINK. We performed TDT analysis using FBAT [50] and found that this effect is very small owing to the large no of families used in our study (see Information S1). Hence, all TDT analyses were performed using PLINK. Four families with large Mendelian errors (>10%) suggestive of sample errors or non-paternity were removed before analysis. Before further analysis, genotype data were filtered to remove families and SNPs with call rates <90%, minor allele frequency <5%, Hardy-Weinberg equilibrium p-value <0.001 and Mendelian errors >5%. The genotype clusters of the SNPs with significant p-values were manually examined. In order to identify large deletions we analyzed families for the presence of clustered Mendelian errors. For a SNP with previous evidence for either autism risk (rs7794745) or its endophenotypes (rs2710102 and rs17236239), a three SNP test and one-sided p-value was used in stage II, but a genome-wide correction and two-sided p-value in the merged stage I and II analysis.

Correction for multiple testing

Given the large number of correlated SNPs tested in our study we used correlation measures based on PCA to estimate the number of independent SNPs to correct for the nominal significance threshold (α = 0.05). We used SNPSpD (see Information S1) to calculate the effective number of SNPs (Meff). Based on this method, we estimated the effective number of SNPs to be 663.12, giving an experiment-wise significance threshold required to keep type-I error rate of 5% as 7.5 x 10-5 (0.05/663.12).

Genotype imputation

To improve the polymorphism content of the CNTNAP2 locus we performed imputation of genotypes not experimentally obtained using BEAGLE (v3.3.0) [51] and MaCH (v1.0) [52]. The imputed genotypes were used to identify potential additional association signals and were not used to fill in missing genotypes of typed markers. Because our samples included admixed individuals, based on self-reports and by principal component analysis, all imputations were performed using combined reference haplotypes from multiple populations since this improves imputation accuracy [52,53]. Additionally, all reference haplotypes were generated to contain only an overlapping panel of SNPs from all populations. The alleles of 497 SNPs from the TG-array panel were flipped to the + strand to ensure consistent alleles between the genotyped and reference samples. Detailed method is described in the Supplementary Methods section (see Information S1).

Imputation quality was assessed by masking a small proportion of the SNPs prior to imputation, and comparing the genotypes at imputed versus genotyped SNPs post-imputation. Of the 909 and 1687 overlapping SNPs, we masked 2.8% (26 and 47 SNPs), 3.6% (34 and 60 SNPs), and 5.7% (51 and 96 SNPs) before imputation. The genotype concordance rate varied depending on the percentage and the identity of the masked SNPs and with maximum and minimum concordance rates of 98.29% and 97.1% respectively. In addition, BEAGLE was marginally better than MaCH at imputation with far fewer Mendelian errors despite individuals being imputed as unrelated in MaCH. Hence all association analyses were performed on genotypes imputed using the BEAGLE program.

Population stratification tests

Principal Component Analysis (PCA) was performed on those multiplex and simplex families that had prior genome-wide SNP data using one affected individual per family. PCA was primarily performed to select individuals of uniform ancestry for association analysis. In the multiplex collection, Affymetrix 500K data were available from the NIMH Center for Collaborative Genetics for 397 out of 408 stage I samples (184/186 families), and 890 out of 1089 stage II samples (501/572 families). For the simplex collection, Illumina 1M and 1MDuo data were available for 317 out of 323 stage I samples only.

HapMap Phase II populations of CEU, YRI, CHB + JPT (270 samples) were used as reference populations. Of the 474 310 SNPs common between HapMap II and NIMH genome-wide data, the following SNPs were filtered out: 395 SNPs with ambiguous chromosome assignment, 77 314 A/T and G/C polymorphisms, 25 monomorphic SNPs and 86 890 SNPs with >5% missing data. The remaining 309 686 SNPs from 270 HapMap II, 685 multiplex samples (184 stage I and 501 stage II), 317 (stage I) simplex samples were merged and further tested for stratification by PCA. All PCAs were performed using the default parameters in the software smartpca from the EIGENSOFT (v3.0) package [54]. The first two principal components explaining 11% of the total variance were plotted using R (v2.10.1). The HapMap reference populations clustered into three distinct clusters and population substructure of the stage I and stage II samples were interpreted by visual inspection of the projection of the top two principal components over the HapMap reference population. All analyses were repeated by using an LD pruned set (r2<0.50) of 99 087 genome-wide SNPs with similar results. PCA indicated that in the multiplex collection, 96% (stage I) and 72% (stage II) were of European ancestry, whereas this was 77% (stage I) in the simplex collection. The rest of the families in both multiplex and simplex collections had either CHB + JPT- or YRI-related ancestry or admixtures between them.

Real time PCR analysis

Total RNA was extracted from fresh frozen postmortem brain tissue using TRIzol reagent (Invitrogen, Carlsbad, CA) according to the manufacturer’s protocol. Complementary DNA (cDNA) was generated using the Superscript III First-Strand Synthesis kit (Invitrogen). cDNA was diluted 1:5 in 10 mM Tris (pH 8.5) and 1 μl of diluted cDNA was used per 20 μl PCR reaction. Quantitative real-time PCR was performed on a CFX96 thermocycler (BioRad, Hercules, CA) using iQ Powermix (Biorad) under conditions optimized for multiplex reactions. Primer and probe sets were optimized until efficiencies were between 95-105% in standard curves of control cDNA dilutions. The following highly efficient primer/probe sets were utilized: CNTNAP2 F5’- CCTCCTTTGCCTTCTTGGTTT G-3’; CNTNAP2 R5’- CTCCAGTTCTTCCGGGCTTG-3’; CNTNAP2 Probe 5’-FAM- AGCCTTCGTTCTCCCTCCGTGCG–BHQ1-3’; MAP2 F5’-TCA CGCACACCAGGCACTCC-3’; MAP2 R5’-TCACTCGGCACCAAGATG-3’; MAP2 Probe 5’-HEX-TCACGCACACCAGGCACTCC-BHQ1-3’; TBP F 5’-TTATGGCACTGGACTGACC-3’; TBP R 5’-CTGCTGCTGTTGCTGTTG-3’; TBP probe 5’-cy5- CTGCTGCTGCTGCTGCTGC-BHQ2-3’. BHQ; black hole quencher. For multiplex quantitative real time PCR, 0.5 μM each F and R primer and 0.2 μM each probe were used in 20 μL reactions. The efficiency of each primer/probe in the multiplex reaction was determined by running a 1:5 serially diluted control cDNA in parallel on every plate. Runs where primer/probe sets were outside of 100 ± 5% efficiency were excluded and performed again with adjusted cycling conditions until satisfactory efficiency was achieved.

For each assay, Ct was determined for experimental and control probes, and relative levels of expression were provided as Δ Ct (experimental Ct - control Ct). The average and standard deviation (SD) across at least three replicate assays was calculated, and any sample with SD ≥0.5 was inspected to determine whether one of the three assays was an outlier, defined as having a Ct difference >1.0 from the other two assays. After outlier removal, the correlation between each of the assays was determined (all r2 ≥0.98), and if necessary, adjusted so that the Y-intercept intersects 0 (we note that while correlations were extremely high, absolute Δ Ct could be off by a constant factor, which has an impact if not all data points are available from replicate assays). Final expression levels were then calculated from the average of the standardized Δ Ct as 2-ΔCt.

Graphical and Statistical analyses

All expression analyses were carried out using R (v2.10.1). Distributions of CNTNAP2 expression were graphically analyzed, as was natural log transformed values, to assess departure from normality. Association of log transformed expression levels with possible covariates (age, age2, sex, PMI, site of collection) was determined in controls using linear regression models. To determine association with autism, we used logistic regression, with expression levels as the independent variable. Bayesian information content (BIC) was used to determine the significance of adding non-linear terms.


Transmission disequilibrium test

The frequency distribution of the custom panel of 3150 SNPs indicates that the CNTNAP2 genomic interval was sufficiently covered with common variants (Figures S1A and S1B in Information S1), with 2215 SNPs having a MAF greater than 5%. Genotypes for 3150 SNPs spanning the CNTNAP2 gene were filtered prior to all family-based association analyses (see Materials and Methods section). Primarily, TDT was performed on parent-offspring trios in both multiplex and simplex families independently and also by merging data from both family types, yielding independent signals of genetic effect at the CNTNAP2 locus.

In the NIMH multiplex collection, we performed TDT on 2209 SNPs in 186 nuclear families comprising of 408 trios. Although no SNP was significant after Bonferroni correction for multiple testing, one SNP in intron 1 showed marked association with ASD (rs17170073, p = 2.4 x 10-4) (Figure 1). The rs17170073 (MAF = 0.07) T allele is over-transmitted with the transmission frequency of τ = 0.69 (Table 1). We also analyzed the parent-of-origin effect of the transmission of rs17170073 alleles, demonstrating no significant transmission disequilibrium (τpat = 0.67, τmat = 0.72, p = 0.60). rs17170073 is in a large intron that has previously been reported to contain a microdeletion in an autistic patient [35]. The association of rs17170073 in the SSC was not significant (τ = 0.49, p = 0.91) (Table 1).

Figure 1. A plot of -log10 p across CNTNAP2 from family-based association test (TDT) results on 186 multiplex families.

The physical position within the CNTNAP2 locus in megabases (Mb) and the -log10 p-values are shown on the x and y-axis, respectively. SNPs with minor allele frequency (MAF) <10% and those with MAF ≥10% are shown as red and blue filled circles, respectively. A horizontal line represents the CNTNAP2 gene, with vertical hash marks indicating exons, and locations of individual SNPs showing region-wide statistical significance in prior studies are indicated. The recombination rate in cM/Mb (right y-axis) is shown in orange. The horizontal grey dashed line represents the -log10 p-value for region-wide significance level (7.5 x 10-5).

(Minor/Major) alleleTriosA1A2τPTriosA1A2τPTriosA1A2τP
rs17170073 (T/C)0.07I40865290.692 x 10-432337380.490.91731102670.607.0 x 10-3
I + II14451861570.540.12
rs2215798 (G/A)0.15I4081051190.470.35323541010.351.6 x10-47311592200.421.7 x 10-3
I + II18024154520.480.21
rs2710093 (G/C)0.17I408851320.391.4 x 10-3323671080.381.9 x 10-37311522400.399.0 x 10-6
I + II14973584190.462.8 x 10-218024084660.474.9 x 10-232997668850.463.4 x 10-3

Table 1. Transmission disequilibrium of CNTNAP2 variants identified in this study.

Stage I samples were genotyped with TG-array and stage II samples were genotyped with either TaqMan or Sequenom assay. Trios: no of trios, A1: minor allele, A2: alternate allele, τ: transmission frequency of the minor allele, MAF: minor allele frequency, and P: p-value for τ
Download CSV

In the SSC, following quality control, 2161 SNPs were analyzed in 323 parent-offspring trios. Like the TDT results in multiplex families, no SNP was significant after Bonferroni correction for multiple testing in the simplex families. However, two SNPs in intron 13 were marginally significant (rs2215798, p = 1.6 x 10-4; rs2708244, p = 3.0 x 10-4) (Figure 2) with an excess transmission from the father compared to the mother that was not statistically significant (rs2215798, τpat = 0.70, τmat = 0.60, p = 0.17; rs2708244, τpat = 0.70, τmat = 0.58, p = 0.12). rs2215798 and rs2708244 (MAF = 0.15) were highly correlated with each other (r2 = 0.98). SNPs rs2215798 (τ = 0.47, p = 0.35) and rs2708244 (data not shown) had no significant genetic effect in the multiplex collection (Table 1).

Figure 2. A plot of -log10 p across CNTNAP2 from family-based association test (TDT) results on 323 simplex families.

TDT was also performed after merging the data from multiplex and simplex families. In the merged data (2148 SNPs in 731 affected offspring trios) two SNPs in intron 14 were significantly associated with autism (rs2710093; τ = 0.39, p = 9.0 x 10-6, rs2253031; τ = 0.39, p = 2.5 x 10-5), even after correcting for the number of SNPs tested by permutation (corrected empirical p = 0.01 and 0.03 respectively) (Figure 3 and Table 1). These two SNPs are highly correlated (r2 = 0.99) and have a data completeness of 99.83% and 99.95%, one and zero observed Mendelian errors, and are in Hardy-Weinberg equilibrium (p = 0.14 and 0.25, respectively). Both SNPs have a minor allele frequency of 0.17. TDT based on the parent-of-origin of the allele did not indicate any transmission disequilibrium at rs2710093 (τpat = 0.65, τmat = 0.57, p = 0.10). A quantile-quantile (QQ) plot of the observed p-values revealed an overall good fit with the null distribution (Figure S2 in Information S1).

Figure 3. A plot of -log10 p across CNTNAP2 from family-based association test (TDT) results on 509 merged multiplex and simplex families.

The bottom panel shows a plot of all pairwise correlations (linkage disequilibrium, D’) between 2148 SNPs in this genomic interval.

In order to validate our finding from this study (rs17170073 in the multiplex, rs2215798 in the simplex and rs2710093 in merged dataset) we performed a stage II analysis by genotyping these three SNPs using TaqMan or Sequenom assays on an independent collection of 572 multiplex (1089 parent-offspring trios) and 1479 simplex (1479 parent-offspring trios) families. In addition, SNPs that were previously associated with strict autism and its endophenotypes (rs7794745, rs2710102 and rs17236239) were also genotyped (Table 2). The marginal genetic effects at rs17170073 (multiplex) and rs2215798 (simplex), and the significant genetic effect at rs2710093 seen in the merged dataset could not be validated in stage II samples (Table 1). Although no significant association was detected by TDT analysis at rs2710102 and rs17236239 genotyped in stage II in both the multiplex and simplex collections, rs7794745 showed marginal association (based on three SNP test) in the stage II multiplex collection (Table 2). This effect in the multiplex collection is in the same direction and similar to what we had previously reported [36]. We investigated if the lack of validation could be explained by diminished power to detect association (Supplementary Methods and Table S1 in Information S1). Table S1 indicates that for SNPs with significant findings in stage I (rs17170073, rs2215798 and rs2710093), there is considerable power in the additional families genotyped from both multiplex and simplex cohorts to detect the same size effect. For example, rs2710093 with GRRAa = 1.6 (240/152, Table 1) and allele frequency of 0.17, has >90% power in both the multiplex and simplex collections at an experiment-wise significance level of α = 7.5 x 10-5. Similarly, rs17170073 and rs2215798 with GRRAa = 2.2 (65/29, Table 1) and GRRAa = 1.8 (101/54, Table 1) respectively also have significant power to detect association in both collections. The failure to find such replicated effects suggests an incorrect hypothesis or heterogeneity.

(Minor/Major) alleleTriosA1A2τPTriosA1A2τPTriosA1A2τP
rs7794745 (T/A)0.39I4081971850.520.543231461360.520.557313433210.520.3900
II10895604520.553.4 x 10-414796546190.510.172568121410710.532.8 x 10-3
I + II14977576370.541.3 x 10-318028007550.510.253299155713920.532.4 x 10-3
rs2710102 (T/C)0.48I4082172170.501.003231761410.564.9 x 10-27313933580.520.20
I + II14977117460.490.3618028668130.520.193299157715590.500.75
rs17236239 (G/A)0.34I4082001860.520.483231421240.530.277313423100.520.21
I + II14976856420.520.2418027207150.500.893299140513570.510.36

Table 2. Transmission disequilibrium of CNTNAP2 variants previously associated with ASD and autism related language traits.

Stage I samples were genotyped with TG-array and stage II samples were genotyped with either TaqMan or Sequenom assay. SNPs rs7794745 and rs17236239 were genotyped only by TaqMan assay as these two SNPs were not in the TG-array panel. One-sided p-values are provided for stage II analysis of SNPs with prior association. Abbreviations are the same as in table 1.
Download CSV

We investigated the role of population substructure in the discordance of genetic effect between stage I and stage II samples. Using PCA analyses of genome-wide data available on the multiplex (stage I and II) and simplex (stage I) families we determined that the majority of the multiplex families (176/184 of the stage I and 360/501 stage II) and simplex families (243/317 of stage I) were of CEU ancestry (Figures S3A-C in Information S1). Consequently, TDT analyses were performed only in families with CEU ancestry (Table S2 in Information S1). These results indicate that in families with uniform ancestry, the strength of the genetic effect becomes stronger for rs7794745 in both multiplex (τ = 0.55 vs 0.54) and simplex (τ = 0.54 vs 0.51) collections; for rs17170073 in the multiplex (τ = 0.60 vs 0.54) collection only; and, for rs2215798 (τ = 0.35 vs 0.48), rs2710093 (τ = 0.40 vs 0.47) and rs2710102 (τ = 0.56 vs 0.52) in the simplex collection only. Although one cannot rule out the influence of random noise in effect size estimation due to changes in the sample size and composition, these associations did not reach region-wide significance and statistical power is an issue in the simplex collection (Table S1 in Information S1) given the smaller number of families analyzed.

We also investigated the possibility that phenotypic differences between stage I and stage II families could explain the discordant results. A strict definition of diagnosis where an offspring was considered autistic if they scored positive for “autism” by both ADI-R and ADOS instruments was used. Although no SNP was significant after corrections for multiple testing, TDT analyses in strictly autistic individuals did strengthen the genetic effect at rs7794745 (τ= 0.57 vs 0.54) in the multiplex, and rs2710093 (τ = 0.37 vs 0.47) and rs2710102 (τ = 0.58 vs 0.52) in the simplex collection (Table S3 in Information S1). Since our analysis of phenotypic effects is grossly underpowered, the effect of random noise on effect size cannot be discounted (Table S1 in Information S1).

In order to evaluate association at untyped markers within this genomic interval we imputed genotypes at 809 and 3708 untyped SNPs using HapMap and 1000 Genomes as references. Following quality control (see Materials and Methods section), TDT analysis on imputed genotypes at untyped markers (455 SNPs from HapMap, 3167 SNPs from 1000 Genomes) did not reveal any additional association that was more significant than the ones we had already identified using genotyped markers, but revealed three additional SNPs with significant transmission disequilibrium (rs2710090; p = 9.5 x 10-5, rs2710091; p = 6.2 x 10-5, and rs1922888; p = 2.6 x 10-5). These three SNPs are in the same LD block and are highly correlated (r2 = 0.96, 0.96, 0.99 respectively) with rs2710093 (Figure S4 in Information S1). This is not surprising given that our dense panel of SNPs afforded complete coverage of the genomic interval for variants with MAF >5%, and hence any existing association should be identified by our TG-array panel of SNPs and imputations only identified correlated SNPs with similar patterns of transmission disequilibrium.

We also performed a stratified TDT analysis by removing families with rare variants identified using published data on CNTNAP2 gene and exome sequencing (see Supplementary Methods and Figures S5A and S5B in Information S1). Although no SNP remained significant after Bonferroni correction, one SNP reached marginal statistical significance in both the multiplex (rs10260544, p = 1.1 x 10-4, MAF = 6%, Missing = 3.7%, and no Mendelian errors) and simplex collection (rs10488072, p = 6.4 x 10-4, MAF = 7.5%, Missing = 0.3%, and no Mendelian errors).

Given the dense panel of SNPs in the custom designed TG-array, it should be possible to use raw intensity information from each SNP for copy number analysis. Our attempts in this exercise proved unsuccessful, owing primarily to the fact that the custom array was designed as a genotyping assay for allelic discrimination, frequently leading to signal saturation and significant variation in intensity between arrays. This prevented us from resolving copy number states accurately. However, we used Mendelian segregation errors to look for deletion polymorphisms in each family. Given a set of genotypes for parents and offspring it is possible to identify deletions and its parental origin by evaluating the presence of contiguous Mendelian errors. Using this method, we did not identify any deletions except in one family in whom the deletion has been previously reported [35].

CNTNAP2 expression is altered in autism-affected cortex

We next assembled a cortical sample series from 39 autism cases and 44 controls from the Autism Tissue Program and the NICHD Brain and Tissue Bank for Developmental Disorders. The series is matched for region (Brodmann area 19 of the occipital cortex), age (mean 22.8±2.9 years cases and 21.1±2.2 years controls, median age 18 and 17 years, case/control respectively), sex (74% male in cases and 75% males in controls) and post-mortem interval (20.9 ± 1.6 hours cases and 17.5±1.0 hours controls, median 21 and 18 hours cases/controls respectively). The BA19 somatosensory/visual association region was selected because the occipital cortex is highly comparable on a gross structural level between autism and control [55], reducing the potential effect of gross tissue heterogeneity on expression levels, and because defects in connectivity have been implicated in this area [56-61].

To determine whether changes in CNTNAP2 expression levels occur in the cortex in autism, we developed a quantitative RT-PCR assay with amplification conditions optimized to allow efficiencies of 1.0 ± 0.05. cDNA library integrity was functionally verified with probe sets that target the 5’ or 3’end of low-abundance mRNA transcripts. Relative expression was determined using the TBP probe (Figure 4A) and the association of log transformed expression levels (Figure 4B) with potential confounders (age, age 2, Sex, PMI, site of collection) was determined in controls using linear regression models. Only age and age2 were significant, with decreased expression associated with increasing age (Figure 4C). Residuals were calculated using β estimates derived from the controls (under the assumption that expression levels in cases are likely to be altered). Using a logistic expression analysis we observed that altered CNTNAP2 expression is highly associated with autism status (p = 1.9 x10-5), with both high and low expression observed in cases, though low levels were more prevalent (Figure 4D). We also assessed normalization to other housekeeping genes including GADPH and neuron-specific MAP2 with similar results.

Figure 4. CNTNAP2 expression in autism and control brains.

(A) CNTNAP2 expression normalized to TBP and (B) log transformed. (C) CNTNAP2 expression as a function of age in cases (red dots) and controls (black dots), and (D) CNTNAP2 expression in autism and control brains adjusted for age and age2 effects.


The existing view of the genetic architecture of autism is that numerous genes are involved, but we do not have enough evidence to distinguish between simple heterogeneity (many genes are involved but each patient harbors mutations at only one gene) and multigenic inheritance (each patient harbors mutations at multiple genes). There are two reasons for this deficiency. First, it is generally believed that different types of sequence variants impart quantitatively different genetic risks and are, therefore, enriched in different types of families: for example, multiplex families are enriched for rare variants of large effect [62,63] whereas simplex families harbor common variants of small effect [64] or de novo mutations [65-67]. Second, there are no comprehensive genome-wide studies of different susceptibility variants (rare/common, SNPs/CNVs, segregating/de novo) in a common but a large set of families to enhance gene discovery and assess the relative weight of each variant type. This study was to correct this second deficiency using CNTNAP2 as an exemplar, since existing studies already suggested that rare deleterious mutations [34], complex chromosomal alterations [34,38,39], and common variants [35-37] of minor genetic effect all predispose to ASD and overlapping [40-43,68,69] neuropsychiatric conditions.

In this study, we evaluated the broader role of common variants in 731 multiplex and simplex families by performing fine-mapping of the CNTNAP2 gene locus at high density using a custom genotyping panel. Our laboratory has reported an association at rs7794745 for ASD, with a significant over-transmission from the mother compared to the father [36]. Our current study, despite genotyping rs7794745 in a larger number of multiplex trios, failed to show a region-wide significant association, although nominal association and transmission disequilibrium that gets stronger after correcting for ancestry and strict autism diagnosis persists. Nevertheless, we demonstrate moderately significant association at rs2710093 in both multiplex and simplex families that survived multiple test corrections in the combined data. However, this genetic effect at rs2710093 could not be validated in an independent collection of 2568 trios. SNPs that were previously associated with autism endophenotypes were not associated with ASD in the current study. In addition, we performed a stratified TDT analyses by removing families based on published CNTNAP2 sequence variants suspected to be mutations. No SNP reached significance in 109 multiplex and 116 simplex families without CNTNAP2 rare variants.

One should question why despite dense common variant coverage at CNTNAP2 we did not identify any association signal that withstood both multiple-testing correction and independent validation. While power may have been an issue in the some stratified analyses, including analyses in families with strict definition of autism and in families with absence of rare variants, it is not the case with the stage II validation cohort. This is particularly intriguing given our gene expression results. It is possible that the original finding represents a false positive finding due to population substructure or diagnostic differences between stage I and stage II samples. However, we found no evidence for either in the non-replication of the original finding. It is likely that biological endophenotypes such as brain size, IQ and prenatal history or behavioral endophenotypes such as language, social interaction and face processing could modify disease association. In fact, it has been known for some time now that the concordance rates of these broader spectrum of phenotypic traits among monozygotic and dizygotic twins is higher than that for the categorical/behavioral diagnosis of “autism”, signifying a greater genetic liability to these endophenotypes [5-7]. Consequently, it is likely that inconsistent replication in independent samples could reflect failure to take various endophenotypes that are indicators of genetic heterogeneity into account [70]. This is a general problem in autism and not unique to this study. If true, this implies that autism genetic studies will be frustrating until we account for significant endophenotypes.

A simpler explanation may suffice. It is likely that CNTNAP2 is involved in autism. Numerous genetic, neurobiological, imaging and mouse model studies have amply clarified that CNTNAP2 plays a crucial role in ASD and other related neurodevelopmental disorders. This study simply shows that its overall burden in ASD cannot be large. We present evidence for the limited role of common variants at CNTNAP2 for ASD in both multiplex and simplex families. With autism being a spectrum disorder with broad range of symptoms and severities, a further detailed analysis of endophenotypes will shed light on the gene-phenotype relationships within each homogeneous subtype, with each subtype having potentially different genetic liabilities.

To further evaluate the role of CNTNAP2 in autism, we performed the first gene expression study in post-mortem brain samples of autistic individuals. In mice, CNTNAP2’s highest expression is in the cerebral cortex with ubiquitous distribution across the cortical mantle (Allen Brain Atlas). We assembled a large series of brain samples from the cortex of autism and controls. In corroboration with the general lack of detection of deletions or other CNVs in autism families, all autism brain samples demonstrated CNTNAP2 expression despite the presumptive unstable locus (FRA7I) spanning the gene. However, CNTNAP2 demonstrated both significantly low- and high-expression in autism cases compared to controls (p = 1.9 x10-5). Whether this altered expression is caused by cis-acting or trans-acting factors in the CNTNAP2 locus was not resolved in this study, nor whether low or high expression might be deleterious, advantageous, or benign, with respect to the autism phenotype. Complete loss of CNTNAP2 expression in mice causes profound neurodevelopmental problems that at least include altered sorting of potassium channel in myelinated neurons. The effects of over-expression of CNTNAP2 have not been evaluated. Our studies provide an intriguing initial insight into CNTNAP2 in that the suggested linkage of the gene to autism likely extends beyond simple haploinsufficiency.

Supporting Information

Information S1.

Combined Supplementary Methods, Supplementary Tables and Supplementary Figures file. Additional details of the methods are provided in Supplementary Methods section. Supplemental tables and figures are provided in Supplementary Tables and Supplementary Figures section.



We thank all patients and their family members who contributed to the three major public resources that were critical to this study.

(1) The SFARI simplex collection (SSC) was designed, sampled and studied by the principal investigators A. Beaudet, R. Bernier, J. Constantino, E. Cook, E. Fombonne, D. Geschwind, D. Grice, A. Klin, D. Ledbetter, C. Lord, C. Martin, D. Martin, R. Maxim, J. Miles, O. Ousley, B. Peterson, J. Piggot, C. Saulnier, M. State,W. Stone, J. Sutcliffe, C. Walsh, E. Wijsman). The DNA samples were obtained from the Simons Foundation and the phenotypic data from[sfari_collection_ v1_3]/ui:view().

(2) The phenotypic data and biomaterials from the National Institute of Mental Health (NIMH) Autism Genetics Initiative ( _data/autism/) were collected with support by NIMH grants MH52708, MH39437, MH00219 and MH00980; National Health Medical Research Council grant 0034328; and by grants from the Scottish Rite, the Spunk Fund, Inc., the Rebecca and Solomon Baker Fund, the APEX Foundation, the National Alliance for Research in Schizophrenia and Affective Disorders (NARSAD), the endowment fund of the Nancy Pritzker Laboratory (Stanford); and by gifts from the Autism Society of America, the Janet M. Grace Pervasive Developmental Disorders Fund, and families and friends of individuals with autism. The Principal Investigators and co-investigators were: Neil Risch, Richard M. Myers, Donna Spiker, Linda J. Lotspeich, Joachim Hallmayer, Helena C. Kraemer, Roland D. Ciaranello, Luigi Luca Cavalli-Sforza (Stanford University, Stanford, CA); William M. McMahon and P. Brent Petersen (University of Utah, Salt Lake City, UT). The Stanford team is indebted to the parent groups and clinician colleagues who referred families and extend their gratitude to the families with individuals with autism who were partners in this research (3). The collection of data and biomaterials from the Autism Genetic Resource and the Autism Genome Research Exchange (AGRE) Consortium ( was supported by the NIMH grant MH64547 and the Cure Autism now Foundation. The Principal Investigator was Daniel H. Geschwind (UCLA). The Co-Principal Investigators include Stanley F. Nelson and Rita Cantor (UCLA), Christa Lese Martin (U. Chicago), T. Conrad Gilliam (Columbia University); co-investigators include Maricela Alarcon (UCLA), Kenneth Lange (UCLA), Sarah J. Spence (UCLA), David H. Ledbetter (Emory University) and Hank Juo (Columbia University). Scientific oversight of the AGRE program is provided by a steering committee (Chair: Daniel H. Geschwind with members: W. Ted Brown, Maja Bucan, Joseph Buxbaum, T. Conrad Gilliam, David Greenberg, David Ledbetter, Bruce Miller, Stanley F. Nelson, Jonathan Pevsner, Carol Sprouse, Gerard Schellenberg, Rudolph Tanzi).

CNTNAP2 gene expression study was supported by a grant from the Simons Foundation (SFARI 137603 to DEA). Tissue was generously provided by the NICHD Brain and Tissue Bank for Developmental Disorders at the University of Maryland, Baltimore, MD, the Autism Tissue Program, and Harvard Brain Tissue Resource Center (supported in part by PHS grant R24 MH068855). The goal of these centers is to distribute tissue, and, therefore, cannot endorse the studies performed or the interpretation of results. We thank the contributing families for their enormous generosity and support of research in autism. We thank Jane Pickett and Ellen Xiu for coordinating efforts for tissue collection.

Author Contributions

Conceived and designed the experiments: ABW DEA AC. Performed the experiments: SS SB SG AO. Analyzed the data: SS SB. Contributed reagents/materials/analysis tools: ABW DEA AC. Wrote the manuscript: SS.


  1. 1. Chakrabarti S, Fombonne E (2005) Pervasive developmental disorders in preschool children: confirmation of high prevalence. Am J Psychiatry 162: 1133-1141. doi:10.1176/appi.ajp.162.6.1133. PubMed: 15930062.
  2. 2. Fombonne E (2009) Epidemiology of pervasive developmental disorders. Pediatr Res 65: 591-598. doi:10.1203/PDR.0b013e31819e7203. PubMed: 19218885.
  3. 3. Center for Disease Control (2012) Prevalence of autism spectrum disorders--Autism and Developmental Disabilities Monitoring Network, 14 sites, United States, 2008. MMWR Surveill Summ 61: 1-19.
  4. 4. Kogan MD, Blumberg SJ, Schieve LA, Boyle CA, Perrin JM et al. (2009) Prevalence of parent-reported diagnosis of autism spectrum disorder among children in the US, 2007. Pediatrics 124: 1395-1403. doi:10.1542/peds.2009-1522. PubMed: 19805460.
  5. 5. Bailey A, Le Couteur A, Gottesman I, Bolton P, Simonoff E et al. (1995) Autism as a strongly genetic disorder: evidence from a British twin study. Psychol Med 25: 63-77. doi:10.1017/S0033291700028099. PubMed: 7792363.
  6. 6. Folstein S, Rutter M (1977) Infantile autism: a genetic study of 21 twin pairs. J Child Psychol Psychiatry 18: 297-321. doi:10.1111/j.1469-7610.1977.tb00443.x. PubMed: 562353.
  7. 7. Steffenburg S, Gillberg C, Hellgren L, Andersson L, Gillberg IC et al. (1989) A twin study of autism in Denmark, Finland, Iceland, Norway and Sweden. J Child Psychol Psychiatry 30: 405-416. doi:10.1111/j.1469-7610.1989.tb00254.x. PubMed: 2745591.
  8. 8. Freitag CM (2007) The genetics of autistic disorders and its clinical relevance: a review of the literature. Mol Psychiatry 12: 2-22. doi:10.1038/ PubMed: 17033636.
  9. 9. Gupta AR, State MW (2007) Recent advances in the genetics of autism. Biol Psychiatry 61: 429-437. doi:10.1016/j.biopsych.2006.06.020. PubMed: 16996486.
  10. 10. O'Roak BJ, State MW (2008) Autism genetics: strategies, challenges, and opportunities. Autism Res 1: 4-17. doi:10.1002/aur.3. PubMed: 19360646.
  11. 11. Bill BR, Geschwind DH (2009) Genetic advances in autism: heterogeneity and convergence on shared pathways. Curr Opin Genet Dev 19: 271-278. doi:10.1016/j.gde.2009.04.004. PubMed: 19477629.
  12. 12. Schaaf CP, Zoghbi HY (2011) Solving the autism puzzle a few pieces at a time. Neuron 70: 806-808. doi:10.1016/j.neuron.2011.05.025. PubMed: 21658575.
  13. 13. Ramoz N, Cai G, Reichert JG, Silverman JM, Buxbaum JD (2008) An analysis of candidate autism loci on chromosome 2q24-q33: evidence for association to the STK39 gene. Am J Med Genet B Neuropsychiatr Genet 147B: 1152-1158. doi:10.1002/ajmg.b.30739. PubMed: 18348195.
  14. 14. Ramoz N, Reichert JG, Smith CJ, Silverman JM, Bespalova IN et al. (2004) Linkage and association of the mitochondrial aspartate/glutamate carrier SLC25A12 gene with autism. Am J Psychiatry 161: 662-669. doi:10.1176/appi.ajp.161.4.662. PubMed: 15056512.
  15. 15. Segurado R, Conroy J, Meally E, Fitzgerald M, Gill M et al. (2005) Confirmation of association between autism and the mitochondrial aspartate/glutamate carrier SLC25A12 gene on chromosome 2q31. Am J Psychiatry 162: 2182-2184. doi:10.1176/appi.ajp.162.11.2182. PubMed: 16263864.
  16. 16. Ashley-Koch A, Wolpert CM, Menold MM, Zaeem L, Basu S et al. (1999) Genetic studies of autistic disorder and chromosome 7. Genomics 61: 227-236. doi:10.1006/geno.1999.5968. PubMed: 10552924.
  17. 17. Barrett S, Beck JC, Bernier R, Bisson E, Braun TA et al. (1999) An autosomal genomic screen for autism. Collaborative linkage study of autism. Am J Med Genet 88: 609-615. doi:10.1002/(SICI)1096-8628(19991215)88:6. PubMed: 10581478.
  18. 18. Consortium IMGSAC (1998) A full genome screen for autism with evidence for linkage to a region on chromosome 7q. Hum Mol Genet 7: 571-578. doi:10.1093/hmg/7.3.571. PubMed: 9546821.
  19. 19. Consortium IMGSAC (2001) Further characterization of the autism susceptibility locus AUTS1 on chromosome 7q. Hum Mol Genet 10: 973-982. doi:10.1093/hmg/10.9.973. PubMed: 11392322.
  20. 20. Philippe A, Martinez M, Guilloud-Bataille M, Gillberg C, Råstam M et al. (1999) Genome-wide scan for autism susceptibility genes. Paris Autism Research International Sibpair Study. Hum Mol Genet 8: 805-812. doi:10.1093/hmg/8.5.805. PubMed: 10196369.
  21. 21. Risch N, Spiker D, Lotspeich L, Nouri N, Hinds D et al. (1999) A genomic screen of autism: evidence for a multilocus etiology. Am J Hum Genet 65: 493-507. doi:10.1086/302497. PubMed: 10417292.
  22. 22. Shao Y, Wolpert CM, Raiford KL, Menold MM, Donnelly SL et al. (2002) Genomic screen and follow-up analysis for autistic disorder. Am J Med Genet 114: 99-105. doi:10.1002/ajmg.10153. PubMed: 11840513.
  23. 23. Lamb JA, Barnby G, Bonora E, Sykes N, Bacchelli E et al. (2005) Analysis of IMGSAC autism susceptibility loci: evidence for sex limited and parent of origin specific effects. J Med Genet 42: 132-137. doi:10.1136/jmg.2004.025668. PubMed: 15689451.
  24. 24. Alarcón M, Cantor RM, Liu J, Gilliam TC, Geschwind DH (2002) Evidence for a language quantitative trait locus on chromosome 7q in multiplex autism families. Am J Hum Genet 70: 60-71. doi:10.1086/338241. PubMed: 11741194.
  25. 25. Molloy CA, Keddache M, Martin LJ (2005) Evidence for linkage on 21q and 7q in a subset of autism characterized by developmental regression. Mol Psychiatry 10: 741-746. doi:10.1038/ PubMed: 15940295.
  26. 26. Alarcón M, Yonan AL, Gilliam TC, Cantor RM, Geschwind DH (2005) Quantitative genome scan and Ordered-Subsets Analysis of autism endophenotypes support language QTLs. Mol Psychiatry 10: 747-757. doi:10.1038/ PubMed: 15824743.
  27. 27. Stone JL, Merriman B, Cantor RM, Yonan AL, Gilliam TC et al. (2004) Evidence for sex-specific risk alleles in autism spectrum disorder. Am J Hum Genet 75: 1117-1123. doi:10.1086/426034. PubMed: 15467983.
  28. 28. Yonan AL, Alarcón M, Cheng R, Magnusson PK, Spence SJ et al. (2003) A genomewide screen of 345 families for autism-susceptibility loci. Am J Hum Genet 73: 886-897. doi:10.1086/378778. PubMed: 13680528.
  29. 29. Consortium IMGSAC (2001) A genomewide screen for autism: strong evidence for linkage to chromosomes 2q, 7q, and 16p. Am J Hum Genet 69: 570-581. doi:10.1086/323264. PubMed: 11481586.
  30. 30. Cantor RM, Kono N, Duvall JA, Alvarez-Retuerto A, Stone JL et al. (2005) Replication of autism linkage: fine-mapping peak at 17q21. Am J Hum Genet 76: 1050-1056. doi:10.1086/430278. PubMed: 15877280.
  31. 31. Kumar RA, Christian SL (2009) Genetics of autism spectrum disorders. Curr Neurol Neurosci Rep 9: 188-197. doi:10.1007/s11910-009-0029-2. PubMed: 19348707.
  32. 32. Peñagarikano O, Abrahams BS, Herman EI, Winden KD, Gdalyahu A et al. (2011) Absence of CNTNAP2 leads to epilepsy, neuronal migration abnormalities, and core autism-related deficits. Cell 147: 235-246. doi:10.1016/j.cell.2011.08.040. PubMed: 21962519.
  33. 33. Strauss KA, Puffenberger EG, Huentelman MJ, Gottlieb S, Dobrin SE et al. (2006) Recessive symptomatic focal epilepsy and mutant contactin-associated protein-like 2. N Engl J Med 354: 1370-1377. doi:10.1056/NEJMoa052773. PubMed: 16571880.
  34. 34. Bakkaloglu B, O'Roak BJ, Louvi A, Gupta AR, Abelson JF et al. (2008) Molecular cytogenetic analysis and resequencing of contactin associated protein-like 2 in autism spectrum disorders. Am J Hum Genet 82: 165-173. doi:10.1016/j.ajhg.2007.09.017. PubMed: 18179895.
  35. 35. Alarcón M, Abrahams BS, Stone JL, Duvall JA, Perederiy JV et al. (2008) Linkage, association, and gene-expression analyses identify CNTNAP2 as an autism-susceptibility gene. Am J Hum Genet 82: 150-159. doi:10.1016/j.ajhg.2007.09.005. PubMed: 18179893.
  36. 36. Arking DE, Cutler DJ, Brune CW, Teslovich TM, West K et al. (2008) A common genetic variant in the neurexin superfamily member CNTNAP2 increases familial risk of autism. Am J Hum Genet 82: 160-164. doi:10.1016/j.ajhg.2007.09.015. PubMed: 18179894.
  37. 37. Vernes SC, Newbury DF, Abrahams BS, Winchester L, Nicod J et al. (2008) A functional genetic link between distinct developmental language disorders. N Engl J Med 359: 2337-2345. doi:10.1056/NEJMoa0802828. PubMed: 18987363.
  38. 38. Rossi E, Verri AP, Patricelli MG, Destefani V, Ricca I et al. (2008) A 12Mb deletion at 7q33-q35 associated with autism spectrum disorders and primary amenorrhea. Eur J Med Genet 51: 631-638. doi:10.1016/j.ejmg.2008.06.010. PubMed: 18675947.
  39. 39. Poot M, Beyer V, Schwaab I, Damatova N, Van't Slot R et al. (2010) Disruption of CNTNAP2 and additional structural genome changes in a boy with speech delay and autism spectrum disorder. Neurogenetics 11: 81-89. doi:10.1007/s10048-009-0205-1. PubMed: 19582487.
  40. 40. Verkerk AJ, Mathews CA, Joosse M, Eussen BH, Heutink P et al. (2003) CNTNAP2 is disrupted in a family with Gilles de la Tourette syndrome and obsessive compulsive disorder. Genomics 82: 1-9. doi:10.1016/S0888-7543(03)00097-1. PubMed: 12809671.
  41. 41. Elia J, Gai X, Xie HM, Perin JC, Geiger E et al. (2010) Rare structural variants found in attention-deficit hyperactivity disorder are preferentially associated with neurodevelopmental genes. Mol Psychiatry 15: 637-646. doi:10.1038/mp.2009.57. PubMed: 19546859.
  42. 42. Friedman JI, Vrijenhoek T, Markx S, Janssen IM, van der Vliet WA et al. (2008) CNTNAP2 gene dosage variation is associated with schizophrenia and epilepsy. Mol Psychiatry 13: 261-266. doi:10.1038/ PubMed: 17646849.
  43. 43. Zweier C, de Jong EK, Zweier M, Orrico A, Ousager LB et al. (2009) CNTNAP2 and NRXN1 are mutated in autosomal-recessive Pitt-Hopkins-like mental retardation and determine the level of a common synaptic protein in Drosophila. Am J Hum Genet 85: 655-666. doi:10.1016/j.ajhg.2009.10.004. PubMed: 19896112.
  44. 44. Poliak S, Salomon D, Elhanany H, Sabanay H, Kiernan B et al. (2003) Juxtaparanodal clustering of Shaker-like K+ channels in myelinated axons depends on Caspr2 and TAG-1. J Cell Biol 162: 1149-1160. doi:10.1083/jcb.200305018. PubMed: 12963709.
  45. 45. Geschwind DH, Sowinski J, Lord C, Iversen P, Shestack J et al. (2001) The autism genetic resource exchange: a resource for the study of autism and related neuropsychiatric conditions. Am J Hum Genet 69: 463-466. doi:10.1086/321292. PubMed: 11452364.
  46. 46. Fischbach GD, Lord C (2010) The Simons Simplex Collection: a resource for identification of autism genetic risk factors. Neuron 68: 192-195. doi:10.1016/j.neuron.2010.10.006. PubMed: 20955926.
  47. 47. the International HapMap Consortium (2005) A haplotype map of the human genome. Nature 437: 1299-1320. doi:10.1038/nature04226. PubMed: 16255080.
  48. 48. de Bakker PI, Yelensky R, Pe'er I, Gabriel SB, Daly MJ et al. (2005) Efficiency and power in genetic association studies. Nat Genet 37: 1217-1223. doi:10.1038/ng1669. PubMed: 16244653.
  49. 49. Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MA et al. (2007) PLINK: a tool set for whole-genome association and population-based linkage analyses. Am J Hum Genet 81: 559-575. doi:10.1086/519795. PubMed: 17701901.
  50. 50. Laird NM, Horvath S, Xu X (2000) Implementing a unified approach to family-based tests of association. Genet Epidemiol 19 Suppl 1: S36-S42. doi:10.1002/1098-2272(2000)19:1+. PubMed: 11055368.
  51. 51. Browning BL, Browning SR (2009) A unified approach to genotype imputation and haplotype-phase inference for large data sets of trios and unrelated individuals. Am J Hum Genet 84: 210-223. doi:10.1016/j.ajhg.2009.01.005. PubMed: 19200528.
  52. 52. Li Y, Willer CJ, Ding J, Scheet P, Abecasis GR (2010) MaCH: using sequence and genotype data to estimate haplotypes and unobserved genotypes. Genet Epidemiol 34: 816-834. doi:10.1002/gepi.20533. PubMed: 21058334.
  53. 53. Huang L, Li Y, Singleton AB, Hardy JA, Abecasis G et al. (2009) Genotype-imputation accuracy across worldwide human populations. Am J Hum Genet 84: 235-250. doi:10.1016/j.ajhg.2009.01.013. PubMed: 19215730.
  54. 54. Patterson N, Price AL, Reich D (2006) Population structure and eigenanalysis. PLOS Genet 2: e190. doi:10.1371/journal.pgen.0020190. PubMed: 17194218.
  55. 55. Schumann CM, Bloss CS, Barnes CC, Wideman GM, Carper RA et al. (2010) Longitudinal magnetic resonance imaging study of cortical development through early childhood in autism. J Neurosci 30: 4419-4427. doi:10.1523/JNEUROSCI.5714-09.2010. PubMed: 20335478.
  56. 56. Brun CC, Nicolson R, Leporé N, Chou YY, Vidal CN et al. (2009) Mapping brain abnormalities in boys with autism. Hum Brain Mapp 30: 3887-3900. doi:10.1002/hbm.20814. PubMed: 19554561.
  57. 57. Damarla SR, Keller TA, Kana RK, Cherkassky VL, Williams DL et al. (2010) Cortical underconnectivity coupled with preserved visuospatial cognition in autism: Evidence from an fMRI study of an embedded figures task. Autism Res 3: 273-279. doi:10.1002/aur.153. PubMed: 20740492.
  58. 58. Léveillé C, Barbeau EB, Bolduc C, Limoges E, Berthiaume C et al. (2010) Enhanced connectivity between visual cortex and other regions of the brain in autism: a REM sleep EEG coherence study. Autism Res 3: 280-285. doi:10.1002/aur.155. PubMed: 20717953.
  59. 59. Noonan SK, Haist F, Müller RA (2009) Aberrant functional connectivity in autism: evidence from low-frequency BOLD signal fluctuations. Brain Res 1262: 48-63. doi:10.1016/j.brainres.2008.12.076. PubMed: 19401185.
  60. 60. Radua J, Via E, Catani M, Mataix-Cols D (2011) Voxel-based meta-analysis of regional white-matter volume differences in autism spectrum disorder versus healthy controls. Psychol Med 41: 1539-1550. doi:10.1017/S0033291710002187. PubMed: 21078227.
  61. 61. Stroganova TA, Orekhova EV, Prokofyev AO, Tsetlin MM, Gratchev VV et al. (2012) High-frequency oscillatory response to illusory contour in typically developing boys and boys with autism spectrum disorders. Cortex 48: 701-717. doi:10.1016/j.cortex.2011.02.016. PubMed: 21458787.
  62. 62. O'Roak BJ, Deriziotis P, Lee C, Vives L, Schwartz JJ et al. (2011) Exome sequencing in sporadic autism spectrum disorders identifies severe de novo mutations. Nat Genet 43: 585-589. doi:10.1038/ng.835. PubMed: 21572417.
  63. 63. Sanders SJ, Ercan-Sencicek AG, Hus V, Luo R, Murtha MT et al. (2011) Multiple recurrent de novo CNVs, including duplications of the 7q11.23 Williams syndrome region, are strongly associated with autism. Neuron 70: 863-885. doi:10.1016/j.neuron.2011.05.002. PubMed: 21658581.
  64. 64. Weiss LA, Arking DE, Daly MJ, Chakravarti A (2009) A genome-wide linkage and association scan reveals novel loci for autism. Nature 461: 802-808. doi:10.1038/nature08490. PubMed: 19812673.
  65. 65. Iossifov I, Ronemus M, Levy D, Wang Z, Hakker I et al. (2012) De novo gene disruptions in children on the autistic spectrum. Neuron 74: 285-299. doi:10.1016/j.neuron.2012.04.009. PubMed: 22542183.
  66. 66. O'Roak BJ, Vives L, Girirajan S, Karakoc E, Krumm N et al. (2012) Sporadic autism exomes reveal a highly interconnected protein network of de novo mutations. Nature 485: 246-250. doi:10.1038/nature10989. PubMed: 22495309.
  67. 67. Sanders SJ, Murtha MT, Gupta AR, Murdoch JD, Raubeson MJ et al. (2012) De novo mutations revealed by whole-exome sequencing are strongly associated with autism. Nature 485: 237-241. doi:10.1038/nature10945. PubMed: 22495306.
  68. 68. Consortium IS; International Schizophrenia Consortium (2008) Rare chromosomal deletions and duplications increase risk of schizophrenia. Nature 455: 237-241. doi:10.1038/nature07239. PubMed: 18668038. PubMed: 18668038.
  69. 69. Petrin AL, Giacheti CM, Maximino LP, Abramides DV, Zanchetta S et al. (2010) Identification of a microdeletion at the 7q33-q35 disrupting the CNTNAP2 gene in a Brazilian stuttering case. Am J Med Genet A 152A: 3164-3172. doi:10.1002/ajmg.a.33749. PubMed: 21108403.
  70. 70. Viding E, Blakemore SJ (2007) Endophenotype approach to developmental psychopathology: implications for autism research. Behav Genet 37: 51-60. doi:10.1007/s10519-006-9105-4. PubMed: 16988798.