m1A in CAG repeat RNA binds to TDP-43 and induces neurodegeneration

Abstract

Microsatellite repeat expansions within genes contribute to a number of neurological diseases^1,2. The accumulation of toxic proteins and RNA molecules with repetitive sequences, and/or sequestration of RNA-binding proteins by RNA molecules containing expanded repeats are thought to be important contributors to disease aetiology^3-9. Here we reveal that the adenosine in CAG repeat RNA can be methylated to N¹-methyladenosine (m¹A) by TRMT61A, and that m¹A can be demethylated by ALKBH3. We also observed that the m¹A/adenosine ratio in CAG repeat RNA increases with repeat length, which is attributed to diminished expression of ALKBH3 elicited by the repeat RNA. Additionally, TDP-43 binds directly and strongly with m¹A in RNA, which stimulates the cytoplasmic mis-localization and formation of gel-like aggregates of TDP-43, resembling the observations made for the protein in neurological diseases. Moreover, m¹A in CAG repeat RNA contributes to CAG repeat expansion-induced neurodegeneration in Caenorhabditis elegans and Drosophila. In sum, our study offers a new paradigm of the mechanism through which nucleotide repeat expansion contributes to neurological diseases and reveals a novel pathological function of m¹A in RNA. These findings may provide an important mechanistic basis for therapeutic intervention in neurodegenerative diseases emanating from CAG repeat expansion.

Fig. 1. m¹A in CAG repeat RNA increases with repeat length and contributes to neurodegeneration in C. elegans.

a–c, The m¹A/rA ratios in ectopically expressed CAG repeat RNAs isolated from HEK293T cells (a; n = 3 biologically independent experiments), ectopically expressed (CAG)₃₈ RNA isolated from HEK293T cells with ectopic co-expression of the indicated RNA demethylases or stable knockdown of m¹A methyltransferases (b; n = 3 biologically independent experiments), and CAG repeat RNA from mouse brain (n = 11 and 12 biologically independent samples for striatum (str) and cortex (ctx) tissues, respectively), Drosophila head (n = 5 biologically independent samples, 10-day old males, 30 heads per sample) and C. elegans (n = 3 biologically independent experiments) (c). d,e, Fluorescence images of representative Q0, Q19 and Q67 nematodes, in Q67 nematodes with ectopic expression of wild-type human ALKBH3 (ALKBH3-WT) or its catalytically inactive mutant (ALKBH3-Mut) (d), and in Q67 nematodes with or without knockdown of W02A11.1 gene (e). f, Ratios of gaps in the nerve cord in Q67 worms with or without expression of human ALKBH3-WT and ALKBH3-Mut (Q0: n = 20 worms; Q19: n = 20 worms; Q67: n = 17 worms; Q67 + ALKBH3-WT: n = 20 worms; Q67 + ALKBH3-Mut: n = 15 worms), and in Q67 nematodes with or without knockdown of W02A11.1 gene (Q67 + vector RNAi: n = 22 worms; Q67 + W02A11.1 RNAi: n = 23 worms). The ratios of gaps were calculated by dividing the total length of gaps over the entire length of the nerve cord. Data are mean ± s.d. P values for the data of mouse brain and Drosophila head samples in c, and between Q67 + vector RNAi and Q67 + W02A11.1 RNAi in f were determined using two-tailed Student’s t-test; all other P values were determined using one-way ANOVA with Tukey’s multiple comparisons test. NS, not significant (P > 0.05). b, *P = 0.021. c, **P = 0.0058 and ***P = 0.0007. ****P < 0.0001. Scale bars, 100 μm. Source Data

Fig. 2. m¹A induces cytoplasmic mis-localization and aggregation of endogenous TDP-43.

a, Representative images of U2OS cells expressing CAG repeat RNA along with ectopic expression of ALKBH3 or shRNA-mediated knockdown of TRMT61A. Scale bars, 10 μm. b, Quantification of area of TDP-43 foci in a. (CAG)₂₂: n = 31; (CAG)₃₈: n = 33; (CAG)₂₂ + ALKBH3: n = 32; (CAG)₃₈ + ALKBH3: n = 30; shControl: n = 33; shTRMT61A-1: n = 31; shTRMT61A-3: n = 32. c, Percentage of TDP-43 co-localized with CAG repeat RNAs. Images are representative of five or six independent frames for each condition. n = 5 for (CAG)₂₂, (CAG)₃₈, (CAG)₂₂ + ALKBH3 and (CAG)₃₈ + ALKBH3; n = 6 for shControl, shTRMT61A-1 and shTRMT61A-3. Data are mean ± s.d. and represent three biologically independent experiments. P values were determined using one-way ANOVA with Tukey’s multiple comparisons test. *P = 0.020, **P = 0.0059 and ***P = 0.0009 between (CAG)₂₂ and (CAG)₃₈; ***P = 0.0003 between (CAG)₃₈ and (CAG)₃₈ + ALKBH3. Source Data

Fig. 3. m¹A enhances the ability of endogenous TDP-43 protein to partition into stress granules.

a, CAG repeat RNA-mediated localization of endogenous TDP-43 into stress granules in U2OS cells with ectopic expression of ALKBH3 or knockdown of TRMT61A. b, FISH and immunofluorescence microscopy were performed to assess the co-localization between G3BP1 and CAG repeat RNAs in U2OS cells with ectopic expression of ALKBH3 or knockdown of TRMT61A. c, Percentages of TDP-43 co-localized with G3BP1. Images are representative of 5 to 7 independent frames for each condition: n = 5 for (CAG)₂₂; n = 7 for (CAG)₃₈ and (CAG)₂₂ + ALKBH3; n = 6 for (CAG)₃₈ + ALKBH3, shControl, shTRMT61A-1 and shTRMT61A-3. d, Percentages of G3BP1 co-localized with CAG repeat RNAs. Images are representative of 6 or 7 independent frames for each condition: n = 7 for (CAG)₂₂, (CAG)₃₈, (CAG)₂₂ + ALKBH3, (CAG)₃₈ + ALKBH3 and shControl; n = 6 for shTRMT61A-1 and shTRMT61A-3. Data are mean ± s.d. and represent three biologically independent experiments. P values were determined using one-way ANOVA with Tukey’s multiple comparisons test. c, *P = 0.033 between (CAG)₂₂ and (CAG)_38; *P = 0.043 between (CAG)₂₂ and (CAG)₂₂ + ALKBH3; **P = 0.0030. d, **P = 0.0072. Scale bars, 10 μm. Source Data

Fig. 4. Synthetic m¹A-containing CAG repeat RNA triggers cytoplasmic redistribution of endogenous TDP-43 and its co-localization with stress granules, and a proposed model illustrating a role of m¹A in CAG repeat length-dependent modulation of biophysical properties of TDP-43.

a, Representative images showing the localization of TDP-43 and G3BP1 in U2OS cells with or without transfection with synthetic CAG repeat RNAs containing zero or three m¹A. Scale bars, 10 μm. b, Sizes of TDP-43 foci induced by synthetic CAG repeat RNA containing zero or three m¹A. n = 45 for (CAG)₇-0m¹A and (CAG)₇-3m¹A; n = 46 for (CAG)₁₆-0m¹A and (CAG)₁₆-3m¹A. Data are mean ± s.d. and represent three biological replicates. P values were determined using one-way ANOVA with Tukey’s multiple comparisons test. c, A model illustrating expanded CAG repeat RNA-induced aberrant phase separation and cytoplasmic redistribution of TDP-43. TDP-43 bound to (CAG)₂₂ RNA maintains liquid-like state and is soluble. (CAG)₃₈ RNA triggers aberrant phase transition of TDP-43 into insoluble inclusions, in which m¹A modification in CAG repeat RNA has an important role in stimulating the liquid-to-gel-like transition of TDP-43 by increasing the local concentration of the protein, thereby promoting its aggregation through its LCD. Source Data

Extended Data Fig. 1. Expression of CAG repeat RNA suppresses the expression of ALKBH3, but not TRMT61A in cells and striatum tissues of mice.

a-c, Western blot images and quantification results showing the relative expression levels of TRMT61A in HEK293T cells with ectopic expression of (CAG)₂₂ and (CAG)₃₈ RNA (a) and in striatum (Str) and cortex (Ctx) tissues of Q7 and Q140 mice (b, c). n = 5 biologically independent samples. Data are mean ± s.d. P values were determined using Two-tailed Student’s t-test. d-e, Western blot images and quantification results showing the relative expression levels of ALKBH3 in HEK293T cells with ectopic expression of (CAG)₂₂ and (CAG)₃₈ RNA. f, RT-qPCR results showing the relative expression levels of ALKBH3 mRNA in HEK293T cells with ectopic expression of (CAG)₂₂ and (CAG)₃₈ RNA, or with empty plasmid control (Contr). Data in e-f are mean ± s.d., and represent three biologically independent experiments. P values were determined using one-way ANOVA with Tukey’s multiple comparisons test. ns, P > 0.05; **P = 0.0041 in e; ***P = 0.0005 between Contr and (CAG)₂₂, and ***P = 0.0001 between (CAG)₂₂ and (CAG)₃₈ in f; ****P < 0.0001. g-j, Western blot images and quantification results showing the relative expression levels of ALKBH3 protein in striatum (g-h) and cortex (i-j) tissues of Q7 and Q140 mice (n = 5 biologically independent samples of striatum or cortex). The loading control was detected in a separate gel in parallel for d and g. Data are mean ± s.d. P values of Str or Ctx were determined using Two-tailed Student’s t-test. ns, P > 0.05; ***P = 0.0041 in h. For gel source data, see Supplementary Fig. 1. Source Data

Extended Data Fig. 2. ALKBH3-WT, but not ALKBH3-Mut extends the lifespan of Drosophila expressing expanded Q78 CAG repeats, and reduces the ratios of m¹A/rA, but not m⁶A/rA in CAG repeat RNA.

a, The mRNA level of wild-type human ALKBH3 (ALKBH3-WT) or its catalytically inactive mutant (ALKBH3-Mut) in Q78 Drosophila was measured by RT-qPCR (n = 6 biologically independent samples, 15 heads per sample). Data are mean ± s.d. P values were determined using one-way ANOVA with Tukey’s multiple comparisons test. ns, P > 0.05; **P = 0.0016; ****P < 0.0001; b, Drosophila lifespan was evaluated with co-expression of wild-type human ALKBH3 or its catalytically inactive mutant in Q78 Drosophila (n = 150 animals per survival curve). c-d, The levels of m¹A (c) and m⁶A (d) in Q78 flies with expression of wild-type human or its catalytically inactive mutant. Data are mean ± s.d., and represent three biologically independent samples (30 heads per sample). Ten-day old male flies on RU486 food were used for the experiments. P values were determined using one-way ANOVA with Tukey’s multiple comparisons test. ns, P > 0.05; *P = 0.012 between Q78 and Q78 + ALKBH3-WT, and *P = 0.027 between Q78 + ALKBH3-WT and Q78 + ALKBH3-Mut in c. Source Data

Extended Data Fig. 3. Electrophoretic mobility shift assay (EMSA) for determining the binding affinities of TDP-43 with rA-, m¹A- and m⁶A-carrying CAG repeat RNA.

a, Construct designs for TDP-43 vectors containing functional (WT) or RNA-binding-defective (5FL) RRMs were used for EMSA. b-i, EMSA for measuring the binding affinities of full-length TDP-43 (b-e) and TDP-43-RRM (f-i) with rA-, m¹A- and m⁶A-carrying RNA probes (5′- CCGUUCCGCCCXGGCCGCGCCCAGCUGGAAUGCA-3′, where ‘X’ is rA, m¹A or m⁶A) (b-c, f-g) or with rA-, and m¹A-carrying CAG repeat RNA probes (5′-CAGCAGCAGCXGCAGCAGCAG-3′, where ‘X’ is rA or m¹A) (d-e, h-i). Protein concentrations ranged from 0 to 1000 nM. Data are mean ± s.e.m., and represent three biologically independent experiments. Shown in (j) and (k) are the gel image and quantification results for the K_d values for wild-type TDP-43 and its RRM mutant (TDP-43-5FL) in binding with (CAG)₇-1 m¹A. For gel source data, see Supplementary Fig. 1. Source Data

Extended Data Fig. 4. m¹A in CAG repeat RNA induces truncation of TDP-43 protein.

a-f, Detergent-solubility fractionation and Western blot for endogenous TDP-43 protein in cells with ectopic expression of CAG repeat RNA (a, b), with or without ectopic expression of ALKBH3-WT or ALKBH3-Mut (c, d), or with or without transfection of synthetic (CAG)₇ or (CAG)₁₆ RNAs containing zero or three m¹A (e, f). Data are mean ± s.d. and represent three biologically independent experiments. P values were determined using one-way ANOVA with Tukey’s multiple comparisons test. ns, P > 0.05; **P = 0.0066 and *P = 0.041 in b; ***P = 0.0004 between (CAG)₂₂ + EV and (CAG)₃₈ + EV, ***P = 0.0003 between (CAG)₃₈ + EV and (CAG)₃₈ + ALKBH3-WT, and ***P = 0.0006 between (CAG)₃₈ + ALKBH3-WT and (CAG)₃₈ + ALKBH3-Mut in d; *P = 0.011, and **P = 0.0065 in f; g-h, Quantification results showing the fractions of truncated TDP-43 protein in striatum (n = 5 biologically independent samples) and cortex (n = 5 biologically independent samples) tissues of mice. Data are mean ± s.d. P values were determined using Two-tailed Student’s t-test. ns, P > 0.05; **P = 0.0013. The loading control was detected in a separate gel in parallel for a, c, e and g. For gel source data, see Supplementary Fig. 1. Source Data

Extended Data Fig. 5. m¹A triggers cytoplasmic mis-localization and aggregation of ectopically expressed EGFP-TDP-43.

a, Representative images of U2OS cells co-expressing EGFP-TDP-43 and CAG repeat RNAs with or without ALKBH3 overexpression. b, Quantification results for the areas of EGFP-TDP-43 foci. (CAG)₂₂: n = 69; (CAG)₃₈: n = 69; (CAG)₂₂ + ALKBH3: n = 68; (CAG)₃₈ + ALKBH3: n = 66. Data are mean ± s.d., and represent three biologically independent experiments. P values were determined using one-way ANOVA with Tukey’s multiple comparisons test. ****P < 0.0001. c, RNA FISH for assessing the co-localization of EGFP-TDP-43 with CAG repeat RNAs. d, Percentages of CAG repeat RNAs co-localized with EGFP-TDP-43. Data are mean ± s.d. and represent three biologically independent experiments. P values were determined using one-way ANOVA with Tukey’s multiple comparisons test. **P = 0.0012 between (CAG)₂₂ and (CAG)₃₈, **P = 0.0033 between (CAG)₂₂ and (CAG)₂₂ + ALKBH3, **P = 0.0074 between (CAG)₂₂ + ALKBH3 and (CAG)₃₈ + ALKBH3; ***P = 0.0006 between (CAG)₃₈ and (CAG)₃₈ + ALKBH3, and ***P = 0.0002 between (CAG)₃₈ and (CAG)₃₈ + TDP-43-5FL. Scale bar, 10 μm. Source Data

Extended Data Fig. 6. (CAG)₃₈ RNA triggers phase transition of TDP-43 protein in cells.

a, FRAP of EGFP-TDP-43 droplets in U2OS cells at 24 h following transfection. Scale bar, 5 μm. Higher magnification is shown in the right panel at different time points following photobleaching at 0 s (scale bar, 2 μm). b, Mean fluorescence intensity of EGFP-TDP-43 in the fully bleached area over time. Data are mean ± s.e.m. from the recovery curves obtained from three biologically independent experiments. c, Representative images of U2OS cells expressing (CAG)₂₂ RNA with or without treatment with 6% 1,6-hexanediol (1,6-HD) for 1 min. Scale bar, 10 μm. d, Partition coefficients of TDP-43 protein in (CAG)₂₂- or (CAG)₃₈-expressing cells with or without treatment with 6% 1,6-HD. The partition coefficient was calculated based on the fraction of EGFP-TDP-43 signal in the droplets over EGFP-TDP-43 signal in immediately adjacent region. n = 27 foci of EGFP-TDP-43 for each group with three biologically independent experiments. Data are mean ± s.e.m. P values were determined using Two-tailed Student’s t-test. ns, P > 0.05; ****P < 0.0001. e, Phase separation of EGFP-TDP-43 and EGFP-TDP-43-ΔLCD proteins in U2OS cells with or without ectopic expression of (CAG)₃₈ mRNA. The experiments were repeated independently three times with similar results. Scale bar, 10 μm. Source Data

Extended Data Fig. 7. m¹A modulates liquid-liquid phase separation of TDP-43 protein.

a, The droplets of EGFP-TDP-43 and EGFP-TDP-43-5FL were detected in the presence of synthetic (CAG)₇ RNAs (50 ng) containing 0, 1 and 3 m¹A residues, or without RNA. b, Partition coefficients of wild-type and mutant TDP-43 proteins in the presence or absence of m¹A-containing CAG repeat RNA. 24 or 25 TDP-43-EGFP droplets observed in three biologically independent experiments for each group were quantified. n = 24 for TDP-43-WT + No RNA and TDP-43-WT + 3m¹A; n = 25 for TDP-43-WT + Control RNA, TDP-43-WT + 1m¹A,TDP-43-5FL + No RNA, TDP-43-5FL + Control RNA, TDP-43-5FL + 1m¹A, and TDP-43-5FL + 3m¹A. Data are mean ± s.e.m. P values were determined using one-way ANOVA with Tukey’s multiple comparisons test. ns, P > 0.05; ***P = 0.0009; ****P < 0.0001. c, The droplets of TDP-43-ΔLCD were detected with CAG repeats RNAs (50 ng) or without RNA. Scale bar, 5 μm. d, TDP-43-EGFP (50 μM) was imaged using confocal microscopy. The fusions of TDP-43-EGFP droplets are also shown in Supplementary Videos 1–4. Scale bar, 5 μm. The experiments shown in c and d were repeated at least three times with similar results. e, Representative examples of FRAP analyses of in vitro phase-separated TDP-43 droplets in the presence of synthetic (CAG)₇ RNAs containing 0, 1 and 3 m¹A residues, or without RNA. FRAP are also shown in Supplementary Videos 5–8. f, FRAP curves displaying the relative TDP-43 intensity of phase-separated droplets in the presence of (CAG)₇ RNAs containing 0, 1 and 3 m¹A residues, or without RNA. The data represented mean ± s.e.m. from the recovery curves obtained from three biologically independent experiments. Scale bar, 2 μm. Source Data