SAF-A Requirement in Anchoring XIST RNA to Chromatin Varies in Transformed and Primary Cells

In mammalian female cells, human/mouse XIST/Xist RNA accumulates and stably localizes to the inactive X chromosome territory in cis (Clemson et al., 1996). A recent study examined how this RNA is anchored to the X chromosome. Hasegawa et al. (2010) reported that knockdown of Scaffold Attachment Factor A (SAF-A) (Helbig and Fackelmayer, 2003) is sufficient to release Xist RNA from the interphase chromosome. This was shown primarily in mouse Neuro2a neuroblastoma cells, which do not exhibit fully normal hallmarks of X chromosome inactivation (XCI). The authors used SAF-A RNAi in mouse embryonic fibroblasts (MEFs) as evidence that the requirement for SAF-A was generalizable. This has led to the perception that SAF-A itself directly connects RNA to DNA through its RNA- and DNA-binding domains (e.g., Nakagawa and Prasanth, 2011; Cerase et al., 2015). We tested the role of SAF-A in localizing human XIST RNA in several different cell types and found that it was generally not essential to maintain XIST RNA localization (or for re-localization following mitosis). Instead, our results suggest that SAF-A is one of multiple redundant anchors in primary somatic cells and that other anchors are often lost or compromised in many transformed cells in both human and mouse. We also found that human SAF-A’s putative RNA-binding (RGG) domain was not required to maintain localized XIST RNA, further suggesting that XIST RNA is not simply tethered to chromatin by a single-molecule SAF-A “bridge.”

To examine whether SAF-A is required for XIST RNA localization, we used SAF-A RNAi and analyzed XIST RNA localization by RNA fluorescence in situ hybridization (FISH). We scored cells according to three distinct phenotypes: localized RNA that is restricted to a condensed or somewhat enlarged chromosome territory, partially released RNA (some dispersal but with a localized RNA territory still evident), or fully released RNA (dispersed with no territory) (Figure S1A). We emphasize that when XIST RNA is no longer anchored to the chromosome, it disperses and is clearly visible throughout the nucleoplasm; hence, if RNA remains within a defined but potentially larger (decondensed) chromosome territory, this is biologically not the same as RNA that has lost its anchors and been released from the chromosome.

We found that SAF-A small interfering RNA (siRNA) (85%–95% depletion quantified by microfluorometry) did not strongly disrupt XIST RNA localization in human female primary Tig-1 lung fibroblasts (interphase or post-mitotic G1 daughter cells) or in primary renal mixed epithelial cells, with only 4% of fibroblasts and 2% of renal cells showing fully released XIST RNA (compared to 0% of controls) (Figures S1B–S1D). In contrast, RNAi for mouse SAF-A fully released Xist RNA in ∼80% of Neuro2a cells, similar to the published results (Hasegawa et al., 2010) (Figures S1B, S1C, and S1E). However, in our MEFs, Xist RNA was never fully released and dispersed, but remained localized to a discrete chromosome territory in 92% of cells (versus 98% in controls) (Figures S1B, S1C, and S1F). This was true even in cells quantified by single-cell microfluorometry to have >90% SAF-A depletion. Results affirm a requirement for SAF-A to localize Xist RNA in Neuro2a cancer cells but challenge the generalizability to normal somatic cell types (human and mouse).

We next tested three transformed or immortalized cell lines—mouse 3X, human HEK293, and human MCF10A cells—and found that in each line, SAF-A RNAi caused only partial release of XIST/Xist RNA in ∼55% of cells and full release in ∼25% (Figures S1B, S1C, and S1G). Importantly, the extent of RNA release in all samples did not correlate with the level of SAF-A knockdown, as quantified by microfluorometry (Figures S1B and S1C). These results show that the requirement for SAF-A to localize XIST/Xist RNA differs by cell type and further show heterogeneity among cell populations of each immortalized line. Furthermore, XIST RNA failed to accumulate and form a localized territory in ∼90% of SAF-A-depleted human induced pluripotent stem (iPS) cells, which have many similarities to cancer cells (Meshorer et al., 2006; Carone and Lawrence, 2013). We found that only XIST RNA transcription foci remained after SAF-A depletion in iPS cells (Figures S1H and S1I), suggesting that the impact was on XIST RNA accumulation or stability, rather than localization. Notably, only Neuro2a cells fully released Xist RNA from the chromosome after SAF-A depletion in most cells. These findings counter the perception that SAF-A is normally the singular anchor for XIST RNA and raise the possibility that immortalized or transformed cells are more dependent on SAF-A to localize XIST RNA than primary cells.

To test whether SAF-A functions as a single-molecule bridge between XIST RNA and DNA, as suggested by Hasegawa et al. (2010), we examined the requirement for SAF-A’s separate RNA- and DNA-binding domains. Importantly, initial studies showed that transient overexpression of full-length SAF-A caused XIST RNA release (Figure S1J). Therefore, we generated stable lines of transformed HEK293 cells carrying dox-inducible Flag-tagged SAF-A or SAF-A mutants, with silent mutations to prevent siRNA targeting of the Flag-tagged transcript (Figure S1K). Whereas stable full-length Flag-tagged-SAF-A rescued XIST RNA localization after endogenous SAF-A knockdown (96% of Flag-positive cells had localized XIST RNA) (Figures S1L and S1M), the G29A DNA binding mutant did not (Figures S1L and S1N), consistent with Hasegawa et al. (2010). Our results further show that a single point mutation in the reported DNA-binding domain is sufficient to release XIST RNA. In this cell line, XIST RNA was localized in only 4% of Flag-positive cells, fewer than after SAF-A knockdown (∼20%) (Figure S1L), raising the possibility that this mutant may have dominant-negative effects. In contrast, SAF-A with a precise deletion of just the RGG domain (ΔRGG-Flag) (Figure S1K) was able to substitute for full-length SAF-A, with XIST RNA localized in 88% of cells expressing ΔRGG-flag following endogenous SAF-A knockdown (Figures S1L and S1O). These findings contrast with published conclusions that the mouse RGG domain is required for Xist RNA localization in Neuro2a cells (Hasegawa et al., 2010). However, in that study, interpretation may be difficult due to overexpression of transiently transfected SAF-A, which we found releases XIST RNA (Figure S1J). Alternatively, the carboxy terminal end of SAF-A may play a role, because this domain was eliminated in the truncated ΔRGG mutant studied by Hasegawa et al. (2010) and was recently shown to have a low-complexity prion-like domain (March et al., 2016). We cannot rule out the possibility that our results differ due to the use of a different RGG deletion.

We found additional evidence for other anchors besides SAF-A from experiments inhibiting Aurora B kinase (AURKB), causing XIST RNA to remain on the mitotic chromosome via anchor factor(s) regulated by AURKB chromatin phosphorylation (Hall et al., 2009). In AURKB-inhibited HEK293 cells, SAF-A releases from mitotic chromosomes as usual, indicating that it is not an “AURKB-dependent” anchor (Figure S1P). Weak SAF-A staining can still be detected with the XIST RNA complex on Xi in some cells, affirming their bona fide in vivo interaction (Figure S1P, arrows). However, SAF-A’s overwhelming detachment supports that other AURKB-controlled anchor(s) maintain XIST RNA localization to the metaphase chromosome.

Based primarily on results in mouse neuroblastoma cells, the prevailing view is that SAF-A/hnRNPU is a single-molecule bridge tethering Xist RNA to DNA during XCI maintenance (Hasegawa et al., 2010, as cited in Nakagawa and Prasanth, 2011; Gendrel and Heard, 2014). This compelling, straightforward model may occur in some contexts, but our studies demonstrate more complexity and support the idea that human SAF-A’s DNA-binding domain contributes to localization of an XIST RNA complex, which may be more embedded than tethered by a single factor (Figure S1Q). Further, we show that the normal binding of an RNA that regulates chromatin may become weakened or compromised in transformed cells, which has significant implications for heterochromatin instability in cancer (reviewed in Carone and Lawrence, 2013). Our findings advance the understanding of how XIST RNA localizes to its parent chromosome and demonstrate unrecognized complexity to the anchoring of chromosomal RNA embedded in the nuclear scaffold. We provide insight into how the fundamental biology of anchoring chromosome-associated XIST RNA likely changes with the epigenetic and neoplastic state.