Whole-genome analysis reveals the contribution of non-coding de novo transposon insertions to autism spectrum disorder

Abstract

Background: Retrotransposons have been implicated as causes of Mendelian disease, but their role in autism spectrum disorder (ASD) has not been systematically defined, because they are only called with adequate sensitivity from whole genome sequencing (WGS) data and a large enough cohort for this analysis has only recently become available.

Results: We analyzed WGS data from a cohort of 2288 ASD families from the Simons Simplex Collection by establishing a scalable computational pipeline for retrotransposon insertion detection. We report 86,154 polymorphic retrotransposon insertions-including > 60% not previously reported-and 158 de novo retrotransposition events. The overall burden of de novo events was similar between ASD individuals and unaffected siblings, with 1 de novo insertion per 29, 117, and 206 births for Alu, L1, and SVA respectively, and 1 de novo insertion per 21 births total. However, ASD cases showed more de novo L1 insertions than expected in ASD genes. Additionally, we observed exonic insertions in loss-of-function intolerant genes, including a likely pathogenic exonic insertion in CSDE1, only in ASD individuals.

Conclusions: These findings suggest a modest, but important, impact of intronic and exonic retrotransposon insertions in ASD, show the importance of WGS for their analysis, and highlight the utility of specific bioinformatic tools for high-throughput detection of retrotransposon insertions.

Keywords: Alu; Autism spectrum disorder; LINE-1; Neurobiology; Polymorphic insertions; Retrotransposons; SVA; Transposable elements; de novo insertions; de novo rates.

Fig. 1

Detection of transposable element insertions (TEIs) in the SSC cohort. A Pipeline and analysis overview. Quad and trio bam files were analyzed for TEIs using a dockerized version of xTea on the cloud in Amazon Web Services (AWS). Candidate TE insertions were filtered using xTea filters, and filters for regions of the genome with reference and known non-reference TEIs for a high confidence set. A custom pipeline for detection of de novo insertions was used, and candidates were manually inspected on the Integrative Genomics Viewer. Enrichment or depletion of TEIs in ASD genes, high pLI genes, genomic regions, and regulatory regions in fetal brain development was tested by simulation analyses. A subset of candidates was validated by full-length PCR. B Mean number of TEIs detected in the SSC cohort with standard deviation. C Percentage of insertions in the SSC cohort that were not found in previous studies (novel) or overlap with TEIs from previous analyses (known) for all TEIs including those in parents and children (left) and Venn diagram showing overlap with other large cohort studies for TEIs detected in unrelated parental samples in our cohort (right). D Cumulative fraction of TEIs in unrelated parental samples which are found at a certain population allele frequency (PAF) within the SSC cohort. 94% L1, 92% Alu, and 95% SVA insertions show < 1% PAF

Fig. 2

Rates of de novo TEIs. A Combined rates of de novo TEIs per birth for ASD and controls compared to previous studies. The adjusted rate in our study accounts for lower sensitivity for detecting TEIs in short-read Illumina data compared to long-read sequencing data. B Rates of de novo TEIs per birth in probands with ASD and unaffected siblings (controls)

Fig. 3

Enrichment of de novo TEIs in SFARI ASD genes. Observed numbers of de novo TEIs in a list of complied ASD genes are marked by red dots. Black dots and lines represent mean numbers and 95% confidence intervals of expected TEIs based on 10,000 random simulations, respectively. More de novo L1 insertions in ASD genes than expected are observed in cases only

Fig. 4

Genomic distribution of polymorphic and de novo TEIs. A 10,000 random simulations were performed for both polymorphic and de novo TEIs based on the observed rates. Log₂ fold change of observed compared to expected counts in different genomic regions are shown for coding and gene regulatory regions. 95% confidence intervals were estimated based on the empirical distribution of the random simulations. Polymorphic TEIs from parental individuals are depleted in exons and regulatory regions in the developing fetal brain. De novo Alu (A) and L1 insertions (B) do not show this depletion compared to 10,000 random simulations. Two-sided empirical p-values and Benjamini–Yekutieli q-values based on multiple correction of all enrichment and depletions performed are represented

Fig. 5

Full-length PCR validations and visual inspection. A Full-length PCR validation of the Alu insertion in CSDE1 and of the de novo L1 insertion in DAB1 in ASD cases. In lymphoblastoid cell line DNA, we validated the insertions in the ASD proband only. NTC: non-template control. B Integrative Genomics Viewer image at the insertion site in gene CSDE1 in an ASD case. For each individual, the sequencing coverage (top) and sequencing reads (bottom) are shown. The insertion shows the canonical signatures of target-primed reverse transcription (TPRT)-mediated retrotransposition: 15 bp target site duplication (TSD) between the two insertion breakpoints, a poly-A tail, supporting clipped reads, and discordant reads with mates mapping to the consensus Alu sequence. The mother has one small clipped read sequence at the breakpoint which has the same sequence as in the proband, suggesting that the insertion could be mosaic at a low allele frequency in the mother’s blood