Autism spectrum disorder (ASD) is a neurodevelopmental disorder characterized by difficulties in social communication and the presence of restricted, repetitive behaviors (Lord et al., 2018). Globally, it is estimated to affect 1%–2% of children (Lyall et al., 2017). Over the past decade, genetic studies have employed genome sequencing (GS) or exome sequencing (ES) on large-scale ASD cohorts, such as Simons Simplex Collection (SSC) and Autism Speaks MSSNG resource (MSSNG). These studies have revealed the significant role of de novo mutations in coding regions and have identified hundreds of ASD susceptibility genes (Iossifov et al., 2014; Satterstrom et al., 2020; Trost et al., 2022). However, the impact of non-coding de novo mutations (ncDNMs) on neurodevelopmental processes remains largely unexplored.

Non-coding variants can modulate the transcriptional activity of regulatory elements in specific cells, potentially altering gene expression and ultimately influencing neurodevelopmental phenotypes (Doan et al., 2016; Tenney et al., 2023; Zhang et al., 2024). Recently, Zhou et al. (2019) reported that ncDNMs in ASD probands from SSC cohorts exhibit a significantly higher functional impact score compared to those in unaffected siblings, as predicted by neural network models. However, several studies incorporating various bioinformatics annotations have failed to establish a significant association between ncDNMs and the risk of ASD (Werling et al., 2018; Trost et al., 2022; Castro et al., 2023). These inconsistent findings highlight the challenges in elucidating the functional roles of ncDNMs in ASD.

One challenge is the functional evaluation of ncDNMs. While bioinformatics functional annotations have facilitated the prioritization of genetic variants, experimental functional evaluation is essential to determine the influence of these prioritized variants on transcriptional activity. Massively Parallel Reporter Assays (MPRAs) provide a high-throughput experimental approach to assess the regulatory functions of non-coding DNA sequences (Melnikov et al., 2012). To date, MPRAs have successfully identified regulatory variants associated with a range of traits and diseases, including cancer, autoimmune disorders, and psychiatric disorders (Tewhey et al., 2016; Ulirsch et al., 2016; Cooper et al., 2022; Guo et al., 2023; Rummel et al., 2023).

Another challenge lies in identifying the target genes of ncDNMs, which is complicated by cell-type-specific regulatory networks that govern gene expression. The prefrontal cortex, a critical brain region involved in social behavior and cognitive functions, is particularly relevant to ASD (Minshew et al., 2007; Xu et al., 2019). Recent advancements in constructing cell-type-specific regulatory networks in the prefrontal cortex, based on single-cell RNA and assay for transposase-accessible chromatin (ATAC) sequencing, provide a strategy to help determine the target genes of regulatory elements (Herring et al., 2022).

In this study, we systematically evaluated the function of all ncDNMs located within regulatory elements in the prefrontal cortex of 1648 ASD probands and 1736 unaffected siblings from the SSC cohort (Zhou et al., 2019), as well as 3025 ASD probands and 231 unaffected siblings from the MSSNG cohort (Trost et al., 2022). We employed a cell-specific variant effect prediction model and MPRA experiments to evaluate the function of ncDNMs. Subsequently, we investigated the association between functional regulatory mutations (FrMuts) and the risk of ASD, characterizing the functional impact of potential ASD-risk ncDNMs (Fig. 1).