Fig 2: Loss of O-GlcNAc modification on JUN and GATAD2A disrupts OIS progression and inhibits SASP-associated tumor growth in vitro.a Representative SA-β-Gal staining of LU-RAS (vehicle), LU-RAS/KD-WT, LU-RAS/KD-AA, SK-RAS (vehicle), SK-RAS/KD-WT, and SK-RAS/KD-AA cells at D10 of OIS. SK-RAS cells and either wild-type (Flag-GATAD2AWT, HA-JUNWT) or mutant (Flag-GATAD2AS182A and HA-JUNS84A) constructs were then reintroduced. Scale bars, 200 μm. b Flow cytometric analysis of the cell cycle distribution (G1, S, and G2 phases) in LU-RAS (vehicle), LU-RAS/KD-WT, LU-RAS/KD-AA, SK-RAS (vehicle), SK-RAS/KD-WT, and SK-RAS/KD-AA cells at D10 of OIS. c Schematic of transwell coculture experiments assessing the effects of SASP factors derived from LU-RAS, LU-RAS/KD-WT, and LU-RAS/KD-AA cells at D10 of OIS on A549 cell growth. d The relative levels of senescence conditioned medium multiple SASP factors at the OIS D10 time point in LU-RAS and its variant cell lines were quantified by ELISA, with the level in vehicle control cells set as 1 (fold change). e LU-RAS and its variant cells were induced to undergo OIS at D10 in the presence of doxycycline (2 μg/mL) under transwell coculture conditions. Whole-cell lysates were extracted and subjected to immunoblot analysis. The immunoblots are representative of three biologically independent experiments with similar results. f Counts of A549 cells cultured alone or cocultured with the indicated cells at D10 of OIS. Scale bars, 200 μm. g Images from the scratch assay of A549 cells at 0 h and 96 h; the cells were cultured alone or cocultured with the indicated cells at D10 of OIS. Scale bars, 200 μm. The data are presented as mean ± SD from four biological replicates with similar results. The p values were calculated using a two-tailed unpaired Student’s t test (a, b, d, f, g). The source data for the values and the graphs in (a, b, d, f, g) and immunoblots in (e) are provided in the Source Data File.

Fig 3: O-GlcNAc modification of JUN and GATAD2A drives chromatin occupancy linked to the OIS phenotype.a Immunoblot analysis of chromatin extracted from LU-RAS/KD-WT and LU-RAS/KD-AA cells at OIS time points D0 and D10, showing chromatin-bound Flag-GATAD2A, HA-JUN, their variants, MBD2, and SNF5. Whole‑cell lysates, cytoplasmic fractions, and soluble nuclear fractions served as controls. n = 3 biologically independent samples were analyzed, with similar results obtained across replicates. b Average enrichment profiles of ChIP-seq reads (n = 2 biological replicates) for Flag-GATAD2AWT, Flag-GATAD2AS182A, HA-JUNWT, and HA-JUNS84A at COGC-seq peaks in LU-RAS/KD-WT and LU-RAS/KD-AA cells at the D0 and D10 OIS time points. c Heatmap showing ChIP-seq signal enrichment for Flag-GATAD2AWT and Flag-GATAD2AS182A within ±2 kb of COGC-seq-annotated enhancer/intergenic peak centers and for HA-JUNWT and HA-JUNS84A within ±2 kb of COGC-seq-annotated promoter peak centers. The color indicates the enrichment level (white: low; green/red: high). d Average ChIP-seq enrichment profiles of MBD2 at the Flag-GATAD2AWT and Flag-GATAD2AS182A ChIP-seq peaks and of SNF5 at the HA-JUNWT and HA-JUNS84A ChIP-seq peaks in LU-RAS cells at the D10 OIS time point. e Integrative Genomics Viewer (IGV) tracks displaying ChIP-seq signals for Flag-GATAD2AWT, Flag-GATAD2AS182A, HA-JUNWT, and HA-JUNS84A at the D0 and D10 OIS time points across representative SASP- and cell cycle-related genes. f Heatmaps showing RNA-seq data for representative SASP- and cell cycle-related genes in LU-RAS/KD-WT and LU-RAS/KD-AA cells at the D10 OIS time point. The data are presented as row Z-scores based on two biologically independent experiments. g, h Validation of O-GlcNAc GATAD2A and JUN binding peaks at the indicated gene loci by ChIP–qPCR (g). The bars show the % of total input chromatin, with IgG used as a negative control. Quantitative PCR (qPCR) analysis of the mRNA levels of the indicated genes in LU-RAS, LU-RAS/KD (vehicle), LU-RAS/KD-WT, LU-RAS/KD-AA and LU-RAS/KD-WT cells treated with OSMI‑4 (5 μM) for 48 h at the D10 OIS time points (h). The data are presented as mean ± SD from four biological replicates with similar results. The p values were calculated using a two-tailed unpaired Student’s t test. The source data for the graphs in (f–h) and immunoblots in (a) are provided in the Source Data File.

Fig 4: O-GlcNAc promotes the assembly of TF-SWI/SNF and NuRD complexes involved in senescence reprogramming.a Workflow for quantitative and site-specific analysis of O-GlcNAc modifications on OCPs during OIS in LU-RAS cells. b Venn diagram showing overlap of high-confidence O-GlcNAc sites across OIS time points. c Mapped O-GlcNAc sites in GATAD2A and JUN. d, e LU-RAS/KD cells stably expressing empty vector (vehicle), wild-type (Flag-GATAD2AWT and HA-JUNWT; LU-RAS/KD-WT cells) or mutant (Flag-GATAD2AS182A and HA-JUNS84A; LU-RAS/KD-AA cells) constructs were analyzed at D0 and D10. Proteins were immunoprecipitated with anti-Flag or anti-HA magnetic beads (d) or pulled down using sWGA lectin (e). f, h Immunoprecipitation of cells in (d) at D0 and D10 followed by immunoblot analysis of NuRD complex components (f) and TF-SWI/SNF complex components (h). g Size exclusion chromatography (Superose 6) of nuclear extracts from the indicated cells, followed by immunoblot analysis of NuRD and SWI/SNF subunits. Fractions 4–13 represent holo-NuRD and SWI/SNF complexes; fractions 13–25 contain unassembled components. i Immunofluorescence images showing nuclear colocalization of Flag-GATAD2A with MBD2 and HA-JUN with SNF5 in LU-RAS/KD-WT and LU-RAS/KD-AA cells at D10. Nuclei were stained with DAPI (blue). Scale bars, 10 μm. A Pearson coefficient of 1 indicates perfect colocalization, 0 indicates none. j Truncated variants of vehicle (vehicle-1 to vehicle-3), Flag-GATAD2A (ΔGATAD2A-1 to ΔGATAD2A-3) and HA-JUN (ΔJUN-1 to ΔJUN-2) were designed (top), and co-IP was performed using Flag and HA tags (bottom). k Simulation of GATAD2A interaction with the NuRD complex (replica 1), with GATAD2A in wheat, MTA1 blue, MBD2 gray, HDAC1 green. Zoomed view shows the O-GlcNAc-S182GATAD2A (red sticks) interaction network, with the P157MBD2–S85HDAC1 distance used to characterize binding. Distance I curves for indicated systems are shown. l Overall model of JUN within SWI/SNF component SNF5 and nucleosome (replica 1), with JUN in wheat, SNF5 sky blue, other proteins gray, DNA green. Zoomed view shows O-GlcNAc-S84JUN interaction with SNF5 and the O-GlcNAc JUN-E59SNF5 distance. Distance curves for indicated systems are shown. Immunoblots (d–h, j) are representative of three independent experiments. Source data for (b) and for immunoblots (d–h, j) are in the Source Data File.

Fig 5: Coinjection of OIS-affected cells lacking O-GlcNAc-modified JUN and GATAD2A with tumorigenic cells fails to promote senescence-associated tumor growth in vivo.a Schematic of the coinjection model of D10 LU-RAS (vehicle), LU-RAS/KD-WT, or LU-RAS/KD-AA cells with A549 cells in nude mice. b, c Fluorescence images of GFP-expressing A549 cells injected alone or coinjected with LU-RAS (vehicle), LU-RAS/KD-WT, or LU-RAS/KD-AA cells (NIR790-stained) at D10 of OIS (n = 6 per group); images were acquired on days 5, 15, and 28. One representative image is shown (b). The relative fluorescence intensity was quantified (c). d Tumors were excised after 28 days. e Tumor volumes were measured. f Immunoblot analysis of senescence markers in xenograft tumor samples. n = 3 biologically independent samples were analyzed, and similar results were obtained across replicates. g, h Xenograft tumor samples were subjected to H&E, SA-β-Gal, IL6, IGFBP2, CCND1, and STAG2 staining. Scale bars, 100 μm. Representative images (g); quantitative analysis of mean IHC staining (h). The data are presented as mean ± SD from six biological replicates with similar results. p values were calculated using a two-tailed unpaired Student’s t-test (c, e, h). The source data for the graphs in (c, e, h) and immunoblots in (f) are provided in the Source Data File.