Regulatory Interactions of Autism-Associated Neuronal Enhancers Implicate Novel Genes in Autism Etiology

Introduction

De novo variants contribute to the majority of sporadic cases of Autism Spectrum Disorders (ASD), yet their involvement is still not well understood. Research has been heavily invested in studying variants in coding regions to identify ASD-risk genes. However, most of those variants occur in the non-coding portion of the genome, which contain regulatory elements that are generally responsible for orchestrating transcriptional activation and repression. Therefore, we hypothesized that de novo variants present in intergenic cis-regulatory elements result in the disruption of their target genes involved in neuronal functions. To pinpoint those target genes, we generated an in-house pipeline to identify neuron-specific functional enhancer-promoter interactions from capture Hi-C experiments.

Materials and Methods

To identify the optimal pipeline for capture Hi-C data analysis, we compared eleven promoter capture Hi-C datasets that were digested with 4- and 6-cutter enzymes, MboI and HindIII enzymes, respectively. Valid read pairs were mapped using either HiCUP or mHiC. Functional interactions were identified using GOTHiC++, CHiCAGO, CHiCANE and CHiCMaxima. To verify those interactions, we tested their reproducibility and their enrichment in DNase I hypersensitive sites or histone posttranslational modifications H3K27ac, H3K4me1 and H3K4me3 peaks. The optimal pipeline was then incorporated to analyze twelve publicly available promoter capture Hi-C datasets derived from astrocytes and three neuronal cell types to define promoter-interacting regions (PIRs). To pinpoint ASD-associated enhancers, we analyzed 584,510 non-coding autosomal single-nucleotide variants from 4,864 ASD cases and 3,865 controls for enrichment in 317,026 cis-regulatory elements that have been identified using ChIP-seq, ATAC-seq and/or Capture Hi-C. To identify altered Transcription Factor Binding Sites (TFBSs), we used FIMO with HOCOMOCO Meme matrices as reference. We analyzed GO enrichment using Enrichr with the reference “GO Biological Process 2021” ontology. To functionally characterize risk genes, we assessed their effect on primary and secondary ASD-associated behaviors, such as social interactions, learning, anxiety and reactivity, using Drosophila mutants with a pan-neuronal knockdown of ASD-risk genes.

Results

Different capture Hi-C analysis pipelines result in diverse interaction sets. The optimal capture Hi-C pipeline is dependent on the project's tolerance of type I and type II errors. We used GOTHiC++ to identify 35,783 PIRs in astrocytes, 21,173 PIRs in hippocampal, 31,547 PIRs in excitatory, 26,606 PIRs in motor neurons. We identified 18 ASD-associated enhancers with 265 and 15 case and control variants, respectively, across 12 chromosomes. They were interacting with 91 novel genes and 4 previously reported ASD-risk genes. We also identified 32 ASD-risk genes that were within 2.5 Mb of each region. ASD-associated cis-regulatory elements had 38 altered TFBSs, four of which are reported in ASD databases. They were enriched for neuron-specific, sensory-specific or diverse glial functions depending on whether the TFs were gained, lost or both. Pan-neuronal knockdown of one novel and one known ASD-risk gene in Drosophila showed abnormal responses to stimulus, anxiety, social interactions and learning.

Conclusion

Cis-regulatory elements contribute to autism etiology. Incorporating chromosome conformation capture technologies is a more accurate approach to define target genes, as they recognize those that are in spatial proximity irrespective of their linear genomic position.