Main

XIST is a long non-coding RNA (lncRNA) that mediates the silencing of gene transcription on the X chromosome during female mammalian development1 via the recruitment of specific protein complexes. These complexes have been identified in studies of the genetic domains involved in XIST silencing2 as well as by recent unbiased proteomic screens that identified direct XIST-binding proteins using zero-distance ultraviolet irradiation-based crosslinking methods3,4. Proteins that are bound directly or indirectly to XIST via protein intermediates have also been identified using crosslinking reagents such as formaldehyde5. These include HNRNPU (also known as SAF-A), which anchors XIST to the X chromosome6, SHARP (SPEN), which recruits HDAC3 (ref. 3), as well as PRC2, which introduces repressive chromatin marks7.

Here we show that XIST-mediated gene silencing requires adenosine methylation, a reversible RNA-modification pathway that forms m6A. Although the m6A modification is well-studied in mRNAs, m6A mapping studies8 have shown that m6A is also present in lncRNAs. Our data show that XIST is highly methylated and that m6A modifications are required for XIST-mediated gene silencing. Formation of m6A in XIST and mRNAs is mediated by two previously unknown components of the m6A methylation complex, RBM15 and RBM15B. These proteins bind and recruit the m6A-methylation complex to specific sites within XIST, leading to m6A formation at adjacent sites. Furthermore, we show that m6A in XIST recruits the m6A reader, YTHDC1 (hereafter DC1), and that the binding of DC1 to XIST promotes XIST-mediated gene repression. These studies reveal a role for m6A and DC1 as mediators of transcriptional repression via the lncRNA XIST.

RBM15 and RBM15B are required for gene silencing

Recent studies have shown that RBM15 binds to XIST3,4. Previously, we found that the knockdown of RBM15 did not block XIST-mediated gene silencing3; however, another study found the opposite to be true9. We therefore considered the possibility that another protein compensated for the function of RBM15 in our RBM15 knockdown experiments. RBM15 possesses notable similarity to another protein, RBM15B, in sequence and domain organization, making it a suitable candidate for compensation of RBM15 function (Extended Data Fig. 1a).

To test the functional redundancy of these proteins, we first investigated whether RBM15 and RBM15B show similar binding patterns in XIST by mapping their binding sites using individual-nucleotide resolution UV crosslinking and immunoprecipitation (iCLIP)10 in human embryonic kidney 293T (HEK293T) cells. For all iCLIP experiments, we examined only the endogenous protein and identified antibodies that selectively precipitated each protein. We also confirmed that there was consistency between the transcriptome-wide iCLIP data set replicates (Extended Data Fig. 1b–g and Supplementary Tables 1, 2).

RBM15 and RBM15B showed a similar distribution of iCLIP tags (that is, processed reads; see Methods for further details) along the length of XIST (Fig. 1a and Extended Data Fig. 1h), including at the A-repeat region, an evolutionarily conserved region in the 5′ region that is essential for the initiation of silencing11. Additionally, RBM15 and RBM15B showed similar distributions of iCLIP tag clusters, which represent regions of enriched binding, and crosslinking-induced truncation sites (CITS), which represent direct contacts with XIST (Supplementary Tables 3, 4).

Figure 1: RBM15 and RBM15B are necessary for XIST-mediated gene silencing.
figure 1

a, RBM15 and RBM15B show similar binding patterns in XIST. Shown is the distribution of normalized RBM15 and RBM15B iCLIP tags (in unique tags per million, uTPM) and statistically significant CITS. Light blue vertical lines, RBM15; dark blue vertical lines, RBM15B; P < 0.0001. b, c, Knockdown of both Rbm15 and Rbm15b (siRBM15/15B) impair XIST-mediated gene silencing. XIST expression was induced by doxycycline, and the X-linked genes Gpc4 (green) and Atrx (red) were quantified by RNA-FISH (b). Representative FISH images are shown with DAPI nuclear counterstain (blue) (c). The number of detected RNA spots for both genes are indicated on each image. Scale bars, 5 μm. Data are mean ± s.e.m. for 50 cells from one experiment. ***P < 0.001, *****P < 0.0001, relative to siControl by unpaired two-sample t-test. NS, not significant.

PowerPoint slide

Full size image

To assess whether RBM15 and RBM15B are required for XIST-mediated gene silencing, we used male mouse embryonic stem (ES) cells that express Xist on the X chromosome in a doxycycline-dependent manner3. Xist-mediated gene silencing is induced by 16 h of doxycycline (Dox)-induced XIST expression and is measured by quantifying the expression of two X-linked genes, Gpc4 and Atrx, using single-molecule RNA fluorescence in situ hybridization (FISH)3. In these assays, we knocked down mRNAs using short interfering RNAs (siRNAs) and confirmed that each examined cell showed successful depletion of both the siRNA-targeted mRNA as well as Dox-induced XIST expression.

In wild-type siRNA-transfected cells, we observed the expected silencing of the X-linked genes. Gpc4 transcript levels decreased from 21 copies (−Dox) to 1 copy (+Dox) per cell and Atrx transcript levels decreased from 17 to 1 copy per cell (Fig. 1b, c and Extended Data Fig. 2a, b). Knockdown of both Rbm15 and Rbm15b, but not knockdown of either gene individually, prevented XIST-mediated gene silencing in these cells (Fig. 1b, c). This was also seen in a female mouse ES cell line that similarly exhibits Dox-inducible XIST expression on one X chromosome (Extended Data Fig. 2c). RBM15 and RBM15B therefore have redundant function in mediating XIST-mediated transcriptional silencing.

RBM15/RBM15B link the methylation complex to XIST

RBM15 and RBM15B were recently identified as high-confidence interactors with Wilms tumour-associated protein (WTAP) in a proteomic analysis12. WTAP binds METTL3 (refs 13, 14, 15), the methyltransferase that mediates methylation of m6A in mRNA16, and is recruited to RNAs via an unknown adaptor protein to trigger m6A formation14.

We therefore investigated whether RBM15 and/or RBM15B is a component of the WTAP–METTL3 complex, targeting it to RNA. Immunoprecipitation of RBM15 or RBM15B from HEK293T nuclear lysates co-precipitated METTL3 (Fig. 2a). Knockdown of WTAP reduced the interaction between METTL3 and both RBM15 and RBM15B (Fig. 2a), indicating that this interaction is mediated by WTAP. A reciprocal immunoprecipitation similarly indicated that METTL3 binds RBM15 and RBM15B in a WTAP-dependent manner (Extended Data Fig. 3a–c).

Figure 2: RBM15 and RBM15B recruit METTL3 to XIST.
figure 2

a, RBM15 and RBM15B interact with METTL3 in a WTAP-dependent manner. RBM15 (left) and RBM15B (right) were immunoprecipitated from HEK293T nuclear extracts. Co-immunoprecipitation of METTL3 was reduced in siWTAP-transfected cells. The IgG heavy chain (H chain) prevents visualization of WTAP; however, knockdown is seen in the input sample. NE, nuclear extracts. b, Quantification of METTL3-bound XIST upon knockdown of methylation machinery components. XIST was quantified by qRT–PCR using regions selected based on the presence or absence of RBM15- and RBM15B-binding sites (indicated with light blue and blue lines, respectively). Data are mean ± s.e.m. from three independent experiments. **P < 0.001, ***P < 0.0001, relative to siControl by unpaired two-sample t-test.

PowerPoint slide

Full size image

To determine whether both RBM15 and RBM15B (RBM15/15B) can recruit WTAP–METTL3 to XIST, we treated HEK293T cells with formaldehyde to crosslink XIST to any bound proteins. We then immunoprecipitated METTL3 from the cell lysates and measured the amount of bound XIST by quantitative reverse transcription PCR (qRT–PCR) at regions with and without RBM15/15B-binding sites. METTL3 immunoprecipitates contained significantly higher levels of XIST than control immunoprecipitates at these binding sites (Fig. 2b and Extended Data Fig. 3d, e). This interaction was impaired after knockdown of WTAP, RBM15, and/or RBM15B, with the greatest loss following knockdown of both RBM15 and RBM15B double knockdown (Fig. 2b). This led us to believe that RBM15/15B is the component of the methylation complex that accounts for its recruitment to XIST.

RNA-anchored methylation complexes

Our initial m6A mapping studies, using methylated RNA immunoprecipitation followed by sequencing (MeRIP–seq), showed XIST contained m6A modifications8, although this approach was at low resolution. More recently, we mapped m6A at single-nucleotide resolution using m6A iCLIP (miCLIP)17. Analysis of the miCLIP data set shows 78 putative m6A residues in XIST, some of which are localized at or near the A-repeat region (Fig. 3a and Extended Data Fig. 4a). To investigate whether RBM15 and RBM15B mediate m6A formation in XIST, we measured m6A levels in XIST in wild-type control and RBM15/15B-deficient cells. Methylated XIST was precipitated with an m6A-specific antibody and XIST levels were quantified from three m6A-containing regions (Fig. 3a). Knockdown of METTL3, RBM15, RBM15B, and both RBM15 and RBM15B resulted in significantly reduced levels of methylated XIST, with the largest reduction in m6A levels following RBM15/RBM15B double knockdown (Fig. 3b and Extended Data Fig. 3d–f). This indicates that RBM15 and RBM15B promote XIST methylation by recruiting WTAP–METTL3.

Figure 3: N6-adenosine methylation is necessary for XIST-mediated gene silencing.
figure 3

a, m6A residues (red lines) identified via miCLIP are broadly distributed along XIST. Normalized miCLIP17 tags are shown in purple. b, Methylation of XIST requires RBM15 and RBM15B. m6A levels in XIST were quantified by m6A-RNA immunoprecipitation followed by qRT–PCR of three m6A regions of XIST. Data are mean ± s.e.m. from six samples coming from three technical replicates of two biological replicates. ***P < 0.0001, **P < 0.001 relative to siControl by unpaired two-sample t-test. c, d, m6A promotes XIST-mediated gene silencing. XIST expression was induced by Dox, and X-linked genes Gpc4 (green) and Atrx (red) were quantified by RNA-FISH (c). Representative FISH images are shown (d). The number of detected RNA spots is indicated on each image. Scale bars, 5 μm. Data are mean ± s.e.m. for 50 cells from one experiment. ****P < 0.005 relative to siControl by an unpaired two-sample t-test.

PowerPoint slide

Full size image

We observed that m6A residues are typically located in the vicinity of RBM15 and RBM15B iCLIP clusters on XIST (Extended Data Fig. 4b). Indeed, the median distance between each RBM15 or RBM15B CITS in XIST and the closest m6A was 45 or 28.5 nucleotides, respectively (Extended Data Fig. 4c). By contrast, the distance between m6A and randomly picked sites along XIST was approximately 70–90 nucleotides (P = 0.0026, RBM15; P = 0.0001, RBM15B). Thus, m6A residues are positioned significantly closer to RBM15 and RBM15B sites than would be expected by chance. This proximity suggests that RBM15/15B recruits the WTAP–METTL3 complex to methylate adenosine bases that lie in proximal m6A consensus sites.

We next asked whether RBM15/15B binds next to m6A bases in mRNA. Using our single-nucleotide-resolution m6A data set in mRNA17, we calculated the spatial relationship of RBM15/15B-binding sites relative to m6A residues. As a control, we measured the binding of RBM15 and RBM15B relative to non-methylated adenosines that fall within the m6A consensus DRACH sequence (where D denotes A/G/U, R denotes A/G and H denotes A/C/U)17. These sites lack miCLIP reads and thus are non-methylated. Transcriptome-wide analysis shows that RBM15/15B-binding sites are significantly enriched on either side of m6A residues, while minimal enrichment is seen at the nearest non-methylated DRACH site (Extended Data Fig. 5a). RBM15/15B-binding sites are characterized by U-rich motifs (Extended Data Fig. 5c–e) that are readily detected adjacent to m6A residues on individual transcripts (Extended Data Fig. 5b).

Notably, knockdown of both RBM15 and RBM15B resulted in a substantial drop in m6A levels in poly(A) RNA (Extended Data Fig. 5f, g), indicating that RBM15 and RBM15B direct methylation of adenosine residues at sites in both mRNA and XIST.

XIST m6A is required for gene silencing

XIST has more mapped m6A residues than any other RNA (Supplementary Tables 5, 6), raising the possibility that m6A may mediate important aspects of XIST function. The role of m6A in XIST-mediated gene silencing cannot be tested in Mettl3−/− mouse ES cells because these cells do not express XIST owing to the persistent expression of XIST-suppressing pluripotency genes18. We thus used the Dox-inducible XIST-expression system to assess the role of METTL3 in XIST-mediated transcriptional silencing. METTL3 knockdown reduces m6A levels across the transcriptome, including in XIST19. In control siRNA-transfected cells, we observed the expected silencing of X-linked genes upon XIST induction (Fig. 3c, d and Extended Data Fig. 2c–e). However, in siMettl3-treated cells, XIST was induced but failed to silence Gpc4 and Atrx expression (Fig. 3c, d and Extended Data Fig. 2d, e). A similar silencing defect was seen in a female mouse ES cell line with Dox-inducible XIST expression (Extended Data Fig. 2c). Therefore, m6A is required for XIST-mediated transcriptional silencing.

DC1 binds XIST to mediate gene silencing

We next investigated the mechanism by which m6A in XIST is recognized in order to mediate transcriptional silencing. m6A residues are recognized by the YTH proteins20 which comprise three members of the YTHDF family (DF1, DF2, and DF3), YTHDC1 (DC1) and YTHDC2 (DC2) (Extended Data Fig. 6a). DF1, DF2, DF3 and DC2 are primarily cytoplasmic21,22,23,24, whereas DC1 is located primarily in the nucleus24.

Using iCLIP, we assessed the transcriptome-wide binding properties of the endogenous YTH proteins and determined whether any interacted preferentially with m6A in XIST (Extended Data Figs 6, 7 and Supplementary Table 1, 2). In this analysis, we quantified the binding of YTH proteins at each of the 78 mapped m6A residues in XIST as well as the other 11,452 mapped m6A residues in the transcriptome. Each m6A residue was assigned an intensity value that was defined as the normalized number of miCLIP tags for each m6A residue17. This value is influenced by both the transcript abundance and the m6A stoichiometry. Next, the binding of each YTH protein to each m6A residue was determined using the normalized number of mapped iCLIP tags at the m6A site. For most m6A residues, the miCLIP intensity value increased with the amount of bound YTH protein (Fig. 4a); however, only DC1 showed clear preferential binding for XIST m6A residues (Fig. 4a, b and Extended Data Fig. 8a–c).

Figure 4: DC1 binds XIST m6A residues and promotes XIST-mediated gene silencing.
figure 4

a, YTH iCLIP tag coverage at 11,530 annotated m6A residues. Correlation coefficients for mRNA m6A (grey) and non-coding RNA (ncRNA) m6A (magenta) are indicated. DF1, DF2 and DF3 show similar correlations between m6A abundance and YTH binding for mRNAs (blue line) and ncRNAs (magenta line). DC1 shows preference for ncRNA m6A, with the top 1% of DC1-bound m6A indicated (dotted ellipse). b, mRNA/ncRNA distribution of the top 1% of DC1-bound m6A sites. Most detected ncRNA m6A are present on XIST (indicated in green). c, Normalized tag distributions for each YTH protein on XIST shows predominantly DC1 binding. High-density m6A regions are indicated by green shading. d, e, Ythdc1 knockdown (siDC1) impairs XIST-mediated gene silencing. XIST was induced by Dox, and X-linked genes Gpc4 (green) and Atrx (red) were quantified by RNA-FISH. Representative FISH images are shown (e). The number of detected RNA spots is indicated on each image. Scale bars, 5 μm. Data are mean ± s.e.m. ****P < 0.005 relative to siControl by an unpaired two-sample t-test.

PowerPoint slide

Full size image

A direct comparison of iCLIP tags on XIST also showed that DC1 is the only YTH protein to exhibit prominent XIST binding (Fig. 4c and Extended Data Fig. 8a–d). Notably, the DC1 iCLIP tag clusters overlap with the XIST m6A miCLIP tag clusters, consistent with the binding of DC1 to m6A residues in XIST (Fig. 4c and Extended Data Fig. 8d).

The binding of DC1 to XIST could also be confirmed through the co-immunoprecipitation of DC1 and XIST using antibodies against DC1, with XIST detected by qRT–PCR using primers that detect either of the two regions with a high DC1 iCLIP signal (Extended Data Fig. 9a). XIST pulldown was reduced following the knockdown of methylation machinery components (METTL3, WTAP, RBM15, RBM15B, and RBM15 and RBM15B double knockdown). Furthermore, DC1 was enriched in the XIST nuclear subcompartment in comparison to autosomal domains as measured by 3D structured illumination super-resolution microscopy (3D-SIM) (Extended Data Fig. 9b–d). This localization was reduced following knockdown of METTL3 or both RBM15 and RBM15B (Extended Data Fig. 9e). Together, these data show that DC1 binds to XIST in an m6A-dependent manner.

We then assessed whether DC1 is required for XIST-mediated transcriptional silencing. Knockdown of DC1 but not DF1, DF2, DF3 or DC2 prevented XIST-mediated gene silencing in cells with Dox-induced XIST expression (Fig. 4d, e and Extended Data Fig. 2f–j) and in differentiating female mouse ES cells (Extended Data Fig. 2i). To determine whether DC1 binding to XIST promotes XIST-mediated gene silencing, we tethered DC1 to XIST using an XIST transcript with three BoxB hairpins appended to the 3′ end (XIST–(BoxB)3) (Fig. 5a). These hairpins bind the λN peptide fused to the C terminus of DC1, allowing the BoxB hairpins to bind the λN peptide. Dox-induced expression of XIST–(BoxB)3 caused transcriptional repression of Gpc4 and this silencing was lost following knockdown of Mettl3 or both Rbm15 and Rbm15b (Fig. 5b, c). However, XIST-mediated gene silencing was rescued when DC1–λN was expressed (Fig. 5b, c). Thus, recruitment of DC1 to XIST is sufficient to induce its repressive function in the absence of the methylation machinery. Taken together, these data suggest that m6A methylation of XIST triggers binding to DC1, which promotes XIST-mediated transcriptional silencing.

Figure 5: m6A-independent tethering of DC1 to XIST is sufficient to exert XIST-mediated gene silencing.
figure 5

a, Schematic of tethering approach. The 3′ end of XIST was genomically modified with three BoxB sequences (XIST–(BoxB)3). m6A-dependent recruitment of DC1 is blocked in methylation-deficient cells; however, artificial tethering can be achieved with DC1–λN, which binds to the BoxB elements in XIST–(BoxB)3. b, c, Dox-induced expression of XIST–(BoxB)3 results in gene silencing in siControl-transfected cells, but not in siMETTL3 or siRBM15 and siRBM15B co-transfected cells. DC1–λN rescued silencing in these cells, suggesting that the primary function of m6A in XIST-mediated gene silencing is to recruit DC1 to XIST. Quantification of Gpc4 expression is shown in b. Representative FISH images showing DAPI-stained nuclei (blue), Gpc4 RNA (green), and XIST (pink) are shown in c. Scale bars, 5 μm. Data are mean ± s.e.m. in b for 50 cells from one experiment. *****P < 0.0001 by unpaired two-sample t-test.

PowerPoint slide

Full size image

Discussion

Although the m6A modification has been well-characterized in mRNA, no function for m6A in lncRNAs has previously been demonstrated. Here we show that m6A functions to enable the transcriptional repression effects of XIST. XIST is highly enriched in m6A throughout its length, enabling the recruitment of the nuclear m6A binding protein DC1. The importance of m6A in XIST function is highlighted by the fact that diverse components of the m6A methylation complex bind XIST and are required for XIST-mediated gene silencing. Together, these discoveries reveal a role for RNA modification in lncRNA function and describe the assembly of XIST into a transcriptionally repressive ribonucleoprotein complex (Extended Data Fig. 10a).

Recent proteomic studies have revealed large numbers of XIST-binding proteins3,4,5,9, several of which we now recognize as contributing to m6A formation or recognition. For example, WTAP was identified in a proteomic analysis of XIST-associated proteins5 and was shown to be required for XIST-mediated gene silencing in a functional screen9. Although WTAP has numerous functions, our data support the idea that its m6A methylation-promoting effects are required for XIST-mediated gene silencing. DC1 was also observed in a proteomic analysis of formaldehyde-crosslinked proteins bound to XIST5.

Similarly, RBM15 was shown to be required for XIST-mediated gene silencing9 and was also identified as an XIST-binding protein3,5,9. Our data suggest that RBM15/15B is a component of the m6A methylation complex that binds XIST, and that it is this methylation role that is essential in bringing about the silencing defect observed when both are knocked down. RBM15 and RBM15B appear to have redundant functions as both need to be knocked down in order to deplete m6A to sufficient levels to impair XIST function. The large number of m6A residues in XIST ensures that at least a few will bind to DC1 to activate gene-silencing mechanisms.

The identification of the WTAP–METTL3 complex13 and its role in m6A formation14 raised several important questions. First, why are some RNAs methylated, while others lack m6A? Second, why are only a subset of DRACH-site adenosine residues selected for methylation, despite the high prevalence of DRACH consensus sites in RNA25? Our data sheds light on these questions. RBM15 and RBM15B, proteins that associate with WTAP–METTL3 and contain RNA-binding domains, enable the binding of WTAP–METTL3 to specific mRNAs, as well as XIST. The localized binding at specific sites in the RNA sequence allows for the selective methylation of adjacent DRACH sites while leaving distant DRACH sites unmethylated. The three-dimensional RNA structure of XIST could promote further adenosine methylation by bringing distant DRACH consensus sites into the proximity of the RBM15/15B-anchored methylation complex.

Our single-nucleotide-resolution map of m6A (ref. 17) showed that RBM15/15B is found adjacent to methylated but not non-methylated DRACH sequences in the mRNA transcriptome. The double knockdown of RBM15 and RBM15B markedly reduce m6A levels in mRNA, supporting the idea that RBM15/15B-binding determines which DRACH sites are methylated in the transcriptome.

How DC1 binding to XIST leads to gene silencing remains unclear. However, a recent proteomics study exploring DC1 binding partners27 may provide initial mechanistic insights. These partners include SHARP, LBR, HNRNPU, and HNRNPK which each have distinct roles in the initiation of transcriptional silencing (Extended Data Fig. 10b–e). Analysis of the DC1 interaction network, based on an independent protein–protein interaction database28, also identifies additional interactions with components of the PRC1 and PRC2 complexes (Extended Data Fig. 10b–e and Supplementary Table 7). Various XIST-interacting gene-silencing proteins may bind to DC1 and utilize the ability of DC1 to bind m6A residues on XIST to achieve additional specificity in the binding of precise locations on XIST. Further experiments are required both to determine whether DC1 directly affects binding of these silencing proteins and to explore the mechanisms used by DC1 to enable m6A-dependent transcriptional silencing.

Methods

No statistical methods were used to predetermine sample size. The experiments were not randomized. The investigators were not blinded to allocation during experiments and outcome assessment.

Cell culture

HEK293T/17 (ATCC CRL-11268) cells were maintained in 1× DMEM (11995-065, Life Technologies) with 10% FBS, 100 U ml−1 penicillin and 100 μg ml−1 of streptomycin under standard tissue culture conditions. Cells were split using TrypLE Express (Life Technologies) according to manufacturer’s instructions. Mouse ES cells expressing Xist RNA from the endogenous locus under a Tet-driven promoter (pSM33 ES cell line) were maintained as previously described29. Cell lines were not tested for mycoplasma contamination.

Generation of female pSM33 cell line

The Tet-regulated promoter was inserted at the promoter region of the endogenous Xist locus of mouse female ES cell line (F1 2-1 line, derived from a 129 × castaneous F1 mouse cross) using CRISPR-mediated homologous recombination. Clonal cell lines derived from single cells were screened for the presence of Tet-inducible promoter by PCR. Promoter integration was confirmed by Sanger sequencing with primers flanking the insertion site. Recombinant Xist alleles were further identified by SNP analysis. A clonal line with promoter insertion in the 129 allele was used for studying Xist-mediated gene silencing.

Insertion of BoxB sequence elements in Xist

Three BoxB sequence elements were inserted at the 3′ end of the endogenous Xist loci in the male pSM33 cell line using CRISPR-mediated homologous recombination. In brief, cells were first co-transfected with a plasmid expressing Cas9 under a CAG promoter, a short guide RNA (Target sequence: 5′-CCTCATCCTCATGTCTTCTC-3′), and a ssDNA ultramer (IDT) containing three BoxB elements (5′-GGGCCCTGAAGAAGGGCCCATGGGCCCTGAAGAAGGGCCCATAGGGCCCTGAAGAAGGGCCC-3′; underlined bases mark the BoxB sequence) flanked by 70-nucleotide-long DNA sequence identical to the upstream and downstream genomic DNA sequence at the point of BoxB insertion. Cells were sorted and single colonies were screened for the insertion of BoxB elements by PCR. Insertion was further confirmed by Sanger sequencing. Recombinant clones were tested for X-chromosome silencing by induction of Xist expression and Gpc4 and Atrx RNA-FISH. A clone showing silencing identical to the non-recombinant cell line was used for DC1–λN–XIST tethering functional assay.

Construction of λN–3×Flag epitope-tagged DC1 expression construct

A human YTHDC1-encoding open reading frame (ORF) was PCR-amplified from oligo-(dT)18-primed HEK293T cDNA using hYTHDC1–EcoRI-F and hYTHDC1–XhoI-R primers (Supplementary Table 8). The PCR fragment was initially cloned in pcDNA3-Flag-HA (1436 pcDNA3-Flag-HA was a gift from W. Sellers; Addgene plasmid 10792) plasmid at EcoRI and XhoI sites. Full-length YTHDC1 was then PCR amplified and subcloned into pCAG-GW-λN-3×Flag-BSD construct using the Gateway entry cloning system (Invitrogen). This plasmid (pCAG-GW-hYTHDC1-λN-3×Flag-BSD) expresses human YTHDC1 protein with a C-terminal λN–3×Flag tag under CAG promoter. We verified that λN–3×Flag-tagged DC1 protein was still functional by ensuring that it could rescue knockdown of the endogenous protein.

Generation of Ythdc1+/− female ES cells

Ythdc1+/− female ES cell line was generated using the CRISPR–Cas9 system. In brief, female ES cells were co-transfected with a Cas9-expressing pCAG plasmid and a pool of short guide RNAs targeting the region around the first codon of the Ythdc1 ORF at the endogenous loci to generate frameshift mutations causing disruption in the reading frame. Target DNA sequences were 5′-AAGCCGGAGGGCAGCCATGG-3′, 5′-GCGGTGGCGGCGGCGGAAGC-3′ and 5′-CGGCGGAAGCCGGAGGGCAG-3′. We screened 24 colonies derived from single cells for the presence of frame-shift mutations at the desired location in Ythdc1 gene using PCR and Sanger sequencing, with primers flanking the target site. No clone showed a homozygous frame-shift mutation, suggesting that homozygous Ythdc1 deletion is lethal. Only clones with heterozygous frame shift mutations were detected. Confirmation of the presence of a heterozygous knockout of Ythdc1 (Ythdc1+/−) was performed by RNA-FISH and immunofluorescence. A clonal cell line showing a 50% reduction in the expression level of Ythdc1 mRNA and protein were used for assaying X-chromosome silencing.

Antibodies

Details of the antibodies used in this study are given in Supplementary Table 1.

siRNA and shRNA transfection

Target sequences of siRNA and short hairpin RNA (shRNA) used in this study are listed in Supplementary Table 9. For validation of antibodies for iCLIP, 20 nM siRNA was transfected using Pepmute transfection reagent (Signagen) and pSuperior–EGFP shRNA plasmid (OligoEngine) was transfected using Fugene HD transfection reagent (Promega) according to the manufacturer’s instructions. Forty-eight hours after the first transfection, a second transfection was performed. Cells were maintained at 70–80% confluency and collected 96 h after the first transfection. Knockdown was confirmed by western blot analysis (list of antibodies and dilutions used are given in Supplementary Table 1).

For studying the effect of Rbm15, Rbm15b, Mettl3, Ythdf1, Ythdf2, Ythdf3, Ythdc1 and Ythdc2 knockdown on XIST-mediated gene silencing, 20 nM of siRNA targeting each gene were transfected into 100,000 pSM33 ES cells using the Neon transfection system (settings: 1,200 V, 40 ms width, 1 pulse; Invitrogen). At the time of XIST induction, the observed knockdown efficiency for all the target genes was greater than 70%. For Mettl3, the efficiency was 95%.

Construction of iCLIP libraries

All iCLIP studies were performed on the endogenous proteins. Previous CLIP-based analyses of YTH proteins used overexpressed proteins. Since this can affect the localization and assembly of proteins into multiprotein complexes, we identified antibodies that bound the endogenous proteins for these studies. iCLIP libraries were constructed as described elsewhere with minor modifications30. To improve the efficiency of cell lysis and dissolution of RNA–protein conjugates, cells were lysed in 1% SDS as described previously31. In brief, 9 × 106 HEK293T cells were seeded per 10 cm dish 12 h before UV irradiation. Media was discarded and 6 ml of ice-cold PBS was gently added to the cells. Cells were maintained on ice and immediately irradiated once with UV at 254 nm (150 mJ cm−2) in a UV crosslinker (Stratagene 2400). Cells were scraped in PBS using a cell scraper and collected by centrifugation at 200g for 10 min at 4 °C. Supernatant was discarded, and cells were gently suspended in 100 μl of 1% SDS with 10 mM DTT and 10× protease inhibitors (EDTA-free cOmplete mini, Roche) and incubated at 25 °C for 10 min to denature the protein complexes. SDS was neutralized with 900 μl of iCLIP lysis buffer (CLB) without SDS (50 mM Tris-HCl pH 7.4, 100 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate). Lysates were sonicated using a Branson Digital Sonifier Model 450 fitted with 3.125 mm tapered microtip probe on ice at 20% amplitude for 30 s with 2 s ON and 10 s OFF cycle. DNase I and RNase I digestion was performed with 2 μl of Turbo DNase I (AM2238, Life Technologies) and 10 μl of different dilutions of RNase I per ml of lysate for 3 min at 37 °C. For validation of antibodies for iCLIP and the construction of iCLIP libraries, 1:5 dilution of RNase I (AM2295, Life Technologies) was used as high (H) and 1:150 dilution was used as low (L) concentration RNase. Antibodies were first bound to CLB-washed Protein A/G beads (88803, Thermo Fisher) in CLB (50 mM Tris-HCl pH 7.4, 100 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS) followed by incubation at 25 °C for 30 min with mixing. Beads were washed twice with CLB.

For validation of antibodies for iCLIP, 500 ng of antibody was used per immunoprecipitation and for the construction of iCLIP libraries, 2–10 μg of antibody was used. Clarified RNase- and DNase-digested lysates were incubated with antibody bound to Protein A/G-beads at 4 °C for 12 h. Further steps of iCLIP library preparation were carried out as described previously30. To avoid cross-contamination of RNA and library PCR products, electrophoresis equipment was treated with 10% commercial bleach for 20 min at 25 °C and thoroughly washed with nuclease-free water before use. Replicates were tagged with unique barcodes using the 5′ Rtclip primer in reverse transcription. Low-, medium- and high-molecular-mass cDNA libraries were mixed at 1:5:5 molar ratio and sequenced on Illumina HiSeq 2500 from a single end for 50 bases.

Analysis of iCLIP sequence data

Low-quality bases, reads with more than two ambiguous base calls, and adaptor sequences were all removed using FLEXBAR tool (–max-uncalled 2 –min-read-length 15 –pre-trim-phred 20, 3′ adaptor: AGATCGGAAGAGCGGTTCAG). Reads were demultiplexed based on 5′ barcodes for individual replicates using an in-house Linux shell script. Reads were processed in pooled or separate replicate modes using the CITS analysis pipeline32. In brief, reads were converted to fasta format using fastq_to_fasta tool from FASTX-toolkit and then collapsed to remove PCR amplified duplicates based on sequence using CIMS/fasta2collapse.pl script. The barcode was stripped and added to the name of the read. Reads were aligned to the human genome (hg19) using Novoalign (v3.02.12, NovoCraft Technologies) (Options: -t 85 -l 16 -s 1 -r None). Further analysis until the identification of CITS (P < 0.0001) was performed as described previously32. Unique sequence reads that are free of PCR duplicates represent unique RNA-protein binding events. These processed reads are referred to as iCLIP/miCLIP tags (or just tags), and the mapped cluster of processed reads are referred to as tag clusters throughout this study.

Motif enrichment analysis

Analysis of motif enrichment was performed on the sense DNA sequence 20 nucleotides up- and down-stream of the called truncation sites using the MEME suite33. For this analysis the top 20% of the sites identified as statistically significant (those with P < 0.0001), with the highest number of crosslinking induced truncations, were used. Since fewer sites were detected for DC2, all of the sites were used for MEME analysis of DC2-binding sites.

Metagene analysis

Metagenes were constructed for the called CITS/miCLIP-identified m6A residues using an in-house Perl annotation pipeline and an R script. In brief, the single-nucleotide sites were mapped to different RNA features (5′ UTR, CDS and 3′ UTR) of the human genome (hg19). The position of the sites was normalized to the median feature length of the transcripts to which the sites mapped. A frequency distribution plot was generated by counting the number of sites in contiguous bins on a virtual mRNA transcript, sites whose feature lengths represent the median feature lengths of transcripts under analysis. A Gaussian estimate of kernel density was then plotted as a metagene. For YTH, RBM15 and RBM15B proteins, all statistically significant CITS (P < 0.0001) were used and for miCLIP m6A, residues identified from poly(A) RNA from ref. 17 were used.

Comparison of iCLIP and miCLIP tag coverage

For comparing iCLIP tags, we calculated normalized tag counts using a previously described approach with minor modifications34. Instead of using read counts per million mapped reads (RPM) normalization to reduce PCR amplification bias, we used unique tag counts obtained from CITS analysis. Each iCLIP tag represents a unique RNA–protein or antibody–m6A binding event. The number of unique events from a million such events is proportional within replicates and also comparable across different CLIP libraries. For this, the number of iCLIP tags per million uniquely mapped tags (unique tags per million, uTPM) was calculated at every coordinate on the human genome using the following formula: where t = number of unique CLIP tags at a base, T = total number of uniquely mapped unique CLIP tags in the whole CLIP library.

For comparing replicates, the normalized mean tag counts (in uTPM) between replicates at randomly selected ten-thousand 100-bp bins on the human genome were compared. For comparing various iCLIP/HITS-CLIP/miCLIP data sets, iCLIP data analysed in pooled mode was used. Here, the normalized total tag count in the 10-bp flanking region of 11,530 miCLIP-identified m6A residues mapping to mRNA and ncRNA (includes snoRNAs, lncRNA and other ncRNAs) were calculated. Only m6A residues in non-BCANN consensus sequence were considered for this analysis. These represent unique sites obtained from merging (mergeBed -s -d 2) of CIMS- and CITS-based m6A site calls from ref. 17. All rRNA, tRNA, and mitochondrial genomic miCLIP sites were removed. Tag counting was performed using the bedtools suite. Tag counts (uTPM + 1) were compared using scatter plots and Pearson correlation coefficients (r) were determined in R.

For identification of DC1-preferred m6A residues, residuals of simple linear regression model were calculated and sorted in R. The top 1% of sites with highest residuals were selected and annotated. HNRNPA2B1 HITS-CLIP data was obtained from a previously published study35 (GEO accession numbers: GSE70061, SRR2071655 and SRR2071656).

For representation of miCLIP tracks in Figs 3a, 4c and Extended Data Figs 4a, b, 5b, 8d, tag counts from miCLIP data sets using poly(A) RNA and miCLIP data sets using total RNA were added at every genomic position (GEO accession number: GSE63753).

Annotation of CITS

Normalized iCLIP tag-abundance was determined in the 20-bp flanking regions of the RefSeq RNA mapping CITS. Sites were then sorted based on tag abundance, and the top 1,000 sites with the highest normalized tag abundance were annotated using the annotatePeaks.pl script from the Homer package36.

Statistical significance of overlap of RNA-binding sites

To determine the statistical significance of overlap of RBM15 and RBM15B CITS (RBM15, n = 37; RBM15B, n = 56; P < 0.0001 for both) on XIST, random sites were generated on the RNA and an overlap with the RBM15 CITS was calculated (± 20 nucleotides) using the bedtools window tool. This was repeated 10,000 times to generate a null distribution for overlap counts. The P-value for the observed overlap between RBM15 and RBM15B was estimated from the null distribution (two-sided). For clusters, random clusters of equal size (median length = 91 nucleotides) were generated on XIST and a similar null distribution to CITS was generated. For both comparisons, the same number of random sites or clusters were generated as in RBM15B data set (n = 30). Clusters showing a minimum overlap of half-cluster length with the RBM15B clusters were counted. All RBM15 clusters (n = 30) overlapped with RBM15B clusters (n = 30) on XIST (P < 0.0001). The RBM15 cluster overlaps with randomly permuted RBM15B clusters, while maintaining the mean cluster size of 91 nucleotides, did not show a similar or greater percentage overlap.

RBM15/15B binding at m6A residues

Unique iCLIP tags were aligned to the hg19 genome using STAR aligner (STAR –outSAMtype BAM SortedByCoordinate –outSAMattributes All –outFilterMultimapNmax 1 –outFilterMismatchNmax 2). For determination of the average RBM15/15B-binding at m6A and non-m6A sites (both of which are DRACH-consensus sequences), sequence alignment (BAM) files were further processed using deepTools37. Methylated DRACH sites (n = 14,209) were obtained by merging miCLIP sites from HEK293 poly(A) and total RNA from ref. 17. A non-methylated DRACH site was identified near each methylated DRACH site within a distance of 20–200 nucleotides in the same transcript in the refseq transcriptome) using an in-house python script. For this purpose, DRACH sites on transcripts with no miCLIP tags were considered to be non-methylated. Heat maps were generated using the plotHeatmap script from the deepTools suite.

Immunoprecipitation of RBM15, RBM15B and METTL3

HEK293T cells were transfected with 10 nM siRNA (Supplementary Table 9) using the Pepmute transfection reagent and then grown to 80% confluency in a 150 mm dish. After 72 h, cells were washed twice with cold PBS, scraped, and collected by centrifugation. The cell pellet was then resuspended in three packed cell volumes of hypotonic buffer (10 mM HEPES pH 7.6, 10 mM KCl, 1 mM EDTA, 0.1 mM EGTA, protease and phosphatase inhibitor cocktail (Pierce)), and incubated on ice for 10 min. Triton X-100 was added to a final concentration of 0.3%, the lysate was briefly vortexed and centrifuged at 15,000g for 1 min at 4 °C. Supernatant (cytoplasm) was discarded, and the nuclear pellet was washed with 3 packed cell volumes of hypotonic buffer and centrifuged as before. The pellet was resuspended in 1 ml NP-40 lysis buffer (50 mM Tris-HCl pH 8.0, 150 mM NaCl, 1% NP-40, protease and phosphatase inhibitor cocktail) and passed through a 21-gauge syringe several times, followed by treatment with 100 U benzonase for 30 min at 37 °C. Nuclear lysates were centrifuged at 21,000g for 15 min at 4 °C. Immunoprecipitations were carried out with 500 μg of nuclear extracts with 5 μg of antibody at 4 °C overnight, followed by a 2 h incubation with 25 μl of Pierce Protein A/G magnetic beads at 4 °C. For the co-immunoprecipitation of METTL3–RBM15B, 250 μg of nuclear lysate was used per 5 μg of the METTL3 antibody. Beads were washed five times with NP-40 lysis buffer and proteins were eluted with 1× Novex Loading buffer with 50 mM dithiothreitol (DTT). The eluent was heat-denatured, electrophoresed, and transferred to a PVDF membrane and probed for different proteins. A list of antibodies and dilutions used for immunoprecipitation and western blot analysis are given in Supplementary Table 1. Quantification of band intensities was performed by the relative quantitation approach using Image Lab software (Bio-Rad, v5.2.1).

RNP immunoprecipitation and quantification of XIST

METTL3/DC1/RBM15/RBM15B-bound XIST RNA was quantified in the immunoprecipitates obtained from formaldehyde-crosslinked cells using a method previously described38 with some modifications. In brief, siRNA-transfected cells were washed with ice-cold PBS and fixed with 1% formaldehyde in PBS for 10 min at 25 °C with gentle rocking. Formaldehyde was quenched by adding glycine to a final concentration of 0.25 M and then incubating at 25 °C for 5 min. Fixed cells were washed three times with ice-cold PBS and resuspended in 0.5 ml of radioimmunoprecipitation (RIPA) buffer (50 mM Tris–HCl pH 7.4, 100 mM NaCl, 1% Igepal CA-630, 0.1% SDS, 0.5% sodium deoxycholate) with protease inhibitors (Roche) and 1 mM DTT per 3 million cells. DNA was sheared by sonication on ice twice at 15% amplitude for 2 s ON, 10 s OFF for a total of 30 s. Lysates were incubated on ice for 10 min, and subjected to DNase I and partial RNase I digestion for 3 min at 37 °C with mixing (2 μl Turbo DNase I and 5 μl of 1 to 25 times diluted RNase I in PBS per 0.5 ml of lysate). Tubes were immediately transferred to ice and incubated for 5 min. Lysates were then clarified by centrifugation at 21,000g at 4 °C for 10 min. Protein (200 μg) was supplemented with SUPERase In RNase inhibitor (100 U ml−1, Thermo Fisher) and then subjected to immunoprecipitation in RIPA buffer. Antibodies targeting METTL3, DC1, RBM15 or RBM15B (2 μg per 10 μl beads; Supplementary Table 1) were first bound to RIPA-buffer-washed Protein A/G magnetic beads (Thermo Fisher). Antibody-bound beads were then washed with RIPA buffer, added to the lysate for immunoprecipitation and incubated at 4 °C for 12 h. Rabbit IgG antibody was used as a control. Beads were washed five times with 500 μl RIPA buffer containing 1 M NaCl and 1 M Urea at 25 °C and resuspended in 100 μl eGFP–RNA (100 pg)-containing RNA elution buffer (50 mM Tris-HCl pH 7.4, 5 mM EDTA, 10 mM DTT, 1% SDS). Formaldehyde-induced crosslinks were reversed by incubation at 70 °C for 30 min with mixing. Supernatant was mixed with Trizol LS (Thermo Fisher) and co-immunoprecipitated RNA was purified according to the manufacturer’s instructions. Glycoblue (Thermo Fisher) was used to visualize the RNA pellet. Purified RNA was then reverse-transcribed with random hexamers using SuperScript III reverse transcriptase. XIST RNA levels were detected by qRT–PCR and normalized to the spike-in eGFP RNA levels. Relative XIST RNA enrichment was calculated as the ratio of normalized XIST RNA levels in protein immunoprecipitation to levels in IgG immunoprecipitates. A very low level of XIST RNA was detected in the immunoprecipitate of non-crosslinked cells compared to the crosslinked cells (<1%). Quantification of XIST was performed using primer pairs directed against three regions in XIST, selected based on the presence of RBM15- and RBM15B-binding sites (see Figs 2b, 3a). These regions were: region 1 (chrX:73,072,444–73,072,560), region 2 (chrX:73,046,651–73,046,776), and region 3 (chrX:73,067,594–73,067,714). Region 1 and 2 contain RBM15/15B-binding sites whereas region 3 lacks RBM15/15B-binding sites. Primers used for quantification are given in Supplementary Table 8. Primer PCR amplification efficiency was between 90 and 100%.

MeRIP qRT–PCR of XIST RNA

Total RNA was isolated from HEK293T cells by Trizol extraction according to the manufacturer’s instructions and poly(A) RNA was isolated using oligo-d(T)25 magnetic beads (NEB). In total, 5 μg of anti-m6A antibody (ab190886, Abcam) was pre-bound to Protein A/G magnetic beads in immunoprecipitation buffer (20 mM Tris-HCl pH 7.5, 140 mM NaCl, 0.05% Triton X-100) for 2 h. A total of 2.5 μg of poly(A) RNA was mixed with 100 pg of non-m6A (eGFP, 0.7 kb) and m6A-containing spike-in RNAs in 400 μl of immunoprecipitation buffer. Protein A/G beads were then added and incubated at 4 °C for 2 h. Samples were washed five times with immunoprecipitation buffer, and RNA was eluted from the beads by incubating with 400 μl of 0.5 mg ml−1 m6ATP for 1 h at 4 °C. Following ethanol precipitation, the input RNA and eluted poly(A) RNA were reverse transcribed with random hexamers and enrichment was determined by qRT–PCR. The spike-in control RNAs were synthesized by in vitro transcription. Non-m6A RNA (eGFP) was transcribed using an eGFP-ORF-containing plasmid in the presence of ATP (no m6ATP). The m6A-containing RNA was transcribed from an artificially synthesized dsDNA template that encoded a 1.6-kb RNA with only one adenosine residue in the presence of m6ATP and no ATP.

X-chromosome silencing assay

For this assay, a previously described method3 was used. In brief, siRNA-transfected male or female pSM33 cells were plated on poly-l-lysine or poly-d-lysine (Sigma-Aldrich) and 0.2% gelatin (Sigma-Aldrich)-coated coverslips in wells of a 24-well plate in 2i media. After 48 h, Xist RNA expression was induced with doxycycline (2 μg ml−1) (Sigma-Aldrich) in fresh media for 16 h. Control cells received only media. Immediately following incubation, cells were fixed for FISH staining.

For inducing differentiation and induction of Xist expression in the female ES cells, 2i media was replaced with MEF media (DMEM, 10% BenchMark FBS; Gemini Bio-products, 1× l-glutamine, 1× NEAA, 1× penicillin and streptomycin; Life Technologies) 12 h after transfection. After another 12 h, cells were treated with 1 μM retinoic acid (Sigma-Aldrich) for 24 h. Untreated cells were maintained in 2i media until fixing.

Cells were then fixed in Histochoice (Sigma-Aldrich) for 10 min, washed with PBS, and subjected to FISH staining and imaging. Atrx, Gpc4, Mettl3, Rbm15, Rbm15b, Xist, Ythdc1, Ythdc2, Ythdf1, Ythdf2, and Ythdf3 RNAs were stained by single-molecule RNA-FISH. They were then imaged and quantified as described in ref. 3. Probe sets and conjugated fluorophores (excitation wavelengths) for FISH probes were TYPE 1-Xist (550 nm), TYPE 4-Gpc4 (488 nm), TYPE 10-Atrx, Rbm15b (740 nm), and TYPE 6-Mettl3, Rbm15, Ythdc1, Ythdc2, Ythdf1, Ythdf2, and Ythdf3 (650 nm). Imaging was performed using Nikon Ti Eclipse microscope with the Nikon CFI Plan Apochromat λ DM 60×/1.40 oil objective. Images were processed in Fiji (ImageJ v1.51d)39. To enhance the FISH spot size, Maximum Filter plugin with a radius of 2.0 pixels was applied to the Gpc4 and/or Atrx channels.

DC1–λN–XIST–(BoxB)3 RNA tether function assay

For this assay, male mouse pSM33 cells expressing Xist–(BoxB)3 RNA under doxycycline control were used. Cells (1.5 × 105) were co-transfected with 20 nM siMETTL3 or siRBM15/15B and 0.75 μg of pCAG-GW-hYTHDC1-λN-3×Flag-BSD plasmid using Neon transfection system (10 μl tip, settings: 1,200 V, 40 ms width, 1 pulse) and seeded on coverslips as described for the X-chromosome silencing assay. For the identification of DC1–λN–3×Flag expressing cells, fixed cells were first subjected to immunofluorescence using mouse anti-Flag antibody (Sigma-Aldrich). Briefly, fixed cells were permeabilized with 0.1% Triton X-100 in PBS at room temperature for 10 min, and blocked with 5% normal goat serum in 0.1% Triton X-100 in PBS at room temperature for 30 min. Cells were then incubated with anti-Flag M2 antibody (Sigma-Aldrich; F3165; dilution 1 to 50) for 1 h at room temperature, followed by washes with 0.1% Triton X-100 in PBS and incubation with secondary antibody (goat anti-mouse IgG antibody-Alexa Fluor 750 conjugate, Thermo Fisher, dilution 1:200) at room temperature for 1 h. The samples were then processed using the RNA-FISH protocol, as described above.

Protein–protein interaction (PPI) network analysis

PINA2 (ref. 28) was used to mine the PPI networks of DC1, its immediate neighbours, the proteins regulating XIST-mediated gene silencing (SHARP, HDAC3, HNRNPK, HNRNPU, NCOR2/SMRT, LBR), and components of PRC (polycomb repressor complexes). Protein sub-networks showing interaction with DC1 and an enrichment of transcription repressor gene ontogology terms (false discovery rate < 0.05, P < 0.05) were curated and filtered for visualization. Networks were imported, visualized, and edited in Cytoscape (v3.3.0)40 for image production. To identify potentially novel interactions between DC1 and the proteins contributing to XIST-mediated gene silencing, publically available mass spectrometry data of DC1-associated proteins (PeptideAtlas accession number PASS00835) from ref. 27 was mined. Peptides were first identified by comparing the mass spectrometry spectra with references from the human proteome database (SwissProt) according to ref. 41 (15 p.p.m. peptide mass tolerance and 20 m.m.u. fragment mass tolerance). Identified peptides with natural log(e) scores below −1 and more than two unique peptides were further mined for peptides from proteins known to regulate XIST-mediated gene silencing. Identified proteins were manually added to the PPI network.

Determination of relative m6A levels by thin layer chromatography

Levels of internal m6A in mRNA were determined by thin layer chromatography (TLC) as previously described42. In brief, poly(A) RNA (100 ng) was digested with 2 U RNase T1 (Thermo Fisher) for 2 h at 37 °C in the presence of RNasin RNase Inhibitor (Promega). Five prime ends were subsequently labelled with 10 U T4 PNK (NEB) and 0.4 mBq [γ-32P]ATP at 37 °C for 30 min followed by removal of the γ-phosphate of ATP by incubation with 10 U apyrase (NEB) at 30 °C for 30 min. After phenol-chloroform extraction and ethanol precipitation, RNA samples were resuspended in 10 μl of water and digested to mononucleotides with 2 U of P1 nuclease (Sigma-Aldrich) for 3 h at 37 °C. Following this, 2 μl of the released 5′ monophosphates from this digest were then analysed by 2D-TLC on glass-backed PEI-cellulose plates (Merck-Millipore). The nucleotides were first separated in the first dimension in isobutyric acid with 0.5 M NH4OH (5:3, v/v), followed by isopropanol, HCl and water at a ratio of 70:15:15 (v/v/v) in the second dimension. Signal acquisition was carried out using a storage phosphor screen (GE Healthcare Life Sciences) at 200 μm pixel size on a Typhoon scanner (GE Healthcare Life Sciences). For quantification, m6A was calculated as a percentage of the total of the A, C and U spots, as described previously42.

Structured Illumination Microscopy (3D-SIM) and image analysis

HEK293T cells were fixed and subjected to immunofluorescence and single-molecule RNA-FISH staining using a protocol from ref. 43 with some modifications. In brief, siRNA-transfected and non-transfected HEK293T cells were seeded on poly-l-lysine-coated no. 1.5 H (170 μm ± 5 μm) coverslips (poly-l-lysine: 3438-100-01, Trevigen; coverslips: 474030-9000-000, Carl Zeiss) in 6-well plates. After 12–24 h of incubation, cells were washed twice with PBS at 25 °C and fixed with 2% methanol-free formaldehyde (28906, Thermo Fisher) in PBS for 10 min at room temperature. Cells were then washed three times with PBS and permeabilized with permeabilization buffer (1% acetylated BSA (Sigma-Aldrich), 0.3% Triton X-100, 2 mM vanadyl ribonucleoside complexes (NEB) in 1× PBS) at 25 °C for 60 min. Following permeabilization, cells were incubated with rabbit anti-YTHDC1 antibody (ab122340, Abcam, dilution 1:1,000) in permeabilization buffer for 2 h at 25 °C in a humidified chamber. Cells were then washed with immunofluorescence-wash buffer (0.5% Tween-20 in PBS) three times at room temperature. Each wash was maintained for 5 min on cells with gentle shaking. Cells were further incubated with donkey anti-Rabbit IgG antibody–Alexa Fluor 488 conjugate (A-21206, Thermo Fisher, dilution 1:1,000) for 30 min at 25 °C in a humidified, dark chamber. Following the incubation, cells were washed as before, fixed with 4% formaldehyde in PBS for 10 min at room temperature, and washed with PBS three times. The second formaldehyde fixation immobilizes the primary and secondary antibodies at the target antigen. This step avoids loss of antibodies during the probe-hybridization step of RNA-FISH. Probe hybridization in RNA-FISH uses organic solvent such as formamide that may alter antibody structure thereby affecting its ability to bind the target antigen.

After PBS wash, cells were equilibrated in FISH-wash buffer (10% formamide in 2× SSC buffer diluted from a 20x stock (S6639, Sigma-Aldrich)) for 10 min at room temperature, and then incubated with fluorescently labelled DNA probes against XIST (Stellaris FISH probes hXIST w/ Q570, SMF-2038-1, Biosearch Technologies) in Hybridization buffer (10% formamide, 10% dextran sulfate in 2× SSC buffer) at a concentration of 100 nM in a humidified chamber at 37 °C for overnight. Following the incubation, cells were washed twice with FISH-wash buffer at 37 °C for 30 min without shaking. Cells were further washed three times with PBS, and then incubated with DAPI (2 μg ml−1 in PBS) for 15 min at room temperature with gentle shaking. Cells were further washed and maintained in PBS until mounting. Coverslips with fixed and stained cells were mounted in mounting media (Prolong Diamond, P36961, Life Technologies) and quickly sealed with a nail polish. After drying of nail polish, the slides were temporarily stored at 4 °C until imaging.

Cells were imaged by super-resolution 3D-SIM on OMX Blaze 3D-SIM super-resolution microscope (Applied Precision) equipped with a 100×/1.40 numerical aperture UPLSAPO oil objective (Olympus), EMCCD cameras (Photometrics), and 405, 488, 568 nm lasers. Fifteen raw images per plane (5 phases at 3 angles) were captured with a Z-spacing of 0.125 μm using an oil with a refractive index of 1.515. To reduce spherical aberrations, an oil of optimal refractive index was first identified. Image reconstruction and registration was performed using SoftWoRx (GE, v6.5) employing channel-specific optical transfer functions (OTFs) and Wiener filter (settings: 0.0020 for red and green channel, 0.0050 for blue channel). Further processing of 32-bit images was performed using Fiji (ImageJ v1.51d) with in-house JavaScript scripts. Images were converted to 16-bit images. A mask for the XIST signal (red) was created on all the slices in Fiji using the thresholding menu option. DC1 (green signal) in the mask was extracted using Fiji’s math menu options. 3D Object Counter plugin was then used to count the green objects (DC1 signal) in the XIST of the nucleus (n = 5, 2 XIST and 2 autosomal domains per nucleus). For autosomal domains, areas showing dense DAPI staining were manually selected at the region of interest, DC1 signal (green) was obtained, and 3D objects were counted. Objects here refer to 3D objects identified based on distribution and centre of mass of red or green signal across contiguous image slices. To calculate the percentage fraction of DC1 signal that is localized in the XIST territory in various knockdowns, total red (XIST) and green (DC1) objects were also counted in each nucleus separately. Percentage DC1 per XIST object was calculated using the following formula:

where  = number of green objects (DC1) in XIST domain,  = total number of red (XIST) objects, and  = total number of green (DC1) objects in the nucleus. Two-tailed Mann–Whitney test was used to calculate statistical significance.

Validation of anti-YTHDC1 antibody for immunofluorescence imaging

HEK293T cells were transfected with pSuperior-EGFP constructs expressing shLacZ or shDC1 shRNA and incubated for 48 h. These cells (20,000 per well) were then seeded on a poly-l-lysine-coated coverslips (coverslips: 1.5 H, 12 mm, round, NC9455457, Fisher Scientific) in 24-well plates. Following a 12-h incubation, cells were processed for immunostaining using the immunofluorescence staining protocol of the 3D-SIM method given above. After the second formaldehyde fixation step, cells were washed three times with 1× PBS and stained with DAPI, washed, and mounted on slides in mounting media following a method similar to the 3D-SIM method. Slides were stored at 4 °C until imaging. DC1 was stained with rabbit anti-YTHDC1 antibody (ab122340, Abcam, 1:1,000) and eGFP (expressed from shRNA expressing plasmid) was stained with chicken anti-GFP antibody (ab13970, Abcam, 1:1,000). Donkey anti-rabbit IgG antibody-Alexa Fluor 568 conjugate (A10042, Thermo Fisher, 1:1,000) and goat anti-chicken IgY antibody-Alexa Fluor 488 conjugate (A-11039, Thermo Fisher, 1:1,000) were used to probe the primary antibodies. Images were captured on a wide-field fluorescence microscope (Nikon Eclipse Ti) using a 60× oil immersion objective. Images were processed on Fiji (ImageJ v1.51d).

Bacterial expression of His6-DF proteins

Full-length DF family cDNA ORFs were PCR amplified from HEK293T oligo-d(T)25-primed cDNA and cloned at NheI and XhoI for DF1 and DF3, NdeI and XhoI for DF2 in pET-28c(+) (Novagen) plasmid. These plasmids were transformed into Rosetta 2(DE3) Singles (Novagen) Escherichia coli cells. Bacteria were grown until they reached an OD600 nm of 0.5 and treated with 0.1 mM IPTG at 18 °C for 1–4 h to allow a comparable level of protein expression. Time points showing a similar level of protein expression for all the DF proteins were only analysed by western blot. DNA oligonucleotides used for amplification of the cDNA ORFs are given in Supplementary Table 8.

Enrichment of DC1-binding RNA motifs in different RNA and genomic features

For this analysis, all the 35,823 CIT sites were used. CITS were first mapped to the different genomic and RNA features in the hg19 genome using the annotation script, annotatePeaks.pl, from the Homer package. Sites mapping to rRNA, tRNA and the mitochondrial genome were discarded. For every site, strand-specific DNA sequence (± 20 nucleotides) was obtained from the hg19 genome. An enrichment of DC1-binding RNA motifs (DRACH, MTTAH, and KTCAHC) in different RNA/genomic features was determined using Centrimo tool in the MEME suite.

Confirmation of X-chromosome silencing by RT–qPCR

Total RNA was extracted and purified from 1 × 106 siRNA-transfected pSM33 cells using RNeasy Mini Kit (Qiagen) and DNA was removed by digestion with RNase-free DNase Set (Qiagen). DNA-free RNA (500 ng) was used to make cDNA with random hexamer using SuperScript III reverse transcriptase (Invitrogen) following the manufacturer’s instructions. Relative expression of genes Gpc4 and Atrx relative to Gapdh was quantified by qPCR using the LightCycler 480 SYBR Green I Master Mix (Roche). Primer information is given in Supplementary Table 8.