Exome sequencing in autism – the big picture and implications for epilepsy genetics

The recent weeks have produced several large studies on exome sequencing in autism and we have previously commented on three publications which appeared back-to-back in Nature. Here I summarize these findings, which advanced our understanding of the genetic architecture of autism  in the light of a recent, equally comprehensive study by Iossifov et al. published in Neuron. The enormous amount of genetic variation and de novo mutations in the human genome reported demonstrate the possibilities and difficulties of exome sequencing studies and require new study design.  Up to 10% of cases of autism appear due to de novo mutations, indicating a significant, but modest contribution comparable to the effect of copy number variations.

Study by Iossifov et al., 2012. The authors performed exome sequencing in families with an affected proband and at least one unaffected sibling (quads).  After stringent filtering they estimate the burden of de novo mutations and indels, finding an elevated rate of likely gene disrupting mutations in probands (17%) compared to unaffected siblings (8%). The rate of all missense mutations in contrast to the gene-disrupting mutations was comparable in probands and siblings. De novo mutations are common in affected and unaffected individuals and additional criteria are required to distinguish causal from benign de novo mutations.  Nevertheless, the rate of gene-disrupting mutations in probands with autism is two-fold higher compared to unaffected siblings, suggesting a modest contribution of these mutations to the overall pathogenesis of autism. The authors further analyze molecular pathways that these genes are implicated in and suggest that many products of genes implicated in autism are functionally associated with the Fragile-X-Mental-Retardation Protein (FMRP).  In line with previous studies, they estimate 300-400 autism susceptibility genes.

Five novel candidate genes. In conjunction with the study by Neale et al., O’Roak et al. and Sanders et al., recurrent gene-disrupting mutations have been identified in 5 genes, namely, SCN2A, CHD8, KATNAL2, DYRK1A and POGZ. Among these five genes, only SCN2A and DYRK1A have previously been implicated in neurodevelopmental disorders.  SCN2A is well known to the epilepsy community and codes for the alpha-2 subunit of the voltage-gated sodium channel. SCN2A is known to cause Benign Familial Neonatal-Infantile Seizures (BFNIS). It should be pointed out that none of the two probands with autism carrying mutations in SCN2A had seizures and the connection between axon initial segment dysfunction and autism in not clear. DYRK1A codes for the dual-specificity tyrosine-(Y)-phosphorylation regulated kinase 1A and represents the main candidate gene for intellectual disability in Trisomy 21. The role of the other three genes is not clear. CHD8 codes for chromodomain helicase DNA binding protein 8 involved in chromatin remodeling. POGZ codes for the pogo transposable element with ZNF domain involved in transcriptional regulation. The role of KATNAL2 coding for the katanin p60 subunit A-like 2 is unknown.  In summary, given the candidate genes from CNV studies in autism including many genes involved in synapse functions, these are not the candidate genes you would expect.  The authors point this out, indicating the lack of overlap in candidate genes from CNV and exome studies in autism.

Common findings in all four studies on exome sequencing in autism. Iossifov et al. summarize that the frequency of de novo mutations is elevated in children with older parents, that de novo gene disrupting mutations are twice as frequent in probands compared to siblings and that the role of all missense mutations in general irrespective of their putative functional role is not significant or marginal.  In summary, the role of de novo mutations in autism can be compared to the role of large Copy Number Variations (up to 10% of cases).  Given the vast expectations put into exome sequencing studies, these results are disappointing. Theoretical predictions suggest a higher rate of de novo mutations in autism and the reasons for this gap are as yet unknown. In addition, these studies highlight the difficulties to distinguish pathogenic from accidental variation, an issue that will be central to the upcoming studies in epilepsy research.

Implication for epilepsy genetics and EuroEPINOMICS. The autism studies provide a template for the EuroEPINOMICS projects, particularly the studies, which sequence probands and parents (trios). In contrast to the “naive” assumption that a causative mutation can be identified in every patient, it will be difficult to tell the pathogenic mutation from benign variation. Assuming that the genetics of various epilepsies is comparable to the genetic architecture of autism, the pathogenic role of identified variants would need to be shown through (a) identification of genes already implicated in neurological disorders, (b) functional studies or (c) statistical evidence. Functional studies will be difficult to implement unless the identified genes fall within pathways for which functional assays are established (e.g. ion channels) and statistical evidence, demonstrating that de novo mutations in a given gene are associated with epilepsy will require large cohorts. From the autism studies, we can expect that using a trio-based design, we will identify likely pathogenic de novo mutations in 10% of probands with epilepsy. Exome sequencing is not the final genetic study that will identify the genetic basis of the majority of patients, but only the beginning for upcoming genome-wide sequencing methods with increasing coverage of the genome that can be interpreted in the presence of a growing body of exome data for neurodevelopmental disorders including autism and intellectual disability. The pick-up rate for other study designs including sequencing of families and sibling pairs is difficult to estimate. However, exome sequencing studies have been particularly successful in this field of genetics.

Some may argue that the outlook for epilepsy studies is rather bleak given the low expected frequency. We don’t. If some similarities between autism, ID and epilepsy are present, exome sequencing studies using a trio-based design have the unique possibility to detect causative mutations in up to 10% of patients and significantly enhance our understanding of various epilepsy syndromes. The prerequisite for these studies, however, is careful phenotyping and biobanking for follow-up studies and further increase in the sequence data generation and processing.

Exome sequencing for neurological syndromes

Model of SCN2A protein from Modbase. Its lower half is the membrane region, which can be seen in the distribution of carbon (white).

SNC2A plays a significant role in autism reported one of three recent exome sequencing studies [1].  We were enthusiastic but others – Mike Eisen in particularly – were critical about the claim. Statistically, his analysis is correct but there is context around SCN2A that should be considered. First, it was already shown to be implicated as early as 2003[2]. The new study reconfirms the finding and provides new evidence corroborating the previous results. Soft factors, particular its known expression in the brain and its role in epilepsy would possibly make it an interesting finding even if it would have been detected in a single case.

Ben Neale, the first author of one of the studies summarized the findings in a recommended blog post. Complex diseases are complex. If years and years of research found a syndrome to be influenced by many factors, is hard to characterize, no sequencing effort no matter how deep find a simple mutation explaining most of the cases – or even a sizable part.

Some of the experiments performed in the EuroEPINOMICS consortium are close in design to the three autism studies. They should provide us with expectations and update us on experimental standards and statistical standards. And teach us some modesty.

SCN2A takes center stage again as an autism gene

SCN2A codes for the alpha-2 subunit of the voltage-gated sodium channel and its mutations are implicated in Benign Familial Neonatal-Infantile Seizures (BFNIS). As BFNIS are rare, SCN2A has served as a model for age-related epilepsies rather than an important gene in clinical practice.  SCN2A has pretty much shaped the concept of dysfunctions of the axon initial segment (AIS) as a core pathology of benign infantile epilepsies in the first year of life. Some studies suggest an interesting mechanism for the striking age-related pattern seen in BFNIS.

Now SCN2A appears in a novel context. Three new studies published in Nature last week used whole-exome sequencing to identify genes with de novo mutations implicated in autism[1,2,3]. Despite over 2000 patients sequenced only few genes were discovered and SCN2A stood out due to independent observations in a Yale led study by Sanders et al, independently confirming earlier hints implicating it in autism spectrum disorders.

Model of SCN2A

Model of about half of the ion channel SCN2A from Modbase based on a bacterial ion channel. Rendered by PyMol.

As of now it is unclear how mutations in SCN2A result in autism but it might be worth noting that the critical period for seizures in BFNIS overlaps with the time period in which neurodevelopmental dysregulations in autism are expected – the first year of life.

Epilepsy research is moving past ion channels to genes previously implicated SCN2A in autism (neurexins, neuroligins, CNTNAP2) and autism research is rediscovering ion channels (SCN1A, em>SCN2A). Time to recycle the equipment for electrophysiology of ion channels.