Repeat E anchors Xist RNA to the inactive X chromosomal compartment through CDKN1A-interacting protein (CIZ1)

Significance

The long noncoding Xist RNA coats and silences one X chromosome in female cells. How Xist localizes in cis to the inactive X compartment is not clear. Here, we reveal a required interaction between CIZ1 protein and Xist Repeat E motifs. Stochastic optical reconstruction microscopy (STORM) shows a tight association of CIZ1 with Xist RNA at the single-molecule level. Deletion of either CIZ1 or Repeat E causes dispersal of Xist RNA throughout the nucleoplasm, as well as loss of the heterochromatin mark H3K27me3 from the inactive X chromosome. We have thus identified a critical factor for stable association of Xist RNA with the inactive X chromosome.

Abstract

X chromosome inactivation is an epigenetic dosage compensation mechanism in female mammals driven by the long noncoding RNA, Xist. Although recent genomic and proteomic approaches have provided a more global view of Xist’s function, how Xist RNA localizes to the inactive X chromosome (Xi) and spreads in cis remains unclear. Here, we report that the CDKN1-interacting zinc finger protein CIZ1 is critical for localization of Xist RNA to the Xi chromosome territory. Stochastic optical reconstruction microscopy (STORM) shows a tight association of CIZ1 with Xist RNA at the single-molecule level. CIZ1 interacts with a specific region within Xist exon 7–namely, the highly repetitive Repeat E motif. Using genetic analysis, we show that loss of CIZ1 or deletion of Repeat E in female cells phenocopies one another in causing Xist RNA to delocalize from the Xi and disperse into the nucleoplasm. Interestingly, this interaction is exquisitely sensitive to CIZ1 levels, as overexpression of CIZ1 likewise results in Xist delocalization. As a consequence, this delocalization is accompanied by a decrease in H3K27me3 on the Xi. Our data reveal that CIZ1 plays a major role in ensuring stable association of Xist RNA within the Xi territory.

X chromosome inactivation (XCI) is one of the most extensively studied epigenetic processes to date. Since its discovery more than 50 years ago, numerous genetic and cellular studies have uncovered several RNA and protein factors to be high-confidence regulators of this process (reviewed in refs. 1–3). Recently, the advent of genomic (4–7) and proteomic (8–10) approaches for studying long noncoding RNAs has brought about a more holistic view of XCI mechanics. Still, how Xist is able to spread across only one of two X chromosomes and be retained within the inactive X (Xi) territory as an Xist cloud (11) remains one of the most challenging questions to address. Despite an intuitive perception that Xist localization must be confined in cis to the allele from which it is transcribed, specific molecular players have yet to be fully elucidated. Although the transcription factor YY1 has been ascribed a role for the nucleation of Xist RNA in cis to the Xi (12, 13), how Xist RNA spreads exclusively along the same chromosome and remains stably associated with it is unknown. The nuclear matrix protein HNRNPU (also known as SAF-A) was also found to be important for Xist RNA localization (13, 14). However, neither protein is particularly enriched on the Xi relative to other chromosomes, and may function indirectly or in a cell-type-specific manner (15, 16).

While performing superresolution stochastic optical reconstruction microscopy (STORM) to investigate candidate protein factors for their ability to colocalize with Xist RNA (17), we came across an ASH2L antibody that piqued our interest. This Trithorax protein, usually associated with active genes, was previously reported to colocalize with Xist RNA in immunofluorescence (IF) experiments (18). Indeed, our initial analysis confirmed this colocalization, with the antibody exhibiting an exceptionally high level of colocalization with Xist RNA in mouse embryonic stem (ES) and immortalized embryonic fibroblast (MEF) cells (Fig. S1 A and B). However, further analysis suggested ASH2L was not the recognized epitope (Figs. S1). IF showed that knockdown of ASH2L by siRNA failed to abolish the Xi-enriched signal, despite effective knockdown at both the protein and mRNA levels (Fig. S1 C and D). In addition, IF using two other commercially available antibodies or an EGFP fusion protein failed to show any sign of ASH2L enrichment on the Xi (Figs. S1E and S2).

Fig. S1.

ASH2L antibody exhibits cross-reactivity to an unknown Xi-localizing factor. (A) STORM imaging of Xist RNA FISH and ASH2L IF, using a specific antibody (A300-107A; Bethyl Laboratories) in female mouse ES cell. Distribution of nearest neighbor’s distance between Xist and ASH2L shows a high degree of proximity, significantly more than randomized control. Boxed area is shown at higher magnification. P values (Kolmogorov-Smirnov test) were calculated by comparing distributions of observed distances with those of control distances on the basis of randomized localizations. (B) STORM imaging of Xist RNA FISH and ASH2L IF in female mouse fibroblast cell. (C) IF, using ASH2L antibody, exhibits minimal difference between control siRNA and ASH2L knock-down cells. (D) Quantitative RT-PCR and immunoblot show effective knock-down of ASH2L mRNA and protein, respectively. (E) IF using two different ASH2L antibodies on female mouse fibroblast cells failed to show any enrichment of ASH2L on the Xi.

Fig. S2.

Screening of EGFP fusion proteins yields CIZ1 as a Xi-localizing factor. CIZ1, HNRNPC, HNRNPM, and ASH2L were expressed as EGFP fusion proteins. Only CIZ1 candidate protein from co-IP/MS colocalizes with XIST RNA in HEK293FT cells, whereas ASH2L is not found enriched on the Xi. Arrow indicates position of XIST clouds.

To identify the epitope associated with the Xi, we performed proteomic analysis of the material immunoprecipitated by the presumptive ASH2L antibody (Table S1). To screen candidates, we constructed and transiently transfected several EGFP-fusion proteins into HEK293FT cells. Among them, CIZ1 arose as a highly enriched factor on the Xi (Fig. 1A and Figs. S2 and S3). CIZ1 was originally identified as a CDKN1A-interacting protein (19) and has been reported to play a role in cell cycle progression (20, 21). It can be found in tight association with the nuclear matrix and is resistant to high salt extraction (22). CIZ1 is also linked to several human diseases, including cervical dystonia (23) and lung cancer (24). Furthermore, although not previously implicated in XCI, CIZ1 did emerge as a potential Xist RNA interactor in one of the recent proteomic studies (8).

Fig. 1.

CIZ1 is a Xi-localizing protein. (A) CIZ1 colocalizes with Xist RNA in transformed tetraploid female mouse embryonic fibroblasts (WT MEF). CIZ1-EGFP colocalizes with Xist in MEFs carrying an EGFP knock-in at the C terminus of the endogenous CIZ1 locus. (B) CIZ1 shows no Xi localization pattern in female MEFs with Xist deleted. (C) Two-color STORM image showing CIZ1 colocalization with Xist RNA in female MEF. The two Xist clouds are boxed, with one shown at higher magnification. The nearest neighbor’s distance measurement shows most of CIZ1 is proximal to Xist RNA, significantly more than randomized control (P values from Kolmogorov-Smirnov test). (D) Stoichiometry of the number of Xist puncta relative to CIZ1 puncta per Xi. *Data for Xist vs. EZH2 are taken from ref. 17 for comparison purposes. Note that each punctum could have multiple molecules of protein or RNA. This analysis is strictly focused on the number of clusters (puncta) of protein or RNA. (E) UV-RIP using CIZ1-EGFP knock-in cell line shows CIZ1-EGFP is in close interaction with Xist RNA. Immnuoblot verifies CIZ1-EGFP was pulled down efficiently compared with 10% input. *Statistically significant difference in paired Student t test. (F) LNA knock-off of Xist RNA from the Xi also displaces CIZ1 within 1 h post transfection while scrambled control LNA (Scr) has no effect. (G) Xist RNA FISH and CIZ1 IF on days 2–14 of female mouse ES cell differentiation. The boxed area in day 2 was enlarged and contrast further adjusted to show weak level of CIZ1 still colocalizing with Xist at this early time. (H) STORM imaging of Xist RNA FISH and CIZ1 IF on the Xi in day 4 differentiating ES cell. (I) Quantitative RT-PCR shows CIZ1 is up-regulated along with Xist during female mouse ES cell differentiation. CIZ1 also shows similar up-regulation in male ES cells.

Fig. S3.

Characterization of CIZ1 antibody. (A) Immunoblot using WT MEF, CIZ1 KO (KO1), and CIZ1-EGFP knock-in (KI) whole-cell lysate. Arrows indicate CIZ1 (WT) and CIZ1-EGFP (KI), respectively. Asterisk denotes nonspecific protein band. Tubulin serves as loading control. (B) CIZ1 IF signal is effectively blocked by antigen at 2.5× molar excess.

We confirmed CIZ1’s localization to the Xi in several ways. First, CIZ1 colocalization with Xist RNA in female mouse cells was examined by IF, using an in-house CIZ1 antibody (Fig. 1A and Fig. S3), as well as by C-terminal knock-in of EGFP at the endogenous CIZ1 locus (Fig. 1A and Fig. S4). Importantly, Xi localization of CIZ1 was not observed in an Xist-deleted female MEF (25), indicating Xist is necessary for CIZ1 recruitment to the Xi (Fig. 1B). To investigate the molecular localization of CIZ1, we performed superresolution imaging analysis using STORM to resolve single Xist particles that were previously deduced to contain one to two molecules of Xist RNA (17). Intriguingly, CIZ1 showed greater proximity to puncta of Xist transcripts than any other previously examined Xi-associated factor (Fig. 1C), including EZH2, H3K27me3, SMCHD1, H4K20me1, and HBiX1 (17). A large fraction of CIZ1-Xist pairs were within 25 nm of each other (Fig. 1C), a distance within the empirical resolution (20–30 nm) of STORM microscopy. Approximately 85% of all pairs showed a separation of <50 nm distance. This distance is significantly different from that which would be derived from a random model (P << 0.001).

Fig. S4.

Generation of CIZ1-EGFP knock-in cell line. (A) Schematic diagram of mouse CIZ1 gene structure (based on mm9 reference genome) and knock-in cassette. A pair of guide RNAs was designed to target near the stop codon. (B) Immunoblot using GFP antibody confirms the presence of CIZ1-EGFP fusion protein. Asterisk denotes nonspecific protein band. (C) CIZ1-EGFP colocalizes with H3K27me3.

These data suggested a very close relationship between CIZ1 and Xist RNA. We counted the number of Xist puncta relative to CIZ1 puncta and observed a similar stoichiometry, with 52.2 ± 11.2 for Xist and 61.9 ± 10.8 for CIZ1 (Fig. 1D). This similarity is consistent with a tight coclustering of Xist and CIZ1, and is also consistent with the nearly equal stoichiometry of Xist to Polycomb repressive complex 2 puncta, as previously reported (17). To probe further whether Xist and CIZ1 might associate with each other at a molecular level, we performed UV-crosslinked RNA immunoprecipitation (UV-RIP), using our CIZ1-EGFP knock-in cell line (Fig. 1E). A clear enrichment over non-UV-crosslinked control or the parental cell line lacking EGFP supported a potential direct interaction between CIZ1 and Xist RNA in vivo. To test this relationship further, we used locked nucleic acid (LNA) antisense oligonucleotides to “knock off” Xist RNA from the Xi (26) and asked whether there were consequences on CIZ1 localization. LNA knock-off caused an immediate delocalization of CIZ1 that was detectable as early as 20 min after transfection, in a manner that was concurrent with loss of Xist RNA (Fig. 1F and Fig. S5A). Recovery of both Xist and CIZ1 occurred very slowly after 4–8 h in a concordant fashion (Fig. S5B). Taken together, these data support a very close association between Xist RNA and CIZ1 in Xi localization.

Fig. S5.

Xist RNA is required for CIZ1 Recruitment to Xi. (A) Administration of LNA antisense to Xist Repeat C motif results in concomitant loss of CIZ1 localization to the Xi within 20 min in MEFs. (B) Slow recovery over 4–8 h was observed, in line with reestablishment of the Xist cloud. (C) CIZ1 does not exhibit any distinctive Xi localization pattern in male MEF. (D) A full-length Xist transgene is able to recruit CIZ1 even when expressed from male MEF. Arrowhead indicates Xist cloud from stably integrated loci.

We then examined the time course of CIZ1 recruitment in differentiating female ES cells: an ex vivo model for X chromosome inactivation. CIZ1 foci were observed as soon as Xist RNA was detected by RNA FISH on day 2 of differentiation, although CIZ1 foci were less intense and appeared somewhat punctate at this early point (Fig. 1G). Between days 4 and 14, CIZ1 signal continued to accumulate coincidentally with Xist RNA in differentiating female ES cells (Fig. 1G). Nearly all Xist foci showed a confidently detectable level of overlapping CIZ1 localization by day 4 (98%, n = 117). STORM imaging of day 4 ES cells further confirmed proximal localization of CIZ1 to Xist RNA (Fig. 1H). Quantitative RT-PCR demonstrated that CIZ1 levels were up-regulated during female cell differentiation with a time course that paralleled Xist’s up-regulation (Fig. 1I). In differentiating male ES cells, CIZ1 was also transcriptionally up-regulated (Fig. 1I), but failed to accumulate on the single active X chromosome, consistent with the absence of XCI in male fibroblasts (Fig. S5C). In contrast, CIZ1 could be recruited ectopically to an induced Xist transgene in male fibroblasts (Fig. S5D), indicating Xist expression is sufficient for CIZ1 recruitment. We conclude that CIZ1 is rapidly recruited to the Xi during XCI, and that Xist RNA is both necessary and sufficient to recruit CIZ1.

To understand CIZ1’s role during XCI, we established two female MEF cell lines harboring small deletions in CIZ1’s exon 5 (present in all splicing isoforms), using the CRISPR/Cas9 system (Fig. 2 A and B). Two knockout (KO) clonal lines were established: KO1 has a frameshift deletion, whereas KO5 has a short (≤16 aa) in-frame deletion (Fig. S6A). Both KO cell lines showed loss of CIZ1 protein in Western blot analysis (Fig. 2B), suggesting the frame-shift and in-frame mutations both produced unstable protein. Intriguingly, loss of CIZ1 protein in both cell lines led to an aberrant pattern of Xist accumulation on the Xi (Fig. S7A). Analysis by 3D STORM superresolution imaging showed poorly localized Xist particles and a gradient of Xist concentration, indicative of diffusion away from the site of synthesis (Xi) (Fig. 2C). Xist RNA FISH and X chromosome paint confirmed that Xist RNA localized beyond the Xi chromosome territory (Fig. 2D). This aberrant localization pattern was not caused by any evident effect on Xist expression (Fig. 2E). Quantitative RT-PCR confirmed that CIZ1 loss did not significantly affect levels of HNRNPU, another factor critical for proper Xist localization (14) (Fig. 2E). We also generated CIZ1 KO female ES cells and observed a similar Xist localization defect (Figs. S6B and S7B). Significantly, the role of CIZ1 in Xist localization is conserved in human, as KO of CIZ1 in HEK293FT cells likewise resulted in dispersal of XIST RNA (Figs. S6C and S7C). Xist delocalization led to a consequent decrease or loss of H3K27me3 on the Xi in KO MEFs (Fig. 2F), consistent with a requirement for Xist in recruiting Polycomb repressive complex 2. Taken together, these data demonstrate that CIZ1 is required for Xist localization.

Fig. 2.

CIZ1 is critical for maintenance of the Xist cloud and Xi chromatin marks. (A) Schematic diagram of murine CIZ1 gene structure (based on mm9 reference genome) and guide RNA target position. Arrow indicates the orientation of transcription, with boxes and dotted lines representing exons and introns, respectively. (B) Immunoblot confirms depletion of CIZ1 protein in KO1 and KO5 cell lines. YY1 was used as a loading control. *Nonspecific protein band. (C) STORM imaging of boxed areas shows Xist particles diffusing away in KO compared with tight cloud in WT cells. Depth in the z-plane is color-coded from red (+400 nm) to green (−400 nm). (D) Xist RNA in KO cells is detected outside the X chromosome territory. Arrows and arrowheads denote the two active and inactive X chromosomes, respectively. (E) CIZ1 depletion has minimal effect on Xist or HNRNPU RNA levels. Two primer sets were used for each gene. Mean ± SD for three replicates is shown. (F) H3K27me3 on the Xi is lost or reduced in a significant fraction of CIZ1 KO cells. Arrowheads indicate H3K27me3 enrichment on Xi.

Fig. S6.

Genotypes of CIZ1 and HNRNPU KO cell lines. (A) Sanger sequencing of MEF CIZ1 KO and mouse ES KO cell lines. (B) Schematic diagram of human CIZ1 gene structure (based on hg39 reference genome) and guide RNA target position. Sanger sequencing is also shown. (C) Schematic diagram of mouse HNRNPU gene structure (based on mm9 reference genome) and guide RNA target position. Sanger sequencing is also shown.

Fig. S7.

CIZ1 is required for Xist/XIST localization in mouse ES cells as well as in human cells. (A–C) Xist/XIST localization was dispersed in mouse fibroblast cells (A), ES cells (B), and human HEK293FT cells (C) harboring CIZ1 KO. Yellow circle indicates nuclear boundary.

We then investigated whether specific motifs in Xist RNA are responsible for CIZ1 recruitment. We first tested female MEFs carrying Xist transgenes with various subdeletions (12). Cell lines with a wild-type Xist transgene or a transgene with a Repeat A deletion were both capable of recruiting CIZ1 (Fig. 3A). In contrast, a transgene containing only exon 1 of Xist failed to recruit CIZ1 (Fig. 3B), arguing that critical CIZ1-interacting domains lie outside of exon 1. To pinpoint required domains, we began by deleting the entire exon 7 (the largest exon after the first), using the CRISPR/Cas9 system and a pair of guide RNAs flanking the endogenous locus in MEFs (Fig. 3C, see “∆Ex7-10”). Because the MEFs are tetraploid with two Xis, we could isolate clones with one Xi containing the desired deletion and the other serving as an internal control within the same nucleus, as well as clones with deletions on both Xis. Deletion of the entire exon 7 not only failed to recruit CIZ1 but also phenocopied the Xist delocalization and loss of H3K27me3, as seen in CIZ1 null MEFs (Fig. 3C and Figs. S8 and S9, see “ΔEx7-10”). This phenotype contrasted sharply with that of wild-type Xist RNA in the same nucleus of each KO cell. UV-RIP-qPCR showed that Xist ∆Ex7-22 (containing the same deletion as ∆Ex7-10, but on both Xis) could no longer be pulled down by CIZ1 protein (Fig. 3D). Likewise, enrichment of H3K27me3 was lost on both Xis (Fig. 3C and Figs. S8 and S9). Thus, exon 7 of Xist RNA is critical for CIZ1 recruitment to the Xi.

Fig. 3.

CIZ1 interacts with Xist RNA through the Repeat E region. (A) Transgenes containing either full-length Xist or Xist lacking the Repeat A region are sufficient to recruit CIZ1. Arrows indicate Xist cloud from endogenous loci; arrowhead indicates overexpressed Xist from stably integrated transgene. (B) A transgene containing only Xist exon 1 is insufficient to recruit CIZ1. (C) Schematic diagram showing several cell lines with endogenous subdeletions within Xist exon 7. Red dotted lines denote deleted segments. Presence or absence of CIZ1 recruitment, normal Xist cloud, and H3K27me3 on Xi is indicated by “+” or “−,” respectively. Only weak CIZ1 recruitment was observed in clone 1-2. At least 50 cells were counted for each genotype. (D) UV-RIP shows CIZ1-Xist interaction requires Repeat E within exon 7. ∆RepE-16 clone contains the entire Repeat E deleted on both Xis. Mean ± SD for three replicates is shown. (E) CIZ1 IF/Xist RNA FISH of representative cells shows loss of CIZ1 recruitment and dispersed Xist RNA away from the Xi in indicated deletion cell lines. Arrow indicates WT Xist; arrowhead indicates Xist with deletion. Contrast was enhanced for Xist RNA FISH to show single particles outside the main cloud. (F) H3K27me3 IF/Xist RNA FISH shows loss of H3K27me3 and dispersed Xist RNA away from the Xi as in E.

Fig. S8.

Various exon 7 deletions result in dispersed Xist cloud and loss of CIZ1. (A) Schematic diagram showing several cell lines with endogenous subdeletions within Xist exon 7. Red dotted lines denote deleted segments. (B) The first 800 nt–1 kb of Repeat E (“1-2” and “∆RepE-4”) is critical for CIZ1 recruitment, with longer deletions able to abolish H3K27me3 (Fig. S9). Minimum of 50 cells were counted for each genotype. (C) CIZ1 IF/Xist RNA FISH of representative cells shows loss of CIZ1 recruitment and dispersed Xist RNA away from the Xi in indicated deletion cell lines. Arrow indicates WT Xist, whereas arrowhead indicates mutant Xist with deletion. Contrast was enhanced for Xist RNA FISH to show single particles outside the main cloud. Clones 3–16 and ∆Ex7-22 have deletions on both Xi’s (−/−).

Fig. S9.

Various exon 7 deletions result in loss of H3K27me3 from Xi. H3K27me3 IF/Xist RNA FISH shows loss of H3K27me3 and dispersed Xist RNA around the Xi. Cell counting results are shown in Fig. S8B.

Within exon 7 lies one of Xist’s many repetitive motifs, Repeat E. Deleting this entire 1.2-kb Repeat E motif resulted in complete loss of CIZ1 and near-complete loss of H3K27me3 IF signal from the Xi, along with severe disruption of the Xist cloud (Fig. 3C and Figs. S8 and S9, see “3-9” and “7a-11”). Finer mapping revealed that deleting the first 800 nucleotides (containing the highly repetitive sequences upstream of a PstI restriction site) was enough to ablate CIZ1 and H3K27me3 IF signal in 50% of the population, along with some disruption of the Xist cloud (Fig. 3 C, E, and F, see “1-2”). A larger deletion encompassing the first 1 kb of Repeat E further reduced CIZ1 IF signal below detection in nearly all cells. This was accompanied by an increased disruption of the Xist cloud and similar loss of H3K27me3 in 50% of cells (Fig. 3 C, E, and F, see “∆RepE-4”). In general, the larger the deletion of Repeat E (beginning at the proximal end), the more pronounced effect on CIZ1/Xist localization and H3K27me3 deposition (Fig. 3C and Figs. S8 and S9). In addition, UV-RIP-qPCR showed that Xist ∆RepE-16 (containing the same deletion as ∆RepE-4, but on both Xis) could no longer be pulled down by CIZ1 protein (Fig. 3D). These data identify the Repeat E in exon 7 of Xist RNA as essential for the recruitment of CIZ1 to the Xi.

HNRNPU had been previously shown to be important as a nuclear matrix factor for the localization of Xist RNA to the Xi chromosomal territory (13, 14). Unlike CIZ1, however, HNRNPU is not seen enriched on the Xi, and may therefore function indirectly as a nuclear matrix factor for the anchorage of heterochromatic factors of various chromosomes. We asked whether CIZ1 and HNRNPU may function together in the same pathway, albeit with CIZ1 being Xi-specific and HNRNPU being more general. To test this, we generated HNRNPU KO cells using CRISPR/Cas9 (Fig. S6C). Surprisingly, HNRNPU KO cells were viable and exhibited two noticeable defects: slow growth and dispersed Xist clouds, consistent with previous experiments using HNRNPU siRNA knockdown (14). Interestingly, CIZ1 IF of HNRNPU KO cells revealed that CIZ1 remained colocalized with Xist RNA, despite Xist particles being dispersed throughout the nucleoplasm (Fig. 4A), with a Pearson’s coefficient of >0.8. Fluorescence intensity showed CIZ1 and Xist signals peaked together nearly perfectly along a linear 8-μm distance, and this tight association was evident by STORM imaging of the same nuclei (Fig. 4A). Thus, CIZ1 interacts with Xist RNA independent of HNRNPU. HNRNPU UV-RIP in CIZ1 KO1 cells demonstrated that, reciprocally, HNRNPU interacts with Xist independent of CIZ1 or Xist exon 7 (Fig. 4B). Taken together, these data argue that both CIZ1 and HNRNPU are necessary for Xist localization to the Xi. However, their interactions with Xist RNA occur independent of each other.

Fig. 4.

CIZ1 interacts with Xist RNA independent of HNRNPU. (A) CIZ1 remains colocalized with Xist RNA in HNRNPU KO cells. Boxed area is enlarged. Pearson coefficient was calculated (O) along with randomized control (R) from the conventional image. Line chart of fluorescence intensity along the yellow line shows CIZ1 signal peaks together with Xist RNA signal (arrows). Two-color STORM image of the same cell shows CIZ1 colocalizes with Xist at the molecular level. Five of 6 colocalizations (black arrows in intensity chart, white arrows in STORM image) were confirmed whereas 1/6 was not seen (gray arrow). (B) HNRNPU UV-RIP using CIZ1 KO and Xist ΔEx7 cell lines suggests HNRNPU can interact with Xist RNA independently of CIZ1. Mean ± SD for three replicates is shown. (C) Overexpression of EGFP-CIZ1 phenocopies CIZ1 KO of dispersed Xist particles away from the Xi, whereas EGPF alone (EGFP EV) has no effect. Immunoblot confirms EGFP-CIZ1 overexpression compared with endogenous levels. (D) Overexpression of EGFP-CIZ1 does not affect Xist or HNRNPU levels. Relative RNA level is normalized to untransfected cells. Mean ± SD for three replicates is shown. (E) Proper Xist localization simultaneously requires at least two independent protein factors, CIZ1 and HNRNPU.

Maintaining physiological levels of CIZ1 seems crucial. Intriguingly, although transient overexpression of EGFP alone did not influence Xist localization, overexpression of EGFP-CIZ1 triggered a phenotype similar to that of CIZ1 depletion (Fig. 4C) without changing Xist or HNRNPU levels (Fig. 4D). We arrived at this conclusion through assessment of the number of delocalized Xist puncta relative to controls. The assessment was performed by two independent scorers (H.S. and D.C.) and by single-blind scoring (H.S.), each yielding similar trends: ∼90% of CIZ1 overexpressed nuclei showed disperse Xist puncta, whereas <12% of wild-type nuclei showed this pattern (Fig. 4C). Thus, anchorage of Xist RNA to the nuclear matrix appears to depend on a fine stoichiometric balance between Repeat E and CIZ1, with too much CIZ1 possibly saturating binding sites in the nuclear matrix and thereby preventing the anchorage of Xist Repeat E. In sum, we suggest that, although both CIZ1 and HNRNPU are required for Xist RNA localization to the Xi territory, they interact with Xist independent of each other (Fig. 4E).

Although much recent attention has been focused on Xist-interacting proteins that are required for transcriptional repression (2, 4, 8–10, 27–29), protein factors responsible for Xist localization have been more difficult to identify. Two previous studies, however, did report Xist exon 7 as an important domain for the spreading and localization process (30, 31). Our present study agrees with the published work on the importance of exon 7 and further identified Repeat E as a critical motif within exon 7. The function of Repeat E has also been analyzed in a recent study of XCI during female ES cell differentiation (32). While our manuscript was in preparation, CIZ1 was also identified by another group as being critical for Xist localization (33). Although there is general agreement on the role of CIZ1, several findings distinguish our study from theirs. For one, whereas our deletional analysis pinpoints the proximal half of Repeat E as being more essential for CIZ1 interaction, Ridings-Figueroa et al. (33) observed that CIZ1 is recruited primarily through the distal half of Repeat E. One possible cause of this difference could be use of inducible Xist deletion transgenes versus endogenous Xist deletions. Regardless, our study furthermore suggests a direct interaction between CIZ1 and Repeat E in vivo, using UV-RIP. This interaction is apparently also critical for the downstream deposition of the H3K27me3 repressive mark on Xi chromatin through Xist being properly localized. The intensity of IF signals and EGFP signals in the CIZ1-EGFP knock-in cell line suggests that although the number of Xist and CIZ1 clusters (puncta) is similar on the Xi (Fig. 1D), the actual molecular stoichiometry of CIZ1 to Xist may exceed one-to-one, with the highly repetitive nature of Repeat E enabling multiple CIZ1 proteins to bind a single Xist transcript. Repeat E is unique to Xist RNA and may provide a high-avidity platform for CIZ1 binding that would be found nowhere else in the transcriptome. A recent study of Xist secondary structure in vivo versus ex vivo showed that the Repeat E region’s accessibility is highly altered in the cellular environment (34), supporting our idea of superstoichiometric binding of CIZ1 protein to this region of Xist RNA in vivo. It may be surprising that CIZ1 mutant mice are viable; however, they have a predisposition toward lymphoproliferative disorders (33, 35), consistent with a loss of Xist function and XCI in blood cells (13, 36). Future work will be directed at a molecular understanding of how CIZ1-mediated Xist localization affects the Xi heterochromatin and gene expression state.

Materials and Methods

Details are found in SI Materials and Methods, which includes detailed methods for cell culture, identification of CIZ1, generation of CIZ1 antibody, Xist oligo preparation, RNA FISH, X chromosome paint, immunofluorescence, microscopy, STORM imaging and analysis, antibodies, LNA transfection, UV-RIP, and generation of knock-in and KO cell lines using CRISPR/Cas9 (37, 38).

SI Materials and Methods

Cell Culture.

Transformed MEF cells were grown on glass coverslips. ES cells were grown on γ-irradiated MEF feeders for 3 d before differentiation. For differentiation, ES colonies were trypsinized and feeders removed. ES cells were allowed to differentiate in suspension, forming embryoid bodies (EBs), for 4 d. EBs were settled down on gelatin-coated coverslips at day 4 and allowed to further differentiate until harvesting.

Identification of CIZ1.

MEF cells were washed in PBS, preextracted in CSK/0.5% Triton X-100 on ice for 5 min and washed again in CSK on ice for an additional 5 min. Cells were crosslinked in 1.5% formaldehyde for 10 min, washed in PBS, resuspended in PBS/0.01% Tween-20, and lysed by sonication using a Covaris S2. ASH2L antibody (A300-107A; Bethyl Laboratories) was conjugated to Protein G Dynabeads (Life Technologies) by 20 mM dimethylpimelimidate in 0.1 M sodium tetraborate and quenched in 0.1 M ethanolamine. Cell lysate was incubated with antibody-bead conjugation overnight at 4 °C. Pellet was washed three times with RIPA buffer (50 mM Tris at pH 8.0, 500 mM NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS), for 10 min at room temperature. To eliminate proteins bound nonspecifically to beads and/or antibody through precipitated DNA, the IP was treated with TurboDNase (Life Technologies) for 1 h at 37 °C and washed again in RIPA buffer followed by PBS. Proteins were eluted from beads in 0.2 M glycine for 5 min at room temperature, and pH was neutralized by addition of Tris at pH 8.0. After briefly running the eluate on a PAGE gel, a small section was excised and proteomic identification was performed (Taplin Mass Spectrometry Facility, Harvard Medical School). cDNA of candidate proteins was cloned into a mammalian EGFP expression vector (a gift from David Spector, CSHL) The subcellular localization of EGFP fusion proteins was assessed in HEK293FT cells after calcium phosphate transfection of plasmid.

Generation of CIZ1 Antibody.

A fragment of murine CIZ1 cDNA (corresponding to amino acids 324–538) was cloned into pET28a vector in frame with a C-terminal 6× His tag. The plasmid was introduced into BL21(DE3) bacteria. Fifty microliters of overnight cultured bacteria was inoculated into 500 mL LB medium and allowed to grow at 37 °C for 2 h. CIZ1(324-538)-His was induced by adding IPTG to 0.5 mM for 3 h at 37 °C. The bacterial pellet was resuspended in denaturing lysis buffer (100 mM NaH₂PO₄, 10 mM Tris, 6 M Guanidine). After brief sonication on Sonifier S-250 (Branson), the lysate was incubated at room temperature for 1 h. The insoluble fraction was removed by centrifugation, and imidazole (final 20 mM) was added to the supernatant to inhibit nonspecific binding of proteins during capture. Ni-NTA beads (Qiagen) were then added and allowed to bind for 2 h at 4 °C. Beads were washed with lysis buffer containing 20 mM imidazole. Bound protein was eluted in lysis buffer containing 250 mM imidazole and dialyzed against PBS overnight, after which an aliquot was sent to Cocalico Biologicals for immunization of rabbit host. For affinity purification of antibody, an overlapping fragment of murine CIZ1 cDNA (corresponding to amino acids 302–538) was cloned into pET28a vector in frame with an N-terminal MBP tag and purified using amylose resin (New England Biolabs), as per manufacturer’s instructions. MBP-CIZ1(302-538) was later conjugated to Affi-Gel 10 (Bio-Rad) as per manufacturer’s instructions. Rabbit serum was incubated with MBP-CIZ1(302-538)/bead conjugate in PBS at 4 °C overnight. After extensive washes with PBS, antibody was eluted in 0.2 M glycine (pH 2.0)/500 mM NaCl and immediately neutralized by addition of 1.5 M Tris (pH 8.0). Antibody was dialyzed against PBS/50% glycerol at 4 °C overnight and stored at −20 °C.

Xist Oligo Probe Preparation.

Xist oligo probes were designed using Primer3 (frodo.wi.mit.edu/primer3/) and synthesized by Integrated DNA Technologies. Amine-ddUTP (Kerafast) was added to 2 pmol pooled oligos by terminal transferase (New England Biolabs) at 37 °C for 4 h. Oligos were purified by phenol/chloroform extraction and concentrated by ethanol precipitation. They were resuspended in 0.1 M sodium borate and labeled with Cy3B (GE Healthcare) or Alexa647 NHS-ester (Life Technologies) at room temperature overnight. After ethanol precipitation of oligos, labeling efficiency was evaluated by absorbance using Nanodrop (Thermo Fisher Scientific).

RNA FISH.

RNA FISH was performed as previously described (17) with minor modifications. Briefly, cells grown on glass coverslips were rinsed in PBS and fixed in 4% paraformaldehyde. After permeabilization in 0.2–0.5% Triton X-100 at room temperature, cells were washed in PBS and dehydrated in a series of increasing ethanol concentrations. Depending on experiment, 3–200 labeled oligo probes were added to hybridization buffer containing 25% formamide, 2× SSC, 10% dextran sulfate, and 1 mg/mL yeast tRNA. RNA FISH was performed in a humidified chamber at 42 °C for 3–4 h. After being washed three times in 2× SSC, cells were mounted for wide-field fluorescent imaging or dehydrated for STORM imaging. Nuclei were counterstained with Hoechst 33342 (Life Technologies).

RNA FISH/X Chromosome Paint.

As described earlier, 1 × 10⁵ cells were cytospun onto glass slides and RNA FISH performed. Slides were mounted and images captured with positions recorded. After imaging RNA signal, coverslips were carefully removed and slides rinsed first in PBS/0.02% Tween-20 and then in PBS to remove mounting medium, treated with RNase A (400 μg/mL in PBS) at 37 °C for 40 min to remove RNA signal, and denatured for DNA FISH in 70% formamide, 2× SSC at 80 °C for 15 min. Slides were quenched in ice cold 70% ethanol and dehydrated in a series of increasing ethanol concentrations, and 1:10 (vol/vol) XMP mouse chromosome paint probe (d-1420–050-FI; MetaSystems) was added to hybridization buffer containing 50% formamide, 2× SSC, 10% dextran sulfate, and 0.2 mg/mL mouse Cot-1 DNA. DNA FISH was performed in a humidified chamber at 42 °C overnight. After being washed three times in 2× SSC, slides were remounted and reimaged at recorded positions. Nuclei were counterstained with Hoechst 33342 (Life Technologies).

Immunofluorescence/RNA FISH.

Immunofluorescence/RNA FISH was performed as previously described (17), with minor modifications. Briefly, cells grown on glass coverslips were rinsed in PBS, fixed in 4% paraformaldehyde, and permeabilized in PBS/0.5% Triton-X 100 for 10 min at room temperature. After being blocked for 20 min in 1% BSA supplemented with 10 mM VRC (New England Biolabs), primary antibodies were added and allowed to incubate at room temperature for 1 h. Cells were washed three times in PBS/0.02% Tween-20. After incubating with secondary antibody for 30 min at room temperature, cells were washed again in PBS/0.02% Tween-20. Cells were fixed again in 4% paraformaldehyde and dehydrated in a series of increasing ethanol concentrations. RNA FISH was performed using a pool of Xist tiling oligonucleotides for 3–4 h at 42 °C in a humidified chamber. After being washed three times in 2× SSC, cells were mounted for wide-field fluorescent imaging or dehydrated for STORM imaging. Nuclei were counterstained with Hoechst 33342 (Life Technologies).

Microscopy.

For wide-field fluorescent imaging, cells were observed on a Nikon 90i microscope equipped with 60×/1.4 N.A. VC objective lens, Orca ER CCD camera (Hamamatsu), and Volocity software (Perkin-Elmer). As previously described (17), STORM imaging was performed on an N-STORM (Nikon) equipped with 100×/1.4 N.A. λ objective lens, ion ×3 EM CCD camera (Andor), and three laser lines (647, 561, and 405 nm). Imaging buffer containing 147 mM βME and GluOX (Sigma) was used to promote blinking and reduce photo-bleaching. For 3D STORM imaging, cylindrical lens was inserted into the optical path to introduce astigmatism. Z calibration was performed using 100 nm TetraSpeck beads (Life Technologies). For two-color STORM imaging, sequential imaging with appropriate emission filters (Cy5 em filter for Alexa 647 and Cy3 em filter for Cy3B) was adapted to suppress any crosstalk. N-STORM module in Element software (Nikon) was used to control microscopes, acquire images, and perform 2D and 3D STORM localizations. Analysis of STORM images was performed as previously described, using in-house Matlab (Mathworks) scripts (17). Briefly, after STORM localizations were binned into each pixel position, Xist cloud was automatically identified using kmeans function. Nearest neighbor’s distance was calculated between Xist localizations within the cloud and any nearest CIZ1/ASH2L localizations. For randomized control calculation of nearest neighbors, CIZ1/ASH2L localizations were randomized within the Xist cloud, and nearest neighbor’s distance was again calculated. For randomization of colocalization analysis between Xist and CIZ1 in HNRNPU KO cells, ImageJ with JACoP plug-in was used.

Antibodies.

The following primary antibodies were used: rabbit H3K27me3 (AM39155; Active Motif and GTX60892; GeneTex), mouse HNRNPU (sc-32315; Santa Cruz), rabbit GFP (ab290; Abcam), rabbit ASH2L (A300-107A; Bethyl Laboratories), rabbit ASH2L (AM39100; Active Motif), and mouse ASH2L (ab50699; Abcam). Dye-conjugated secondary antibodies were purchased from Life Technologies. For two-color STORM, Cy3B was conjugated to anti-rabbit secondary antibody (Jackson ImmunoResearch), using NHS-Cy3B (GE Healthcare) as previously described (17).

LNA Transfection.

MEF cells were transfected as previously described (26). Briefly, 2 × 10⁶ MEF cells were resuspended in 100 μL Nucleofector MEF II solution (Lonza) containing 2 μM Xist anti-Repeat C LNA. After electroporation was performed as per manufacturer’s instructions, cells were allowed to settle on 0.2% gelatin-coated glass coverslips. Complete media were added and cells cultured normally until harvest at various points.

UV-RIP (UV Crosslinked RNA Immunoprecipitation).

UV-RIP was performed as previously described, with minor modifications (12). One 15-cm plate of female MEF cells per IP was UV-crosslinked at 254 nm (200 mJ/cm²) in 5 mL cold PBS and collected by scraping. Cells were incubated in nuclear isolation buffer (10 mM Hepes at pH 7.5, 1.5 mM MgCl2, 10 mM KCl, 0.5 mM DTT) for 30 min on ice. Nonidet P-40 was added to 0.1% for 10 min on ice. Cells were incubated in lysis buffer (0.5% Nonidet P-40, 0.5% sodium deoxycholate, 200 U/mL RNase Inhibitor [Roche], and protease inhibitor mixture [Roche] in PBS at pH 7.4) at 4 °C for 25 min with rotation, followed by DNase treatment (30 U of TURBO DNase, 15 min at 37 °C). SDS was added to 0.1% and lysate was further incubated at 4 °C for 15 min. After centrifugation, the supernatant was incubated with 2 mg GFP or 1 mg CIZ1 antibodies for 2 h at 4 °C. Ten to 20 μL Dynabeads Protein G was added for 1 h at 4 °C. Beads were washed three times with 1× PBS supplemented with 1% Nonidet P-40, 0.5% sodium deoxycholate, and additional 350 mM NaCl (total 500 mM Na⁺), and DNase-treated (10 U) for 30 min at 37 °C. Beads were washed three times with the same wash buffer supplemented with 10 mM EDTA, followed by three washes with low-salt buffer (50 mM Tris at pH 7.5, 50 mM NaCl, 10 mM EDTA, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS). Crosslinked proteins were digested in 100 mM Tris⋅HCl at pH 7.4, 50 mM NaCl, 10 mM EDTA, 0.1 mg/mL Proteinase K (Roche), and 0.5% SDS for 30 min at 55 °C. RNA was recovered using TRIzol (Life Technologies).

Generation of Knock-In and KO Cell Lines.

All guide RNAs were designed using tools available online (www.rgenome.net/). CIZ1 KO in MEF and CIZ1-EGFP knock-in guide RNA (gRNA) sequences were cloned into “gRNA Cloning Vector” and introduced into cells alongside vector expressing hCas9_D10A (37). All other gRNA sequences were cloned into pSpCas9(BB)-2A-GFP or pSpCas9(BB)-2A-Puro (38). Guide RNA/Cas9 plasmid delivery into MEF cells was performed using Nucleofector MEF II solution (Lonza), as per manufacturer’s instructions; delivery into ES cells was by electroporation (Bio-Rad), delivery into HEK293FT cells was by calcium phosphate transfection. After delivery of guide RNA/Cas9 plasmids, cells were cultured for 5–8 d before clonal selection. For CIZ1-EGFP knock-in, an EGPF-3×HA insertion cassette flanked by 1-kb homologous groups was cloned into pBluescript II vector and included with the guide RNA/Cas9 plasmid delivery. The following guide RNA sequences were used (PAM sequences in italics):

CIZ1KOs and knock-ins:

• CIZ1 KO in MEF (a pair of 17-nt short gRNAs):

• Left: GGGGTGACCATCTGGGG TGG

• Right: GCAGCAGTTCTTTCCCC AGG

• CIZ1-EGFP knock-in in MEF (a pair of 17-nt short gRNAs):

• Left: GAGCGCCGAAGGGGAGG AGG

• Right: GCCTCAAAACCTGATAG AGG

• CIZ1 KO in mouse ES cell:

• CAGCCTTACACCACCCCAGA TGG

• CIZ1 KO in HEK293FT:

• GCCACTCGCCAGTCCTTGCT GGG

• HNRNPU KO in MEF:

• CTCGGGAGCGGGCCTAGAGC AGG

Xist exon 7 deletions:

• Clones ΔEx7-10 and ΔEx7-22:

• Upstream: TTCATCTTCCATCGTGTACA TGG

• Downstream: ACAAGATGGCGTCTGTAACT TGG

• Clones ΔRepE-4 and ΔRepE16:

• Upstream: AGAATTAGACACACAGACCA AGG

• Downstream: ATACATCATTCCGTCCGGTC AGG

• Clone 7a-11:

• Upstream: ATACATCATTCCGTCCGGTC AGG

• Downstream: TTACGTATTCATACGTTTCC TGG

• Clone 1–2:

• Upstream: AGAATTAGACACACAGACCA AGG

• Downstream: AGAACATGCAGGAGAAACAT GGG

• Clones 3–9 and 3–16:

• Upstream: ATTCTAAAGTAATCCTTTCT TGG

• Downstream: ATGGCCCGTCTTGAGCTGGC TGG