Xist Exon 7 Contributes to the Stable Localization of Xist RNA on the Inactive X-Chromosome

To equalize X-linked gene dosage between the sexes in mammalian females, Xist RNA inactivates one of the two X-chromosomes. Here, we report the crucial function of Xist exon 7 in X-inactivation. Xist exon 7 is the second-largest exon with a well-conserved repeat E in eutherian mammals, but its role is often overlooked in X-inactivation. Although female ES cells with a targeted truncation of the Xist exon 7 showed no significant differences in their Xist expression levels and RNA stability from control cells expressing wild-type Xist, compromised localization of Xist RNA and incomplete silencing of X-linked genes on the inactive X-chromosome (Xi) were observed in the exon 7-truncated mutant cells. Furthermore, the interaction between the mutant Xist RNA and hnRNP U required for localization of Xist RNA to the Xi was impaired in the Xist exon 7 truncation mutant cells. Our results suggest that exon 7 of Xist RNA plays an important role for stable Xist RNA localization and silencing of the X-linked genes on the Xi, possibly acting through an interaction with hnRNP U.


Introduction
In eukaryotes, an overwhelming majority of genomes are transcribed as non-protein-coding RNAs (ncRNAs) [1,2]. One of the major classes of ncRNAs is long ncRNAs (lncRNAs), which vary in length from a few hundred bases to tens of kilobases. A number of lncRNAs play an important role in transcriptional regulation through their interaction with chromatinmodifying enzymes, which direct them to specific target genes [3,4]. LncRNAs are also known to be involved in various biological processes such as the regulation of the cell cycle [5], cellular differentiation and development [6], the regulation of metabolism [7] and disease pathogenesis [8,9].
X inactive-specific transcript (Xist) RNA is one such lncRNA, which regulates chromatin organization and transcriptional gene silencing in one of the two X-chromosomes to equalize the X-linked gene dosage between males and females [10]. In the epiblast lineage, either the paternal or maternal X is randomly inactivated; this is referred to as random X-inactivation. In random X-inactivation, several non-coding genes in the X-inactivation center (Xic) are known to interact to initiate X-inactivation in one of the two X-chromosomes in females [11,12]. Xist RNA is exclusively expressed from the Xic of the future inactive X-chromosome (Xi) at the onset of X-inactivation and has a pivotal role in initiating X-inactivation in cis [13]. Highly expressed Xist RNA coats the entire length of Xi [14] and recruits silencing factors such as the polycomb repressive complex 2 (PRC2) of lysine methyltransferase for H3K27me3 to the Xi [15,16]. Xist RNA induces gene silencing on the Xi by a cascade of epigenetic modifications that are maintained through multiple rounds of cell division [17]. There is evidence that the overall X-inactivation can be maintained in the absence of XIST/Xist [18,19]; however, recent evidence has shown that Xist deletion from the Xi induces the partial de-repression of the Xlinked genes [17,20]. Indeed, a recent paper has shown that the depletion of Xist in murine hematopoietic stem cells after the establishment of X-inactivation leads to a genome-wide aberration in gene expression, especially in the expression of X-linked genes, and an induction of highly aggressive myeloproliferative neoplasm and myelodysplastic syndrome in a female-specific manner [21]. This finding suggests a critical role of Xist in the maintenance phase of Xinactivation to prevent cancer transformation and progression. Therefore, proper regulation of Xist is critical in both the initiation and maintenance phases for cell survival, cellular differentiation and development, and the prevention of cancer pathogenesis in mammalian species.
Xist RNA has multiple functional domains and directly or indirectly interacts with various proteins such as transcription factors, chromatin modifying enzymes and scaffold proteins [22]. Comprehensive functional analysis using a series of deletions of Xist RNA based on the inducible Xist transgene has identified the functional domain of Xist RNA for gene silencing, and broad redundant region for Xist RNA localization on the Xi and formation of macrochromatin bodies (MCB) associated with histone variant macroH2A1 [23]. This approach successfully demonstrated that repeat A of the 5´region of Xist RNA is crucial for X-linked gene silencing and that the redundant region contributes to stable Xist RNA localization on the Xi. Although PRC2 binds promiscuously to a variety of RNA molecules, PRC2 exhibits preferential binding to repeat A of the Xist RNA in vivo and in vitro [24][25][26][27][28][29][30][31]. Several reports have indicated that Xist-specific repeat motifs contained within exon 1 and conserved in various eutherian mammalian species are involved in the localization of Xist RNA on the Xi [32][33][34]. Transcriptional factor YY1 interacts with both Xist RNA and the Xist gene body through repeat C of Xist RNA and YY1 binding sites near repeat F in Xist, respectively; it is believed to anchor the Xist RNA to the Xist gene on the Xi as a nucleation center for Xist RNA spreading [34]. Moreover, the interaction between Xist RNA and the Xi can be blocked by targeting Xist repeat C with peptide nucleic acids (PNAs) or locked nucleic acids (LNAs). This fact contrasts with findings that a lack of repeat C does not affect the localization of Xist RNA on the Xi of the transgene assay [23,32,33].
Heterogeneous nuclear ribonucleoprotein U (hnRNP U; also known as SAF-A or SP120) has an essential role for Xist RNA localization on the Xi [35]. Interestingly, hnRNP U accumulation on the Xi is also dependent on the Xist RNA, indicating an interdependent relationship between Xist RNA and hnRNP U for their specific localization on the Xi [35,36]. hnRNP U was originally identified as a nuclear matrix-or scaffold-attachment region (MAR or SAR)associated protein [37][38][39]. It is known to have a wide variety of functions such as gene expression, telomere regulation and nuclear organization [40][41][42][43]. Because hnRNP U has a unique molecular structure with a SAF-Box MAR binding domain and an arginine-glycine-glycine (RGG) RNA-binding domain at its N-terminal and C-terminal, respectively [44,45], hnRNP U has been proposed to bridge Xist RNA and the Xi [35].
While most Xist RNA functional domains for RNA localization and gene silencing are mapped within exon 1 of Xist RNA, the function of exon 7 in Xist RNA is unknown in spite of the presence of a well-conserved repeat E among the eutherian mammals [46]. To elucidate the role of the Xist exon 7 in X-inactivation, we generated Xist exon 7-truncated mutant ES cells by inserting tandem polyadenylation (tpA) signals. In this paper, we report our novel findings regarding the interplay between the Xist exon 7 and hnRNP U for stable localization of Xist RNA on the Xi and X-linked gene silencing.

Results
A truncation mutation of Xist RNA at the end of exon 6 does not affect the expression or stability of Xist RNA during X-inactivation To investigate the role of exon 7 in mouse Xist RNA, we inserted two tandem polyadenylation signals (2xtpA) into the end of exon 6 of Xist coupled with splice acceptor (SA)-internal ribosomal entry site (Ires)-hygromycin (Hyg)-pA for Tsix truncation to induce non-random Xinactivation of the mutant X. This resulted in the simultaneous truncation of both Xist and Tsix (Xist delEx7 Tsix TST6 ) ( Fig 1A and S1 Fig). We used the 16.7 mouse female ES cell line carrying one Mus musculus 129SvJ (129) X-chromosome and one Mus castaneous (Cast) X, which enabled us to perform allele-specific analysis based on single nucleotide polymorphisms (SNPs) between these mouse strains [47,48]. As a control, we also established a Tsix truncation mutant by inserting SA-Ires-Hyg-pA at Xist intron 6, resulting in Tsix TST6 female ES cells ( Fig  1A and S1 Fig). Allele-specific quantitative RT-PCR (RT-qPCR) analysis of the Tsix expression revealed that the 129 X-chromosome was targeted by the truncation mutation ( Fig 1B). While the Tsix expression upstream of the truncation site could be detected by using both the 129 and Cast allele-specific primer sets (T1) in control Tsix TST6 and two Xist delEx7 /Tsix TST6 mutant cells, the Tsix transcript was efficiently truncated downstream (T2, Tsix 129 exon 4) of the pA insertion site. These mutant ES cells allow us to address the effect of the Xist mutation on Xinactivation because the Tsix mutation caused a non-random inactivation of the mutant 129 X-chromosome [47].
First, to determine if the 2xtpA insertion was sufficient for the truncation of the Xist RNA, we performed 129 allele-specific RT-qPCR for the Xist exon 7 (Fig 1C). In two independently isolated Xist delEx7 Tsix TST6 mutant clones, the Xist exon 7 on the 129 Xi was found to not be expressed upon ex vivo differentiation. This suggests that the insertion of the 2xtpA cassettes efficiently truncates the Xist transcripts during X-inactivation. Next, to investigate the Xist expression in the Xist delEx7 Tsix TST6 mutant cells upon differentiation, the 129 allele-specific RT-qPCR was carried out using a primer set designed to amplify Xist exon 1 through exon 3 Role of Exon 7 of Xist RNA in X-Inactivation ( Fig 1D). This RT-qPCR showed that Xist expression was upregulated in both Tsix TST6 and Xist delEx7 Tsix TST6 mutant cells upon differentiation, comparable to the Xist expression level in wildtype female cells. We also examined the ratio of Xist RNA expressed from mutant 129 Xi to that from Cast wild-type active X-chromosome (Xa) by allele-specific RT-PCR analysis using a polymorphic restriction enzyme digestion with wild-type, control Tsix TST6 and Xist de-lEx7 Tsix TST6 mutant cells (Fig 1E and 1F). As expected, due to the Tsix mutation, the Xist was nearly exclusively expressed in the 129 X-chromosomes of both the Tsix TST6 and Xist delEx7 Tsix TST6 mutant cells in contrast to the wild-type cells (70-80% of the 129 X was chosen as the Xi) [48]. Furthermore, because the Xist truncation mutant deletes approximately 40% of full length Xist RNA (7.5 kb out of 17.8 kb), we investigated whether an Xist truncation mutation lacking exon 7 affects the stability of the RNA by measuring the half-lives of 5-ethynyl uridine (EU) pulse-labeled Xist RNAs (Fig 1G). The half-life assay identified no significant differences in the Xist RNA stability between the control and two mutant clones. The half-life of the fulllength wild-type Xist RNA in Tsix TST6 ES cells and the exon 7-truncated mutant Xist RNA in the two Xist delEx7 Tsix TST6 ES cells was approximately 2 hours. This half-life shorter than that observed in previous reports, which found a half-life of~3.5 hours in the Xist RNA half-life assay of the Tsix knockout mutant female ES cells using RNA polymerase II inhibitor. This differences can likely be explained by the different approaches we used [49]. Similar to the halflife of the wild-type and exon 7-truncated mutant Xist RNA in ES cells, the half-life of the wildtype and mutant Xist RNA in EB upon differentiation on day 8 was approximately 2~2.5 hours (Fig 1G). These results suggest that the truncation of the Xist exon 7 does not alter the Xist expression or the stability of the Xist RNA during X-inactivation. Because a shorter isoform (Sisoform) of the Xist RNA lacks a large region of exon 7 in a long isoform (L-isoform) except for a 1.1 kb repeat E and 0.5 kb region of its 3´end by alternative splicing, the S-isoform Xist RNA is similar to the mutant Xist delEx7 RNA lacking the entire exon 7. We examined whether the Land S-isoforms Xist RNA have a different stability. The L-isoforms of the Xist RNA exhibited a slightly longer half-life in wildtype 16.7 and cells TsixTST6 mutant (S2 Fig), suggesting that alternative splicing might slightly affect the stability of Xist RNA.
Because the Xist expression is tightly linked to the differentiation status, we also performed RT-qPCR analysis of the pluripotent cell markers, Nanog and Oct3/4, to clarify whether the Xist truncation mutation influences the ES cell differentiation (S3A Fig). RT-qPCR showed that the Nanog and Oct3/4 expression levels in both the control Tsix TST6 and Xist delEx7 Tsix TST6 mutant cells were similarly decreased as the embryoid body (EB) differentiation progressed. Moreover, the outgrowth of the Xist delEx7 Tsix TST6 mutant cells was similar to that of the control Tsix TST6 mutant cells at day 8 upon differentiation, with a slightly slower growth (S3B Fig). These results suggest that EB differentiation was not significantly affected by the Xist exon 7 truncation mutation.
Xist exon 7 is essential for the stable localization of Xist RNA on the Xi Having shown that neither the control Tsix TST6 nor the Xist delEx7 Tsix TST6 mutation affected the expression of Xist or the stability of Xist RNA and that neither mutant exhibited severe growth defects in EB differentiation during X-inactivation, we next examined the chromosome-wide gene silencing induced by the mutant Xist RNA during X-inactivation. We performed RNA fluorescence in situ hybridization with immunofluorescence (immuno-RNA FISH) to observe the accumulation of Xist RNA and a hallmark of facultative heterochromatin marker, H3K27me3, on the Xi upon differentiation. In the Tsix TST6 mutant cells, the number of robust (strong) Xist RNA cloud-and H3K27me3-positive cells gradually increased upon differentiation. This is greater than that in the wild-type 16.7 ES cells at each time point, likely due to faster Xist induction by the Tsix mutation in Tsix TST6 mutant cells [48]. In contrast to the strong focal staining of the Xist RNA and H3K27me3 in the control Tsix TST6 mutant cells, the immuno-RNA FISH results in the Xist delEx7 Tsix TST6 mutant cells indicated a significantly reduced number of Xist RNA cloud-and H3K27me3-positive cells, despite the comparable levels of Xist expression between the mutant and control Tsix TST6 cells (Figs 1D and 2). Instead of strong focal Xist RNA clouds, the Xist delEx7 Tsix TST6 mutant cells often exhibited a faint (weak) Xist signal. While the Tsix TST6 mutant cells exhibited strong Xist clouds associated with H3K27me3 staining in approximately 65~70% of the cells on both day 8 and day 12, the colocalization of strong Xist clouds and H3K27me3 in the Xist delEx7 Tsix TST6 mutant cells was less than 20% and 14% on days 8 and 12, respectively. The difference in terms of the percentages of the Xist cloud-positive nuclei between the Tsix TST6 and Xist delEx7 Tsix TST6 mutant cells became more significant as differentiation progressed (Fig 2B and 2C). Notably, the total percentage of the Xist cloud-positive (strong and weak) cells in the Xist delEx7 Tsix TST6 mutant cells on days 8 and 12 were significantly decreased from 45% to 24% (clone #1) and from 37% to 19% (clone #2), respectively. In contrast, the Tsix TST6 cells maintained a high percentage of Xist cloud-positive cells (over 75% Xist cloud-positive) during the transitional period from day 8 to day 12. Our results suggest that exon 7 of the Xist RNA is essential for stable Xist RNA localization, especially in the maintenance of Xist RNA localization and associated H3K27me3 modifications on the Xi.

Xist exon 7 is required for the silencing of X-linked genes during Xinactivation
Intriguingly, the Xist delEx7 Tsix TST6 mutant clones exhibited impaired localization of the Xist RNA and H3K27me3 on the Xi. This observation prompted us to investigate whether the compromised Xist RNA and H3K27me3 localization on the Xi affects the expression status of Xlinked genes on the Xi during EB differentiation using two different approaches (Fig 3): RT-qPCR analysis using allele-specific primers ( Fig 3A) and allele-specific RT-PCR using a polymorphic restriction enzyme digestion (Fig 3B and 3C). The allele-specific RT-qPCR analysis using 129 allele-specific primer sets can discriminate the Pgk1 and Mecp2 expression on the mutant 129 X-chromosome from their expression on the Cast X-chromosome. Thus, we can only detect Pgk1 and Mecp2 expression on the mutant 129 X-chromosome. In the control Tsix TST6 EB cells on days 8 and 12 of differentiation, the X-linked Mecp2 and Pgk1 expression from the 129 Xi gradually decreased upon differentiation and remained less than 13% and 12% compared to that of the undifferentiated ES cells (day 0), respectively. In contrast to the control cells, the mutant 129 allele-specific RT-qPCR for Mecp2 and Pgk1 in the Xist delEx7 Tsix TST6 mutant clones showed that the partial silencing of Mecp2 (40% and 25% in clones #1 and #2, respectively) and Pgk1 (55% and 47% in clones #1 and #2, respectively) occurred on day 8 instead of on day 0. The partial silencing of Mecp2 and Pgk1 in the Xist delEx7 Tsix TST6 mutant cells was then followed by the reactivation of Mecp2 (98% and 89% in clones #1 and #2, respectively) and Pgk1 (64% and 58% in clones #1 and #2, respectively) on day 12 upon differentiation, although the initial phase of silencing in Mecp2 and Pgk1 on day 4 was similar between the Tsix TST6 and Xist delEx7 Tsix TST6 mutant cells. Additionally, to confirm the compromised Mecp2 and Pgk1 expression from the 129 Xi relative to that of the wild-type Cast Xa in the Xist-delEx7 Tsix TST6 mutant cells, allele-specific RT-PCR analysis based on a polymorphic restriction enzyme digestion was performed (Fig 3B and 3C). The ratios of the X-linked gene expression from the 129 Xi were markedly reduced from differentiation day 8 in the Tsix TST6 cells. However, in the mutant clones, the ratios of X-linked gene expression from the 129 Xi remained high during the late stages of X-inactivation. Combined with the data in Fig 2, these results suggest that the Xist delEx7 Tsix TST6 mutant Xist RNA can partially induce the silencing of Xlinked genes, but cannot sustain transcriptional repression on the Xi. This is likely due to the loss of the Xist RNA clouds and the H3K27me3 on the Xi.

hnRNP U directly interacts with exon 1 and 7 regions of Xist RNA
We previously reported that the nuclear matrix protein hnRNP U directly interacts with the Xist RNA via the RGG RNA-binding domain and is involved in the chromosomal localization of Xist RNA along the entire Xi region [35]. To further investigate the interaction between the Xist RNA and the hnRNP U protein across the Xist gene in detail, especially in exon 7 of Xist RNA, UV-crosslinking RNA-immunoprecipitation (UV-crosslinking RIP) analysis was carried out using Neuro2a cells as previously reported (Fig 4A). Multiple primer pairs were designed for the quantitative analysis to examine the binding affinity of hnRNP U to specific regions of the Xist RNA. The UV-crosslinking RIP showed that exons 1 and 7 of the Xist RNA could be significantly co-immunoprecipitated with the FLAG-hnRNP U, unlike the control 7SK, U2, beta-actin and Gapdh RNA. Within exon 1 of the Xist RNA, the second half of the exon (primer positions 4-7) exhibited preferential binding with hnRNP U over the first half of the exon (primer positions 1-3). Interestingly, the hnRNP U also bound to exon 7 of the Xist RNA with as high affinity at primer positions 13-18, with the exception of the repeat E region at primer positions 10-12. We also explored how hnRNP U interacted with the XIST RNA in human cells. Consistent with the data showing that the knockdown of human hnRNP U by siRNA induces the dissociation of XIST RNA from the Xi [43], we confirmed that the hnRNP U knockdown in human HEK293T cells led to the disperse localization of histone macroH2A, which accumulates on the Xi in an Xist RNA-dependent manner (S4 Fig) [19]. These data suggest that hnRNP U is essential for XIST RNA localization on the Xi in both humans and mice.
were counted and classified based on the Xist RNA and H3K27me3 signal. Robust and faint Xist RNA FISH signals were classified as "strong" and "weak" Xist RNA, respectively. (C) The graph shows the mean ± SD of nuclei with Xist RNA clouds from three independent experiments. P-values were derived from an unpaired t-test between control Tsix TST6 and Xist delE7 Tsix TST6 cells on the same day upon differentiation (*p<0.05, **p<0.01, ***p<0.001).
doi:10.1371/journal.pgen.1005430.g002 Similar to the interaction between mouse Xist RNA and hnRNP U, the UV-crosslinking RIP analysis revealed that the human hnRNP U preferentially bound to XIST RNA broadly in the second half of the first exon and the last exon, except for the repeat E region (Fig 4B). Overall, our results suggest that exon 7 of the Xist RNA could contribute to the stable localization of Xist RNA on the Xi through interaction with hnRNP U.
The Xist exon 7 deletion alters the interaction between Xist RNA and hnRNP U Next, to investigate whether the unstable Xist RNA localization and X-linked gene silencing on the Xi observed in the Xist delEx7 Tsix TST6 mutant cells was due to the disturbed interaction between hnRNP U and the Xist RNA lacking exon 7, we performed UV-crosslinked RIP using Tsix TST6 and Xist delEx7 Tsix TST6 mutant ES cell lines expressing a FLAG-HA epitope tag hnRNP U. To establish the ES cell lines expressing the FLAG-HA tagged hnRNP U from the endogenous hnRNP U loci, we used the CRISPR/Cas system. With this system, we obtained 2 and 3 FLAG-HA biallelic knock-in ES cell lines derived from the Tsix TST6 and Xist delEx7 Tsix TST6 mutant ES cells, respectively (Fig 5A and 5B). By RT-qPCR and western blotting using anti-FLAG and anti-hnRNP U antibodies, we confirmed that the FLAG-HA knock-in ES cell lines expressed similar levels of FLAG-HA-hnRNP U to those of the parental ES cell lines (the Tsix TST6 and Xist delEx7 Tsix TST6 mutant ES cell lines) on day 12 upon differentiation (Fig 5C-5E). We also confirmed that the FLAG-HA knock-in at the endogenous hnRNP U did not alter the kinetics of Xist expression upon differentiation and X-linked gene silencing compared to the parental ES cell lines (Figs 1D and 3A, S5A and S5B Fig). Furthermore, upon differentiation on day 12, the Tsix TST6 and Xist delEx7 Tsix TST6 mutant EB cells expressing the FLAG-HA-hnRNP U from the endogenous loci exhibited a similar proportion of Xist RNA-and H3K27me3-positive cells to those observed in their parental cell lines (Fig 2 and S5C Fig).
Using EB cells expressing FLAGHA-hnRNP U at day 12 upon differentiation, UV-crosslinking RIP analysis followed by quantitative RT-PCR of the Xist RNA was carried out ( Fig  5F). Tsix TST6 cells expressing wildtype Xist RNA exhibited a similar pattern of hnRNP U binding across Xist RNA, as observed in the Neuro2A cells showing preferential binding of hnRNP U to the 3´regions of exon 1 and exon 7 of Xist RNA (with the exception of repeat E) (Fig 4A). Approximately 2.5-18% of the wild-type Xist RNA was precipitated by FLAG-hnRNP U immunoprecipitation at all primer pair positions in the differentiating EB of the Tsix TST6 and Xist delEx7 Tsix TST6 mutant ES cell lines, while approximately 2-25% of the Xist RNA was recovered by FLAG-hnRNP U IP at a similar position in the Neuro2A cells (Figs 4A and 5E). The slightly lower Xist RIP efficiency that was observed in the EB cells could be due to the incomplete induction of X-inactivation (less than 80% on day 12) (Fig 2B) or to insufficient UVirradiation on the inside of the sphere-shaped differentiating EB cells. Interestingly, the prominent binding of FLAG-HA-hnRNP U at the 3' region of exon 1 in Xist RNA was impaired bỹ 30% in the Xist delEx7 Tsix TST6 mutant cells compared to that of the wild-type Xist RNA in the Tsix TST6 cells (Fig 5F). These data indicate that exon 7 of Xist RNA could be required for the interaction between the Xist RNA and hnRNP U or maintenance of the Xist RNA-hnRNP U RNP complex. Overall, these data suggest that exon 7 of Xist RNA is essential for stable Xist RNA localization and X-linked gene silencing, possibly acting through an interaction with hnRNP U.

Discussion
In this study, we provided novel evidence that Xist exon 7 is necessary for the localization of Xist RNA during X-inactivation. Our results show that the aberrant localization of mutant Xist RNA lacking exon 7 on the Xi leads to the loss of the Xist cloud and the H3K27me3 modification on the Xi and impairs X-linked gene silencing due to the impaired binding affinity of Xist RNA to hnRNP U which bridges the Xist RNA and the Xi.
Previously, to examine the effects of Xist mutations on chromosome silencing and coating, tetracycline (tet)-inducible mutant Xist RNA constructs containing exon 7 specific deletions were generated and introduced into the Hprt locus of male ES cells [23]. The results showed that a large deletion of the 3' end of the Xist RNA that included exon 7 resulted in weak defects in Xist RNA localization on the Xi, X-linked gene silencing and MCB formation. In contrast to these results, our Xist truncation mutant at the endogenous locus exhibited a severe phenotype in terms of X-inactivation (Figs 2 and 3). Exon 7 truncated Xist RNA lost its ability to localize to the Xi and failed to induce or maintain X-linked gene silencing. Similar phenotypic differences were observed in the repeat C study in the Xist RNA [23,32,33]. Two independent studies The protein coding or 5´UTR regions are shown as a gray or white box, respectively. The FLAG-HA tag is shown as a red box. The first ATG sequence of the translation start is labeled in blue. The protein coding or 5´UTR regions are capitalized or lowercased, respectively. The sgRNA sequence is labeled in green. The protospacer-adjacent motif (PAM) sequence is labeled in red. (B) PCR genotyping analysis of the FLAG-HA tag knock-in hnRNP U allele. Genomic PCR using primers hnRNPU-utrF and hnRNPU-codeR amplified 212 or 284 bp PCR products from the wild-type or the FLAG-HA knock-in alleles, respectively. All FLAG-HA targeted knock-in ES cells shown in this work are homozygous knock-in clones. (C) Quantitative RT-PCR analysis of FLAGHA-hnRNP U using hnRNPU-F and hnRNPU-R in the parental and targeted ES cell lines at day 12 upon differentiation. The mean ± SEM from three independent experiments is shown. (D) A representative western blotting analysis using anti-FLAG, hnRNP U and tubulin antibodies using antisense PNA or LNA oligonucleotides to block the repeat C function of the Xist RNA revealed a crucial function of the repeat C in Xist RNA for its localization to the Xi [32,33]. On the other hand, a tet-inducible Xist transgene assay showed that the repeat C deletion in the Xist RNA affects neither the Xist RNA localization on the Xi nor the silencing of adjacent selection markers [23]. These differences might reflect differences in the systems used to elucidate the functional domain of the Xist RNA. One possibility is that the tet-inducible promoter used in the tet-inducible transgene assay might induce much higher levels of Xist expression [23]. Therefore, an abundant mutant Xist RNA could compensate for the defects of the mutant Xist RNA in the Xist RNA localization and X-linked gene silencing.
Interestingly, the tet-inducible XIST transgene assay in human cells showed that XIST RNA lacking a large part of its 3' region (from the end of the exon 1 through the last exon) displayed dispersed localization of XIST RNA on the targeted locus, yet could still induce adjacent EGFP silencing [50]. Furthermore, the deletion of the central region of XIST exon 1, including repeat C in the XIST transgene, did not affect the XIST localization [50]. Whereas human XIST RNA has only one poorly conserved repeat C in exon 1, there are 14 copies of a 120 bp repeat C in mice, which are crucial for Xist RNA function [32,33,51]. Thus, while both the first and last exons of Xist RNA are indispensable for its localization onto the Xi in mice, it is possible that the role of the second largest exon in human XIST, exon 6, might be more crucial for XIST RNA localization on the Xi than exon 1 in human cells.
To further explore the detailed mechanisms of how Xist RNA is stably anchored onto the Xi, we extended our previous approach [35] to clarify the detailed interaction between Xist RNA and hnRNP U. Because hnRNP U has an RGG RNA-binding domain required for the interaction with Xist RNA and a SAF domain for SAR/MAR attachment and DNA binding [40], it is likely that hnRNP U anchors Xist RNA onto the Xi to induce and maintain X-linked gene silencing during X-inactivation. The UV-crosslinking RIP analysis using a growing number of primers across the Xist gene showed that hnRNP U broadly bound not only to exon 1 of Xist RNA but also to exon 7 (Figs 4A and 5F). We also showed that the truncation mutation of Xist RNA lacking exon 7 results in an impaired interaction of the mutant Xist RNA with hnRNP U, indicating that exon 7 of Xist RNA is essential for the stable interaction with hnRNP U (Fig 5F). These results raise the possibility that stable Xist RNA localization on the Xi is mediated through the binding of hnRNP U to both exon 1 and 7 of the Xist RNA.
Previous work has shown that an absence of hnRNP U leads to the loss of the L-isoform of Xist RNA which contains the full length of exon 7, suggesting that an interaction of hnRNP to exon 7 of Xist RNA might regulate the stability of L-isoform of Xist RNA [35]. These observations suggest that the interaction between hnRNP U and exon 7 of Xist RNA is crucial for both the stability of Xist RNA and the localization of the Xist RNA on the Xi. Alternatively, hnRNP U might regulate the splicing of Xist RNA because hnRNP U is known to regulate the global control of alternative splicing of noncoding RNAs [52]. Although hnRNP U might regulate the splicing of nascent Xist RNAs transcribed at the endogenous Xist locus, it is unlikely that the hnRNP U accumulated across the entire Xi plays a role in the splicing of Xist RNA because the majority of Xist RNA is the spliced mature RNA [35]. Furthermore, a previous report has shown that even without splicing, inducible Xist cDNA transgenes can induce hnRNP U and Xist RNA localization to the Xi following gene silencing [36], indicating that Xist RNA-with the parental and FLAG-HA hnRNP U knock-in ES clones at day 12 upon differentiation. (E) Quantification analysis of the hnRNP U western blotting, including Fig 5D. Each was normalized to the parental TST6 cells (set to 1). The mean ± SD values from three independent experiments are shown. (F) UVcrosslinking RIP analysis using Tsix TST6 and Xist delEx7 Tsix TST6 mutant ES cell lines expressing FLAG-HA-tagged hnRNP U upon differentiation at day 12. The mean ± SD bar from three independent experiments is shown with an unpaired t test P values (*p<0.05, **p<0.01).
doi:10.1371/journal.pgen.1005430.g005 dependent hnRNP U accumulation on the Xi is not related to the splicing of Xist RNA. On the other hand, because the S-isoform of Xist RNA lacks a large region of exon 7 except for a 1.7 kb repeat E and 0.4 kb region of its 3´end, it lacks a large hnRNP U binding region similar to the mutant Xist delEx7 RNA lacking the entire exon 7. It would be interesting to delineate the distinct roles of the L-and S-isoforms of Xist RNA in X-inactivation.
How does exon 7 of Xist RNA contribute to the stable interaction with hnRNP U and localization of Xist RNA on the Xi? One possibility is that multiple hnRNP U binding across exons 1 and 7 of Xist RNA might be essential for the stable localization of Xist RNA on the Xi. Thus, the mutant Xist delEx7 RNA lacking multiple hnRNP U binding sites within exon 7 of Xist RNA results in its unstable localization on the Xi. Alternatively, exon 7 of the Xist RNA might contribute to the stable localization of Xist RNA on the Xi through the RNP complex formation with hnRNP U. Because it is known that hnRNP U can form nucleic acid-mediated aggregation [53], multiple hnRNP U binding to both exon 1 and exon 7 within one Xist RNA molecule might facilitate the formation of an aggregated hnRNP U complex and stabilize the functional RNP complex. Without exon 7 of the Xist RNA, an insufficient number of hnRNP U might not be able to maintain the aggregated hnRNP U-Xist RNA complex, resulting in inefficient Xist RNA binding to the Xi.
To date, many chromatin modification enzymes have been known to interact with lncRNAs [54,55] but it remains unclear how these chromatin modification enzymes are recruited to their specific target loci to modify the chromatin structure. As a representative example, Xist RNA interacting with multiple chromatin modifying enzymes is targeted to the Xi to establish a multi-layer repressive epigenetic landscape [22]. While we note the potential role of hnRNP U to target Xist RNA to the Xi through interactions with specific RNA regions in this work, multiple recent reports have suggested that interactions between hnRNP proteins and lncRNAs play a pivotal role in numerous physiological processes, such as local gene regulation, nuclear organization, and the immune response [43,[56][57][58]. Although the role of hnRNP proteins in each physiological process remains to be elucidated, these findings suggest that hnRNPs may be crucial players in the recruitment of lncRNA and its associated factors, such as chromatin modifying enzymes, onto their target loci. The interaction between Xist RNA and hnRNP U would be a great model to uncover how RNP complexes of lncRNAs, hnRNPs and chromatinmodifying enzymes are established and recruited to their specific target sites to regulate local gene regulation. Further studies of the interaction between the hnRNPs and lncRNAs will be necessary to shed light on the detailed molecular mechanism underlying the biological processes mediated by lncRNAs.

ES cells and cell culture
16.7 wild-type female ES cells [47] and their derivatives were maintained on irradiated male mouse embryonic fibroblast (MEF) feeder cells in Dulbecco´s modified Eagle´s medium (DMEM; Life Technologies) supplemented with 15% fetal bovine serum (FBS; Hyclone), 25 mM Hepes (pH 7.2-7.5) (Life Technologies), 1% MEM non-essential amino acid (Life Technologies), 1% GlutaMAX-I (Life Technologies), 100 units/ml Penicillin-Streptomycin (Life Technologies), 0.1 mM β-mercaptoethanol (Life Technologies), and 500 units/ml leukemia inhibitory factor (LIF). For the embryoid body (EB) differentiation [59], the ES cells were grown on feeder cells for 3 days and then partially trypsinized using 0.05% Trypsin-EDTA (Life Technologies). The trypsin reaction was stopped by adding differentiation media (ES media without LIF) and the ES cells were incubated for 30 min in a CO 2 incubator to remove the feeder cells. The floating ES colonies were then transferred into bacterial plates and grown as a suspension culture in a CO 2 incubator for the first 4 days. They were then attached onto gelatin-coated tissue culture plates.

Targeted truncation of Xist and Tsix
The targeting vectors for the Xist/Tsix and Tsix truncation mutation were constructed using homologous recombination in bacteria [60]. A 9.3 kb XhoI-ClaI fragment from exon 2 to 7 of Xist from the sx9 P1 clone [61] was cloned into the XhoI-ClaI of pGEM7-DTAR [62], yielding pGEM-Xex2-7-DTAR. The pGEM-Xex2-7-DTAR was transfected to SW106 for recombination in bacteria. To generate the promoterless Hyg selection cassette, the BamHI fragment of SA-tpA from pSS-SAtpA1lox-DTAF [48] was subcloned into a pGEM vector derivative in which the EcoRI and HindIII sites in pGEM11-Zf(-) were replaced by SpeI, yielding pGEM--SAtpA-1. BamHI (Klenow-filled)-EcoRV Ires-Hyg fragment of pQCXIH (Clontech) was inserted into the Klenow-filled HindIII site between SA and tpA of pGEM-SAtpA-1 in the same direction. The BamHI SA-Ires-Hyg-tpA cassette was transferred into the pBluescript II SK-(pBS-SKII-) plasmid at the BamHI site, yielding pBS-SA-Ires-Hyg-tpA. To create a selection marker flanked by two FRT derivative F3 sites for the selection of homologous recombination in bacteria, the adaptor (F3-F and F3-R primers) was ligated into the HindIII/EcoRI of pBS-SKII-(pBS-2xF3), and a NheI-EcoRI zeocin (Zeo) cassette of pSKY [63] was cloned into the pBS-2xF3 at NheI and MfeI site between two F3 sites, yielding the pBS-2xF3-Zeo plasmid. For the construction of the Tsix-truncation targeting vector, the right arm (annealed XTST-F9L and XTST-R9L primers) for homologous recombination in bacteria was inserted into the BstBI/SacI of pBS-2xF3-Zeo, yielding pBS-2xF3-R85. The SpeI (Klenow-filled)-NotI SA-Ires-Hyg-tpA of pBS-SA-Ires-Hyg-tpA and the left arm (annealed TST-F8 and TST-R8 primers) were inserted into the HindIII (Klenow-filled)-XhoI of pBS-2xF3-Zeo-R85. tpA was replaced by an Eco53kI-NotI beta-actin pA (bpA) fragment of pGEM11-DTAL. For the Xist/ Tsix truncation, the SA-tpA cassette was cloned into pBS-SKII(-). SA was replaced with an additional tpA fragment which was amplified with pA-BamXmaI-F and pA-BglII-R, yielding pBS-2xtpA-F. The left arm of the Tsix-truncation targeting construct for bacterial recombination was replaced with the left arm for Xist/Tsix-truncation (adaptor with XST-F7 and XST-R7) and EcoRV-NotI 2xtpA cassette from pBS-2xtpA-F. Selection cassettes with left and right homology arms for bacterial recombination were released by SalI-Bsu36I digestion and inserted into pGEM-Xex2-7-DTAR at position chrX: 103,468,781 and chrX: 103,468,781-103,468,821 in GRCm38/mm10 (UCSC genome browser) for the Tsix-and Xist/Tsixtruncation constructs, respectively. The Zeo selection marker was finally removed from the targeting constructs for recombination in bacteria by arabinose-inducible Flpe recombinase in the SW105 E. coli. The Tsix-and Xist/Tsix-truncation vector was linearized by XhoI digestion for homologous recombination in the ES cells.
ES cell targeting was performed as described previously [47]. Briefly, an ES cell suspension in ice-cold PBS with 3x10 6 cells and 40 μg of the linearized targeting vector was used for electroporation using the BioRad GenePulser (240 V, 500 μF). 250 μg/ml hygromycin was added at 24 hours after transfection, and hygromycin-resistant colonies were picked on 8-9 days after electroporation. Colonies were screened by genomic PCR using 7.2-F and Hyg-F or SA-R and 7.2In-R primer pairs (S1 Fig and S1 Table).

Allele-specific RT-PCR
Allele-specific RT-PCR was performed as described [64]. Briefly, after the first strand synthesis described above, PCR was performed using Maxima Hot Start PCR Master Mix (Thermo Scientific), and an aliquot of the first round of PCR was cycled once in the second round of PCR. The PCR products were purified by ethanol precipitation, digested with restriction enzymes, transferred to Hybond N (GE Healthcare) and hybridized with a 32 P-end-labeled oligonucleotide probe. The primers, probes and restriction enzymes were: Xist (XA-F and XA-R; XSP1 [62]; ScrFI); Mecp2 (MeA-F and MeA-R; NS65 [62]; DdeI); Pgk1 (Pgk1-F and Pgk1-R; Pgk1-P; MseI). Typhoon phosphorimager (GE Healthcare) was used to obtain quantitative measures. The data were quantified using GelEval (v1.37, Frog Dance Software) or ImageQuant (GE Healthcare).

Half-life assay
The half-life assay for Xist RNA was performed with Click-iT Nascent RNA Capture Kit (Life Technologies) following the manufacturer's instructions. Briefly, undifferentiated ES cells or EB were seeded to gelatinized 6-cm dishes and cultured with ES medium containing 0.1 mM 5-ethynyl uridine (EU) for 22 hrs. The next day, the EU-containing medium was replaced with EU-free medium, and cells were harvested at 0, 40, 80, 120, 180 and 240 min after changing the medium. 5 μg of total RNA extracted as above was biotinylated via the Click-iT reaction, precipitated by ethanol precipitation and dissolved in 50 μl of water. 1 μg of the biotinylated RNA was bound to Dynabeads MyOne Streptavidin T1 magnetic beads (Life Technologies) for 30 minutes at room temperature, and washed with 5x Click-iT reaction buffer 1 and 5x Click-iT reaction buffer 2. The biotinylated RNA-bound beads were resuspended with 12 μl of 5x Click-iT reaction buffer 2 and used as a template for RT-qPCR. RT-qPCR was performed as described above, with the 129 allele-specific primer set for Xist exon1-3 (X1), the S-isoform Xist specific primer set (XiI7SRT-F and XiI7SRT-R) or the L-isoform Xist specific primer set (XiI7LRT-F and XiI7LRT-R).

Immuno-fluorescence in situ hybridization (Immuno-FISH) and immunofluorescence
Immuno-FISH was performed as described [65] with some modification. All ES and EB culture were dispersed by Accutase (Innovative Cell Technologies) and cyto-spinned onto slides at 1,500 rpm for 10 min. After Triton X-100 permeabilization in the cytoskeleton (CSK) buffer, the cells were fixed by 4% paraformaldehyde for 10 min at room temperature. An immunofluorescence assay using a mouse anti-H3K27me3 antibody (Active Motif, #61017) and an Alexa Fluor 555-labeled secondary antibody against mouse IgG (Life Technologies) was followed by 4% paraformaldehyde fixation for 10 min. RNA FISH was carried out with a 25 nM oligonucleotide probe cocktail, which hybridized from exon 1 to 6 of Xist RNA, in the hybridization buffer (10% formamide, 2x SSC, 2 mg/ml BSA, 10% dextran sulfate) at 37°C overnight. The next day, the slides were washed in wash buffer (10% formamide and 2x SSC), wash buffer with 5 ng/ml DAPI, and 2x SSC for 5 min each at 37°C. They were then treated with an antifade solution.

UV-crosslinking RNA immunoprecipitation (RIP) analysis
UV-Crosslinking RIP was performed following a previously described procedure [35] with some modifications. Briefly, mouse Neuro2a or ES cells stably expressing Flag-tagged hnRNP-U or human HEK293T cells were rinsed with PBS and UV-irradiated with 400 mJ/cm 2 at 254 nm. The cells were lysed in SDS buffer (50 mM Tris-HCl [pH 8.0], 1 mM EDTA, 150 mM NaCl, 1 mM DTT, 1% SDS, 1% Triton X-100) and incubated on ice for 10 min. Samples were sheared for 6 cycles of 30 sec ON at high power/30 sec OFF with the Bioruptor (Diagnode). Sonicated lysates were diluted 10-fold with dilution buffer (50 mM Tris-HCl [pH 8.0], 1 mM EDTA, 150 mM NaCl, 1 mM DTT, 1% Triton X-100, 1× protease inhibitor cocktail, RNase inhibitor). The soluble fraction was obtained by centrifuging at 15,000 rpm for 15 min at 4°C. One-tenth of the lysates for immunoprecipitation was kept as the "10% of input". Anti-Flag M2 beads (Sigma), or protein G beads (Millipore) conjugated with 3 μg of control IgG (Jackson or Millipore) or anti-hnRNP-U (3G6, Santa Cruz) were added to the cell lysates and rotated at 4°C for 2 h. The beads were washed twice with high salt buffer (20 mM Tris-HCl [pH 8.0], 1 mM EDTA, 500 mM NaCl, 1 mM DTT, 0.1% SDS, 1% Triton X-100) and three times with low salt buffer (20 mM Tris-HCl [pH 8.0] 1 mM EDTA, 150 mM NaCl, 1 mM DTT, 0.1% SDS, 1% Triton X-100). The washed beads were treated with proteinase K at 37°C for 1 h, and the RNA was extracted with Trizol LS Reagent (Life Technologies) according to the manufacturer's instructions. Purified RNA treated with DNase I was dissolved in 20 μl nuclease free water, and cDNA was synthesized using 2 μl of the RNA samples. Quantitative PCR was performed using the primers listed in S1 Table. Targeted knock-in of the FLAG-HA epitope tag to hnRNP U by the CRISPR/Cas system pSpCas9(BB)-2A-Puro [66] (pX459, Addgene plasmid #48139) was used for CRISPR/Casmediated FLAG-HA knock-in to the 5´-end of the hnRNP U coding region. To avoid the premature termination of sgRNA and to improve the sgRNA-Cas9 assembly, mutations were introduced to pX459 [ , each of which were phosphorylated, denatured and annealed before ligation, yielding pX459FE-hnRNPU. 5x10 6 ES cells were plated on 6-well plates with feeder cells 8 hours before transfection. Cells were transfected with FuGENE HD (Roche) according to the manufacturer's instructions, with 1 μg pX459FE-hnRNPU vector and 1 μg single-stranded oligodeoxynucleotide (hnRNPU-FLAGHA-KI) for the homology directed repair (HDR)-mediated FLAG-HA knock-in. 16 hour after transfection, 2 μg/ml of puromycin was added to the ES media and ES cells were cultured for 24 hours. After puromycin selection, cells were trypsinized and 1/5 of cells were plated onto a 10 cm dish with feeder cells without any selection drug. Colonies were picked on 8 days after plating.

Western blotting
Differentiated EB cells were lysed on day 12 with RIPA buffer and sonicated using a Bioruptor (Diagenode) set on "HIGH" for 4 cycles of 30 sec ON and 30 sec OFF. Equal amounts of total protein (20 μg per lane) were applied and run on a 4-15% SDS-polyacrylamide gel and transferred to PVDF membranes (Millipore). Membranes were blocked by 5% skim milk in TBST (50 mM Tris-HCl [pH 7.5], 150 mM NaCl, 0.1% Tween-20) at room temperature for 1 h, and incubated using primary antibodies against FLAG M2 (mouse 1:2,000, Sigma), hnRNP U (clone 3G6, mouse 1:1,000, Millipore) and alpha Tubulin DM1A (mouse 1:5,000, Santa Cruz) overnight at 4°C. After incubation with the secondary antibody conjugated to horseradish peroxidase (HRP) (anti-mouse IgG 1:10,000, Jackson ImmunoResearch) at room temperature for 1 h, immune complexes were visualized using an enhanced chemiluminescence (ECL) detection system (Thermo Scientific) or Immobilon Western chemiluminescent HRP substrate (Millipore), and quantitated by AlphaView (Alpha Innotech). RT-qPCR analysis for the expression of pluripotent markers (Nanog and Oct3/4). Each Nanog and Oct3/4 expression level was normalized to that of differentiation day 0 (set to 1) and Gapdh, which was used as an internal control. The mean ± SD bar for three independent experiments is shown. (B) Representative phase contrast images of the differentiated EB cells on differentiation at day 8. knock-in ES cells using primers that extend through Xist exons 1 to 3. The expression value was normalized to that of Tsix TST6 at day 0 (set to 1,) and Gapdh was used as an internal control. The graph shows the mean ± SD bar from three independent experiments.

Supporting Information
(B) 129 allele-specific RT-qPCR analysis of the X-linked Pgk1 gene upon differentiation, normalized to undifferentiated cells (set to 1), with Gapdh used as an internal control. The mean ± SD bar from three independent experiments is shown. P values were calculated using an unpaired t test ( Ã p<0.05, ÃÃ p<0.01, ÃÃÃ p<0.001). (C) Top: Immuno-FISH for Xist RNA (green) and H3K27me3 (red) in FLAG-HA targeted knock-in ES cells after differentiation (day 12). Nuclei were counterstained with DAPI. Scale bar, 10 μm. Bottom: frequency of the Xist cloud-and H3K27me3-positive cells on day 12 upon differentiation. More than 300 nuclei in each ES cell line at each time point from three independent experiments were counted and classified based on the Xist RNA and H3K27me3 signal. (AI) S1 Table. The list of primers used in this study. (DOCX)