M6A-dependent RNA condensation underlies FUS autoregulation and can be harnessed for ALS therapy development

Abstract

Mutations in the FUS gene cause aggressive amyotrophic lateral sclerosis (ALS-FUS). Beyond mRNA, FUS generates partially processed transcripts retaining introns 6 and 7. We demonstrate that these FUSint6&7-RNA molecules form nuclear condensates, scaffolded by the highly structured intron 7 and associated with nuclear speckles. Using hybridization-proximity labeling proteomics, we show that the FUSint6&7-RNA condensates are enriched for splicing factors and the N6-methyladenosine (m6A) reader YTHDC1. These ribonucleoprotein structures facilitate posttranscriptional FUS splicing and depend on m6A/YTHDC1 for integrity. In cells expressing mutant FUS, FUSint6&7-RNAs become hypermethylated, which in turn stimulates their condensation and splicing. We further show that FUS protein is repelled by m6A. Thus, ALS-FUS mutations may cause abnormal activation of FUS posttranscriptional splicing through altered RNA methylation. Notably, ectopic expression of FUS intron 7 sequences dissolves endogenous FUSint6&7-RNA condensates, down-regulating FUS mRNA and protein. Our findings reveal a condensation-dependent mechanism regulating FUS splicing, with possible therapeutic implications for ALS.

Fig. 1.. Regulation of FUS RNA with retained introns 6 and 7.

(A) Splice sites flanking the retained region in FUSint6&7-RNA are weaker compared to other splice sites in the FUS gene. MaxEntScore was used to determine the splice site strength. All splice sites are plotted, and E6D and E8A are given in red. (B) FUS mislocalization in FUSΔNLS cell lines used in the study. ho, homozygous; het, heterozygous. (C and D) Reduced retention of introns 6 and 7 in FUS RNA demonstrated by PCR (C) and qRT-PCR (D). WT NH–WT cell samples not heated (NH) during RNA extraction. A combination of three primers, FUS_ex6_for, FUS_int6_rev, and FUS_ex8/9_rev, was used to detect intron 6 inclusion. *P < 0.05, N = 3 to 4, Kruskal-Wallis with Dunn’s post hoc test. M, DNA molecular weight marker. (E) FUS overexpression does not restore FUS intron 6 retention in FUSΔNLS lines, whereas expression of mutant FUS down-regulates FUSint6&7-RNA in WT cells. GFP-tagged FUS and its ALS-linked variants R522G (mainly cytoplasmic) and R518K (mainly nuclear) were used. N = 3 to 5. *P < 0.05, Mann-Whitney U test, as compared to WT cells expressing WT FUS-GFP. n.s., not significant. (F to H) FUSint6&7-RNA level remains unchanged in FUS KO cells with intact FUS transcription but undetectable FUS protein. FUS locus CRISPR-Cas9 editing schematic (F), FUSint6&7-RNA analysis by qRT-PCR (G), and RNA sequencing (H) are shown. *P < 0.05, N = 3, Mann-Whitney U test.

Fig. 2.. FUS RNA with retained introns forms dynamic nuclear foci.

(A) FUSint6&7-RNA forms nuclear foci. Representative images for RNA-FISH with a FUS intron 6–specific probe in HeLa cells are shown. Three types of foci, based on their size, are labeled. Arrows indicate the large (type 3) foci, which likely assemble near the sites of transcription. (B) Large foci (type 3) frequency per cell in different cell lines corresponds to the cell line ploidy. (C) FUSint6&7-RNA foci visualized using a FUS intron 6 RNAscope-ISH probe with chromogenic detection. Representative images for HeLa and SH-SY5Y cells are shown. (D to F) FUSint6&7-RNA foci formation relies on ongoing transcription and is sensitive to changes in RNA metabolism. HeLa cells were treated with actinomycin D (actD), CHX, or pladienolide B (plad.b) for 4 hours or with IFN-β for 24 hours. FUSint6-positive foci were quantified by a high-content imaging assay (D) and analyzed by qRT-PCR (E) and by high-resolution imaging (F). In (D), data are for six individual wells, from a representative experiment, *P < 0.05 and **P < 0.01, Kruskal-Wallis with Dunn’s test. In (E), N = 4 to 5, **P < 0.01, Kruskal-Wallis with Dunn’s test. (G) FUSint6&7-RNA foci form in cultured human motor neurons. Day 36 neurons were used for RNA-FISH and RNAscope-ISH (FUS intron 6–specific probes). DAPI, 4′,6-diamidino-2-phenylindole.

Fig. 3.. FUSint6&7-RNA foci are phase-separated condensates nucleated by intron 7 and associated with splicing speckles.

(A) FUSint6&7-RNA foci are associated with splicing speckles. PNN was a speckle marker; n = 20 cells. (B) RNase, but not DNase, treatment eliminates FUSint6&7-RNA foci. (C) FUSint6&7-RNA is semiextractable. qRT-PCR was performed in samples with or without heating before RNA isolation. N = 3, *P < 0.05, Mann-Whitney U test. (D) FUSint6&7-RNA is relatively stable. FUSint6&7-RNA and FUS mRNA levels were analyzed by qRT-PCR after a 4-hour actinomycin D treatment. GAPDH was used for normalization. N = 3 to 4, *P < 0.05, Mann-Whitney U test. h, hours. (E) FUSint6&7-RNA foci are sensitive to an LLPS-disrupting agent. 1,6-Hexanediol (1,6-HD) treatment was performed in semipermeabilized cells. Five to six fields of view were analyzed from a representative experiment; **P < 0.01, Mann-Whitney U test. (F and G) Comparable expression of exogenous FUS introns. Construct schematics (F) and qRT-PCR analysis (G) are shown. N = 3 to 4. OE, overexpression. (H) FUS intron 7, but not intron 6, forms compact, dense nuclear condensates upon ectopic expression. (I) FUS intron 7 is more structured than intron 6, with lower ensemble diversity. Minimum free energy structure and base-pairing probabilities heatmaps were generated by RNAfold. (J and K) Molecular dissection of FUS intron 7 condensation. Full-length (FL) intron without splice sites was used as a control. Regions analyzed (J) and representative images and quantification (K) are shown. (L to O) CRISPR-mediated deletion of a middle portion of FUS intron 6 destabilizes FUSint6&7-RNA and leads to FUS mRNA and protein depletion. Positions of sgRNAs (L), levels of FUS RNA species (M), FUSint6&7-RNA condensate analysis (N), and FUS protein levels (O) in homozygous Δint6 clones (cl.1 and 2) are shown. N = 3 to 4, *P < 0.05, Mann-Whitney U test. SH-SY5Y was used in (A) and HeLa in other panels. Scale bars, 2 μm in (A); 10 μm in [(B), (E), (H), (K), and (N)]; and 20 μm in (O).

Fig. 4.. FUSint6&7-RNA condensates regulate FUS mRNA levels.

(A) Ectopic expression of FUS intron 7 dissolves endogenous FUSint6&7-RNA condensates. Single FUSint7 oligonucleotide probe recognizing exoFUSint7 condensates but not endogenous FUSint6&7-RNA condensates was used in combination with the Stellaris FUSint6 probe pool. Representative images and quantification of endogenous FUSint6&7-RNA condensates are shown. Cells with high, medium, and low exoFUSint7 expression are indicated. Inset shows fusion of a FUSint6&7-RNA condensate with an exoFUSint7 condensate. Number of endogenous condensates quantified in exoFUSint7 condensate–containing cells (int7) versus nontransfected cells (NT) in the same field of view (FOV) is indicated inside the bars. Correlation between the area of exoFUSint7 signal in individual nuclei and the number of endogenous FUSint6&7-RNA condensates are also shown (25 cells). Scale bar, 10 μm. (B) Ectopic expression of exoFUSint7 or exoFUSint6 + 7 leads to FUS mRNA down-regulation. qRT-PCR analysis of FUS total (FUS mRNA + FUSint6&7-RNA) and mRNA levels is shown. N = 3 to 5, *P < 0.05, Kruskal-Wallis with Dunn’s test. (C) Ectopic expression of exoFUSint6 + 7 leads to FUS protein down-regulation. Representative Western blot and quantification for HeLa cells are shown. N = 6, **P < 0.01, Kruskal-Wallis with Dunn’s test. (D and E) FUS intron retention is responsive to cellular stress. FUS RNA levels were analyzed in cells recovering from NaAsO₂ stress (1-hour stress + 3-hour recovery) by PCR with a triple primer combination (D) and qRT-PCR (E). N = 5 to 7, **P < 0.01, Mann-Whitney U test. (F and G) ExoFUSint7 expression down-regulates FUS mRNA and protein in an ALS-FUS cell model. FUSΔNLS lines were analyzed by qRT-PCR (F) and Western blot (G). N = 3 to 4, *P < 0.05, Mann-Whitney U test. In (F) and (G), ΔNLS10 and ΔNLS4 lines were used, respectively. In (F), data for intron 7–specific primer were included to confirm successful exoFUSint7 overexpression. a.u., arbitrary units; vec, vector.

Fig. 5.. Proteomic analysis of FUSint6&7-RNA condensates by HyPro-MS.

(A) Experimental pipeline for HyPro-MS analysis of FUSint6&7-RNA condensates. (B) Efficient labeling of the endogenous FUSint6&7-RNA condensates using HyPro probes. Stellaris FUSint6–specific probe was used for costaining. Arrows indicate the foci labeled by both Stellaris and HyPro probes. Scale bar, 10 μm. (C) Principal component analysis (PCA) demonstrating clustering of triplicated HyPro-MS samples for both probes and the no-probe control. (D) Volcano plot for FUSint6&7-RNA condensates versus no-probe control (Ctrl). Proteins with Padj < 0.05 are labeled in black, and proteins with Padj < 0.05 and involved in RNA splicing and/or implicated in neurodegeneration are labeled in red. (E) Volcano plot for FUSint6&7-RNA condensates versus ACTB probe. Proteins with Padj < 0.1 are labeled in black, and the top 15 hits are labeled in red. (F) Dot plot of Gene Ontology (GO) Biological Process term enrichment analysis for nuclear proteins identified in FUSint6&7-RNA condensates and significantly enriched as compared to no-probe control. (G) Dot plot of GO Biological Process term enrichment analysis for nuclear proteins identified in FUSint6&7-RNA condensates and significantly enriched as compared to ACTB probe control. (H) Validation of HyPro-MS proteins enriched in FUSint6&7-RNA condensates, as well as FUS and other proteins previously shown to bind FUSint6&7-RNA. Cells expressing exoFUSint7 were analyzed by RNA-FISH and immunofluorescence with appropriate antibodies. ab1,2 - different FUS antibodies. Representative images are shown. Scale bar, 15 μm. Top graph: Twenty-eight to 35 transfected cells with condensates were analyzed per protein. Bottom graph: Forty-one and 27 individual condensates were analyzed for YTHDC1 and FUS enrichment, respectively, ****P < 0.0001, two-tailed unpaired t test.

Fig. 6.. m6A/YTHDC1 maintain FUSint6&7-RNA condensates.

(A and B) YTHDC1 depletion destabilizes FUSint6&7-RNA condensates. Representative images and quantification of condensates (A) and high-resolution images (B) are shown. One hundred seven and 65 cells (four to five FOVs) were analyzed for scrambled (scrmbl) and YTHDC1 siRNA, respectively. **P < 0.01, Mann-Whitney U test. (C) YTHDC1 depletion does not affect the FUSint6&7-RNA level but down-regulates FUS mRNA. N = 5 to 7, *P < 0.05, Mann-Whitney U test. (D and E) FUS introns 6 and 7 are extensively methylated. m6A-Atlas 2.0 was used for mapping methylation sites (D) and calculating the m6A mark density (E). Colored dots indicate cell lines. (F) A high-confidence DRACH motif in FUS intron 7 shown in a structural context. (G and H) Pharmacological inhibition of a m6A writer METTL3 destabilizes FUSint6&7-RNA condensates (G) without changing levels of this RNA (H). In (G), cells were treated for 16 hours, and >200 cells (five FOVs) were analyzed per condition, **P < 0.01, Mann-Whitney U test. In (H) (qRT-PCR), N = 4. (I) Pharmacological inhibition of a m6A eraser FTO promotes FUSint6&7-RNA condensate assembly. Seventy-nine, 106, and 117 cells (five FOVs) were analyzed for dimethyl sulfoxide (DMSO), 4- and 16-hour FB23-2 treatments, respectively. *P < 0.05, Kruskal-Wallis with Dunn’s test. (J) FUSint6 RNAscope-ISH reveals partial cytoplasmic redistribution of FUSint6&7-RNA in STM2457-treated cells. Arrows indicate cytoplasmic accumulations of this RNA; nucleus is circled. (K) Preserved or enhanced formation of FUSint6&7-RNA condensates in FUSΔNLS lines. Two hundred thirty-four, 239, 234, 141, and 62 cells were analyzed for WT, ΔNLS4, ΔNLS7, ΔNLS10, and ΔNLS11 lines, respectively (four to nine FOVs). *P < 0.05 and **P < 0.01, Kruskal-Wallis with Dunn’s test. (L) m6A RNA immunoprecipitation (MeRIP) coupled with qRT-PCR demonstrates increased FUSint6&7-RNA methylation in FUSΔNLS lines. N = 4, *P < 0.05, Kruskal-Wallis with Dunn’s test. (M) Enhanced association of FUSint6&7-RNA with YTHDC1 in FUSΔNLS lines, as demonstrated by RNA immunoprecipitation (RIP). Data were combined for three ΔNLS lines. *P < 0.05, Mann-Whitney U test (WT versus ΔNLS). Scale bars, 5 μm in (B) and 20 μm in other panels.

Fig. 7.. m6A has a repelling effect on FUS protein.

(A) Ectopically expressed GFP-tagged FUS is not enriched in the endogenous FUSint6&7-RNA condensates. FUSint6 Stellaris probe was used in HeLa cells. Note FUS protein accumulation in paraspeckles (arrows). Scale bar, 5 μm. (B) FUS overexpression does not affect the FUSint6&7-RNA condensate size. NT, nontransfected (cells analyzed in the same FOV). Condensate size was measured in 15 FOVs per condition. Scale bar, 10 μm. (C) M6A limits FUS condensation in vitro. Recombinant human his-tagged FUS and synthetic Cy5-labeled RNA oligonucleotides, fully methylated or unmodified, were used. Representative images of FUS condensates and quantification of average condensate size and fusion events are shown. Five to eight FOVs were analyzed per condition from a representative experiment. *P < 0.05 and ***P < 0.001, Kruskal-Wallis with Dunn’s test. (D) FUS-RNA complex formation is diminished by methylation, as analyzed by electrophoretic mobility shift assay. Recombinant FUS protein and RNA oligonucleotides as in (C) were used. Recombinant YTHDF1 protein was used in parallel as a positive control. Complex formation in the indicated area (blue line) was quantified by densitometry (5.0 and 7.5 μM data points combined). N = 3, *P < 0.05, Mann-Whitney U test. (E) Confocal nanoscanning (CONA) assay demonstrates reduced binding of FUS to methylated RNA, as compared to its unmodified counterpart. Cy5-labeled RNA oligonucleotides as above and lysates of cells expressing GFP-tagged FUS were used. GFP ring fluorescence intensity was normalized to Cy5 fluorescence–RNA coating. Eighty to 100 beads were analyzed per condition. ****P < 0.0001, Kruskal-Wallis with Dunn’s test. (F) AlphaFold3 modeling confirms the repelling properties of m6A in FUS binding to RNA. FUS RRM is in blue, and RNA is in orange (unmodified) or green (methylated). Interacting interfaces of FUS protein and the DRACH motif–containing RNA are boxed. Amino acid (aa) sequence interacting with the UUGG motif in the oligonucleotide is also indicated.

Fig. 8.. Regulation of FUS RNA condensation and posttranscriptional splicing in WT and ALS-FUS cells and the effect on exogenous FUS introns.

(A) Regulation of FUS expression by RNA condensation in healthy cells. (B) Dysregulated FUS RNA condensation in ALS-FUS cells. (C) Approach to FUS expression regulation based on the intron-mediated condensation mechanism.