Complementary Alu sequences mediate enhancer–promoter selectivity

Enhancers determine spatiotemporal gene expression programs by engaging with long-range promoters1–4. However, it remains unknown how enhancers find their cognate promoters. We recently developed a RNA in situ conformation sequencing technology to identify enhancer–promoter connectivity using pairwise interacting enhancer RNAs and promoter-derived noncoding RNAs5,6. Here we apply this technology to generate high-confidence enhancer–promoter RNA interaction maps in six additional cell lines. Using these maps, we discover that 37.9% of the enhancer–promoter RNA interaction sites are overlapped with Alu sequences. These pairwise interacting Alu and non-Alu RNA sequences tend to be complementary and potentially form duplexes. Knockout of Alu elements compromises enhancer–promoter looping, whereas Alu insertion or CRISPR–dCasRx-mediated Alu tethering to unregulated promoter RNAs can create new loops to homologous enhancers. Mapping 535,404 noncoding risk variants back to the enhancer–promoter RNA interaction maps enabled us to construct variant-to-function maps for interpreting their molecular functions, including 15,318 deletions or insertions in 11,677 Alu elements that affect 6,497 protein-coding genes. We further demonstrate that polymorphic Alu insertion at the PTK2 enhancer can promote tumorigenesis. Our study uncovers a principle for determining enhancer–promoter pairing specificity and provides a framework to link noncoding risk variants to their molecular functions.