Acute METTL3 depletion accelerates Xist-mediated gene silencing

In recent work, we analyzed the role of mA in regulating nascent RNA processing, making use of the dTAG degron system to rapidly deplete METTL3, the catalytic subunit of the mA writer complex, in iXist-ChrX (clone C7H8) mouse embryonic stem cells (mES cells). We used MeRIP-seq (mA RNA immunoprecipitation and sequencing) to show that acute depletion of METTL3 for 2 h, followed by 24 h of Xist induction with continued METTL3 depletion, results in a rapid transcriptome-wide loss of mA, including at major peaks in proximity to the A-repeat (exon 1) and E-repeat (exon 7) of Xist. To further verify this finding, we performed an extended METTL3 depletion (24 h) before Xist induction (Extended Data Fig. 1a). Both conditions result in a near complete loss of the majority of mA peaks across the transcriptome, including those on Xist (Extended Data Fig. 1b-e).

Here, we used the same degron strategy to investigate how acute loss of METTL3 and mA affects Xist-mediated chromosome silencing. iXist-ChrX mES cells have an interspecific Mus castaneus × 129 strain background with a stable XX karyotype and are engineered with a TetOn promoter for induction of Xist expression specifically from the Mus castaneus X chromosome. Accordingly, we are able to isolate chromatin-associated RNA and subject it to sequencing (ChrRNA-seq) to accurately determine the relative expression level of active X (Xa) and Xi alleles for a large number of X-linked genes that have informative single-nucleotide polymorphisms (SNPs). In addition to two previously described lines with a METTL3 C-terminal FKBP12 tag, we derived two independent lines with METTL3 tagged with FKBP12 on the N terminus (Fig. 1a) in the iXist-ChrX background. In all FKBP12-tagged cell lines METTL3 depletion occurred rapidly, within 2 h of adding the dTAG-13 reagent (Fig. 1b and Extended Data Fig. 2a), and moreover resulted in strongly reduced levels of METTL14, a subunit of the mA complex that heterodimerizes with METTL3 (Fig. 1b and Extended Data Fig. 2b). We noted a reduction in levels of N-terminal FKBP12-METTL3, indicating that insertion of the tag affects translation or stability of METTL3 protein (Extended Data Fig. 2c).

We went on to quantify Xist-mediated silencing following METTL3 depletion as outlined in Fig. 1c. We analyzed silencing at an early time point, after 24 h Xist induction (with or without prior dTAG-13 treatment for 2 h), to minimize potential indirect effects of mA depletion (Xist-mediated silencing in the iXist-ChrX mES cells occurs progressively over a period of around 6 days). The results are summarized in Fig. 1d. In the absence of dTAG-13 reagent, silencing levels in FKBP12-tagged lines were very similar to those seen in control (C7H8) cells, indicating that there are no effects attributable to addition of the degron at the C or N terminus of METTL3. Interestingly, addition of dTAG-13 resulted in a highly reproducible enhancement or acceleration of Xist-mediated silencing, evident in all four independently derived clones (Fig. 1d and Extended Data Fig. 1f). Of note, accelerated silencing is the converse of the silencing deficiency reported in prior work using constitutive knockout or knockdown of METTL3 and other subunits of the mA writer complex. This difference likely reflects that acute METTL3 depletion is less influenced by indirect effects compared to chronic knockout experiments. Indeed, global gene expression differences in our study correlate better with mA-modified mRNAs compared to published studies that used long-term knockout approaches (Extended Data Fig. 2d). Principal component analysis (PCA) comparing silencing in untreated cells with acute depletion of METTL3 after 24 h of Xist induction (Fig. 1e), together with silencing analysis for previously defined gene categories (Extended Data Fig. 2e-h), indicated that all X-linked genes are equally affected by accelerated silencing.

To determine whether the observed acceleration of Xist-mediated silencing is attributable to loss of METTL3 catalytic function, we performed complementation experiments using ectopic expression of GFP-METTL3 transgenes. Transgene constructs were integrated under the control of the Rosa26 constitutive promoter into the H5 clone with C-terminal degron-tagged METTL3 using CRISPR-Cas9-facilitated homologous recombination. In parallel, we established lines using a transgene encoding GFP-METTL3-D395A, a substitution that ablates METTL3 catalytic activity. Constitutive expression of GFP-METTL3 transgenes was maintained in both the presence and the absence of dTAG-13 (Extended Data Fig. 3a-c).

Both transgene encoded proteins reversed the observed reduction in METTL14 protein levels (Extended Data Fig. 3b,c versus Fig. 1b and Extended Data Fig. 2b), indicating the formation of stable GFP-METTL3/METTL14 heterodimers. Compared to wild-type (WT) GFP-METTL3, levels of GFP-METTL3-D395A were reduced (Extended Data Figs. 2c and 3b,c). This effect was not seen in the presence of dTAG-13 and is not linked to transcriptional levels (Extended Data Fig. 3b-d). A possible explanation is that METTL3-dependent mA autoregulates the complementary DNA-derived ectopic METTL3 transcript through RNA degradation. We went on to analyze Xist-mediated silencing in the transgene complementation lines. ChrRNA-seq analysis showed that ectopic expression of WT GFP-METTL3 fully complements the accelerated silencing phenotype observed following dTAG-13 treatment, whereas expression of GFP-METTL3-D395A has no effect (Fig. 1f and Extended Data Fig. 3e).

To confirm that accelerated Xi gene silencing is attributable to METTL3 catalytic activity, we made use of a recently developed pharmacological METTL3 inhibitor, STM2457 (ref. ). STM2457 treatment of parental C7H8 XX mES cells resulted in changed levels of YTHDC1 and WTAP and altered ratios of alternative splicing (Extended Data Fig. 4a,b), as occurs following acute depletion of METTL3 (ref. ). We then analyzed Xi gene silencing following treatment of cells with either DMSO or STM2457 for 6 h, followed by 24 h of Xist induction under continued treatment (Extended Data Fig. 4c). ChrRNA-seq analysis revealed that low-dose STM2457 treatment results in a modest increase in Xist RNA levels and accelerated XCI dynamics, with higher doses eliciting a clearly enhanced effect (Extended Data Fig. 4d,e). Collectively, these results confirm the role of METTL3 in regulating the rate of Xist-mediated silencing and further support that METTL3 function in this scenario is through mA catalysis.

We noted a close correlation between accelerated Xist-mediated silencing and levels of Xist RNA, as determined from ChrRNA-seq datasets. Specifically, increased Xist RNA levels up to approximately twofold were apparent following acute METTL3 depletion in the independent N-terminally and C-terminally tagged cell lines (Fig. 2a and Extended Data Fig. 1f). Additionally, complementation with WT GFP-METTL3 but not GFP-METTL3-D395A restored Xist RNA levels to those seen in untreated cells (Fig. 2b and Extended Data Fig. 3f).

To further investigate the effect of acute METTL3 depletion on Xist RNA, we used super-resolution three-dimensional structured illumination microscopy (3D-SIM) imaging to assay features of Xist RNA domains at the single-cell level. For these experiments, we engineered the METTL3 C-terminal degron into previously described iXist-ChrX XX mES cells in which the TetOn-inducible Xist allele has a Bgl stem-loop array integrated into Xist exon 7, allowing detection of Xist RNA molecules through binding of a BglG-HaloTag fusion protein labeled with fluorescent Halo dyes (Fig. 2c and Extended Data Fig. 2a). In prior work using this system, we reported that Xist RNA accumulates to maximal levels of around 50-100 molecules per cell over a period of 1.5-5 h, referred to as the expansion phase. A time point of 24 h was previously selected where Xist RNA levels have attained a steady state. Using these parameters, we observed an increase in the number of Xist molecules following acute METTL3 depletion (Fig. 2d,f), in agreement with the ChrRNA-seq data. In addition, the volume encompassing Xist centroids (overall Xist domain size) was also significantly increased at 1.5 h (expansion phase) and 24 h (steady-state phase) of Xist induction (Fig. 2e,f). These observations confirm that acute depletion of METTL3 and mA leads to increased Xist RNA levels and an enlarged Xist domains corresponding to the Xi territory.

We also investigated whether other well-characterized nuclear long noncoding RNAs (lncRNAs) that are mA modified are similarly affected by acute METTL3 depletion. Accordingly we examined levels of Neat1, Malat1 and Kcnq1ot1 RNAs, all of which are expressed in mES cells and have high levels of METTL3-dependent mA (Extended Data Fig. 5a-c). Levels of Neat1 and Malat1 were unaffected by acute METTL3 depletion; however, similarly to Xist RNA, Kcnq1ot1 levels increased approximately 1.5-2-fold (Extended Data Fig. 5d-f). The effect on Kcnq1ot1 levels was dependent on METTL3 catalytic activity (Extended Data Fig. 5f, right).

Increased abundance of Xist RNA following acute depletion of METTL3 could result from changes in the rate of Xist RNA transcription and/or RNA turnover. To investigate these possibilities, we applied RNA-SPLIT (sequential pulse localization imaging over time) coupled to super-resolution 3D-SIM microscopy to differentially label successive waves of Xist transcripts (presynthesized and newly synthesized) before fixation. The labeling regimen for determining turnover rates is shown in Fig. 3a. Experiments were performed at both the expansion phase and the steady-state phase using a 20-min interval. A 2-h dTAG-13 treatment was performed before Xist induction to minimize secondary or indirect effects. Example images in Fig. 3b are from the expansion phase. As shown previously, turnover of Xist RNA occurs within 140 min at the expansion phase and 220 min at the steady-state phase (Fig. 3b,c). In marked contrast, following acute METTL3 depletion, there was little Xist RNA turnover detectable across the entire time course of the experiment (220 min) during both the expansion and the steady-state phases (Fig. 3b,c and Supplementary Fig. 1). Reduced turnover of Xist transcripts was also demonstrated using an orthogonal approach, SLAM-seq, based on transient 4sU labeling of newly synthesized RNA (Extended Data Fig. 6a-c). Allele-specific analysis using these sequencing data confirmed accelerated silencing and increased Xist RNA levels following acute METTL3 depletion (Extended Data Fig. 7).

We further applied RNA-SPLIT to measure Xist transcription rates in the presence and absence of mA, achieved by quantifying foci for newly synthesized Xist RNA over time during expansion phase. As shown in Fig. 3d,e, there is a significantly reduced transcription rate in the acute METTL3 depletion condition compared to WT cells. This finding is consistent with prior work demonstrating a feedback mechanism that links Xist transcription and turnover. Accordingly, we conclude that loss of mA results in accelerated X-chromosome silencing because of overaccumulation of Xist transcripts.

The cellular functions of mA are mediated by reader proteins that can bridge to downstream pathways. YTHDC1 is the best-characterized protein that directly recognizes mA in the nucleus. Interestingly, YTHDC1 coimmunoprecipitation experiments revealed an association with ZCCHC8, a core subunit of the NEXT complex that targets nonpolyadenylated transcripts in the nucleus for degradation. Both YTHDC1 and ZCCHC8 have a role in degrading nonpolyadenylated chromatin-associated regulatory RNAs (carRNAs), for example, PROMPTs and eRNAs. Consistent with this finding, ZCCHC8 has been reported to interact with YTHDC1 in experiments using stable isotope labeling in cell culture and mass spectrometry. The YTHDC1-RNA exosome axis has also been implicated in the degradation of other nuclear RNAs, for example, SμGLT lncRNA and C9ORF72 repeat RNA.

To investigate whether YTHDC1 is important for regulating Xist RNA turnover, we used CRISPR-Cas9 facilitated genome editing to establish XX mES cell-derived lines with the FKBP12 degron tag inserted into the gene encoding YTHDC1 (Fig. 4a). YTHDC1 depletion on addition of dTAG-13 reagent was validated by western blot analysis (Fig. 4b,c) and examination of the effects on Tor1aip2 alternative last exon splicing, which is significantly affected by METTL3 and mA and conditional YTHDC1 knockout (Extended Data Fig. 8a,b). We went on to assay Xist-mediated silencing and Xist RNA levels following YTHDC1 depletion as described above. We observed no increase in the silencing rate or in the levels of Xist RNA in two independent degron-tagged cell lines (Fig. 4d,e and Extended Data Fig. 8c,d). If anything, both Xist RNA levels and the silencing efficiency of X-linked genes appeared modestly reduced compared to controls. However, this reduction was notably less pronounced than that observed with acute depletion of known XCI regulators, for example PCGF3/5 (ref. ) (Fig. 4d). Additionally, there was little or no effect on Xist RNA stability by acute depletion of YTHDC1, as determined by SLAM-seq (Extended Data Fig. 6c). Similarly, Kcnq1ot1 RNA, levels of which increase following METTL3 depletion, were unaffected by YTHDC1 depletion (Extended Data Fig. 5g-i, left).

We went on to investigate the role of the NEXT complex in Xist RNA turnover by inserting the FKBP12 degron tag into the gene encoding the core subunit ZCCHC8 in XX mES cells (Fig. 5a and Extended Data Fig. 9a,b). As an additional control, we established XX mES cell lines with the FKBP12 degron tag inserted into the gene encoding ZFC3H1, an essential subunit of the poly(A) tail exosome targeting (PAXT) complex (Fig. 5b and Extended Data Fig. 9a,c). PAXT mediates an alternate pathway for degradation of polyadenylated RNA in the nucleus, potentially functioning as a timer to remove aberrant RNAs that are not efficiently exported. dTAG-13 treatment led to rapid and complete depletion of the FKBP12-tagged proteins within 2 h (Fig. 5a,b). We noted that mA-dependent alternative last exon splicing of the Tor1aip2 gene was not affected by acute depletion of ZCCHC8 or ZFC3H1, indicating that neither NEXT nor PAXT is required for mA deposition on target mRNAs (Extended Data Fig. 9d,e). Acute ZCCHC8 depletion was further validated by monitoring upregulation of PROMPTs (Extended Data Fig. 10a,b).

We went on to perform ChrRNA-seq to assay Xist-mediated silencing after 24 h of Xist induction, in either the presence or the absence of NEXT or PAXT complexes using the approach described above for analysis of METTL3 and YTHDC1. As shown in Fig. 5c, acute depletion of ZCCHC8 resulted in strongly accelerated silencing, whereas depletion of ZFC3H1 resulted in a marginal reduction in gene silencing (Fig. 5d). Consistent with these observations, Xist RNA levels were elevated following depletion of ZCCHC8 but not of ZFC3H1, where Xist levels were slightly lower if anything (consistent with marginally reduced gene silencing) (Fig. 5e,f). The higher Xist RNA levels and stability following acute ZCCHC8 depletion are correlated with an even more marked acceleration of silencing than that was seen with acute METTL3 depletion (Fig. 1d versus Fig. 5c). Upregulation of Xist RNA upon acute depletion of ZCCHC8 and NEXT is evident across the entire transcript (Extended Data Fig. 10c). PCA indicates that accelerated silencing affects X-linked genes equivalently across the X chromosome, which contrasts with the silencing deficiency observed upon knockout of the key silencing factor SPEN or deletion of B/C-repeats in Xist (Extended Data Fig. 10d). Levels of Kcnq1ot1 RNA were similarly elevated following acute depletion of ZCCHC8 but not ZFC3H1 (Extended Data Fig. 5g-i, right). SLAM-seq analysis indicates that elevated Xist RNA levels following acute ZCCHC8 depletion are attributable to increased transcript stability (Extended Data Fig. 6c). Taken together these results suggest that mA promotes Xist RNA turnover by the NEXT complex independently of YTHDC1.