O-GlcNAcylation promotes the cytosolic localization of the m6A reader YTHDF1 and colorectal cancer tumorigenesis

O-linked GlcNAc (O-GlcNAc) is an emerging post-translation modification that couples metabolism with cellular signal transduction by crosstalk with phosphorylation and ubiquitination to orchestrate various biological processes. The mechanisms underlying the involvement of O-GlcNAc modifications in N6-methyladenosine (m6A) regulation are not fully characterized. Herein, we show that O-GlcNAc modifies the m6A mRNA reader YTH domain family 1 (YTHDF1) and fine-tunes its nuclear translocation by the exportin protein Crm1. First, we present evidence that YTHDF1 interacts with the sole O-GlcNAc transferase (OGT). Second, we verified Ser196/Ser197/Ser198 as the YTHDF1 O-GlcNAcylation sites, as described in numerous chemoproteomic studies. Then we constructed the O-GlcNAc-deficient YTHDF1-S196A/S197F/S198A (AFA) mutant, which significantly attenuated O-GlcNAc signals. Moreover, we revealed that YTHDF1 is a nucleocytoplasmic protein, whose nuclear export is mediated by Crm1. Furthermore, O-GlcNAcylation increases the cytosolic portion of YTHDF1 by enhancing binding with Crm1, thus upregulating downstream target (e.g. c-Myc) expression. Molecular dynamics simulations suggest that O-GlcNAcylation at S197 promotes the binding between the nuclear export signal motif and Crm1 through increasing hydrogen bonding. Mouse xenograft assays further demonstrate that YTHDF1-AFA mutants decreased the colon cancer mass and size via decreasing c-Myc expression. In sum, we found that YTHDF1 is a nucleocytoplasmic protein, whose cytosolic localization is dependent on O-GlcNAc modification. We propose that the OGT–YTHDF1–c-Myc axis underlies colorectal cancer tumorigenesis.


Reviewed by members of the JBC Editorial Board. Edited by Robert Haltiwanger
O-linked GlcNAc (O-GlcNAc) is an emerging post-translation modification that couples metabolism with cellular signal transduction by crosstalk with phosphorylation and ubiquitination to orchestrate various biological processes. The mechanisms underlying the involvement of O-GlcNAc modifications in N 6 -methyladenosine (m 6 A) regulation are not fully characterized. Herein, we show that O-GlcNAc modifies the m 6 A mRNA reader YTH domain family 1 (YTHDF1) and finetunes its nuclear translocation by the exportin protein Crm1. First, we present evidence that YTHDF1 interacts with the sole O-GlcNAc transferase (OGT). Second, we verified Ser196/ Ser197/Ser198 as the YTHDF1 O-GlcNAcylation sites, as described in numerous chemoproteomic studies. Then we constructed the O-GlcNAc-deficient YTHDF1-S196A/S197F/ S198A (AFA) mutant, which significantly attenuated O-GlcNAc signals. Moreover, we revealed that YTHDF1 is a nucleocytoplasmic protein, whose nuclear export is mediated by Crm1. Furthermore, O-GlcNAcylation increases the cytosolic portion of YTHDF1 by enhancing binding with Crm1, thus upregulating downstream target (e.g. c-Myc) expression. Molecular dynamics simulations suggest that O-GlcNAcylation at S197 promotes the binding between the nuclear export signal motif and Crm1 through increasing hydrogen bonding. Mouse xenograft assays further demonstrate that YTHDF1-AFA mutants decreased the colon cancer mass and size via decreasing c-Myc expression. In sum, we found that YTHDF1 is a nucleocytoplasmic protein, whose cytosolic localization is dependent on O-GlcNAc modification. We propose that the OGT-YTHDF1-c-Myc axis underlies colorectal cancer tumorigenesis.
The N 6 -methyladenosine (m 6 A) modification is quite abundant on internal mRNAs, and its function and regulation have caught a wave of intense investigations (1,2). Its numerous writers, erasers, and readers are under stringent control (3), and one of the readers is YTH domain family 1 (YTHDF1) (4), which promotes translation efficiency during arsenite recovery (5). YTHDF1 enhances translation in adult mouse dorsal root ganglions during injury recovery and augments axonal regeneration (6). YTHDF1 fuels translation upon neuronal stimuli, which is conducive to learning and memory (7). YTHDF1 also recognizes m 6 A-marked lysosomal protease mRNAs, thus mediating the decay of neoantigens and bolstering tumor suppressive immunotherapy (8). Recently, YTHDF1 and YTHDF3 are also found to promote stress granule formation, as m 6 A mRNAs are found to be enriched in stress granules (9).
The interconnection between m 6 A mRNA and cancer are being revealed (10)(11)(12), and m 6 A participates in many aspects of tumor biology: cancer stem cell, tumor cell proliferation, or oncogene expression. YTHDF1, in particular, has been found at the nexus of multiple tumorigenic pathways. YTHDF1 binds the m 6 A-modified mRNA of c-Myc, whose enhanced translation promotes glycolysis and cancer cell proliferation (13). In non-small cell lung cancer, YTHDF1 upregulates the translation efficiency of CDK2, CDK4, and cyclin D1, and YTHDF1 is also elevated in people who live at high altitudes, possibly through the hypoxia Keap1-Nrf2-AKR1C1 pathway (14). In gastric cancer, YTHDF1 enhances the expression of frizzled 7, a key Wnt receptor that hyperactivates the Wnt-β-catenin pathway (15). In ovarian cancer, YTHDF1 promotes the translation of eukaryotic translation initiation factor 3 subunit C (EIF3C), a component of the protein translation initiation factor EIF3 complex (16). In cervical cancer, YTHDF1 elevates the translation of hexokinase 2 via binding with its 3 0 -UTR, thus promoting the Warburg effect (17). All these findings suggest that YTHDF1 binds with its targets via m 6 A mRNA and is fundamental during human carcinogenesis.
Investigations show that some of the m 6 A regulators are subject to post-translational modifications. YTHDF2, another m 6 A reader that mediates mRNA decay (18), is subject to SUMOylation at K571 upon hypoxia stress (19). SUMOylation alters the binding affinity of YTHDF2 with m 6 A, thus deregulating the downstream target genes, leading to lung cancer progression (19). An m 6 A writer, methyltransferase-like 3 (METTL3), is modified by lactylation at its zinc-finger domain, changing its RNA capturing capacity and regulating immunosuppression of tumor-infiltrating myeloid cells (20). METTL3 is also acetylated, which regulates its localization and cancer metastasis (21).
The O-linked GlcNAc (O-GlcNAc) glycosylation occurs intracellularly (22,23). Functioning as a rheostat to environmental stress or cellular nutrient status, O-GlcNAc monitors transcription, neural development, cell cycle, and stress response (22,23). However, whether it plays a role in m 6 A regulation has remained enigmatic. Historically O-GlcNAc studies have been strenuous due to technical impediments. Recent years have witnessed the combined methodology of chemoenzymatic labeling, bioorthogonal conjugation, and electron-transfer dissociation mass spectrometry (MS), which have smoothened the way for biological investigations. Previously, an isotope-tagged cleavable linker together with chemoenzymatic labeling screen identified the O-GlcNAc sites of YTHDF1 to be S196 and S198 (24). A second enrichment strategy using Gal labeling followed by chemical oxidation points the YTHDF1 O-GlcNAcylation region as . In another isotope-targeted glycoproteomic study in T cells, YTHDF1 O-GlcNAcylation occurs on several residues, including Ser196, Ser197, and Ser198 (26). In this manuscript, we first confirmed that YTHDF1 O-GlcNAcylation occurs on Ser196, Ser197, and Ser198. Then we found that YTHDF1 is a nucleocytoplasmic protein with exportin 1 (CRM1) mediating its cytoplasmic shuttling. We further presented evidence that O-GlcNAcylation promotes YTHDF1 cytosolic localization, thus enhancing downstream target expression, such as c-Myc. Our results were further correlated with The Cancer Genome Atlas analysis combined with mouse xenograft models. Our data highlight the significance of glycosylation in m 6 A regulation and tumorigenesis.

YTHDF1 is O-GlcNAcylated at Ser196, Ser197, and Ser198
As YTHDF1 has been reproducibly identified in O-GlcNAc profiling screens (24,27,28), we first assessed the binding affinity between YTHDF1 and the sole O-GlcNAc writer-O-GlcNAc transferase (OGT). 293T cells were transfected with Flag-YTHDF1 and HA-OGT plasmids, and the two overproduced proteins coimmunoprecipitate (Fig. 1A). When the endogenous proteins were examined, YTHDF1 proteins were also present in the anti-OGT immunoprecipitates (Fig. 1B). Pull-down assays were then utilized to evaluate the physical association. 293T cells were transfected with HA-OGT, and the cell lysates were incubated with recombinant glutathione-S-transferase-YTHDF1 proteins, which pulled down overproduced OGT proteins (Fig. 1C). When pull-down assays were carried out between recombinant OGT and YTHDF1, again glutathione-S-transferase-YTHDF1 pulled down His-OGT (Fig. 1D), suggesting that OGT and YTHDF1 directly interact in vivo and in vitro.
Then we assessed the O-GlcNAcylation of YTHDF1. 293T cells were enriched for O-GlcNAc by supplementing the media with glucose and Thiamet-G (TMG, the OGA inhibitor) (TMG + Glu) as previously described (29,30). The endogenous YTHDF1 proteins were immunoprecipitated from the lysates, and RL2 antibodies detected a crisp band upon O-GlcNAc enrichment (Fig. 1E), suggesting that YTHDF1 is indeed O-GlcNAcylated. We then decided to mutate the three Ser, as several chemoproteomic studies have identified YTHDF1 O-GlcNAcylation sites to be Ser196-198 (24,27,28). Thus we generated a YTHDF1-S196AS197FS198A (AFA) mutant. When we transfected the WT and AFA mutant into cells, the AFA mutant significantly diminished YTHDF1 O-GlcNAcylation levels ( Fig. 1F), suggesting that they are the main glycosylation sites.

Crm1 mediates the nuclear cytoplasmic shuttling of YTHDF1
To investigate the potential functions of YTHDF1 O-GlcNAcylation, we first employed an immunoprecipitation (IP)-MS analysis. Flag-YTHDF1 plasmids were transfected into cells, and the lysates were immunoprecipitated with anti-Flag antibodies. Interestingly, the MS results revealed many importins and exportins (Table S1). When we did literature research, YTHDF1 was among the binding partners of exportin 1 (Crm1) in a recent proteomic study (31). Therefore, we suspect that YTHDF1 might be a nuclear cytoplasmic protein and Crm1 might mediate the process.
To test this possibility, we first assessed the association between YTHDF1 and Crm1. We also utilized KPT-330, a Crm1 inhibitor. Overexpressed YTHDF1 coimmunoprecipitates with Crm1, and KPT-3301 markedly reduced the interaction ( Fig. 2A). Moreover, endogenous YTHDF1 interacts with Crm1 (Fig. 2B), suggesting that YTHDF1 could be a nuclear cytoplasmic protein. We then utilized the nuclear cytoplasmic fractionation assay, and fractionation results revealed a nuclear portion of YTHDF1 (Fig. 2C). We further adopted KPT-330 in the fractionation assay and found that it significantly enhanced the nuclear fraction of YTHDF1 (Fig. 2D). Furthermore, in immunofluorescence staining samples, both endogenous YTHDF1 and overproduced YTHDF1 manifested significant upregulation of nuclear staining signals (Fig. 2, E and F), suggesting that Crm1 could export YTHDF1 to the cytosol.
Then we directly measured the effect using the AFA mutant. YTHDF1-AFA displayed marked reduction in association with Crm1 (Fig. 3D). In the fractionation analysis, AFA again manifested a much higher portion in the nucleus (Fig. 3E). Lastly, we employed fluorescence microscopy to visualize whether O-GlcNAcylation affected YTHDF1 localization. As shown in Figure 3, F and G, both 5S-G treatment and the AFA mutant enhanced nuclear YTHDF1 staining, probably by blocking its nuclear export via Crm1. These The cell lysates were subject to immunoprecipitation and immunoblotting with the antibodies indicated. B, HeLa cell lysates were immunoprecipitated with anti-OGT antibodies and immunoblotted with the indicated antibodies. C, 293T cells were transfected with HA-OGT, and the cell lysates were incubated with recombinant GST-YTHDF1. D, recombinant His-OGT and GST-YTHDF1 proteins were incubated and subject to pull-down assays. E, cells were treated with the OGA inhibitor Thiamet-G(TMG) and glucose to enrich for O-GlcNAcylation as described previously (29). Then the cell lysates were immunoprecipitated with anti-YTHDF1 antibodies and immunoblotted with anti-O-GlcNAc RL2 antibodies. F, YTHDF1-S196A, S197F, S198A, -S196 (AFA) mutants were constructed, and the cells were transfected with HA-OGT together with SFB-YTHDF1-WT and SFB-YTHDF1-AFA mutants.

O-GlcNAcylation of YTHDF1
assays suggest that YTHDF1 O-GlcNAcylation promotes the binding between YTHDF1 and Crm1 and the resultant nuclear export.

A potential NES lies in proximity to YTHDF1 O-GlcNAcylation sites
We were curious why O-GlcNAcylation has such a conspicuous effect on YTHDF1 localization and looked for potential nuclear export signals (NESs) surrounding the S196S197S198 region. As NES consists of the Φ1-X(2-3)-Φ2-X(2-3)-Φ3-X-Φ4 motif (Φ: hydrophobic amino acid) (33), we found a potential NES juxtaposing the 196-198 Ser cluster (Fig. 4A). We mutated the corresponding hydrophobic amino acid to Ala and generated 4A (Fig. 4A), as previously described for the NES of the cyclic GMP-AMP synthase (34). When we examined for YTHDF1-Crm1 association, the 4A mutant significantly downregulated binding with Crm1 (Fig. 4B). In the fractionation studies, 4A also elevated nuclear YTHDF1 localization (Fig. 4C). In the immunofluorescent staining experiments, 4A also had a more prominent nuclear localization pattern than the control (Fig. 4D). Combined, these data suggest that O-GlcNAcylation might boost the association of the neighboring NES with Crm1.

O-GlcNAcylation of YTHDF1 MD simulations suggest that S197 O-GlcNAcylation increases the interaction between NES and Crm1 via hydrogen bonds
We then explored deeper as why O-GlcNAcylation increases binding with Crm1. Recently, a structural study focusing on the interface between NES and CRM1 found that many NESs might form hydrogen bonds with CRM1 (35); therefore, we wondered whether hydrophilic O-GlcNAc could enhance the interaction by increasing hydrogen bonding. And we utilized the molecular dynamics (MD) simulation approach and constructed the system. Since the AlphaFold Protein Structure Database cannot clearly predict the NES domain of YTHDF1 (pLDDT < 50) (36, 37), the ColabFold web server was used to build the initial structure of a short fragment (residues 182-210) including the NES domain (Fig. 5A) (38). The initial structure was further optimized for 300 ns with MD simulations (Fig. 5, B and C). The binding domain of CRM1 (residues 362-645, Fig. 5D) was cropped from the crystal structure of the PKI NES-CRM1-RanGTP nuclear export complex (PDB ID: 3NBY) (39). The Rosetta Docking protocol (version 3.12) was applied to build the YTHDF1 NES and CRM1 complex (40)(41)(42). The NES fragment was set as the input structure with 10 Å translation and 360º rotation. One hundred poses were created after the docking process (Fig. 5E) and only two obtained reasonable relative positions (with the NES domain close to the CRM1binding domain) (Fig. 5F). After 500 ns of MD simulations, only one complex maintained a reasonable interaction (Fig. 5G). The last frame of this complex was chosen as the initial structure for further analysis.
The RMSD values indicated that both systems can reach stable states in 200 ns (Fig. 5H). The trajectory of the last 100 ns was extracted for further analysis. The binding energy of glycosylated fragment to CRM1-binding domain was −118.04 ± 0.32 kcal/mol, which is lower than that of the unglycosylated fragment to the CRM1-binding domain (−116.21 ± 0.24 kcal/mol) (Fig. 5I). The number of hydrogen bonds between the fragment and CRM1 was increased when S197 was glycosylated (3.70 ± 0.06 in the glycosylated system versus 3.32 ± 0.04 in the nonglycosylated system, Fig. 5J) because the glycan at S197 frequently forms hydrogen bonds

O-GlcNAcylation of YTHDF1
with H577 and D535 in CRM1 to pull the NES domain to the CRM1-binding domain (Fig. 5K). Taken together, MD simulations suggest that O-GlcNAc might increase hydrogen bonding between YTHDF1 and Crm1.

YTHDF1 O-GlcNAcylation promotes downstream target expression (e.g. c-Myc)
Recently, many YTHDF1-mediated m 6 A mRNA targets have been identified, such as the protein translation initiation factor EIF3 (16), the key Wnt receptor frizzled 7 (15), and c-Myc (13). We focused on c-Myc, as m 6 A-modified c-Myc mRNA has been demonstrated to recruit YTHDF1 (13). We reasoned that YTHDF1 O-GlcNAcylation would promote c-Myc expression as there is more cytosolic YTHDF1. We first carried out The Cancer Genome Atlas analysis and found that in colon adenocarcinoma and rectal adenocarcinoma samples, both YTHDF1 and c-Myc are overexpressed in the tumor samples (Fig. 6, A and B), indicative of a positive correlation between YTHDF1 and c-Myc in colorectal cancer. We therefore generated stable YTHDF1-knockdown SW620 cells using shYTHDF1, and c-Myc protein levels were attenuated upon YTHDF1 downregulation (Fig. 6C). When the knockdown cells were rescued with YTHDF1-WT or YTHDF1-AFA plasmids, c-Myc expression was comparable to the control in the YTHDF1-WT-rescued cells, but not in the YTHDF1-AFA-rescued cells (Fig. 6D). The stable SW620 cells were then utilized in mouse xenograft experiments, and the tumor size and weight were monitored (Fig. 6, E-G). As expected, the YTHDF1-WT-rescued cells produced much larger tumors than the AFA mutants, suggesting that YTHDF1 O-GlcNAcylation promotes colorectal cancer, probably via c-Myc.

Discussion
In this work, we first confirmed that YTHDF1 O-GlcNAcylation occurs on Ser196/197/198, and then we identified that glycosylation promotes shuttling of YTHDF1 to the cytoplasm by CRM1. Consequently, cytosolic YTHDF1 will upregulate its downstream target expression (e.g., c-Myc), and then tumorigenesis will ensue.
Our work is in line with the observation that elevated O-GlcNAcylation correlates with different cancer types, such as breast cancer, prostate cancer, and bladder cancer (43). In colon cancer, both O-GlcNAc and OGT abundance increased in clinical patient samples (44). Here, we found that YTHDF1 O-GlcNAcylation boost the expression of c-Myc, at least in SW620 cells. In xenograft models, the O-GlcNAc-deficient YTHDF1-AFA mutants attenuated tumor progression, suggesting that OGT could regulate many more downstream substrates to promote cancer growth.
Of the many m 6 A readers, YTHDF1-3 has been considered as cytosolic proteins (45). We show here that YTHDF1 is partly localized to the nucleus, and we found a potential NES in YTHDF1. Incidentally, the NES neighbors the O-GlcNAcylation sites, suggesting that O-GlcNAcylation might promote the interaction between NES and Crm1. MD simulations suggest that the hydrophilic O-GlcNAcylation might increase the binding between NES and Crm1 through hydrogen bonding.
A great many investigations have shown that O-GlcNAcylation alters protein localization, such as pyruvate kinase M2 (PKM2) (46) and serine/arginine-rich protein kinase 2 (SRPK2) (47). PKM2 O-GlcNAcylation at Thr405/Ser406 promotes ERK-dependent phosphorylation of PKM2 at Ser37, which is required for PKM2 nuclear translocation (46,48). And PKM2-T405A/S406A attenuates the interaction with importin α5 (46). SRPK2 is O-GlcNAcylated at Ser490/ Thr492/Thr498, which is close to a nuclear localization signal (47); this nuclear localization signal mediates SRPK2 nuclear localization by importin α (47). Indeed, a general mechanism has been proposed that at least some O-GlcNAcylated proteins are imported to the nucleus by importin α (47). Our work here suggests that maybe in some other cases, O-GlcNAcylation might shuttle the O-GlcNAcylated proteins to the cytoplasm by exportin.
The intertwined relationship between RNA and glycosylation is just emerging. Recently, a "glycoRNA"concept has been coined as small noncoding RNAs are found to be decorated with sialylated glycans (49). As far as m 6 A is concerned, many readers have been identified in O-GlcNAc chemoproteomic profiling works (25)(26)(27)(28)50), including YTHDF1, YTHDF3, and YTHDC1. Upon hepatitis B virus infection, YTHDF2 O-GlcNAcylation is found to be increased, which enhances its protein stability (51). In a recent investigation from our group (https://doi.org/10.11 01/2022.09.03.506498), we found that YTHDC1 O-GlcNAcylation is induced upon DNA damage and takes part in homologous recombination by enhancing binding with m 6 A mRNA. Here we show that YTHDF1 O-GlcNAcylation mediates its localization by promoting binding with exportin. We envision that O-GlcNAcylation is likely to be found in many more aspects of RNA metabolism.
Peroxidase-conjugated secondary antibodies were from Jackson ImmunoResearch. Blotted proteins were visualized using the ECL detection system (Amersham). Signals were detected by a LAS-4000 and quantitatively analyzed by densitometry using the Multi Gauge software (Fujifilm). All Western blots were repeated at least three times.

MD simulations
The O-glycan (β-N-Acetyl-D -Glucosamine) at S197 of the YTHDF1 fragment was built using the Glycan Reader & Modeler module (55). The role of O-glycosylation in the YTHDF1 fragment interacting with CRM1 was investigated via MD simulations with the GROMACS (version 2021.2) software package (56,57). Two systems (unglycosylated fragment and O-GlcNAcylated fragment at S197 in complex with CRM1-binding domain, respectively) were neutralized and solvated by 150 mM KCl and TIP3P water molecules. The systems were minimized and equilibrated using default equilibration inputs from the CHARMM-GUI webserver (58) with the CHARMM36m force field (59,60). In brief, the systems were equilibrated in the isothermal-isobaric (NPT) ensemble for 200 ns. The pressure was set at 1 atm maintained by the Parrinello-Rahman barostat (61), and the temperature was maintained at 310.15 K with the Nosé-Hoover thermostat (62). Periodic boundary conditions were applied throughout the simulations. The SHAKE algorithm was used to constrain all bonds with hydrogen atoms (63). The particle-mesh Ewald summation method was applied to treat long-range electrostatic interactions (64).
Analysis of MD trajectory data was performed through MDAnalysis (65). The binding energy (enthalpy) and perresidue energy contributions were calculated by the molecular mechanics/Poisson-Boltzmann (generalized-Born) surface area method with the gmx_MMPBSA tool (66,67). Interactions between the YTHDF1 fragment and the CRM1binding domain were displayed by PyMol (68).

Mouse xenograft
A total of 1 X 10 6 control, YTHDF1 shRNA, YTHDF1 shRNA; YTHDF1-WT or YTHDF1 shRNA;YTHDF1-AFA stable SW620 cells were resuspended in Matrigel (Corning) and then injected into the flanks of nude mice (4-6 weeks old). Tumor volumes were measured from day 3 to 9 after injection. At 9 days after the injection, tumors were dissected. The mice were obtained from the Animal Research and Resource Center, Yunnan University (Certification NO. SCXK[Dian]K2021-0001). All animal work procedures were approved by the Animal Care Committee of the Yunnan University (Kunming, China).

Data availability
All data are contained within the manuscript.
Supporting information-This article contains supporting information.