O-GlcNAcylation determines the translational regulation and phase separation of YTHDF proteins

Abstract

N⁶-methyladenosine (m⁶A) is the most abundant internal mRNA nucleotide modification in mammals, regulating critical aspects of cell physiology and differentiation. The YTHDF proteins are the primary readers of m⁶A modifications and exert physiological functions of m⁶A in the cytosol. Elucidating the regulatory mechanisms of YTHDF proteins is critical to understanding m⁶A biology. Here we report a mechanism that protein post-translational modifications control the biological functions of the YTHDF proteins. We find that YTHDF1 and YTHDF3, but not YTHDF2, carry high levels of nutrient-sensing O-GlcNAc modifications. O-GlcNAcylation attenuates the translation-promoting function of YTHDF1 and YTHDF3 by blocking their interactions with proteins associated with mRNA translation. We further demonstrate that O-GlcNAc modifications on YTHDF1 and YTHDF3 regulate the assembly, stability and disassembly of stress granules to enable better recovery from stress. Therefore, our results discover an important regulatory pathway of YTHDF functions, adding an additional layer of complexity to the post-transcriptional regulation function of mRNA m⁶A.

Figure 1.. YTHDF1 and YTHDF3 are modified by O-GlcNAc.

(a) The three YTHDF paralogs have low sequence conservation (above) and similar intrinsically disorder regions (bottom) along the effector regions instead of YTH domains. By LC-MS/MS identification, YTH domains are modified by similar modifications in all three YTHDFs, but effector domains of YTHDF1 and YTHDF3 are modified by O-GlcNAc. PTM sites identified by LC-MS/MS of YTHDF proteins from HEK293T cells are highlighted with different shapes. Effector domain, P/Q/N-rich domain and YTH domain are annotated. Blue indicates sequence conservation sites. Green indicates predicted disorder regions. (b) Strategy of O-GlcNAcylation analysis using two-step labelling protocol of chemo-enzymatic labelling and CuAAC or SPAAC. First, protein with O-GlcNAcylation is labelled by Gal-T1(Y289L) and conjugated with GalNAz containing azide handler on GlcNAc moiety. Then, the azide group is reacted with biotin-alkyne or DBCO-PEG5K by CuAAC or SPAAC. Finally, biotin signal is detected by streptavidin blot and O-GlcNAcylation stoichiometry is determined with mass-tag by western blot. Besides, shifted band number indicates O-GlcNAcylated site number. (c) O-GlcNAcylation analysis of immuno-precipitated flag tagged YTHDF1, YTHDF2 and YTHDF3 from HEK293T cells. SA denotes streptavidin blot by streptavidin-HRP. (d) O-GlcNAcylation stoichiometry analysis of YTHDF1, YTHDF2 and YTHDF3 in HEK293T and HeLa with indicated antibodies for western blot. The shifted bands pointed by arrowheads are target proteins with O-GlcNAcylation. Two shifted bands indicate O-GlcNAcylation occurs on two sites. (e) Quantification result of O-GlcNAcylation level of YTHDF1 and YTHDF3 in (d). Data are presented as mean ± SD (n = 3 biologically independent replicates). Significance was calculated using two-sided t-test. (f) O-GlcNAcylation analysis of total endogenous O-GlcNAcylation levels in HEK293T and HeLa. Coomassie staining is performed as loading control. (g) O-GlcNAcylation stoichiometry analysis of transfected flag tagged YTHDF1, YTHDF2 and YTHDF3 in HEK293T and HeLa. Flag* denoted samples which were performed O-GlcNAcylation stoichiometry analysis and immunoblotting by anti-flag antibodies. (h) Expression level analysis of OGT and OGA by western blot with actin as the loading control in HEK293T and HeLa cell lysate. (i) Quantification result of expression level of OGT and OGA between HEK293T and HeLa in (d). Data are presented as mean ± SD (n = 3 biologically independent replicates). Significance was calculated using two-sided t-test.

Figure 2.. Reversible O-GlcNAcylation sites of YTHDF1/3.

(a) The O-GlcNAc modified region in the effect domain is highlighted by a red rectangle with peptide sequence displayed. In the figure, blue indicates sequence conservation and the modified sites are highlighted by red boxes. (b) O-GlcNAcylation stoichiometry analysis of WT-YTHDF1 and YTHDF1-mutations. A representative mass spectrum of YTHDF1 peptide with S196 (c) or S157 (d) modified by O-GlcNAc. The spectra of unmodified peptides were used for side-by-side comparisons. The arrows pointed to the signature b/y-ions in modification spectrum. O-GlcNAcylation stoichiometry analysis of YTHDF1 in HEK293T treated with or without 50 μM OGT inhibitor OSMI-1 for 5 hr (e) or HeLa cells treated with or without 10 μM OGA inhibitor Thiamet G for 5 hr (f). O-GlcNAcylation stoichiometry analysis of YTHDF1 in HEK293T (g) or HeLa cells (h) with different glucose concentrations. Flag* and YTHDF1* denote samples which are performed O-GlcNAcylation stoichiometry analysis and immunoblotting by the respective antibodies. Arrowheads indicates the O-GlcNAc modified proteins.

Figure 3.. O-GlcNAcylation of YTHDF1/3 repressed their protein interaction.

(a) Enrichment analysis of proteins enriched in WT-YTHDF1, YTHDF1-Mut and Both groups by immunoprecipitation experiments and label-free quantification with Flag tagged WT-YTHDF1 and YTHDF1-Mut in HEK293T cells. The numbers of enriched proteins are annotated at the bottom. Size of circle denotes gene ratio. Color of circle denotes adjusted p value. (b) Interaction network of enrichment groups highlighted by red in (a) and exclusively enriched in the YTHDF1-Mut group. Proteins involved in translation, protein folding and mRNA processing are annotated by the color. The translation-related proteins are highlighted by a red oval. (c) Pairwise comparison of label-free quantification of co-immunoprecipitated proteins interacting with WT-YTHDF1 and YTHDF1-Mut from HEK293T cells. And the position of the x and y axes reflected the average LFQ intensities in the WT-YTHDF1 and YTHDF1-Mut, respectively. YTHDF1 protein partners were plotted in green (40S ribosomal proteins), blue (60S ribosomal proteins), red (translation initiation factors, EIF), and orange (Heterogeneous nuclear ribonucleoprotein, HNRNP), respectively. EIFs enriched in YTHDF1-Mut group were annotated. The diagonals denoted the fold change threshold of −0.5, 0 and 0.5 (Log₁₀(WT-YTHDF1 / YTHDF1-Mut)). (d) Co-IP analysis of RPS6, EIF2A and EIF2S3 by N-WT-YTHDF1 and N-YTHDF1-Mut from HEK293T cells. Flag* denotes samples which are performed O-GlcNAcylation stoichiometry analysis and immunoblotting by flag antibodies. Arrowheads indicates the O-GlcNAc modified proteins. (e) Enrichment analysis of proteins enriched in WT-YTHDF3, YTHDF3-Mut and Both groups by immunoprecipitation experiments and label-free quantification with Flag tagged WT-YTHDF3 and YTHDF3-Mut in HEK293T cells. (f) Pairwise comparison of label-free quantification of co-immunoprecipitated proteins interacting with WT-YTHDF3 and YTHDF3-Mut from HEK293T cells. The diagonals denoted the fold change threshold of −0.5, 0 and 0.5 (Log₁₀(WT-YTHDF3 / YTHDF3-Mut)). (g) Distribution of PAR-CLIP peaks across the length of mRNA. (h) Binding motif of WT-YTHDF1/3 (left) and YTHDF1/3-Mut (right) from PAR-CLIP of HEK293T cells.

Figure 4.. O-GlcNAcylation of YTHDF1 repressed target mRNA translation efficiency.

(a) Schematic of the tethering reporter assay. 3×BoxB sequence is inserted at 3’ UTR of firefly luciferase mRNA reporter (F-luc-3BoxB). The N-terminal domain of WT-YTHDF1/3 and YTHDF1/3-Mut are fused with λ peptide used to recognize BoxB RNA sequence. The expression of luciferase mRNAs is examined in HEK293T and HeLa cells, and the construct containing λ peptide alone is used as a negative control. The F-luc signal is normalized by R-luc in the same plasmid. Translation efficiency analysis of N-WT-YTHDF1-λ, N-YTHDF1-Mut-λ and λ in HEK293T (b) or HeLa cells (c). Error bar, mean ± SD, n = 3 biologically independent repeats. Significance was calculated using two-sided t-test. (d) Translation efficiency analysis of N-WT-YTHDF1-λ and λ treated with or without OGA inhibitor Thiamet G in HeLa. Thiamet G: 5 μM, 5 hr. Error bars, mean ± SD, n = 3 biologically independent repeats. Significance is calculated using two-sided t-test. (e-h) Cumulative distribution of translation efficiency for YTHDF1/3 target mRNA and non-target mRNA. Ythdf1/3 knock down HEK293T (e-f) and HeLa (g-h) are rescued by WT-YTHDF1/3 (e, g) or YTHDF1/3-Mut (f, h). Distribution with boxplot bounds depict quartile 1, median and quartile 3, with whiskers at 1.5× interquartile range and outlier points. The p values were calculated using two-sided t-test (right boxes).

Figure 5.. Dynamic O-GlcNAcylation of YTHDF1/3 regulated m⁶A mRNA translation efficiency in a cell cycle dependent manner.

(a) Strategy of dataset analysis for m⁶A containing mRNA translation efficiency in M and S phase. The mRNA translation efficiency, YTHDF1 PAR-CLIP and IP, and YTHDFs iCLIP datasets are acquired and combined. Distribution of mRNA translation efficiency in different groups is presented and compared. (b) Overlap of mRNAs of translation efficiency dataset and YTHDF1 target mRNAs dataset. The dataset of m⁶A-modified mRNA in HeLa cells is generated by PAR-CLIP and/or the immuno-purified ribonucleoprotein complex of YTHDF1. The dataset of the translation efficiency of mRNA in HeLa cells is generated by ribosome profiling in the M and S phases, respectively. The mRNAs in green and red are used for subjected analysis. (c) Cumulative distribution log₂-fold changes of translation efficiency of YTHDF1 target mRNAs between M and S phase (left box). Distribution with boxplot and p values are calculated using two-sided t-test (right box). Boxplot bounds depict quartile 1, median and quartile 3, with whiskers at 1.5× interquartile range and outlier points. (d) Overlap of mRNAs of translation efficiency dataset and YTHDFs iCLIP for m⁶A site numbers dataset. mRNAs in green and red are used for subjected analysis. (e) Cumulative distribution log₂-fold changes of translation efficiency of mRNAs containing different m⁶A numbers between M and S phase (left box). The dataset of m⁶A-modified mRNA in HeLa cells is generated by iCLIP. Distribution with boxplot and p values are calculated using two-sided t-test (right box). Boxplot bounds depict quartile 1, median and quartile 3, with whiskers at 1.5× interquartile range and outlier points. (f) O-GlcNAcylation stoichiometry analysis for YTHDF1/3 in HeLa at S and M phase. Cyclin A2 and H3pS10 are used as cell cycle marker of S and M phase. YTHDF1* and YTHDF3* denote samples which were performed O-GlcNAcylation stoichiometry analysis and immunoblotting by respective antibodies. Arrowheads indicates the O-GlcNAc modified proteins. (g) Quantification analysis of O-GlcNAcylation stoichiometry analysis for YTHDF1 in HeLa at different cell cycle stages by double thymidine release. Grey shading indicates the dynamic change window shown in (h). (h) Monitor of luciferase expression in HeLa at the indicated cell cycle stages by nocodazole release. Y axis, d(F-luc/R-luc)/dt, indicating the changing rate of protein production. Error bar, mean ± SD, n = 3 biologically independent repeats. (i) Expression level comparison of OGT and OGA between S and M phase in HeLa by western blot. Cyclin A2 and H3pS10 were used as cell cycle marker of S and M phase. Actin was used as loading control. (j) Quantification analysis of expression level of OGT and OGA in (i). Error bar, mean ± SD, n = 3 biologically independent repeats. (k) Interacting protein analysis of flag-YTHDF1 between S and M phase in HeLa by immunoprecipitation and LC-MS/MS. The EIFs were presented, and the outer layer of the circle was colored based on the fold change of label-free quantification. The proteins involved in translation regulation and RNA-binding were colored in inner layer of the circle by yellow and green respectively.

Figure 6.. O-GlcNAcylation of YTHDF1/3 increased their dynamic nature in stress granules.

(a) Colocalization of EGFP-YTHDF1 and G3BP1 under stress in HEK293T and HeLa. Scale bars: 10 μm. (b) Strategy of O-GlcNAcylation stoichiometry analysis inside or outside of stress granules. HEK293T cells are treated with NaCl stress. Then, the lysate is centrifuged for SGs isolation and subjected to O-GlcNAcylation stoichiometry analysis for YTHDF1/3. (c) O-GlcNAcylation stoichiometry analysis of YTHDF1/3 in HEK293T. G3BP1 is used as stress granules marker. SG, stress granule fraction. non-SG, lysate without stress granule fraction. YTHDF1* and YTHDF3* denote samples were performed O-GlcNAcylation stoichiometry analysis and immunoblotting by respective antibodies. Arrowheads indicates the O-GlcNAc modified proteins. (d) Quantification result of the fraction of YTHDF1 in stress granules, in HEK293T overexpressing EGFP-WT-YTHDF1 treated with or without OSMI-1, or in HeLa overexpressing EGFP-WT-YTHDF1 treated with or without Thiamet G, n = 30. Representative immunostaining result is shown in Fig S9a–b. Error bar, mean ± SD. Significance was calculated using two-sided t-test. (e) Immunostaining of Ythdf1/3 knock down HeLa rescued by Flag-WT-YTHDF1 or Flag-YTHDF1-Mut and treated with 0.6 M sorbitol stress. Scale bars: 10 μm. (f) Quantification result of fraction of YTHDF1 in stress granules of (e), n = 69, 55. Error bar, mean ± SD. Significance is calculated using two-sided t-test. (g-i) FRAP analysis of EGFP-WT-YTHDF1 and EGFP-YTHDF1-Mut treated with 0.6 M sorbitol stress in HeLa. Scale bars: 1 μm. Error bar, mean ± SD, n = 3. Significance was calculated using two-sided t-test. (j-k) Time lapse imaging of EGFP-WT-YTHDF1 and EGFP-YTHDF1-Mut released from 0.6 M sorbitol stress in living HeLa cells. Scale bars: 10 μm. Error bar, mean ± SD, n=3. Significance is calculated using two-sided t-test. (l) Monitor of luciferase expression in HEK293T after 1 hr sorbitol stress treatment. Grey shading indicates the sorbitol treatment window. Y axis, d(F-luc/R-luc)/dt, indicating the changing rate of protein production. Error bar, mean ± SD, n = 3 biologically independent repeats.

Figure 7.. A proposed model of regulation of YTHDF1/3 by O-GlcNAcylation.

(a) YTHDF1/3 instead of YTHDF2 can be reversibly modified by protein O-GlcNAcylation. YTHDF1/3 mediates enhanced translation of m⁶A-modified mRNA. However, the enhanced translation is suppressed by O-GlcNAcylation. In contrast, YTHDF2 reduces the stability of m⁶A-modified mRNA. (b) O-GlcNAcylation of YTHDF1/3 is reversibly regulated in different cell lines, culture conditions and cell cycle state. High O-GlcNAcylation of YTHDF1/3 represses m⁶A-modified mRNA translation promotion. Meanwhile, low O-GlcNAcylation of YTHDF1/3 promotes m⁶A-modified mRNA translation by recruiting eIF, ribosome and other translation related proteins. O-GlcNAcylation of YTHDF1/3 regulate the assembly, stability, and disassembly of stress granule for a better recovery from stress.