Brain specimens are a rare resource in neurodevelopmental disorders but can uniquely provide insight into the genetic mechanisms of disease through the ability to directly study the very tissues that underlie the manifestations of these disorders [9,10,11,12,13,15]. This includes the ability to identify the tissue of specific post-zygotic mutations. Such mutations have for a long time been recognized as an important etiological factor in focal cortical dysplasias and epilepsy syndromes [13]. However, more recently, their role in other neurodevelopmental disorders, including ASD, has emerged [11,12,15]. Much of the evidence is based on the detection of alternate allele fraction (AAF) from the deep sequenced coverage of peripherally derived specimens. In the current study, we examined donated brain tissue from patients with ASD and identified PZMs in several genes that may be aetiologically implicated in their ASD diagnoses. Although none of the genes habouring predicted-damaging PZMs was previously implicated in ASD, intellectual disability (ID), or other neurodevelopmental disorders, two, TRAK1 and CLSTN3, are widely expressed in the brain.

TRAK1 is closely associated with DISC1, itself implicated in schizophrenia and neural development [29,38]. Mutations in DISC1 have been well documented in association with schizophrenia but not ASD. Additionally, one predicted pathogenic mutation in TRAK1 has also been described in the literature in a subject with schizophrenia [39]. The implications, therefore, for neurodevelopmental disorders such as ASD are unclear, although the genetic relationship between ASD and schizophrenia is now well-established [40]. Stronger evidence for a possible role in ASD phenotypes exists for CLSTN3, a calsyntenin, which as a group function as synaptogenic adhesion molecules [30,41]. Specifically, CLSTN3 is thought to be important for the normal development and function of the GABAergic and glutaminergic synapses, and although not specifically identified as an ASD-implicated gene, a de novo predicted damaging mutation in this gene has previously been described in an ASD proband [42]. Rodent models of CLSTN3 have identified the impact of deletion on impairing cognitive function [43], offering a possible pathophysiological mechanism.

In contrast to our search among the PZMs, we did identify several germline mutations that are likely implicated in the ASD diagnoses of these subjects. For example, a mutation in PTEN was identified in one subject. PTEN is an overgrowth gene, with a variable phenotype that may include macrocephaly, tissue proliferation, including tumors, and neurodevelopmental disorders [31]. Indeed, research suggests that as many as 20% of individuals with autism who also are macrocephalic may have a PTEN mutation [44]. In another subject, a predicted damaging mutation in SCN1A was identified. SCN1A encodes the α-1 subunit of the NaV1.1 sodium channel and is strongly associated with epilepsy as one of the most important channelopathies [32,33]. Its most notable association is Dravet Syndrome, but other epilepsy and movement disorders have also been described, as has ASD, although the exact prevalence is unclear, and may be in the region of 24–40% [45]. We do not know if individual AN16115 had epilepsy during their lifetime.

We also identified variants impacting CDH13 and CACNA1C. CDH13 is a Cadherin gene, a group of genes which encode a class of calcium-dependent transmembrane protein. Their embryonic expression earmarks the critical role they play in axonal growth and synapse development. Although there are many members of this protein superfamily, a small number have been described in association with neurodevelopmental phenotypes, including ASD and ADHD [34]. CACNA1C is another calcium channel protein, encoding the α-1c subunit of the L-type voltage-gated calcium channel. Common variants in this gene have been identified in association with several different neuropsychiatric phenotypes [35,36], including ASD [37]. Moreover, mouse models have suggested a role for this gene in potentially related cognitive endophenotypes that include fear conditioning [46]. It is quite possible, therefore, that these mutations also played an etiological role in the ASD diagnosis in these cases and further reinforce the importance of cadherins and a channelopathy-mediated pathogenesis in ASD.

This current study was conducted to show the potential utility of studying patterns of mutation in brain tissue and specifically to identify post-zygotic mutations that may be of aetiological relevance. Although we failed to identify predicted damaging mutations in any known ASD or other neurodevelopmental genes, our study has successfully identified post-zygotic mutations, including a small number in genes that warrant further investigation. Our results may alternatively suggest a less widespread role for PZMs than previously predicted. However, considering the emerging evidence from epilepsy and ASD discussed above, this seems unlikely. The small sample size of the current study will have impacted on the power of this study to identify these variants. It is also possible that PZMs are of aetiological significance only for particular phenotypes, such as severe ASD, nonverbal ASD, or ASD with co-morbidities such as epilepsy. In the absence of phenotype information, we are unable to further comment on this; however, moving forward, any brain tissue resource will need to also include detailed phenotype information to facilitate the unravelling of these brain-behaviour connections. Moreover, we are not able to indicate whether undissected brain regions in these specimens are also negative for PZMs. As discussed in Wintle et al. [15]), the BA17 visual cortex was used for this current study, in part due to its easy dissection and availability. Ideally, the search for PZMs will be predicated on the identification of histopathological abnormality, which will then motivate the decision on which regions of the brain to dissect.

We hope that our results will additionally provide a resource of genetic results on postmortem tissue in ASD patients. The nature of these mutations, including their mechanism, the genes impacted, and the potential effects on brain development will only be truly resolved with greater emphasis on studies such as this one that are also able to evaluate histology and cellular function in the tissues impacted. There are several ways in which studies such as this can make further progress. The establishment of tissue resources that are available to the scientific community are making samples available, allowing increased sample sizes, and these are often from patients who have detailed phenotype information. In considering sample size, there are a number of operational complexities in establishing a tissue resource. Samples may have to be collected from different geographic locations, and these may have been stored under different conditions. Detailed information will be needed on factors that may impact the results and their interpretation. Moreover, whereas on the one hand the tissue may comprise samples resected during surgery, for other families the tissue may have been donated after the death of a loved one. This requires a sensitive and ethical process to be in place. Consequently, we anticipate that sample sizes will remain modest in comparison to other methodological approaches. This notwithstanding, the strength of this approach is the ability to directly identify PZMs in brain tissue as well as more closely study the impact of mutations in the tissue itself. In the future, it will be important to more carefully examine genetic architecture across different regions; this will be facilitated by using single cell genomic approaches. Additionally, it will be important to stratify samples by sex and developmental age, as well as to compare the genetic architecture between histologically normal and abnormal tissues.