Back

A multi-looping chromatin signature predicts dysregulated gene expression in neurons with familial Alzheimer's disease mutations

Chandrashekar, H.; Simandi, Z.; Choi, H.; Ryu, H.-S.; Waldman, A. J.; Nikish, A.; Muppidi, S. S.; Gong, W.; Paquet, D.; Phillips-Cremins, J. E.

2024-02-27 molecular biology
10.1101/2024.02.27.582395 bioRxiv
Show abstract

Mammalian genomes fold into tens of thousands of long-range loops, but their functional role and physiologic relevance remain poorly understood. Here, using human post-mitotic neurons with rare familial Alzheimers disease (FAD) mutations, we identify hundreds of reproducibly dysregulated genes and thousands of miswired loops prior to amyloid accumulation and tau phosphorylation. Single loops do not predict expression changes; however, the severity and direction of change in mRNA levels and single-cell burst frequency strongly correlate with the number of FAD-gained or -lost promoter-enhancer loops. Classic architectural proteins CTCF and cohesin do not change occupancy in FAD-mutant neurons. Instead, we unexpectedly find TAATTA motifs amenable to binding by DLX homeodomain transcription factors and changing noncoding RNAPolII signal at FAD-dynamic promoter-enhancer loops. DLX1/5/6 mRNA levels are strongly upregulated in FAD-mutant neurons coincident with a shift in excitatory-to-inhibitory gene expression and miswiring of multi-loops connecting enhancers to neural subtype genes. DLX1 overexpression is sufficient for loop miswiring in wildtype neurons, including lost and gained loops at enhancers with tandem TAATTA arrays and singular TAATTA motifs, respectively. Our data uncover a genome structure-function relationship between multi-loop miswiring and dysregulated excitatory and inhibitory transcriptional programs during lineage commitment of human neurons homozygously-engineered with rare FAD mutations.

Matching journals

The top 4 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.