Jpx RNA Activates Xist by Evicting CTCF Sha Sun,1,2,3 Brian C. Del Rosario,1,2,3 Attila Szanto,1,2,3 Yuya Ogawa,1,2,3,4 Yesu Jeon,1,2,3 and Jeannie T. Lee1,2,3,* 1Howard Hughes Medical Institute, Boston, MA 02114, USA 2Department of Molecular Biology, Massachusetts General Hospital, Boston, MA 02114, USA 3Department of Genetics, Harvard Medical School, Boston, MA 02114, USA 4Present address: Department of Pediatrics, University of Cincinnati College of Medicine, Cincinnati Children’s Hospital Medical Center, Cincinnati, OH 45229, USA *Correspondence: [email protected] http://dx.doi.org/10.1016/j.cell.2013.05.028 SUMMARY (Scute, SisA, Runt, Upd) activates the master regulator Sxl, and autosomal factors, such as Deadpan, together with nuclear In mammals, dosage compensation between XX volume, oppose the X-linked signals (Barbash and Cline, 1995; and XY individuals occurs through X chromosome Cline and Meyer, 1996; Salz and Erickson, 2010). inactivation (XCI). The noncoding Xist RNA is ex- In mammals, the X:A ratio currently exists only as a genetic pressed and initiates XCI only when more than one concept. Evidence derives from studies of aneuploids and poly- X chromosome is present. Current models invoke a ploids indicating that XXX females establish two inactive X’s (Xi) dependency on the X-to-autosome ratio (X:A), but when diploid but inactivate only one Xi when tetraploid; further- more, XXX triploid females establish one or two Xi, consistent molecular factors remain poorly defined. Here, we with their intermediate X:A ratio (Brown and Chandra, 1973; Gar- demonstrate that molecular titration between an tler et al., 2006). Two conceptual frameworks have been instruc- X-encoded RNA and an autosomally encoded pro- tive. Early thinking favored the ‘‘Blocking Factor Hypothesis,’’ in tein dictates Xist induction. In pre-XCI cells, CTCF which a single complex of autosomal factors binds to and blocks protein represses Xist transcription. At the onset of one Xic and protects that X chromosome from silencing (Ohno, XCI, Jpx RNA is upregulated, binds CTCF, and extri- 1969; Lyon, 1971; Kay et al., 1994; Starmer and Magnuson, cates CTCF from one Xist allele. We demonstrate that 2009). Remaining Xs are unprotected and become silenced by CTCF is an RNA-binding protein and is titrated away default. The alternative ‘‘Two Factors Model’’ postulates the ex- from the Xist promoter by Jpx RNA. Thus, Jpx acti- istence of not only a blocking factor to protect one X but also a vates Xist by evicting CTCF. The functional antago- ‘‘competence factor’’ to trigger silencing on additional Xs (Gartler nism via molecular titration reveals a role for long and Riggs, 1983; Lee and Lu, 1999; Lee, 2005). Thus, regardless of details, all current models imply a molecular titration of noncoding RNA in epigenetic regulation. X-linked and autosomal factors. Consistent with their separate evolution, fly and worm regula- INTRODUCTION tors are not apparently utilized in the mouse. Unlike the inverte- brate systems, mammalian dosage compensation is allelically Dosage compensation balances X chromosome content controlled and involves regulatory switches defined by long non- between females and males in organisms with the XY mecha- coding RNA (lncRNA) within the master ‘‘X-inactivation center’’ nism of sex determination. In fruitflies, the single X chromosome (Xic). Xist RNA emanates from the Xic, coats the future Xi in in males is transcriptionally upregulated 2-fold; in roundworms, cis, and deposits silencing complexes along the chromosome expression from two X chromosomes is halved in hermaphro- (Brockdorff et al., 1992; Brown et al., 1992). Xist is expressed dites; and in mammals, one of two X’s is transcriptionally only when X:A R1.0, and the number of Xist RNA foci follows silenced in females during X chromosome inactivation (XCI) the ‘‘n-1’’ rule in diploid cells (n = X chromosome number). (Cline and Meyer, 1996; Lucchesi et al., 2005; Payer and Lee, Recent work shows that Xist is both negatively and positively 2008; Wutz, 2011). Although dosage compensation is achieved regulated. It is repressed by the antisense Tsix RNA and the non- differently in various organisms, all three mechanisms depend coding Xite locus (Lee and Lu, 1999; Sado et al., 2001; Ogawa on the ‘‘X-to-autosome ratio’’ (X:A), which triggers the epigenetic and Lee, 2003) but activated by the long noncoding Jpx RNA process only when X:A = 0.5 in fruitflies or X:A R 1.0 in round- (Tian et al., 2010) and the E3 ubiquitin ligase, RNF12 (Jonkers worms and in mammals. Failure to achieve dosage compensa- et al., 2009)(Figure 1A). tion results in early embryonic death. In roundworms, X-linked The identities of X-encoded ‘‘numerators’’ and autosomally factors (SEX-1, CEH-39) and autosomal signals (SEA-1, SEA-2) encoded ‘‘denominators’’ of the X:A ratio have been elusive. In antagonize each other at the master regulatory gene, xol-1 principle, X-linked and autosomal regulators must converge (Meyer, 2010). In fruitflies, accumulation of X signal elements at the Xic—potentially at the Tsix, Xite,orXist locus. Indeed, Cell 153, 1537–1551, June 20, 2013 ª2013 Elsevier Inc. 1537 A Negative B regulation Cnbp2 Ftx Jpx Xist Xite Tsx Xite Positive Xpct regulation Tsix Tsix Jpx/Enox 10 Kb RepA Probe 2 Probe 1 (BAC5) Tg(Jpx) Xist α Rex1 Rnf12 Tg(EF1 :Jpx) XCI C 5 Tg(Jpx) 2 Tg(Jpx) E 4 Cells with Cells with Jpx Cells with 2 Xist clouds n Jpx 1.5 1 Xist cloud 3 Xist 1 Xist cloud n Xist Ctrl 7.5% 0% 53 1 2 Ctrl 0% >200 E2 8.6% 1.2% 81 1.8% 168 0.5 Bs2 E1 12.3% 1.4% 212 1 B2 3.5% 115 Ef1 6.8% 5.0% 161 0 0 Ef2 6.3% 7.0% 158 Level Log2(Tg/Ctrl) E2 E1 -1 Bs2 B2 -0.5 D Xist Xic 2 μ Ctrl Tg(Jpx) Bs2 Tg(Jpx) B2 Tg(EF1α:Jpx) Ef1 Ctrl Tg(Jpx) E2 Tg(Jpx) E1 Tg(EF1α:Jpx) Ef2 F Xist Xic 2 μ Tg(Jpx) JTB1 Tg(EF1a:Jpx) JTF4 Tg(Jpx) ETB1 Tg(EF1a:Jpx) ETF2 H G 6 1.5 JTB1 ETB1 JTF4 5 ETF2 5 Cells with Cells with Cells with 4 1.0 1 Xist cloud n 1 Xist cloud 2 Xist clouds n 4 1.0 3 3 TsixTST/Y 5.0% 139 TsixTST/+ 44.2% 4.6% 129 2 0.5 JTB1 7.7% 104 ETB1 47.7% 12.2% 172 2 0.5 JTF4 15.4% 123 ETF2 47.9% 12.9% 194 1 1 Level Log2(Tg/Ctrl) 0 0.0 0 0.0 Jpx Xist Jpx Xist (legend on next page) 1538 Cell 153, 1537–1551, June 20, 2013 ª2013 Elsevier Inc. integrating extra Xic copies into either a male or female genome lates XCI during cell differentiation. Quantitative RT-PCR mimics the presence of supernumerary Xs and triggers ectopic confirmed overexpression of Jpx RNA between 0.5- to 2-fold XCI (reviewed in Starmer and Magnuson, 2009; Lee, 2011). in multiple independent clones on day 4 of differentiation (d4; One study implicated the Tsix/Xite loci as binding sites for de- Figure 1C), and RNA-DNA fluorescence in situ hybridization nominators without identifying specific autosomal factors (Lee, (FISH) showed ectopic Xist clouds in differentiating embryonic 2005). Another study showed that XCI is sensitive to dosage of stem (ES) cells at d4 (Figures 1D and 1E). Consistent with autosomal OCT4 protein (Donohoe et al., 2009). The X-encoded elevated steady-state levels of Xist shown by qRT-PCR (Fig- E3 ubiquitin ligase, RNF12, was also proposed as a candidate ure 1C), 1%–4% of transgenic male cells possessed an Xist numerator, as excess RNF12 triggers ectopic Xist expression cloud, and 1%–2% of diploid female cells exhibited a second (Jonkers et al., 2009) and RNF12-mediated ubiquitylation of Xist cloud. (Note: tetraploid cells were excluded by performing REX1 occurs at the initiation of XCI (Gontan et al., 2012). DNA FISH using probe 2; Figure 1D.) Ectopic Xist clouds were RNF12’s candidacy as numerator may be complicated by its neither observed in control male nor female cells carrying vector catalytic nature, however. The necessity of precise X:A titration sequences (ctrl). Thus, whereas reducing Jpx expression by half renders catalytic factors conceptually problematic because cat- blocks Xist activation (Tian et al., 2010), 2-fold overexpression of alytic factors with rapid enzymatic rates are unlikely to be limited Jpx caused ectopic Xist expression. by 2-fold molar differences. Indeed, deleting a single allele of The magnitude of overexpression was, however, small. Noting Rnf12—a state that mimics the XY state—delayed but did not a general correlation between Jpx levels and degree of Xist up- abrogate Xist upregulation in mice (Jonkers et al., 2009; Shin regulation (Figures 1C–1E), we asked if increasing Jpx overex- et al., 2010; Barakat et al., 2011). pression caused greater Xist induction. By retrofitting the strong In molecular titration models, numerators are more easily en- Elongation Factor 1a (EF1a) promoter into the Jpx transgene visioned as stoichiometric than as catalytic factors. Candidate (Tg(EF1a:Jpx); Figure 1B), we further enhanced Jpx activity numerators must in theory satisfy several experimental criteria. and observed greater Xist induction. RNA-DNA FISH indicated First, it must be X-linked and escape XCI in order to provide that up to 7% of cells displayed two Xist clouds (Figures 1D numerical information. Second, it should elicit discrete XCI phe- and 1E). Correlation of Jpx and Xist overexpression was also notypes in response to changes in gene copy number.