m6A-binding YTHDF proteins promote stress granule formation

Abstract

Diverse RNAs and RNA-binding proteins form phase-separated, membraneless granules in cells under stress conditions. However, the role of the prevalent mRNA methylation, m⁶A, and its binding proteins in stress granule (SG) assembly remain unclear. Here, we show that m⁶A-modified mRNAs are enriched in SGs, and that m⁶A-binding YTHDF proteins are critical for SG formation. Depletion of YTHDF1/3 inhibits SG formation and recruitment of mRNAs to SGs. Both the N-terminal intrinsically disordered region and the C-terminal m⁶A-binding YTH domain of YTHDF proteins are important for SG formation. Super-resolution imaging further reveals that YTHDF proteins appear to be in a super-saturated state, forming clusters that often reside in the periphery of or at the junctions between SG core clusters, and potentially promote SG formation by reducing the activation energy barrier and critical size for SG condensate formation. Our results suggest a new function of the m⁶A-binding YTHDF proteins in regulating SG formation.

Fig. 1 |. m⁶A-modified mRNAs are enriched in stress granules (SGs) in U-2 OS cells under oxidative stress.

a, Immunofluorescence staining of mRNA m⁶A (with polyclonal anti-m⁶A antibody from Abcam) in U-2 OS cells treated with 0.5 mM NaAsO₂ for 30 min shows enrichment of m⁶A signal in SGs and P-bodies. SG marker G3BP1 and P-body marker DCP1A are detected using immunofluorescence and polyA signal is detected using FISH. Images are representative examples from three independent experiments. The bottom right panel shows box plots of the enrichment of m⁶A signal and polyA signal in SGs. The box plots show the median (middle lines), 25%-75% quartiles (boxes), and standard deviation (error bars). n =55 cells, from 3 independent experiments. P value was determined using unpaired Mann–Whitney U test, two-sided. b, c, Transcriptome-wide analysis of the fractions of mRNAs that are enriched in SGs (magenta), depleted in SGs (cyan), and neither enriched nor depleted in SGs (yellow), as a function of the m⁶A ratio for individual mRNAs that are longer than 3000 nt (b), or shorter than 3000 nt (c). The m⁶A ratio is defined as the percentage of transcripts that contain m⁶A. n = 9049 genes in total. d, Examples of smFISH images showing that the mRNA with a higher m⁶A ratio tends to have a higher degree of enrichment in SGs. smFISH signals (magenta) were overlaid with SG-marker G3BP1 (green) and P-body-marker DCP1A (blue). CDKN1A, which has a low m⁶A ratio, shows a relatively low degree of colocalization with G3BP1, whereas GPR75, which has a higher m⁶A ratio, shows a higher degree of colocalization with G3BP1. Cells treated with 0.5 mM NaAsO₂ for 30 min to induce oxidative stress. Images are representative examples from three independent experiments. e, The fraction of mRNA localized in SG measured using smFISH versus the m⁶A ratio for individual mRNAs. Error bars represent mean ± SEM. n = 32 cells (RB1), 60 cells (PRPF8), 50 cells (CDKN1A), 70 cells (TFRC), 24 cells (EIF3A), 32 cells (EFR3A), 13 cells (SPTBN1), 87 cells (MCL1), 159 cells (FAT1), 43 cells (PANX1), 51 cells (GPR75), from 3 independent experiments each.

Fig. 2 |. YTHDF proteins promote SG formation.

a, Two-color immunofluorescence images of YTHDF1 and G3BP1 show the disappearance of large SGs upon YTHDF1 and YTHDF3 double siRNA knockdown. The upper panels show the images for cells treated with control (scrambled) siRNA and the lower panels show the images of the YTHDF1/3 double knockdown cells. Images are representative examples from three independent experiments. b, Quantification of the fraction of G3BP1 in SGs for U-2 OS cells treated by control siRNA, single knockdown cells treated with YTHDF1, YTHDF2 or YTHDF3 siRNA, YTHDF1/3 double knockdown cells, and YTHDF1/2/3 triple knockdown cells. Oxidative stress in these cells was induced by 0.5 mM NaAsO₂ treatment for 30 min. n = 234 cells (control siRNA), n = 242 cells (YTHDF1 siRNA), n = 111 cells (YTHDF2 siRNA), n = 204 cells (YTHDF3 siRNA), n = 357 cells (YTHDF1/3 siRNA), n = 118 (YTHDF1/2/3 siRNA), from 3 independent experiments each. c, Fraction of polyA (black) and m⁶A (orange) signals in SGs in cells treated with control siRNA as well as in YTHDF1/3 double knockdown cells. n = 256 cells (control siRNA), n = 203 cells (YTHDF1/3 siRNA), from 3 independent experiments each. d,e, Overexpression of full-length YTHDF1, YTHDF2, and YTHDF3 proteins rescues the SG formation in YTHDF1/3 knockdown cells. All constructs are tagged with SNAP at the C-terminal end. Overexpressed proteins were imaged using a fluorescent dye that labels the SNAP-tag. d, Two-color images of SNAP-tag, detected by dye molecules conjugated to SNAP, and G3BP1, detected by immunofluorescence, for cells expressing a control SNAP-tag plasmid that does not contain YTHDF (upper panels) and for cells expressing SNAP-YTHDF1 (lower panels). Cells were treated by 0.5 mM NaAsO₂ for 25 min to induce oxidative stress. Images are representative examples from three independent experiments. e, Quantification of the fraction of G3BP1 in SGs for YTHDF1/3 double knockdown cells overexpressing the control SNAP-tag, and for YTHDF1/3 double knockdown cells overexpressing SNAP-YTHDF1, SNAP-YTHDF2 or SNAP-YTHDF3. n = 125 cells (Control), n = 46 cells (YTHDF1), n = 156 cells (YTHDF2), n = 12 cells (YTHDF3), from 3 independent experiments each. For b,c,e, the boxplots show the median (middle lines), 25%-75% quartiles (boxes), and Tukey-style whiskers extend to the most extreme datapoint within 1.5 × interquartile ranges (IQR) beyond the box. P values were determined by comparing with controls; Unpaired Mann–Whitney U test, two-sided.

Fig. 3 |. Both the N-terminal intrinsically disordered region (N-IDR) and the C-terminal YTH domain are important for YTHDF’s role in promoting SG formation.

a, Amino acid composition and predictions of IDRs, prion-like domains (PLDs), and secondary structures in YTHDF1 protein. The likelihood scores for being disordered predicted by PONDR-VSL2 (red) and being PLD-like predicted by PLAAC (cyan) for each amino acid are plotted in the upper panel. The secondary structure prediction by NetSurfP-2.0 is shown immediately below the disordered and PLD-like likelihood scores. Locations for several conserved amino acids in all three YTHDF proteins that are important for IDR/PLD properties are marked. b, Differently truncated YTHDF1 constructs tagged with SNAP at the C-terminal end. c,d, Overexpression of different YTHDF protein fragments in YTHDF1/3 knockdown cells showing that truncated YTHDF constructs lacking the N-IDR or YTH domain cannot rescue SG-formation. c, Two-color images of SNAP-tag, detected by dye molecules conjugated to SNAP, and G3BP1, detected by immunofluorescence, for cells expressing a YTHDF1 fragment lacking the YTH domain (upper panels) or the N-IDR (lower panels). The cells are treated with 0.5 mM NaAsO₂ for 25 min to induce oxidative stress. Images are representative examples from three independent experiments. d, Quantification of the fraction of G3BP1 in SGs for YTHDF1/3 double knockdown cells overexpressing the full-length YTHDF1 (black), or YTHDF1/3 double knockdown cells overexpressing YTHDF1/2/3 fragments lacking the C-terminal YTH domain (magenta), YTHDF1 fragment lacking the N-IDR, or YTHDF1 fragment lacking both N-IDR and P/Q-PLD (orange). The boxplots show the median (middle lines), 25%-75% quartiles (boxes), and Tukey-style whiskers extend to the most extreme datapoint within 1.5 × IQR beyond the box. P values were determined by comparing with full-length YTHDF1; Unpaired Mann–Whitney U test, two-sided. n = 46 cells (YTHDF1-FL), n = 14 cells (YTHDF1-N), n = 21 cells (YTHDF2-N), n = 19 cells (YTHDF3-N), n = 33 (YTHDF1-C1), n = 33 cells (YTHDF1-C2), from 3 independent experiments each.

Fig. 4 |. Inhibiting m⁶A-binding of YTHDF proteins partially impairs SG formation.

a, A construct that contains CRY2olig, a blue-light inducible oligomerization domain, and BFP (upper), and a construct that contains CRY2olig, BFP, and a mutated YTHDF1 fragment (YTHDF1(D401N)-C), which harbors a single amino acid mutation D401N in the RNA-binding YTH domains and lacks the N-IDR (lower). This mutation increases the m⁶A binding affinity of the YTH domain by 10-fold. b, c, d, Formation of SGs is reduced in cells overexpressing construct containing the YTHDF1(D401N)-C mutant. b, Two-color images of BFP and G3BP1 for cells expressing a control plasmid that contains CRY2olig-BFP but does not contain the YTHDF1(D401N)-C mutant (upper panels) and for cells expressing CRY2olig-BFP-YTHDF1(D401N)-C (lower panels). Cells are treated with 0.5 mM NaAsO₂ for 25 min to induce oxidative stress. Images are representative examples from three independent experiments. c, d, Quantification of the fraction of G3BP1 in SGs (c) and number of SGs per cell (d) for U-2 OS cells overexpressing CRY2olig-BFP and U-2 OS cells overexpressing CRY2olig-BFP-YTHDF1(D401N)-C. The cells were treated with 0.5 mM NaAsO₂ for 25 min. n = 72 cells (CRY2olig-BFP), n = 62 cells (CRY2olig-BFP-YTHDF1(D401N)-C), from 3 independent experiments each. e, Light-induced oligomerization of YTHDF1(D401N)-C construct could not rescue SG formation in YTHDF1/3 double knockdown cells. Three-color images of BFP-tag, detected by immunofluorescence, G3BP1, detected by immunofluorescence, and DCP1A, detected by immunofluorescence, for U-2 OS cells treated by control siRNA with overexpression of CRY2olig-BFP (upper panels), by YTHDF1/3 siRNA with overexpression of CRY2olig-BFP (middle panels) and by YTHDF1/3 siRNA with overexpression of CRY2olig-BFP-YTHDF1(D401N)-C (lower panels). Although we observed slightly enriched G3BP1 signal colocalizing with BFP signals in cells treated with YTHDF1/3 siRNA and CRY2olig-BFP-YTHDF1(D401N)-C, they largely represent DCP1A-positive P-bodies. Images are representative examples from three independent experiments. f, g, Quantification of the fraction of G3BP1 in SGs (f) and number of SGs per cell (g) for YTHDF1/3 double knockdown U-2 OS cells overexpressing CRY2olig-BFP or CRY2olig-BFP-YTHDF1(D401N)-C construct (orange) shown in comparison with cells treated by control siRNA with overexpression of CRY2olig-BFP (black). The small fractions of G3BP1 that colocalize with DCP1A were included in the SG quantifications in f and g. The cells were treated with 0.5 mM NaAsO₂ for 30 min and with three repeating cycles of 5 min blue-light-on and 5 min blue-light-off. n = 62 cells (control siRNA + CRY2olig-BFP), n = 37 cells (YTHDF1/3 siRNA + CRY2olig-BFP), n = 170 cells (YTHDF1/3 siRNA + CRY2olig-BFP-YTHDF1(D401N)-C), from 3 independent experiments each. For c,d,f,g, the boxplots show the median (middle lines), 25%-75% quartiles (boxes), and Tukey-style whiskers extend to the most extreme datapoint within 1.5 × IQR beyond the box. P values were determined by comparing with control siRNA + CRY2olig-BFP; Unpaired Mann-Whitney U test, two-sided.

Fig. 5 |. YTHDF protein reduces the critical size and activation energy barrier for SG condensate formation.

a, STORM imaging of G3BP1 protein in U-2 OS cells treated with control siRNA or YTHDF1/3 siRNA. The left panel shows an image of an unstressed cell treated by control siRNA, the middle panel shows an image of a NaAsO₂-treated cell treated with control siRNA, and the right panel shows the image of a NaAsO₂-treated, YTHDF1/3 knockdown cell. Insets are zoom-in images of regions in yellow boxes. Images are representative examples from three independent experiments. b, Distribution of cluster radius for G3BP1 protein under the three different conditions described in (a). More than 160000 clusters from ~60-80 cells from three independent experiments were pooled and analyzed for each condition. Counts of clusters with a radius larger than 50 nm are displayed. c, Diagram of Gibbs free energy change (ΔG) for cluster formation as a function of the cluster radius (R) for sub-saturated (b > 0), saturated (b = 0), and super-saturated (b < 0) states in the classical nucleation theory. ΔG contains two terms - a surface energy term and a bulk energy term: ΔG = aR² + bR³. When molecules are in the super-saturated state, a critical radius (R_c = −2a/3b) exists, beyond which the cluster continues to grow in size irreversibly. E_a, the value of ΔG at R_c, represents the activation energy barrier for super-critical cluster (condensate) formation. d, Log-log plots of normalized −Log(P) vs. R calculated from the size distribution of G3BP1 clusters in unstressed cells (black), NaAsO₂-treated cells (red) and NaAsO₂-treated, YTHDF1/3 knockdown cells (magenta). P is the probability density of the clusters with radius R. In a steady-state system, P of sub-critical clusters (R < R_c) follow Boltzmann distribution: $P=A{e}^{-∆G/k_{B}T}$. Thus, $\Delta G (in k$_{B$T)=-\mathrm{L}\mathrm{o}\mathrm{g}\left(P\right)-c$}, where $c=-\mathrm{L}\mathrm{o}\mathrm{g}A$. By fitting the data of −Log(P) versus R to the equation of $-Log\left(P\right)=a{R}^2+b{R}^3+c$, we obtained the values of a, b, from which we calculated the corresponding R_c and E_a. The normalized values of −Log(P), which equal to −Log(P) − c and are equivalent to ΔG for sub-critical clusters, are plotted here. Data for clusters with radii between 50 nm and 350 nm are shown. The plot from one of the three independent experiments is shown here, and all three independent experiments show similar plots. Mean values of calculated critical radius R_c are shown as dashed lines with the shaded areas representing SEM for the 3 independent experiments. e, R_c and E_a for super-critical cluster formation of G3BP1 derived from the measured cluster size distributions for the three conditions. Mean ± SEM are shown (n = 3 independent experiments for each condition). P values were determined by unpaired two-tailed Student’s T-Test. f, Single-color STORM imaging of YTHDF1 and two-color STORM imaging of YTHDF1 and G3BP1 in unstressed (upper panel) and NaAsO₂-treated cells (lower panels). YTHDF1 forms clusters, which often reside on the periphery of G3BP1 clusters or in between G3BP1 clusters in SGs. Images are representative examples from three independent experiments. g, Log-log plots of normalized −Log(P) vs. R calculated from the size distribution of YTHDF1 clusters in unstressed (black) and NaAsO₂-treated cells (red). The plot from one of the two independent experiments is shown here, and all independent experiments show similar plots. h, The b values in ΔG for YTHDF1 (Mean, n = 2 independent experiments) and G3BP1 (Mean ± SEM, n = 3 independent experiments) clusters in unstressed U-2 OS cells. Dot plots of individual data points are overlaid on bar graphs. For all data in this figure, cells were treated with 0.5 mM NaAsO₂ for 30 min to induce oxidative stress.