Autism spectrum disorder (ASD) comprises a group of neurodevelopmental disorders characterized by impairments in social interaction and communication, restricted interests and repetitive behaviors (Bourgeron, 2015) that affect around 1% (Zeidan et al., 2022) and 2.3% (Maenner, 2021) of the population worldwide. ASD displays great clinical heterogeneity and shows high comorbidity rates with neuropsychiatric and other medical conditions (Tye et al., 2019). Epilepsy is a severe neurological condition characterized by recurrent seizures with 0.76% of lifetime prevalence across the worldwide population (Fiest et al., 2017).

The co-occurrence of ASD and epilepsy has been well documented at the epidemiological level: between 11% and 39% of ASD patients have epilepsy (Canitano, 2007; Lukmanji et al., 2019), while around 15%–47% of people with epilepsy also have ASD (Clarke et al., 2005; Lukmanji et al., 2019). This comorbidity is also associated with intellectual disability and greater severity of the ASD symptomatology (Ko et al., 2016; Lee et al., 2020). Beyond clinical comorbidity, a clear pattern of biological overlap between ASD and epilepsy has been demonstrated (Lee et al., 2015). At the cellular level, several studies have suggested that an alteration of the excitatory—inhibitory (E/I) balance could underlie this co-occurrence (Frye and Rossignol, 2016; Bozzi et al., 2018).

From the genetic perspective, both ASD and epilepsy have been described as complex disorders with considerable heritability estimates (around 80% for ASD (Bai et al., 2019) and 32% for epilepsy (Chen et al., 2017)), which denotes a solid genetic base. A consistent polygenic contribution from both common (Abou-Khalil et al., 2018; Grove et al., 2019) and rare protein-disrupting variation (Feng et al., 2019; Satterstrom et al., 2020; Motelow et al., 2021) has been reported for both ASD and epilepsy. Previous studies have evidenced the genetic overlap between both disorders at the rare variant level; ASD patients with rare protein-disrupting variants, especially affecting neurodevelopmental genes, show increased likelihood of comorbid epilepsy (Bishop et al., 2021). This has been observed for some genetic syndromes such as Phelan-McDermid (Bozzi et al., 2018) or the Fragile X syndrome (Chen et al., 2017) but it has also been described at a genome-wide level for de novo disrupting mutations (Heyne et al., 2018; Satterstrom et al., 2020; Mahjani et al., 2021). In fact, the diagnostic yield of reportable pathogenic variants in ASD is increased from 5%–19% to 15%–60% when comorbid epilepsy is present (Bishop et al., 2021).

By contrast, at the common genetic variation level, recent studies have suggested a limited shared genetic risk from common polygenic contributions across both disorders (The Brainstorm Consortium et al., 2018). However, existent genetic correlation between complex traits may reside in specific pathways and remain undetected when analyzing the whole common genetic variation across the genome (Colbert et al., 2021; Perry et al., 2022). In the past few years, new genomic tools have been developed in order to study the shared common variation in specific genome regions or pathways. Methods such as p-HESS (Shi et al., 2017) or GNOVA (Lu et al., 2017) allow to estimate partial genetic correlation and annotation-stratified covariance to unravel potential specific mechanisms involved in the comorbidity between disorders.

Dysregulation of microRNAs (miRNAs), non-coding RNAs that regulate gene expression by repressing translation via destabilization of mRNA species, has been consistently described in both ASD and epilepsy (Issler and Chen, 2015; Toma et al., 2015; Wu et al., 2016; Wang and Zhao, 2021). MiRNAs regulate the expression of key genes during neurodevelopment and adult brain structure and function, by acting on processes such as synaptogenesis, neurogenesis, neuronal differentiation, and neuroplasticity. Some specific miRNAs have been described to be down- or up-regulated both in ASD and epilepsy. For instance, miR-146a has been found to be upregulated in ASD children and patients with epilepsy (Wang et al., 2015; An et al., 2016; Nguyen et al., 2018). Another example is miR-199, which has been found to be upregulated in patients with temporal lobe epilepsy with hippocampal sclerosis (Antônio et al., 2019), but downregulated in ASD (Sarachana et al., 2010). Since many of these miRNAs are dysregulated both in ASD and other neurodevelopmental disorders (Hicks and Middleton, 2016; Hu et al., 2017), a combination of miRNAs may be potentially used as novel biomarkers for ASD diagnosis or even to describe subgroups of ASD (Hicks et al., 2020; Cui et al., 2021; Salloum-Asfar et al., 2021). Similarly, the potential use of miRNAs as biomarkers and therapeutic targets for epilepsy has been suggested (Henshall et al., 2016; Wang and Zhao, 2021).

Collectively, miRNAs are predicted to regulate around the 60% of the protein-coding genes, by establishing complex regulatory pathways for each miRNA (Friedman et al., 2009). The implication of common genetic variation within genes targeted by some miRNAs has been described for psychiatric disorders (Toma et al., 2015; Hauberg et al., 2016) and available evidence supports that the study of common genetic variation across genes targeted by miRNAs dysregulated in different conditions or traits might be useful to disentangle the genetic correlation structure between them. For example, genetic variation within MIR19A/MIR19B has been implicated in the genetic overlap of educational attainment with ASD and attention hyperactivity disorder (ADHD) (Verhoef et al., 2021).

The current study was designed to assess the shared genetic contribution to both ASD and epilepsy within target genes of miRNAs dysregulated in both conditions. To achieve this aim, we first identified the target genes of miRNAs previously described to be altered in both conditions in more than one study. MiRNA target predictions from various in silico sources were used to ensure a high consistency of the selected genes. Then, we evaluated partial genetic covariance between ASD and epilepsy across these miRNAs’ targets and compared it against their covariance across genes not targeted by these miRNAs. Given the available evidence indicating shared miRNAs’ dysregulation in both conditions, we hypothesize that common genetic variation within these miRNAs’ targets will inform about the genetic correlation structure between both phenotypes. Finally, using polygenic transmission disequilibrium test (pTDT), a method previously performed in a larger ASD cohort (Weiner et al., 2017), we analyzed ASD and epilepsy polygenic transmission using an ASD cohort of 233 trios and analyzed differences in these polygenic contributions to ASD with and without comorbid epilepsy.

This study suggests the presence of a negative genetic relationship between ASD and epilepsy across common genetic variation within target genes of miRNAs involved in both conditions. Moreover, we described a lower transmission of ASD risk alleles from variants within target genes of these miRNAs in ASD patients with comorbid epilepsy than expected by chance. Our results support the idea of a genetic divergence between ASD and epilepsy in terms of common variation, which is more marked when analyzing pathways potentially involved in both conditions such as some miRNA regulatory networks. These results contrast with the direct genetic overlap found between both phenotypes across rare de novo disrupting variation and contribute to conceptualizing the complexity of the genetic relationships between these conditions.

From the genomic perspective, while rare de novo variation affecting neurodevelopmental related genes has been previously described as a predisposing factor to both ASD and epilepsy (Todd and Bassuk, 2018; Hiraide et al., 2019; Mahjani et al., 2021), no clear link between both disorders at the level of common genetic variation has been found (Abou-Khalil et al., 2018; The Brainstorm Consortium et al., 2018). This is important since most ASD genetic risk resides in common polygenic contribution (Gaugler et al., 2014). Here in this study, by restricting our analysis to the regulatory networks of 14 miRNA involved in both ASD and epilepsy, we described a negative genetic correlation between both disorders. Although these results may appear counterintuitive, since these 14 miRNAs have been described to be affected in both phenotypes, several recent studies have described patterns of genetic relationships between ASD and comorbid phenotypes at the level of common genetic variation that differ from those found for disrupting rare variation. For instance, whereas ASD subjects with low IQ have a significantly increased number of disrupting de novo mutations (Iossifov et al., 2014), a positive genetic correlation between ASD and IQ has been described (Zhang et al., 2021). Moreover, ASD polygenic scores predict higher cognitive performance in the global population (Clarke et al., 2016). Another rare-common disparity in genetic relationships was described for schizophrenia; while there is a consistent overlap across genes disrupted by de novo rare variation in schizophrenia and ASD (Kenny et al., 2014; Satterstrom et al., 2020; Singh et al., 2020), shared common variation between both phenotypes has been reported to be chiefly restricted to high functioning autism (González-Peñas et al., 2020).

The negative significant covariance reported here between ASD and epilepsy at the selected miRNA regulatory pathways depict a scenario in which common predisposing variation within miRNA contributes to both conditions in opposite directions. This could be due to the fact that common ASD predisposing variation underpin high functioning autism (González-Peñas et al., 2020) while rare variation is mainly involved in more severe forms of ASD, which are more frequently associated with intellectual disability and epilepsy (Strasser et al., 2018). In this sense, when considering haploinsufficient genes, which have been related with both ASD and epilepsy related genes in terms of rare genetic variation (Katayama et al., 2016; Carvill et al., 2021), we described a clear underrepresentation across target genes of the miRNAs here studied. Also, across the reduced set of haploinsufficient target genes (n = 313), a positive correlation between ASD and epilepsy was described, reinforcing the existence of different genetic contributions from genes primarily affected by rare or common genetic variation.

The lower transmission of ASD risk variants that we observed in ASD patients with comorbid epilepsy supports the idea of common variation differently contributing to phenotypic variation across the autism spectrum (Heyne et al., 2018; González-Peñas et al., 2020; Zhang et al., 2021).

In a recent study, Antaki et al. (2022). have described a negative correlation between transmitted ASD PRS and de novo protein truncating mutations, consistent with a liability threshold model by which the genetic load needed to meet a diagnostic criterion is reached by the additive contribution of both polygenic background and rare high impact genetic variation, in combination to environmental risk factors (Falconer, 1967). Our results describing a lower ASD PRS transmission in ASD patients with epilepsy than expected by chance suggests this liability threshold model is also observable at a pathway specific level. Therefore, focusing on shared altered pathways between ASD and comorbid conditions such as miRNAs may help to disentangle the genetic variation underlying the observed phenotypic heterogeneity of ASD.

All these findings support that the diagnostic category ASD, as described in the fifth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5) could encompass distinct genetically divergent subcategories (Li et al., 2019) that reflect phenotypic variation within the spectrum. Our results support the observed heterogeneity of rare and common variation contribution to complex disorders and highlight the likely underlying genetic divergence between ASD with and without comorbid epilepsy (Burgess et al., 2019). These results contribute to a better understanding of the biological characteristics of the ASD clinical subtypes and may guide patient stratification by focusing on specific pathways or mechanisms of gene expression regulation.

Our work was subject to several limitations. First, the discovery sample size of the latest GWAS of epilepsy used here was modest. Since various types of conditions exist under the epilepsy umbrella, larger and stratified case-control cohorts would be required to reach sufficient power and comparable to ASD and to validate the results presented here. Second, our independent cohort for pTDT analyses, although carefully phenotyped, had also a limited sample size. These findings need to be replicated in a larger sample. Finally, we here selected a list of miRNAs for which dysregulation is observed in both conditions through a comprehensive literature search. Although we only considered miRNAs with significant findings in at least two independent studies, there was heterogeneity among studies in terms of the technique performed and tissue used for the analysis and there could also be miRNA selection or reporting biases. In vivo studies incorporating more reliable methods that enable direct detection of miRNA targets could lead to more powerful outcomes.

In summary, our results provide further evidence about the genetic complexity of ASD, suggesting miRNA dysregulation as one important pathway to explain the genetic complexity in the relationship between ASD and epilepsy. A clear understanding of ASD genetic heterogeneity could be useful to identify risk groups and trajectories and pave the way for precision medicine approaches to neurodevelopmental conditions (Table 2).