Z-DNA underlies the target choice of Aire by facilitating promoter poising

Abstract

Aire is an unconventional transcriptional regulator that upregulates the expression of thousands of genes encoding peripheral tissue antigens in medullary thymic epithelial cells (mTECs), which promotes clonal deletion or phenotypic diversion of self-reactive T cells. Unlike classical transcription factors’ predictable behaviors in regulating fixed subsets of target genes, Aire-induced gene expression exhibits ‘ordered stochasticity’ on the single cell level while ‘skewed diversity’ on mTEC populations from different mouse strains. The molecular basis or logic underlying this unique targeting behavior of Aire remains largely unknown as Aire does not bind to a particular DNA-sequence motif like typical transcription factors do. Here, we implemented two orthogonal approaches to investigate Aire’s cis-regulatory mechanisms: construction of a convolutional neural network and leveraging natural genetic variation via analysis of F1-hybrid mice. Both approaches nominated Z-DNA and Nfe2•Maf as putative positive regulators of Aire’s target choices. Genome-wide mapping studies showed that Z-DNA-forming and Nfe2l2-binding motifs were positively associated with the inherent ability of a gene’s promoter to generate DNA double-strand breaks (DSBs), and promoters with strong DSB generation were more likely to enter a poised state with accessible chromatin and already-assembled transcriptional machinery. Consequently, Aire preferentially targeted genes with poised promoters. We propose a model whereby Z-DNA anchors the Aire-mediated transcriptional program by increasing DSB generation and promoter poising. Our work not only expands our understanding about the biology of Aire but also gives insights into the fundamental yet elusive mechanisms of Z-DNA-mediated transcription.