Structural modularity of the XIST ribonucleoprotein complex

Abstract

Long noncoding RNAs are thought to regulate gene expression by organizing protein complexes through unclear mechanisms. XIST controls the inactivation of an entire X chromosome in female placental mammals. Here we develop and integrate several orthogonal structure-interaction methods to demonstrate that XIST RNA-protein complex folds into an evolutionarily conserved modular architecture. Chimeric RNAs and clustered protein binding in fRIP and eCLIP experiments align with long-range RNA secondary structure, revealing discrete XIST domains that interact with distinct sets of effector proteins. CRISPR-Cas9-mediated permutation of the Xist A-repeat location shows that A-repeat serves as a nucleation center for multiple Xist-associated proteins and m6A modification. Thus modular architecture plays an essential role, in addition to sequence motifs, in determining the specificity of RBP binding and m6A modification. Together, this work builds a comprehensive structure-function model for the XIST RNA-protein complex, and suggests a general strategy for mechanistic studies of large ribonucleoprotein assemblies.

Independent domains for recruitment of PRC1 and PRC2 by human XIST

“…Similar to mouse, PRC1 recruitment required the BC region, which is considerably smaller in humans with only part of a single C repeat and a separation of the B into two clusters of repeats [3]. The D repeat region was also required for human PRC1 recruitment, consistent with structural analysis suggesting a large protein-interaction domain containing from B-C to beyond the D repeats in exon 1, as well as HNRNPK binding to the human D repeats [35]. Thus, the B-C-D region may recruit PRC1, likely through HNRNPK, with weighting in humans towards the requirement including the D repeat region as B-C is smaller than in mouse.…”

“…LncRNAs have been suggested to function as modular scaffolds for protein binding and the internal repeats of XIST/Xist have exemplified the concept [35,36]. In this study we sought to characterize domains of the human XIST, focussing not only on the repeat regions, but also interrogating the non-repetitive regions of the lncRNA.…”

“…One could envision specific RNA secondary structures formed through long range interactions that may allow for complexes to bind to XIST, or distinct protein interactions with different RNA regions that are required to assemble essential protein complexes. While not exclusive of each other, the latter explanation may fit with evidence of five hubs of prolific protein binding, around the A repeats, F repeats, B-C-D repeats, E repeats and 3' end of XIST [35]. Studies in mouse models have argued that the multivalency of the E repeat region allows homo-and hetero-typic interactions with partial redundancy that function in promoting phase separation and formation of the inactive X compartment [38,39].…”

“…Each feature required two or three regions of XIST separated by kilobases of RNA, that when independently deleted had no impact upon recruitment of the feature. Previous work had proposed that long-range interactions were endemic across XIST through the use of RNA duplex mapping approaches [35,37]. One could envision specific RNA secondary structures formed through long range interactions that may allow for complexes to bind to XIST, or distinct protein interactions with different RNA regions that are required to assemble essential protein complexes.…”

“…Again, in mouse models, it has been shown that silencing is necessary for the spread of UbH2A into genes [31], and it is possible that we only detect UbH2A when it has spread beyond initial deposition. The A repeats are also critical for binding of RBM15 for m6A deposition at the 5' and extreme 3' ends of XIST [35], so it is possible that modification of the RNA affects the ability of XIST to bind PRC1 [26,40,41].…”