Autism spectrum disorders (ASD) and epilepsy frequently occur together. Approximately 5% to 40% of individuals with autism have epilepsy 1 , whereas the prevalence of epilepsy in the general population is about 0.5% to 1%. Conversely, 5% to 30% of individuals with epilepsy have ASD 2 3 4 5 , whereas the prevalence of ASD in the general population is about 1/150; although recent studies reached even higher rates: 1/80 6 . The variability of prevalence rates have been attributed to the heterogeneity of samples with respect to age, gender, comorbidity, subtype of ASD, intellectual disability and risk factors 7 8 . It may also result from the criteria used to diagnose epilepsy. Finally, the genetic background of autism may have a considerable impact on epilepsy rates in autism.

Recently, a meta-analysis of 24 reports on autism and epilepsy confirmed intellectual disability and female gender as major risk factors of epilepsy in autism 9 . The pooled prevalence of epilepsy was 21.4% in individuals with autism and intellectual disability versus 8% in individuals with autism without intellectual disability, and the more severe the intellectual disability, the more prevalent the epilepsy. With a male:female ratio of 2, the risk of epilepsy was found significantly higher for females 9 . However, this might reflect the known circumstance that females with autism tend to be more severely mentally disabled 10 .

Autism and epilepsy are both complex and heterogeneous disorders. Recent literature reviews underline the important role of genetic factors in the etiology of ASD 11 12 13 . Rare chromosomal abnormalities (for example maternally inherited duplications of chromosome 15q11–q13) and single gene syndromes (for example fragile X, tuberous sclerosis) are important risk factors of ASD. More recently, individual genes of major effect have been identified by resequencing or array-based methods, but variants at these and other loci are present in no more than 1% to 2% of children with ASD 14 . Some of these genes with major effects have been associated with ASD, intellectual disability and/or seizures, suggesting that the three disorders may share common genetic risk factors (Table 1). Recent whole genome DNA microarray studies have identified submicroscopic duplications and deletions (copy number variations (CNVs) located in many loci and mostly including de novo events) in the same developmental disorders (ASD, intellectual disability, epilepsy), and in schizophrenia as well 15 16 17 (Table 2).

Adapted from Betancur 15 and Amiet 65 . ASD, autism spectrum disorders; E, epilepsy; GABA, γ-aminobutyric acid; GLUT1, glucose transporter type 1; ID, intellectual disability; nACh, nicotinic acetylcholine; RasGAP, Ras GTPase activating protein; SCZ, schizophrenia.

Adapted from Betancur 15 , Johnson and Shorvon 17 , and Levinson et al. 120 . ASD, autism spectrum disorders; CNV, copy number variation; E, epilepsy; ID, intellectual disability; SCZ, schizophrenia.

However, all together, an identified genetic etiology is estimated to account for 10% to 20% of ASD cases 15 . Therefore, in the majority of epilepsy cases, most autism appears to be a multifactorial hereditary disorder, depending on polygenic heredity and environmental factors Environmental factors associated with autism are mainly pre-, peri- and postnatal, such as in utero exposure to teratogenic medications (for example thalidomide, valproate) 121 , prematurity, neonatal encephalopathy and hyperbilirubinemia 122 .

Most of the samples studied in autism are composed of autistic individuals without information about whether autism affects only one individual in the family (simplex autism) or multiple siblings in the family (multiplex autism). This is a major issue since higher genetic risk may be hypothesized in multiplex autism. Some recent studies have suggested disparate clinical and genetic settings depending on simplex or multiplex autism. More de novo rare mutations in genes of major effect have been associated with simplex autism 108 , while an interactive effect of multiple common susceptibility alleles have long been hypothesized as risk factors of familial autistic syndromes 123 124 . Whether some clinical variables may help to delineate more homogeneous subgroups in multiplex autism remains an open question. Regarding epilepsy and ASD, only one study specified that part of the sample included individuals with multiplex autism 125 . However, there was no separate analysis of these two groups except to observe that a similar incidence of electroencephalogram (EEG) abnormalities was found. Banach et al. examined gender differences and intellectual quotients (IQs), and found no difference in non-verbal IQ between females and males in multiplex families, while females had lower non-verbal IQ scores than males in simplex families 126 . However, the prevalence of major effect genetic risk factors as a function of epilepsy, intellectual disability and gender in multiplex families remains unknown.

The aims of the current study are: 1) to assess the prevalence of seizures in multiplex autism and compare it to the prevalence of siblings without ASD. Since autism and epilepsy share common genetic risk factors, we hypothesize a higher risk of epilepsy in individuals with autism than siblings without ASD; 2) to compare the prevalence of major effect genetic risk factors in multiplex autism with and without epilepsy. We hypothesize the group with epilepsy to have an increased risk; 3) to assess whether autism and epilepsy co-segregate within multiplex autism families; and 4) to assess whether higher prevalence of epilepsy in multiplex autism is mediated by intellectual disability and gender as it has been shown in the literature 9 .

Primary analyses were performed on 664 children from 290 families who had informative data about presence or absence of epilepsy: 417 (62.8%) children with autistic disorder, 61 (9.2%) children with PDD-NOS and 186 (28%) children without autism (Table 3). Secondary analyses were performed on 586 children from 257 families after exclusion of children with strong risk factors for autism and seizures.

Performed on 664 children and 290 families. ASD, autism spectrum disorders; PDD-NOS, pervasive developmental disorder-not otherwise specified.

Epilepsy in autism has been a subject of increasing interest, and reported prevalence rates of seizures in this condition vary from 5% to 40% 1 . The variability of rates have been attributed to the heterogeneity of samples with respect to age, gender, comorbidity, subtype of pervasive developmental disorder (PDD), intellectual disability and risk factors 9 . But most of the studies have been conducted in samples with simplex autism; for example the prevalence found in the Simons Simplex Collection (a large multisite phenotypic and DNA collection that includes families (n = 2,644) in which there is only one child (aged 4 to 17 years) with ASD) is 2.4% 135 . In multiplex families, we found that 13.4% of children with autism and 8.2% of children with PDD-NOS had epilepsy. These results are in the range of prevalence rates described in the literature. In these multiplex families, we found that 2.2% of the siblings without autism had epilepsy. This prevalence rate is high compared to the general population where 0.7% to 0.8% of children up to the age of 15 years have repeated seizures 136 . Moreover, the permutation tests performed in this study show significant evidence that epilepsy phenotypes co-segregated within families.

Studies suggest that there are two peaks of onset of seizures in autism, one in early childhood (before the age of 5 years old), and one near adolescence 137 . In the sample studied here, the admixture analysis confirms this two-peak model: 62% of the children had their first seizures before the age of 5 years. Mean age of onset for the second peak is in pre/early adolescence as only 9% of the children had their first seizure after the age of 12 years. However, with a mean age of 9.2 years, the children studied are relatively young and many of them had not yet reached adolescence. Therefore, it is likely that our estimate of the age of onset for the later peak may be spuriously low as we cannot rule out that epilepsy could begin later for some individuals. Unfortunately, conducting a subanalysis with individuals aged 16 years and older was not possible given the sample size.

The risk of epilepsy in individuals with simplex autism was significantly associated with intellectual disability and gender. In a meta-analysis of 17 studies, the pooled prevalence of epilepsy was 21.5% in individuals with autism and intellectual disability, against 8% in individuals with autism and normal intelligence 9 . Of note, the increased risk with gender in simplex autism may reflect the known circumstance that females with autism tend to have more severe intellectual disability: the more severe the intellectual disability, the lower the male:female ratio 10 . In the present study, the respective prevalence rates of epilepsy was 41.2% in individuals with autism and intellectual disability, against 11.1% in individuals with autism and normal intelligence (P = 0.005), and the more the adaptive behavior level was impaired, the higher the prevalence of epilepsy (Figure 3). However, there was no increase of epilepsy linked to gender in individuals with ASD. Furthermore, no significant difference in IQ among males and females was found in the AGRE sample. A similar result was suggested in previous reports of multiplex families 126 138 .

Identification or strong suspicion of a risk factor for autism (genetic or not) tended to increase the prevalence of epilepsy. The non-genetic risk factors identified were mainly perinatal fetal distress, the implication of which for autism, epilepsy and intellectual disability is well known. The prevalence of epilepsy remained significantly higher even when only identified or suspected genetic risk factors were considered. Genetic risk factors may be more often implicated when autism and epilepsy are associated 139 . This is also the case in multiplex families.

Interpretation of the informative families presented in Table 5 requires consideration of the AGRE database limitations, which include: excluding individuals with major perinatal risk factors or with known genetic abnormalities (for example fragile X syndrome) (but some siblings with such risk factors may have been included); and in many individuals, data are unknown or missing (for example not all individuals underwent CNV search; see Methods). Also, large family histories including epilepsy and ASD statuses of all members were not available. Careful scrutiny of Table 5 shows that several combinations of risk can lead to a multiplex pedigree: 1) the existence of a common risk factor for epilepsy and autism leading to the suggestion of screening for known genetic risk factors of epilepsy in AGRE families AU25, AU51, AU176, AU275, AU461, AU548 and AU922; and 2) the co-occurrence of two different risk factors (genetic and non-genetic ones) leading to a false ‘multiplex pedigree’, such as in families AU731, AU123, AU765 and AU1208; and 3) combinations of major and minor risk factors as it cannot be excluded that a) in some families interpretation of genotype/phenotype relationship may be particularly complex (for example AU25, AU51) and b) the same ASD genetic risk factor may be responsible for increased risk of prematurity (for example AU731).

In terms of pathophysiological mechanisms, several hypotheses have been proposed to explain the comorbidity between autism and epilepsy: 1) accidental co-occurrence given the high frequency of both conditions; 2) altered internal organization of minicolumns in the cortex of autistic individuals may be associated with a defect in inhibitory local circuit projection, and the defect in γ-aminobutyric acid (GABA)ergic fibers may correlate with the increased prevalence of seizures in autism 140 141 ; 3) common genetic or neurodevelopmental risk factors; and 4) epilepsy by itself may induce the development of autistic symptoms 142 . Indeed, early-life seizures may affect synaptic plasticity through alteration of neurotransmitter receptor systems and molecules essential for intrinsic neuronal function, and changes that may contribute to an enhanced risk of autism 143 144 . Supporting this hypothesis, infantile spasms and/or early-onset seizures have been significantly associated with autism 145 146 147 . In the present study, few children had seizures in their first year of life (11/63, 17%) and none had infantile spasms. Thus, this mechanism may not account for much of the high prevalence found. As stated previously, in individuals with autism from multiplex families, common pathophysiological mechanisms appear to be more prevalent and may be genetic. If both autism and epilepsy can be considered as disorders of synaptic plasticity 143 144 , autism and epilepsy may be the consequences of the same pathophysiological mechanisms, resulting in abnormal plasticity of genetic origin. Numerous genetic conditions that associate autism and epilepsy in their phenotypic features are caused by mutations in genes implicated in synaptic development, functioning and/or plasticity, such as Angelman syndrome 148 , fragile X syndrome 149 , Rett syndrome 150 and tuberous sclerosis 120 151 . Furthermore, missense and nonsense mutations of several genes involved in neurodevelopment have been identified in individuals with epilepsy, autism or both (Table 1). Most of those genes are involved in cerebral development pathways and functioning: synaptic formation (NLGN4, NRXN1, CDH8, PCDH19, SHANK3), ionic channel subunits (SCN1A, SCN2A, SCN1B, KCNMA1, KCNQ2, CACNA1A) and neurotransmitter receptors (GABRG2, CHRNA4). Similarly, CNVs such as 1q21.1, 3q29, 7q11.23, 15q11–13, 15q13.3, 16p11.2, 16p13.11, 17q12 and 22q11.2 have been implicated in different developmental disorders, such as ASD, intellectual disability and epilepsy 15 17 152 , as well as schizophrenia 16 152 153 (Table 2). Autism, epilepsy and intellectual disability may share alteration of common pathways involved in neurodevelopmental maturation and neural functioning. We propose that these genes need to be more systematically investigated as major and minor risk factors.

Although the current study was conducted in the largest available sample of multiplex families, limitations should be listed. First, we cannot exclude biases in recruitment due to exclusion criteria as evidenced by the very low rate of infantile spasm in the sample. Furthermore, of the 664 children included in this analysis, 35 children (31 children with autistic disorder, two children with PDD-NOS and two children without autism) had strong risk factors for autism and seizures, such as pre- or perinatal insult/prematurity before 35 weeks (n = 19), chromosomal abnormality (n = 8), or abnormal neurological or brain imaging examination (n = 7). However, when analyses were performed excluding these children and their siblings, the results remained unchanged (see also Additional file 1). Second, individuals’ ages were significantly different in children with autism with epilepsy, and children with autism without epilepsy, and the small number of adolescents may have reduced the prevalence of late onset seizures. Third, given the heterogeneity of intellectual functioning in autism and the heterogeneity of intellectual disability measures, we used two standardized tools for the cognitive and adaptive assessment: RCPM and VABS. Assessing a broad sample of 38 autistic children, Dawson et al. found scores on the RCPM to be, on average, 30 percentile points higher than their scores on the Wechsler scale of intelligence 154 . The VABS assess the adaptive functioning with a parent interview. Strong positive relationships have been found between Vineland composite and Vineland subscales (communication, socialization, daily living skills) and IQ, particularly when considering low functioning individuals 155 . Therefore, the two scales can be considered as complementary for the cognitive assessment of children with autism. The fact that we found similar results in the AGRE sample strengthens this view. Fourth, we cannot exclude that the absence of association with gender was related to our final sample size and a lack of statistical power. To assess this issue, we decided to perform a meta-analysis in a larger sample including other studies reporting IQ in multiplex pedigrees 126 138 . The standardized mean difference between mean non-verbal IQ in males versus females from multiplex pedigrees was 0.01 (95% CI: –0.17; 0.19), meaning that there is no difference between mean non-verbal IQ according to gender (Additional file 2: Figure S1). The meta-analysis confirmed our results.

We conclude that epilepsy in multiplex autism may define a different subgroup in terms of clinical characteristics and genetic risk: association with intellectual disability, no association with gender and possible common risk factors. Future studies should be encouraged to increase knowledge of common genetic risk factors between ASD and epilepsy.

aCGH: Array comparative genomic hybridization; ADI-R: Autism diagnostic interview-revised; AGRE: Autism genetic resource exchange; ASD: Autism spectrum disorders; CI: Confidence interval; CNV: Copy number variation; EEG: Electroencephalogram; GABA: γ-aminobutyric acid; GLUT1: Glucose transporter type 1; GWA: Genome-wide association; ID: Intellectual disability; IQ: Intellectual quotient; nACh: Nicotinic acetylcholine; NQA: Not quite autism; OR: Odds ratio; PCR: Polymerase chain reaction; PDD: Pervasive developmental disorder; PDD-NOS: Pervasive developmental disorder-not otherwise specified; RasGAP: Ras GTPase activating protein; RCPM: Raven’s colored progressive matrices; SCZ: Schizophrenia; SD: Standard deviation; SNRPN: Small nuclear ribonucleoprotein protein N; VABS: Vineland adaptive behavior scales.

CA and BG are salaried employees of IntegraGen (Évry, France). DC and ST are compensated consultants for IntegraGen. IGA, CL, EL and NB declare that they have no competing interests.

CA, DC, CL, ST and EL designed the study. CA, BG and IGA extracted the data and performed the quality assessment. CA, IGA, EL and CL critically reviewed the literature. NB, BG, CA, CL and DC performed the statistical analysis. NB, CA and DC performed the meta-analysis. CA, CL, IGA and DC wrote a preliminary draft. All authors read, critically modified and approved the final manuscript.