Multiple distinct domains of human XIST are required to coordinate gene silencing and subsequent heterochromatin formation

Abstract

Background Mammalian dosage compensation is achieved by the inactivation of one X chromosome in XX individuals. In eutheria this process is initiated early in development by the long non-coding RNA XIST. Studies of the initiation of silencing by XIST have focussed on mouse models, so the domains of XIST required to induce silencing in humans, and their relationship with domains required to establish heterochromatin remain to be determined. Methods We have previously established an inducible XIST cDNA in somatic cells and shown it can induce silencing and recruit heterochromatic features. We now assess a series of deletions across the transgene for the ability to induce silencing and integrate these results with time-course and chromatin-remodelling inhibitor treatments to follow the steps of XIST-induced silencing and heterochromatinization. Discussion We find that in addition to the previously reported necessity of the 5’ A repeat region for XIST-induced silencing, the 1 kb around the small F repeat region and a non-repetitive region at the 3’ end of the RNA are also required to silence genes. Silencing of genes up to 17 Mb from the XIST integration occurs within 2 days, while formation of a Cot-1 depleted domain is slower, and more dependent on the region encompassing Repeat F. The role of this region encompassing Repeat F in both the silencing of actively transcribed genes, the spread of H3K27me3 and the formation of a transcriptionally inert domain suggests a role in a pathway crucial for the spread of XIST across the chromatin to target distal regions of inactivation. Histone deacetylation requires only the A repeat region, with HDAC3 inhibition showing limited effect on silencing, but an impact on H3K27me3 recruitment, and as a result the recruitment of MacroH2A. Global HDAC inhibition impacted silencing in both a distance and dose-dependent fashion. The E repeat region was required for CIZ1 and H4K20me1 recruitment as well as H3K27me3; however, these appeared to act relatively independently. The H3K27me3 mark established by PRC2 integrated silencing and many of the heterochromatic features, while the PRC1 mark ubH2A appeared to be downstream of silencing in these human somatic cells.

Derivation of a minimal functional XIST by combining human and mouse interaction domains

“…Induction of an XIST cDNA integrated into an FRT site on chromosome 8p results in both silencing of flanking 8p genes and recruitment of heterochromatin marks ( 43 ). Assessment of deletions spanning this full XIST reinforced that the 5′ end of XIST including the conserved A repeats was critical for silencing ( 44 ). Although the 5′ A repeats are essential for silencing, they are only sufficient to induce local silencing ( 45 ).…”

“…Thus, on the basis of the current knowledge about the different functions of each XIST domain, we tried to generate a smaller transgene that could recapitulate all XIST functions. Domain A is crucial for silencing by recruiting proteins at the initiation of XCI in both mice and humans ( 5 , 6 , 15 , 44 ), and domain E is implicated in XIST/Xist localization, also through protein interactions ( 12 , 28 , 29 , 48 ); however, our recent results suggested that the functional pathways utilized in human somatic cells may differ from those described in mouse differentiation [reviewed in Patrat et al. ( 7 )].…”

“…In our recent study, we identified that the 3′ end of XIST was necessary for silencing and also for UbH2A and SMCHD1 recruitment ( 44 ). On the basis of previous functional constructs lacking exon 8 (see Supplementary Material, Fig.…”

“…A smaller XIST could improve its utility for chromosome therapy, as the current construct is an unwieldy 14 kb. Furthermore, as a follow-up to deletions of individual domains ( 18 , 44 ), these constructs lead to a better understanding of the functional requirement for individual XIST domains. Generating a minimal construct entirely from human regions would be preferable; however, given the small size of the PID, we started with a hybrid mouse/human construct and future work would delineate whether human Bh/B functions similarly.…”

Population variability in X-chromosome inactivation across 9 mammalian species

“…In addition to stochasticity, genetic effects can influence the choice of allelic inactivation and contribute to population variability in XCI ratios. Allelic inactivation during XCI is mediated by the cis-acting long non-coding RNA XIST 30 , which silences its corresponding X-allele through epigenetic modifications 31,32 . Heterozygous variants affecting XIST expression can bias allelic inactivation 15 .…”