LINC01963 promotes pancreatic ductal adenocarcinoma proliferation via METTL3/IGF2BP2 axis-mediated m⁶A modification of c-Myc

Abstract

Background

C-Myc overexpression is an important molecular hallmark of pancreatic ductal adenocarcinoma (PDAC), but directly targeting c-Myc is extremely challenging. Identifying key upstream factors involved in c-Myc overexpression provides promising indirect targets for c-Myc.

Methods

Public transcriptomic and clinical datasets, including TCGA, GEO, were integrated to identify c-Myc-associated long noncoding RNAs in PDAC, with LINC01963 selected for further investigation. The functional roles of LINC01963 were validated using human PDAC cell lines and in vivo proliferation models. The molecular mechanisms underlying c-Myc regulation by LINC01963 were explored using RNA pull-down, RIP-seq, RIP-qPCR, Co-IP, mass spectrometry, ubiquitination assays, truncation and site-directed mutagenesis analyses, and dual-luciferase reporter assays. Survival associations were evaluated using Kaplan–Meier analysis and Cox proportional hazards regression.

Results

Here, the long noncoding RNAs (lncRNAs) highly expressed in PDAC and significantly correlated with c-Myc expression were identified using RNA sequencing datasets. Among them, LINC01963 was found to interact with c-Myc, as confirmed by RNA pull-down and RIP-qPCR assays. Furthermore, high LINC01963 expression was correlated with poor PDAC prognosis, and functional studies demonstrated that its knockdown inhibited PDAC cell proliferation and xenograft tumor growth. Mechanistic studies identified LINC01963 as a key regulator of c-Myc stability, consequently affecting cell cycle through the c-Myc/p21-related signaling pathways. Further investigation revealed that LINC01963 enhanced N6-methyladenosine (m⁶A) modification of c-Myc mRNA by protecting methyltransferase-like 3 (METTL3) protein from KDM1B-mediated K48-linked ubiquitination and proteasomal degradation. Intriguingly, LINC01963 also stabilized c-Myc mRNA by facilitating the formation of a ternary complex with insulin-like growth factor 2 mRNA-binding protein 2 (IGF2BP2) and m⁶A-modified c-Myc.

Conclusions

Our study reveals that LINC01963 promotes PDAC tumorigenesis through METTL3/IGF2BP2 axis-coordinated regulation of c-Myc, suggesting a new strategy for indirectly targeting c-Myc.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13046-026-03650-5.

Introduction

Pancreatic ductal adenocarcinoma (PDAC), the most common type of pancreatic cancer, is lethal with a 5-year survival rate of approximately 10% [1, 2]. Most PDAC patients are diagnosed at advanced stages and ultimately die of tumor progression due to chemoradiotherapy resistance [3, 4]. Despite significant breakthroughs in targeted therapies for solid tumors in recent years [5, 6], effective targeted strategies for PDAC are still lacking.

The oncoprotein c-Myc, a master transcription factor, plays a prominent role in key cellular processes such as proliferation and protein synthesis [7]. Aberrant c-Myc expression is a critical molecular hallmark of various cancers, including PDAC [8–10]. Preclinical studies have demonstrated that c-Myc inactivation induces profound changes in both tumor cells and the tumor microenvironment (TME), leading to rapid tumor regression [11–14]. These findings have positioned c-Myc as a promising therapeutic target for multiple malignancies [14]. However, to date, no specific inhibitor directly targets c-Myc due to its “undruggable” properties: lack of a stable conformation and small molecule binding pockets [15–17]. Therefore, great efforts are undertaken to understand the complex regulatory networks governing c-Myc expression, developing indirect targeting approaches.

Long noncoding RNAs (lncRNAs) are regulatory transcripts (> 200 nt) that modulate gene expression through four distinct mechanisms: transcriptional, post-transcriptional, translational, and post-translational regulation [18]. M⁶A RNA methylation represents a key post-transcriptional mechanism by which lncRNAs regulate the expression of oncogenes like c-Myc [19–22], and it is dynamically regulated by m⁶A regulators (m⁶A methyltransferases, demethylases and readers). Overexpression of the m⁶A methyltransferase METTL5 enhances c-Myc translation activity [23]. Besides, m⁶A readers such as the IGF2BP family proteins, upregulate c-Myc levels by stabilizing c-Myc mRNA and promoting its translation [24]. All these studies indicate that dysregulation of m⁶A regulators is a key factor in the overexpression of c-Myc in PDAC. Recent studies have highlighted the role of m⁶A-modified lncRNAs, which are governed by m⁶A regulators, in influencing the expression of target genes. In pancreatic cancer, m⁶A reader (SNIP1) binds to the m⁶A modification site on lncRNA BCAN-AS1, recruiting c-Myc protein to form a ternary complex [19]. This complex prevents c-Myc ubiquitination and degradation, demonstrating that m⁶A regulators can stabilize c-Myc by bridging their interaction with m⁶A-modified lncRNAs. Whereas research on how lncRNAs interact with m⁶A regulators to modify c-Myc m⁶A methylation remains lacking and requires further exploration. Therefore, a deeper understanding of the precise mechanisms by which lncRNA and m⁶A regulators regulate c-Myc overexpression is crucial for developing effective strategies to indirectly target c-Myc in PDAC treatment.

In this study, we identified a c-Myc-associated lncRNA named LINC01963, as an oncogenic molecule in PDAC. LINC01963 silencing significantly inhibited c-Myc expression and xenograft tumor growth, affecting PDAC cell cycle and proliferation through the c-Myc/p21-related signaling pathways. We have found that LINC01963 enhances m⁶A modification of c-Myc mRNA by preventing methyltransferase-like 3 (METTL3) protein from Lysine Demethylase 1B (KDM1B)-mediated ubiquitin–proteasome degradation. Furthermore, LINC01963 stabilizes c-Myc mRNA by acting as a scaffold to facilitate the formation of a ternary complex with the m⁶A reader insulin-like growth factor 2 mRNA-binding protein 2 (IGF2BP2) and m⁶A-modified c-Myc. Our findings reveal that LINC01963 promotes PDAC tumorigenesis by coordinating the METTL3/IGF2BP2 axis to regulate c-Myc, suggesting its therapeutic potential as an indirect c-Myc target in PDAC.

Methods

Cell lines and reagents

Human normal pancreatic ductal cells (HPNE) and human PDAC cell lines (PANC-1 and MIA PaCa-2) were obtained from the National Biomedical Cell Resource Center (Beijing, China) and verified by STR profiling (D2081-2084), and tested negative for contamination. Cells were maintained at 37 °C in a humidified 5% CO₂ incubator. HPNE cells were grown in 75% Dulbecco’s modified Eagle’s medium (DMEM) without glucose (Sigma-Aldrich, Cat# D5030) and 25% Medium M3 Base (Incell Corp; Cat#M300A-100), supplemented with 1% penicillin and streptomycin (Gibco, Cat# 15,140,122), 5% fetal bovine serum (Corning, Cat# MT35016CV), 5.5 mM D-glucose (Sigma, Cat# G8270), 10 ng/ml human recombinant EGF (Gibco, Cat# AF10015100) and 750 ng/ml puromycin (Gibco, Cat# A1113803). The remaining cancer cells were maintained in DMEM or RPMI 1640 supplemented with 10% FBS, and 1% penicillin and streptomycin.

RNA extraction and quantitative real-time PCR analysis (qRT-PCR)

Total RNA was isolated from PDAC cell lines or tissues using TRIzol reagent (Invitrogen, Cat# 10,296,010). Reverse transcription of complementary DNA (cDNA) was performed using a PrimeScript Reverse Transcription Kit (Takara, Cat# RR036A). cDNAs were analyzed by qRT-PCR on a Roche LightCycler 480 II using the SYBR-Green method. GAPDH served as an endogenous control for normalization of RNA levels. The relative levels of RNAs were calculated using the comparative Ct method. Three replicates were performed for each experiment. Gene-specific primer sequences are shown in Supplementary Table 1.

Cytoplasmic and nuclear fractionation

Cytoplasmic and nuclear fractions from PANC-1 and MIA PaCa-2 cells were extracted using the NE-PER Nuclear and Cytoplasmic Extraction Reagent (Thermo Fisher Scientific, Cat# 78,833). LINC01963 expression was detected in each fraction with qRT-PCR assays. GAPDH and U6 were used as cytoplasmic and nuclear controls, respectively.

RNA stability assay

RNA stability was estimated using actinomycin D treatment (ActD, Selleck, Cat# S8964). Twenty-four hours after PDAC cells were transfected, ActD (5 μM) was added to the medium to disrupt RNA synthesis. Total RNA was extracted at the indicated time points (0, 30, and 120 min). The c-Myc mRNA levels were detected by qRT-PCR assay. The experiments were independently replicated three times.

Western blotting analysis

Protein extracts from cells, RIP, and RNA pull-down samples were prepared using 1X radioimmunoprecipitation assay buffer (RIPA, Solarbio, Cat# R0020) with 1% PMSF (Solarbio, Cat# P0100). The concentration of protein was quantified using the Pierce™ BCA Protein Assay Kit (Thermo Fisher Scientific, Cat# 23,225). Equal amounts of protein samples were separated by 10% SDS-PAGE, transferred to PVDF membranes (MilliporeSigma, Cat# IPVH08100), and visualized by immunoblotting with ECL reagent (MilliporeSigma, Cat# WBKLS0050). Information on antibodies can be found in Supplementary Table 2.

Protein stability assay and ubiquitination assays

For the protein stability assay, transfected PDAC cells were treated with 20 μM cycloheximide (CHX; MilliporeSigma, Cat# C4859), an inhibitor of de novo protein synthesis, for the indicated times and harvested. Protein expression of c-Myc was then determined by western blotting analysis.

For the ubiquitination assays, PDAC cells were co-transfected with the indicated constructs. To block proteasomal degradation, cells were treated with MG132 (20 µM, 4 h; MCE, Cat# HY-13259) prior to lysis when required. Cell lysates were prepared in IP lysis buffer and subjected to immunoprecipitation using the Pierce™ Crosslink Magnetic IP/Co-IP Kit (Thermo Fisher Scientific, Cat# 88,805) following the manufacturer’s protocol. Anti-FLAG or anti-HA antibodies were pre-crosslinked to Protein A/G magnetic beads and incubated with clarified lysates. After extensive washing, bound proteins were eluted and analyzed by SDS-PAGE and western blotting. Ubiquitination levels were detected using antibodies against ubiquitin or HA to visualize HA-K48-Ub-conjugated proteins. Input fractions were examined to confirm equal expression of METTL3, KDM1B, and loading controls.

RNA pull-down assay

RNA pull-down assays were performed using the Pierce™ Magnetic RNA–Protein Pull-Down Kit (Thermo Fisher Scientific, Cat# 20,164) according to the manufacturer’s protocol with minor optimization. Biotin-labeled full-length, truncated, or antisense DNA probes targeting LINC01963 were synthesized by RiboBio and Gzscbio Co., Ltd. (Guangzhou, China); probe sequences are listed in Supplementary Table 1. The bound proteins were analyzed by western blotting or Coomassie blue staining, followed by Liquid Chromatography-Tandem Mass Spectrometry (LC–MS/MS) identification. Detailed experimental procedures are described in the “Supplementary Materials and Methods”.

Coomassie blue staining and LC–MS/MS

After SDS-PAGE separation, gels were stained using the PageBlue™ Protein Staining Solution (Thermo Fisher Scientific, Cat# 24,620) according to the manufacturer’s protocol. Protein bands of interest were excised for in-gel digestion and analyzed by LC–MS/MS at the Biomarker Research Platform, Translational Medicine Center, Peking Union Medical College Hospital. Proteins identified by LC–MS/MS analysis were listed in Supplementary Table 3. Detailed experimental procedures are described in the “Supplementary Materials and Methods”.

Plasmid construction

Full-length human IGF2BP2 (NM_006548, amino acids 1–599), together with a series of truncated and domain-deletion variants designed to map functional regions, were constructed by GeneChem (Shanghai, China). Full-length METTL3 was cloned into an expression vector containing an N-terminal 3 × FLAG tag (METTL3-FLAG) by GeneChem (Shanghai, China). A panel of METTL3 truncation constructs and site-directed mutants, including Mut-KL-FLAG and Mut-328-FLAG, was generated by GeneCloud Biotechnology Co., Ltd. (Guangzhou, China) according to the AlphaFold-guided structural design described in this study. The HA-tagged K48-only ubiquitin plasmid (HA-K48Ub) was constructed by GeneChem (Shanghai, China). In this construct, lysine 48 (K48) was retained as the sole ubiquitin linkage site, whereas all other lysine residues were mutated to arginine, thereby restricting polyubiquitin chain formation exclusively to K48-linked ubiquitination. A shRNA-resistant LINC01963 construct (LINC01963-res) was generated by GeneChem (Shanghai, China) by deleting a 19-nucleotide region encompassing the shLINC01963 target sequence.

Unless otherwise indicated, all expression constructs were subcloned into the CV702 vector, which contains a CMV enhancer/promoter, an N-terminal 3 × FLAG tag, and an SV40-driven puromycin resistance cassette. All plasmids were verified by Sanger sequencing prior to use. PDAC cells were transiently transfected with the indicated plasmids using Lipofectamine 3000 (Invitrogen) according to the manufacturer’s instructions. Expression of FLAG-tagged or HA-tagged proteins was confirmed by western blotting before downstream assays.

Methylated RNA immunoprecipitation sequencing (MeRIP-seq) and RNA immunoprecipitation followed by qPCR (RIP-qPCR)

MeRIP-seq was conducted using the NEBNext® Magnesium RNA Fragmentation Module (NEB, Cat# E6150), Monarch® RNA Cleanup Kit (NEB, Cat# T2030) and EpiMark® N6-Methyladenosine Enrichment Kit (NEB, Cat# E1610S) following the manufacturer’s protocols. Sequencing libraries were prepared from the enriched RNA and sequenced on an Illumina NovaSeq 6000 platform (Novogene, Beijing). For RIP-qPCR, total RNA without fragmentation was immunoprecipitated using the same NEB-based procedure with antibodies against METTL3 (Abcam, Cat# ab195352), IGF2BP2 (Abcam, Cat# ab124930), m⁶A (Abcam, Cat# ab151230), FLAG (Abcam, Cat# ab125243) or Normal Rabbit IgG (Cell Signaling Technology, Cat# 2729). The recovered RNA was reverse-transcribed and analyzed by qRT-PCR. Primer sequences are provided in Supplementary Table 1. Further details are provided in the “Supplementary Materials and Methods”.

RNA FISH and immunofluorescence staining (RNA-FISH-IF)

A Cy3-labeled probe specific to LINC01963 was constructed by RiboBio, and an RNA fluorescence in situ hybridization (RNA-FISH) assay was performed using a Fluorescent in Situ Hybridization Kit (RiboBio). RNA-FISH-IF of IGF2BP2 and METTL3 were performed to analyze the co-localization of LINC01963, IGF2BP2 and METTL3. Briefly, PDAC cells were fixed with 4% paraformaldehyde and permeabilized with 0.3% Triton X-100 (Solarbio, Cat# T8200). Then, cells were blocked and incubated with primary antibodies against IGF2BP2 and METTL3, followed by secondary fluorescent antibody staining. After fixation again, cells were washed and hybridized with FISH probes at 37 °C overnight in a dark and humidified chamber. Cell nuclei were counterstained with DAPI. Fluorescent images were captured with a confocal laser scanning microscope (Olympus).

Dual luciferase reporter assays

To evaluate the stability of c-Myc mRNA, we constructed the wild-type (WT) 3'UTR vector of c-Myc and a mutated 3'UTR vector with a single m⁶A site mutation in the GV272 plasmid (GeneChem; Shanghai, China). MIA PaCa-2 cells were seeded in triplicate in 24-well plates. Then, 500 ng reporter plasmids with above WT or mutant c-Myc and 20 ng Renilla luciferase (Rluc) control plasmids (pRL-TK) were co-transfected with or without LINC01963/IGF2BP2/METTL3 expression using Lipofectamine 3000. firefly luciferase (Fluc) and Rluc activities were measured 24 h later with the Dual Luciferase Reporter Assay System Kit (Promega, Cat# 1500) according to the manufacturer’s instructions. Fluc activity was normalized to the corresponding Rluc activity. The experiments were independently replicated three times.

Cell proliferation and colony formation assays

For cell proliferation assays, transfected PANC-1 and MIA PaCa-2 cells were seeded in 96-well plates (2,000 cells per well). After cell attachment and growth for 12 h, 10 μL of Cell Counting Kit-8 (CCK-8, APExBIO, Cat# K1018) reagent was added to each well containing fresh culture medium to reach a final volume of 100 μL. The cells were incubated for 1 h at 37 °C in 5% CO₂ incubator, and the absorbance at 450 nm was measured using an EPOCH2T microplate reader (BioTek Instruments). The experiments were independently replicated three times, with six replicates per group.

For colony formation assays, transfected PANC-1 and MIA PaCa-2 cells were seeded in 6-well plates (1000 cells per well). After incubation for one week with culture medium changes every three days, the cells were fixed in 4% paraformaldehyde for 15 min and subsequently stained with 1% crystal violet for 30 min. The cell colonies then photographed and quantified using ImageJ software.

Ethynyl deoxyuridine (EdU) cell proliferation assay

Cell proliferation was also measured by EdU Incorporation Assays (RiboBio, Cat# C10310) following the manufacturer’s instructions. Briefly, PANC-1 and MIA PaCa-2 cells seeded into 96-well plates were transfected for 48 h and incubated with 50 μM EdU per well for an additional 2 h at 37 °C. Next, cells were fixed with 4% paraformaldehyde in PBS, then stopped by the addition of glycine, and stained with 1 × Apollo solution. Finally, the cells were subjected to nuclear staining with 1 × Hoechst33342 solution and visualized under a Zeiss fluorescence photomicroscope. The experiments were independently replicated three times.

Cell cycle assay

For cell cycle assay, transfected cells were collected and fixed with 70% ice-cold ethanol overnight at 4 °C. Then the cells were resuspended and incubated in 500 µL of propidium iodide by the Cell Cycle and Apoptosis Analysis Kit (Beyotime, Cat# C1052) according to the manufacturer’s instructions and analyzed by FACScan (BD Biosciences). The percentage of cells in G0-G1-, S, and G2-M phases was analyzed using FlowJo software. The experiments were independently replicated three times.

In vivo xenograft model

Stable LINC01963-knockdown cell lines were generated by lentiviral infection. Briefly, MIA PaCa-2 and PANC-1 cells were infected with lentiviruses carrying shLINC01963 or control shNC, and selected with puromycin to establish stable pools. Knockdown efficiency was validated by qRT-PCR.

For orthotopic implantation, stable MIA PaCa-2/shNC, MIA PaCa-2/shLINC01963, PANC-1/shNC, or PANC-1/shLINC01963 cells (1 × 10⁶ cells in 50ul PBS) were injected into the surgically exposed pancreatic tail of nude mice under isoflurane anesthesia. After implantation, the abdominal wall and skin were sutured under sterile conditions. Twenty-four BALB/c nude mice (4 weeks old; Vital River Laboratory Animal Technology Co., Ltd.) were used for the MIA PaCa-2 model (n = 12 per group), and eight BALB/c nude mice for the PANC-1 model (n = 4 per group).

Tumor growth was monitored weekly by bioluminescence imaging (IVIS Lumina III). Mice were injected intraperitoneally with D-luciferin (150 mg/kg, Abmole, Cat# M9053) 10 min before imaging. MIA PaCa-2 tumors were harvested on Day 39 and PANC-1 tumors on Day 28, followed by measurements of body weight, tumor weight, and tumor volume . Tumor tissues were fixed in 10% neutral-buffered formalin, paraffin-embedded, and processed for H&E and immunohistochemistry.

Single-cell RNA-seq analysis of LINC01963

Single-cell resolution profiling of LINC01963 expression in PDAC was performed using the Seurat analytical pipeline (v4.1.1; https://satijalab.org/seurat/). The cohort included 2 uninvolved pancreas (UNIN), 8 intraductal papillary mucinous neoplasms (IPMN), and 6 PDAC tissue samples, together with 4 UNIN samples from public datasets [25, 26]. Batch effects across patients were mitigated using Harmony algorithm (v.0.1.0; https://github.com/immunogenomics/harmony/). Principal component analysis (PCA) was first used for dimensionality reduction, followed by Uniform Manifold Approximation (UMAP) for visualization. Cell type annotation was assigned according to canonical marker genes and cluster-specific differentially expressed genes. LINC01963 expression was quantified across all annotated cell types and specifically in ductal cells at different disease stages, with results visualized as violin plots.

Public data processing

We used data from TCGA-PAAD (The Cancer Genome Atlas-Pancreatic Adenocarcinoma), GEO (Gene Expression Omnibus), TARGET (Therapeutically Applicable Research to Generate Effective Treatments), CPTAC (Clinical Proteomic Tumor Analysis Consortium), QCMG (Queensland Centre for Medical Genomics) and GTEx (Genotype-Tissue Expression Program). Analyses were performed in R software 4.0.1 to assess the differential expression, prognostic value and correlation between LINC01963 and c-Myc.

Statistical analysis

Statistical analyses were performed using GraphPad Prism 9.0 and R 4.0.1. Data are presented as mean ± SD from at least three independent biological replicates unless otherwise stated. For comparisons between two independent groups, two-sided unpaired Student’s t-test was used for normally distributed data with homogeneous variances, and Welch’s t-test was applied when variances were unequal; when normality assumptions were not met, the Wilcoxon rank-sum tests was used. For comparisons among multiple groups, one-way or two-way ANOVA was performed followed by Tukey’s or Sidak’s post-hoc tests as indicated. Time-course CCK-8 proliferation assays and longitudinal bioluminescence measurements were analyzed by two-way ANOVA with Sidak’s correction.

Overall survival was analyzed by Kaplan–Meier curves with two-sided log-rank tests. Hazard ratios (HRs) with 95% confidence intervals (CIs) were estimated using Cox proportional hazards models with all analyses performed in R 4.0.1 along with Zstats v1.0 (www.zstats.net). Correlations between continuous variables were assessed using Pearson’s or Spearman’s correlation coefficients as appropriate. Differential m⁶A peak enrichment from MeRIP-seq was evaluated with Benjamini–Hochberg procedure to control the false discovery rate. All data are presented as the mean ± SD unless otherwise stated. Statistical significance was set at a threshold of P value less than 0.05. The exact statistical tests, n values, and multiple-testing corrections used for each experiment are specified in the corresponding figure legends.

Results

LINC01963 is overexpressed and correlated with c-Myc expression in PDAC

Using TCGA-PAAD dataset (n = 179), we identified 135 lncRNAs that were highly expressed in PDAC and positively correlated with c-Myc mRNA (Spearman’s R > 0.3, p < 3.5 × 10^–5) (Fig. 1A, Supplementary Fig. 1 A and Supplementary Table 4). In the GEO cohort GSE183795 (244 PDAC tumor-normal pairs), 4173 transcripts were upregulated in PDAC tissues (normal vs tumor, log2 fold change < −0.1, p < 7 × 10^–4) (Fig. 1A, Supplementary Fig. 1B and Supplementary Table 5). Intersecting these sets revealed 6 c-Myc-associated lncRNAs that were stably overexpressed in PDAC (Fig. 1B). Among them, LINC01963 showed significantly higher expression in PDAC patients with poor versus good prognosis in GSE42952 (Fig. 1C).

Fig. 1.

LINC01963 is overexpressed and correlated with c-Myc expression in PDAC. A Venn diagram displaying six overlapping transcripts between 4173 upregulated transcripts and 135 upregulated c-Myc-associated lncRNAs (TCGA-PAAD, n = 179; GSE183795, n = 244). B Six lncRNAs satisfied both criteria: upregulated in PDAC and positively correlated with c-Myc. C In GSE42952 (n = 33), high LINC01963 expression associated with poor prognosis; groups were stratified into poor- (OS < 19 months and DFS < 7 months) and good-prognosis (OS and DFS > 50 months) groups following the original criteria. Overall survival, (OS). Disease-Free Survival (DFS). D Kaplan–Meier curves of OS in CPTAC-PDAC cohort (n = 135) stratified by LINC01963 expression. Confidence interval, (CI). E–F LINC01963 levels in stage II-IV (n = 87) vs. stage I (n = 9) samples from the QCMG-PAAD dataset (E), and in PDAC tumor vs. normal tissues across TCGA, TARGET, and GTEx datasets (n = 345) (F). G LINC01963 levels in PDAC cell lines (PANC-1 and MIA PaCa-2) and human normal pancreatic ductal cells (HPNE) by qRT-PCR assays (n = 3). H Subcellular localization of LINC01963 determined by nuclear/cytoplasmic fractionation followed by qRT-PCR (n = 3). I Representative RNA-FISH images showing LINC01963 localization in PDAC cells. Scale bar, 10 μm. J-K CPAT and CPC2 analyses indicated that LINC01963 lacks protein-coding potential. MYC and GAPDH served as positive controls, while lncRNA XIST and HOTAIR were used as negative controls. Data are displayed as mean ± SD in C, E–H. RNA-FISH-IF experiments were independently repeated at least twice with similar results. Statistical significance among the indicated groups was assessed by Welch’s t-test (C, E–F), two-sided log-rank test (D) and one-way ANOVA followed by Tukey’s post-hoc test (G). * P < 0.05, ** P < 0.01, *** P < 0.001

Notably, high expression of LINC01963 was significantly associated with poor overall survival (OS) in the CPTAC-PDAC cohort (Fig. 1D) and remained an independent prognostic factor in univariate and multivariate Cox regression analyses (Supplementary Table 6). In QCMG-PAAD, LINC01963 expression was higher in advanced disease (stage II-IV) than in early disease (stage I) (Fig. 1E). Consistently, analysis of multiple datasets demonstrated that LINC01963 was significantly upregulated in pancreatic tissue samples compared to normal pancreas or adjacent non-tumor tissues (Fig. 1F). LINC01963 levels were also higher in PDAC cell lines (PANC-1 and MIA PaCa-2) than in human normal pancreatic ductal cells (HPNE) (Fig. 1G). Collectively, these cross-cohort data nominate LINC01963 as a c-Myc-associated lncRNA that is overexpressed in PDAC and linked to advanced stage and poorer survival, warranting further mechanistic investigation.

We then determined the cell-type specificity of expression, subcellular localization and coding potential of LINC01963 in PDAC. Analysis of our single-cell sequencing (scRNA-seq) data from 20 pancreatic samples revealed that LINC01963, a modestly conserved intergenic lncRNA (Supplementary Fig. 1 C), was predominantly expressed in malignant ductal epithelial cells (Supplementary Fig. 1D-E). Subcellular fractionation analysis (Fig. 1H) and RNA-FISH assays (Fig. 1I) demonstrated that LINC01963 exhibited a dual localization in both the nucleus and cytoplasm, with a predominant nuclear distribution. The coding potential of LINC01963 was further evaluated using the CPC2.0 and CPAT databases, both of which consistently supported that LINC01963 possessed negligible coding potential (Fig. 1J-K).

LINC01963 promotes PDAC cell proliferation in vitro and in vivo

Among the two independent siRNAs tested (001–002), siLINC01963−001 exhibited the most efficient knockdown and caused the strongest suppression of cell proliferation, as determined by RT-qPCR and CCK-8 assays (Supplementary Fig. 1F-G). Thus, siLINC01963−001 was selected for subsequent experiments. Consistent results were observed across both PANC-1 and MIA PaCa-2 cell lines, where CCK-8, colony formation and EdU assays collectively demonstrated that LINC01963 knockdown markedly suppressed PDAC cell proliferation, while its overexpression exerted the opposite effect (Fig. 2A-F).

Fig. 2.

LINC01963 promotes PDAC cell proliferation in vitro. A-B Knockdown and overexpression efficiency of LINC01963 in both PDAC cell lines analyzed by qRT-PCR (n = 3). C-D Cell proliferation rates of PDAC cells after LINC01963 knockdown/overexpression were evaluated using CCK-8 proliferation assays at the indicated time points (n = 6 per time point). E Representative images (left) and quantification of colony formation assays (right) in MIA PaCa-2 cells after LINC01963 knockdown or overexpression (n = 3). F Representative images and quantification of EdU assays in both PDAC cell lines after LINC01963 knockdown or overexpression (n = 3). Scale bar, 200 μm. All data are presented as the mean ± SD from three independent biological replicates. Statistical significance was calculated by two-sided unpaired Student’s t-test (A-B, E–F) and two-way ANOVA followed by Sidak’s post-hoc test for time-course proliferation assays (C-D). * P < 0.05, ** P < 0.01, *** P < 0.001. **** P < 0.0001

To further characterize the functional relevance of LINC01963, we first evaluated its role in HPNE. LINC01963 overexpression markedly increased its transcript levels (Supplementary Fig. 1H) but had no detectable effect on HPNE cell viability (Supplementary Fig. 1I), indicating that oncogenic function of LINC01963 likely depends on the malignant cellular context. To further exclude off-target effects, we generated stable LINC01963-knockdown (shLINC01963) and overexpression (LV-LINC01963) cell lines using lentiviral transduction and confirmed efficiency by qRT-PCR (Supplementary Fig. 1 J-K). In addition, a shRNA-resistant LINC01963 construct (LINC01963-res) was engineered by deleting the endogenous 19-nt shLINC01963 targeting sequence (Supplementary Fig. 1L). Re-expression of LINC01963-res in shLINC01963 cells restored LINC01963 levels (Supplementary Fig. 1 M), confirming that on-target specificity of the knockdown phenotype.

We next evaluated the in vivo relevance of LINC01963 in orthotopic pancreatic xenograft models (Fig. 3A). In MIA PaCa-2 xenografts, LINC01963 knockdown significantly suppressed tumor growth, as shown by reduced bioluminescence signals (Fig. 3B-C) and significantly smaller tumor volumes at the endpoint (Fig. 3D), without affecting body weight (Fig. 3E). H&E staining confirmed successful tumor engraftment (Fig. 3F), and Ki-67 staining revealed reduced proliferative activity in shLINC01963 tumors (Fig. 3G).

Fig. 3.

LINC01963 promotes PDAC growth in vivo in both MIA PaCa-2 and PANC-1 orthotopic models. A Schematic diagram of the experimental workflow. Stable MIA PaCa-2/shNC and MIA PaCa-2/shLINC01963 cells (1 × 10.⁶ cells per mouse) were orthotopically implanted into the pancreas of nude mice, followed by weekly IVIS (In Vivo Imaging System) bioluminescence imaging and tumor harvest on Day 39. B Tumor growth curve of orthotopic xenograft model was assessed by weekly IVIS bioluminescence imaging (n = 11–12 mice per group). C Representative IVIS images on Days 5, 31 and 39. D Left, representative photographs of excised tumors at the endpoint (Day 39). Right, quantification of tumor volumes (n = 11–12 mice per group). E Body weights of mice on Day 39. F Representative hematoxylin and eosin (H&E) staining showing low-magnification overviews (scale bar, 1 mm) and high-magnification views highlighting tumor (T) and adjacent normal tissue (N) (scale bar, 250 μm). G Representative H&E and Ki-67 immunohistochemical staining of tumor sections (scale bar, 100 μm). For the PANC-1 orthotopic model, stable PANC-1/shNC and PANC-1/shLINC01963 cells generated by lentiviral infection (1 × 10⁶ cells per mouse) were similarly implanted into the pancreas of nude mice, followed by weekly IVIS imaging and tumor collection on Day 28. H Representative IVIS images on Days 6, 14, and 27. I Bioluminescence-based tumor growth curves (n = 4 mice per group). J Representative photographs of excised tumors at the endpoint. K-L Quantification of tumor volume and tumor weight (n = 4 mice per group). M Body weights at the endpoint. All quantitative data are presented as mean ± SD. Statistical significance was analyzed by two-way ANOVA followed by Sidak’s post-hoc test for longitudinal tumor growth curves (B, I), and two-sided unpaired Student’s t-test (D-E, K-M). * P < 0.05, ** P < 0.01

To substantiate these findings, we established an independent PANC-1 orthotopic xenografts. Consistent with the results in MIA PaCa-2 xenografts, LINC01963 silencing in PANC-1 tumors led to markedly reduced tumor burden, as shown by longitudinal bioluminescence imaging (Fig. 3H-I), tumor volume, and tumor weight (Fig. 3J-L), again without changes in body weight (Fig. 3M). Taken together, these results demonstrate that LINC01963 functions as an oncogenic lncRNA that drives PDAC growth in vitro and in vivo.

LINC01963 exerts an oncogenic role via mediating c-Myc/p21-related signaling pathways

The regulatory relationship between LINC01963 and c-Myc was further investigated in PDAC cells. qRT-PCR revealed that c-Myc mRNA was significantly reduced upon LINC01963 knockdown in PANC-1 and MIA PaCa-2 cells (Fig. 4A). However, the knockdown of c-Myc did not significantly affect the expression of LINC01963 (Fig. 4B). Western blotting also showed that LINC01963 knockdown decreased while LINC01963 overexpression increased the protein level of c-Myc (Fig. 4C). Collectively, all these results establish LINC01963 as an upstream regulator of c-Myc, enhancing its expression at both mRNA and protein levels.

Fig. 4.

LINC01963 exerts oncogenic role via mediating c-Myc/p21-related signaling pathways. A qRT-PCR quantification of c-Myc mRNA after LINC01963 knockdown in PDAC cells (n = 3). B LINC01963 levels in PANC-1 cells measured by qRT-PCR after altering c-Myc expression (n = 3). C c-Myc protein levels in PDAC cells analyzed by western blotting after altering LINC01963 expression. D Colony formation assays showing the effects of c-Myc knockdown or overexpression on LINC01963-induced cell proliferation in MIA PaCa-2 cells. Representative images (left) and quantification (right, n = 3). (E) CCK-8 assays showing the effect of c-Myc modulation on LINC01963-induced proliferation in MIA PaCa-2 cells (n = 3). (F) Flow cytometry (left) and quantification (right, n = 3) showing the effects of c-Myc knockdown or overexpression on LINC01963-induced cell cycle progression. G CDKN1A mRNA levels in MIA PaCa-2 cells analyzed by qRT-PCR after LINC01963 overexpression (n = 3). H-I Western blotting of p21 (CDKN1A) protein levels following LINC01963 knockdown/overexpression (H) or combined modulation with c-Myc (I). Band intensities were measured in ImageJ and expressed as (target/GAPDH) ratios relative to control (set to 1.00). Western blot experiments were independently repeated at least twice with similar results. All quantitative data are displayed as mean ± SD from three independent experiments. Statistical significance was assessed by two-sided unpaired Student’s t-test (A, G), one-way ANOVA followed by Tukey’s post-hoc test (B, D) and two-way ANOVA with Sidak’s post-hoc test (E). * P < 0.05, *** P < 0.001, **** P < 0.0001

To elucidate the role of c-Myc in LINC01963-induced PDAC proliferation, we first analyzed the effects of c-Myc on cell growth. Ectopic expression of c-Myc markedly increased its transcript levels in both PANC-1 and MIA PaCa-2 cells (Supplementary Fig. 2 A). Three independent siRNAs were then designed to silence c-Myc, among which sic-Myc−001 achieved the most efficient knockdown and strongest suppression of cell viability (Supplementary Fig. 2B-D). This siRNA was therefore used in subsequent experiments. Silencing c-Myc significantly suppressed, while overexpression promoted PDAC cell proliferation and cell-cycle progression (Supplementary Fig. 2E-H). Notably, c-Myc overexpression effectively rescued LINC01963 knockdown-impaired cell proliferation, whereas c-Myc silencing blocked LINC01963-induced proliferation (Fig. 4D-E). Thus, c-Myc is a critical downstream effector of LINC01963 in regulating PDAC cell proliferation.

Given the crucial role of c-Myc in cell cycle regulation, we next speculated that LINC01963 affects cell cycle progression in a c-Myc-dependent manner. Flow cytometry analysis demonstrated that LINC01963 knockdown induced cell cycle arrest at the G1 phase, while LINC01963 upregulation promoted G1-to-S-phase transition (Supplementary Fig. 2I). Furthermore, overexpression of c-Myc alleviated the G1 arrest induced by LINC01963 knockdown. In contrast, c-Myc knockdown markedly inhibited the cell cycle progression induced by LINC01963 overexpression (Fig. 4F). Collectively, these findings demonstrate that LINC01963 promotes PDAC cell proliferation by modulating the cell cycle in a c-Myc-dependent manner.

To identify the key c-Myc-regulated genes involved in LINC01963-mediated cell cycle, we screened potential candidates and observed that upregulation of LINC01963 significantly inhibited CDKN1A mRNA levels (Fig. 4G). Western blotting confirmed that the protein levels of p21 (encoded by CDKN1A), a cell cycle inhibitor, were also reduced following LINC01963 upregulation (Fig. 4H). Consistent with previous reports, silencing c-Myc significantly increased CDKN1A expression at both mRNA and protein levels (Supplementary Fig. 2 J-L). Notably, the enhanced p21 expression induced by LINC01963 knockdown was reversed upon c-Myc overexpression, while the inhibitory effect of LINC01963 overexpression on p21 was abolished by c-Myc knockdown (Fig. 4I). Therefore, these results indicate that LINC01963 promotes PDAC cell proliferation via mediating c-Myc/p21-related signaling axis.

LINC01963 preserves METTL3 stability by antagonizing KDM1B-mediated K48-linked ubiquitination to sustain m6A-dependent c-Myc regulation

To elucidate the molecular mechanism underlying LINC01963-mediated regulation of c-Myc expression, we first investigated its effect on c-Myc mRNA stability. RNA stability assays showed that LINC01963 knockdown significantly decreased the half-life of c-Myc mRNA (Fig. 5A). Given the well-documented role of m⁶A modification in regulating c-Myc mRNA stability [24, 27], we further explored whether LINC01963 stabilizes c-Myc mRNA through an m⁶A-dependent mechanism. MeRIP-qPCR demonstrated that LINC01963 silencing significantly reduced the m⁶A modification of c-Myc mRNA (Fig. 5B). Consistent with this finding, MeRIP-seq analysis showed a notable reduction in the m⁶A peak abundance of c-Myc mRNA upon LINC01963 knockdown (Fig. 5C). To further confirm the target specificity of LINC01963 at the transcriptome level, we performed a global MeRIP-seq analysis. The overall distribution of m⁶A methylation remained largely unchanged between siLINC01963 and siNC cells (Supplementary Fig. 3 A), indicating that LINC01963 knockdown does not induce widespread m⁶A alterations. Differential Peak analysis identified significantly altered m⁶A peaks with false discovery rate (FDR) < 0.05 and |log₂fold change|> 0.20 (Supplementary Fig. 3B, Supplementary Table 8). The c-Myc peak exhibited a pronounced site-specific loss of m⁶A enrichment in siLINC01963 cells. These results collectively demonstrate that LINC01963 enhances c-Myc mRNA stability by selectively modulating its m⁶A methylation.

Fig. 5.

LINC01963 stabilizes c-Myc mRNA by protecting METTL3 from KDM1B-mediated K48-linked ubiquitination. A PDAC cells transfected with siNC or siLINC01963 were treated with actinomycin D (ActD, 5 µM) to block transcription. c-Myc mRNA abundance was quantified by qRT-PCR at 0, 30, and 120 min and expressed relative to the 0-min level (n = 3). B MeRIP-qPCR showing enrichment of m⁶A-modified c-Myc mRNA in PDAC cells with or without LINC01963 knockdown (enrichment relative to input) (n = 3). C MeRIP-seq profiles of c-Myc in MIA PaCa-2 cells transfected with siNC or siLINC01963. Compared with siNC, LINC01963 knockdown reduces c-Myc transcript levels (input, blue lines) and m⁶A enrichment (IP, red lines), with normalized IP/Input analysis confirming decreased m.⁶A modification per transcript. D RNA pull-down coupled western blotting analysis showing that METTL3 binds to LINC01963. LacZ RNA served as an unrelated negative control, GAPDH as the loading control. E RIP-qPCR showing the interaction between METTL3 protein and LINC01963 in PANC-1 cells (n = 3). F Representative RNA-FISH-IF images showing colocalization of LINC01963 with METTL3, predominantly in the nucleus in both PDAC cell lines. Scale bar, 5 μm. G–H Western blot showing METTL3 expression in PDAC cells after LINC01963 knockdown or overexpression, with or without MG132 treatment (20 μM, 4 h). I METTL3 protein turnover in PDAC cells after LINC01963 knockdown or overexpression. Cells were treated with cycloheximide (CHX) for the indicated times. Right, quantification of remaining METTL3 normalized to time 0 (n = 3). (J) Ubiquitination assay of METTL3 in PANC-1 cells stably expressing shLINC01963 or shNC. Cells were co-transfected with FLAG-METTL3 and HA-K48Ub plasmids, and treated with or without MG132 (20 μM, 4 h) before lysis. FLAG-METTL3 was immunoprecipitated and immunoblotted with anti-Ub (total ubiquitination) or anti-HA antibody (K48-linked ubiquitination). K48-linked ubiquitin chains represent the canonical ubiquitin linkage that targets substrates for proteasome-dependent degradation. K Western blotting of METTL3 in PDAC cells after silencing predicted E3 ligases (BARD1, STUB1, KDM1B, PRKN, MID1). L Co-IP demonstrating the interaction between METTL3 and KDM1B in PDAC cells after LINC01963 knockdown or overexpression. Cells were treated with MG132 (20 μM, 4 h) before lysis. Lysates were immunoprecipitated with anti-KDM1B and immunoblotted for METTL3. Western blot and RNA-FISH-IF experiments were independently repeated at least twice with similar results. All quantitative data are displayed as mean ± SD from three independent experiments. Statistical significance was assessed by two-sided unpaired Student’s t-test, except for (A, I), which used two-way ANOVA with Sidak’s post-hoc test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001

To further determine the key m⁶A regulators in LINC01963 mediated m⁶A modification of c-Myc mRNA, we performed RNA pull-down assays followed by western blotting, RIP-qPCR and RNA-FISH-IF assays. These experiments revealed that METTL3, an m⁶A writer, specifically binds to LINC01963 and co-localizes with it in PDAC cells (Fig. 5D-F). Based on these findings, we first investigated METTL3’s role in m⁶A-dependent c-Myc regulation. MeRIP-qPCR assays indicated that METTL3 knockdown significantly reduced m⁶A modification on c-Myc mRNA, whereas overexpression enhanced c-Myc expression (Supplementary Fig. 3D-F). Notably, we constructed luciferase reporter plasmids containing either wild-type (c-Myc WT) or mutant m⁶A site of the c-Myc 3’UTR (c-Myc MUT). The mutation site was selected based on the intersection of differential motif enrichment analysis from MeRIP-seq (Supplementary Fig. 3G) and SRAMP-predicted m⁶A modification sites in c-Myc (data not shown). Intriguingly, METTL3 knockdown significantly decreased Fluc activity in the c-Myc WT reporter, but not in the MUT variant (Supplementary Fig. 3H-I). Collectively, these findings establish METTL3-mediated m⁶A modification regulates c-Myc mRNA stability in PDAC.

To investigate whether LINC01963 exert its regulatory effects by modulating METTL3 expression, we performed qRT-PCR and found neither overexpression nor knockdown of LINC01963 significantly altered METTL3 mRNA levels (Supplementary Fig. 4 A). However, LINC01963 overexpression significantly increased, while its knockdown decreased METTL3 protein (Fig. 5G), indicating post-transcriptional regulation. Treatment with the proteasome inhibitor MG132 reversed the effects of LINC01963 knockdown and overexpression on METTL3 protein levels (Fig. 5H). Consistent with this, CHX chase assays showed that silencing LINC01963 significantly shortened, while overexpression prolonged METTL3 protein half-life (Fig. 5I). Given the proteasome-dependent regulation of METTL3 by LINC01963, we focused on K48-linked ubiquitination, the canonical proteasome-targeting ubiquitin linkage formed through lysine 48 of ubiquitin [28]. Silencing LINC01963 elevated K48-linked ubiquitination of METTL3, while MG132-mediated proteasome blockade stabilized these ubiquitinated METTL3 in both groups and attenuated the apparent between-group difference (Fig. 5J, Supplementary Fig. 4B). Therefore, LINC01963 stabilizes METTL3 by suppressing K48-linked ubiquitination and proteasomal degradation, with concomitant changes in c-Myc and p21 protein expression (Supplementary Fig. 4C).

To elucidate the mechanism by which LINC01963 regulates METTL3 ubiquitination, we screened potential E3 ubiquitin ligases targeting METTL3, including top candidates predicted by UbiBrowser 2.0 (Supplementary Fig. 4D) and the previously reported METTL3 E3 ligase STUB1. Initial screening in PDAC cell lines revealed that knockdown of either STUB1 or KDM1B significantly enhanced METTL3 protein levels (Fig. 5K). However, Co-IP assays showed that only the METTL3-KDM1B interaction was specifically modulated by LINC01963. LINC01963 depletion strengthened the METTL3-KDM1B interaction, while its overexpression attenuated this binding (Fig. 5L, Supplementary Fig. 4E-F). Consistently, KDM1B positively regulated METTL3 ubiquitination, as KDM1B overexpression increased, whereas its knockdown reduced METTL3 ubiquitination in PDAC cells (Supplementary Fig. 4G). Under proteasome-inhibited conditions, silencing KDM1B significantly reduced K48-linked ubiquitination of METTL3 (Supplementary Fig. 4H-I). These results indicate that KDM1B is a crucial E3 ubiquitin ligase mediating LINC01963-associated K48-linked ubiquitination and proteasomal degradation of METTL3.

Structurally, METTL3 comprises an N-terminal leading helix (LH), a nuclear localization signal (NLS), a zinc finger domain (ZFD) containing two CCCH-type motifs (ZnF1 and ZnF2), and a C-terminal methyltransferase domain (MTD) [29, 30] (Supplementary Fig. 5A). AlphaFold-guided structural modeling predicted that KDM1B and LINC01963 interact with METTL3 through partially overlapping, multi-residue interfaces, with Tyr328 positioned at the intersection of the two interfaces (Supplementary Fig. 5B). To delineate these interactions, we generated FLAG-tagged METTL3 full-length, domain truncations, and mutant constructs, including a multi-site interface mutant (Mut-KL) and a Tyr328 point mutation (Mut-328) (Supplementary Fig. 5 C). Co-IP assays revealed that KDM1B interaction with METTL3 was markedly reduced by the Mut-KL interface mutations, whereas the Tyr328 point mutation caused only a partial reduction (Supplementary Fig. 5D). Consistently, RIP-qPCR assays demonstrated that LINC01963 interaction with METTL3 was strongly impaired in the Mut-KL construct, while the Tyr328 mutation exerted a substantially weaker effect (Supplementary Fig. 5E). Importantly, LINC01963 knockdown enhanced KDM1B-METTL3 association only in the wild-type METTL3 context, but not in the Mut-KL mutant (Supplementary Fig. 5 F). Together, these results demonstrate that LINC01963 stabilizes METTL3 by competing with KDM1B through a distributed, multi-site interaction interface, rather than a single dominant residue.

LINC01963 specifically interacts with the KH1-2 domain of IGF2BP2 in PDAC cells

To investigate whether m⁶A readers participate in LINC01963-regulated c-Myc stability, we performed RNA pull-down assays followed by mass spectrometry. Consistently, IGF2BP2, a key m⁶A reader, was identified as a stable binding partner of LINC01963 in replicate experiments (Fig. 6A). This interaction was further validated by RNA pull-down coupled with western blotting (Fig. 6B), RIP-qPCR (Fig. 6C) and RNA-FISH-IF assays (Fig. 6D). In addition, catRAPID predictions indicated a high binding probability between LINC01963 and IGF2BP2, supporting the robustness of their interaction (Fig. 6E). Collectively, these results indicate that LINC01963 stably interacts with IGF2BP2.

Fig. 6.

LINC01963 specifically interacts with the KH1-2 domain of IGF2BP2 in PDAC cells. A Left, schematic diagram of LINC01963 RNA pull-down coupled with LC–MS/MS using 3’-biotinylated antisense tiling probes against LINC01963 (even/odd-numbered probe sets) versus non-specific control probes (LacZ) in MIA PaCa-2 cells. Right, Coomassie-stained SDS-PAGE of pull-down eluates showing specifically enrichment of IGF2BP2 (red box). B RNA pull-down followed by western blotting confirming IGF2BP2 binding to LINC01963. GAPDH served as the loading control. C RIP-qPCR in MIA PaCa-2 cells showing a significant enrichment of LINC01963 in IGF2BP2 immunoprecipitates relative to IgG control (normalized to input) (n = 3). D Representative RNA-FISH-IF images showing colocalization of LINC01963 and IGF2BP2, predominantly in the cytoplasm of PANC-1 cells. Scale bar, 5 μm. E Predicted interaction parameters from catRAPID indicate a high interaction propensity for IGF2BP2-LINC01963 pair. F catRAPID residue-nucleotide interaction map showing multiple predicted binding hotspots, primarily within IGF2BP2 amino acids (aa) 200–370 (y-axis) and corresponding clusters along the LINC01963 transcript (x-axis). G Schematic diagram of FLAG-tagged full-length IGF2BP2 (WT) and truncated or domain-deleted constructs. RRM, RNA recognition motif; KH, K-homology domain. Numbers indicate amino acid positions. H RIP-qPCR assays showing significant enrichment of LINC01963 by full-length IGF2BP2 (aa 1–599) and by the isolated KH1-2 domain (aa 195–352) in MIA PaCa-2 cells. NC, empty-vector control (n = 3). I Deletion of IGF2BP2 KH1-2 domain (Δ195–352) abrogates LINC01963 binding, reducing RIP-qPCR signal to background levels (n = 3). J Schematic diagram of full-length LINC01963 and three truncated fragments (#1-#3) designed based on starBase and catRAPID predictions for RNA pull-down assays. K Biotinylated RNA pull-down followed by western blotting in MIA PaCa-2 cells revealed that IGF2BP2 binds three distinct LINC01963 regions. “Sense” represents the 3’-biotinylated full-length LINC01963 RNA. “NC1/NC2” are 3′-biotinylated antisense RNAs targeting LINC01963, used as negative controls. GAPDH served as a non-binding control. Western blot experiments were independently repeated at least twice with similar results. All quantitative data are displayed as mean ± SD from three independent experiments. Statistical significance was assessed by two-sided unpaired Student’s t-test. ****P < 0.0001

To delineate the specific binding regions, we first analyzed the structural domains of IGF2BP2, which comprises two RNA recognition motifs (RRM1-2) and four K-homology domains (KH1-4) [31]. Based on catRAPID predictions, we constructed FLAG-tagged plasmids encoding full-length IGF2BP2 and various truncations targeting individual domains (Fig. 6F-G). RIP-qPCR revealed that KH1-2 region is essential for LINC01963 binding (Fig. 6H-I). Similarly, we designed truncated LINC01963 fragments based on integrated predictions from starBase and catRAPID (Fig. 6J). In vitro RNA pull-down assays identified the 900-1101nt, 1346-1566nt, and 2035-2258nt regions of LINC01963 as critical for IGF2BP2 binding, suggesting multiple binding interfaces (Fig. 6K).

LINC01963 stabilizes c-Myc mRNA by cooperating with IGF2BP2

Given that IGF2BP2 stabilizes m⁶A-modified mRNAs and directly binds LINC01963 [32, 33], we hypothesized that IGF2BP2 is required for LINC01963-induced c-Myc mRNA stabilization. mRNA stability assays showed that IGF2BP2 silencing significantly blocked LINC01963-mediated c-Myc mRNA stabilization (Fig. 7A), demonstrating the indispensable role of IGF2BP2 in maintaining c-Myc stability. Intriguingly, LINC01963 depletion significantly reduced IGF2BP2 binding to m⁶A-modified c-Myc, which could not be rescued by METTL3 overexpression (Fig. 7B). These findings highlight the critical role of LINC01963 in stabilizing c-Myc mRNA by facilitating IGF2BP2-mediated recognition of m⁶A-modified c-Myc. To further explore the regulatory relationship between LINC01963 and IGF2BP2, we analyzed TCGA-PAAD and GETx datasets, along with qRT-PCR and western blotting. Integrated analyses showed that LINC01963 did not regulate IGF2BP2 expression at either the mRNA or protein level (Fig. 7C-E). Taken together, these findings suggest that LINC01963 functions as a scaffold that forms a ternary complex with IGF2BP2 and m⁶A-modified c-Myc to stabilize c-Myc mRNA in PDAC cells.

Fig. 7.

LINC01963 stabilizes c-Myc mRNA by cooperating with IGF2BP2. A IGF2BP2 knockdown PDAC cells with ectopically expressed LINC01963 were treated with 5 µM ActD to block transcription. c-Myc mRNA abundance was quantified by qRT-PCR and statistical significance was calculated at the final time point (120 min) (n = 3). B RNA immunoprecipitation with anti-IGF2BP2 or control IgG followed by qRT-PCR to quantify c-Myc mRNA in siNC, siLINC01963, and siLINC01963 + METTL3 overexpression cells (n = 3). Enrichment is reported as IP/Input and shown relative to the IgG control. C Spearman’s correlation between IGF2BP2 mRNA and LINC01963 levels across PDAC tumors (TCGA-PAAD) and normal pancreas samples (GTEx). D-E IGF2BP2 expression in MIA PaCa-2 cells upon LINC01963 knockdown or overexpression was assessed by qRT-PCR (n = 3) and western blotting. F Schematic illustration of LINC01963–mediated cell cycle and cell proliferation in PDAC. This figure was created by Biorender. Western blot experiments were independently repeated at least twice with similar results. All quantitative data are displayed as mean ± SD from three independent experiments. Statistical significance was assessed by two-way ANOVA with Sidak’s post-hoc test (A-B), Pearson correlation analysis (C) and two-sided unpaired Student’s t-test (D). * P < 0.05, ** P < 0.01, **** P < 0.0001

Discussion

Across multiple independent cohorts, we identified LINC01963 as a c-Myc-associated lncRNA that is consistently overexpressed in PDAC, enriched in advanced-stage disease, and associated with poor patient survival. Mechanistically, our study revealed LINC01963 regulates c-Myc expression by stabilizing METTL3 and facilitating IGF2BP2-mediated recognition of m⁶A-modified c-Myc mRNA. Functionally, LINC01963 inhibition induces cell cycle arrest via the c-Myc/p21-related signaling, suppressing PDAC cell proliferation and xenograft tumor growth (Fig. 7F). Collectively, these findings elucidated the molecular mechanisms by which LINC01963 regulates c-Myc, highlighting the LINC01963/METTL3/IGF2BP2 axis as a promising therapeutic target for indirectly modulating c-Myc in PDAC.

c-Myc, a critical transcription factor regulating proliferation and cell cycle, is overexpressed in PDAC and drives tumor progression [34, 35]. However, its “undruggable” nature due to structural limitations highlights the need to identify potential regulators of c-Myc [36–38]. Therefore, we performed comprehensive bioinformatics analyses and identified six lncRNAs that are significantly correlated with c-Myc expression and overexpressed in PDAC tumors. Among these, LINC01963 showed consistent associations with advanced disease stage and unfavorable prognosis across multiple cohorts, suggesting a potential oncogenic role in PDAC. LINC01963 was previously reported to decrease chemosensitivity of prostate cancer cells to docetaxel via the miR-216b-5p/TrkB axis [39]. In pancreatic cancer, LINC01963 was identified as a component of a prognostic risk score model [40]. These findings collectively highlight the multifaceted role of LINC01963 in cancer progression. Our findings further validate LINC01963 as an oncogenic lncRNA that promotes c-Myc expression and PDAC progression, as demonstrated by both in vitro and in vivo experiments. Mechanistically, LINC01963 regulates PDAC cell cycle and proliferation via the c-Myc/p21-related pathways. These results suggest that inhibiting LINC01963 could serve as a promise therapeutic strategy for PDAC. In addition, we observed that LINC01963 promotes PDAC cell motility in vitro, as evidenced by wound healing and Transwell assays (data not shown). However, in vivo metastasis models were not included in the present study, whether LINC01963 functionally contributes to PDAC metastasis warrants future investigation using in vivo metastasis models.

Since LINC01963 mediated c-Myc stability in an m⁶A-dependent manner, we next sought to identify the m⁶A writer involved in this process and validated METTL3 as a key LINC01963-interacting protein by LC–MS/MS analysis. Consistently, silencing either LINC01963 or METTL3 significantly reduced the m⁶A modification of c-Myc. Mechanistically, our data support a model that LINC01963 associates with the c-Myc-METTL3 complex and competitively interferes with the binding of the E3 ubiquitin ligase KDM1B, thereby attenuating K48-linked ubiquitination and proteasomal degradation of METTL3. This notion is further supported by the observation that LINC01963 depletion enhances, whereas its overexpression weakens, the METTL3-KDM1B interaction. KDM1B is well recognized not only as a histone demethylase but also as an E3 ubiquitin ligase that suppresses cancer cell growth by directly ubiquitinating and promoting the proteasome-dependent degradation of oncogenic substrates, such as O-GlcNAc transferase [41–43]. Notably, our study identifies METTL3 as a previously unrecognized ubiquitination substrate of KDM1B, extending the functional scope of KDM1B beyond histone demethylation. This observation raises the possibility that KDM1B-mediated ubiquitin-dependent degradation of METTL3 may be mechanistically linked to KDM1B demethylase activity, a relationship that remains to be explored.

Cellular proteins can undergo mono-, multi-, or polyubiquitination, and polyubiquitin chains linked through distinct lysine residues confer diverse functional outcomes [44, 45]. Among these, K48-linked polyubiquitin chains represent the most canonical ubiquitin linkage type for proteasome-mediated protein degradation [46–48], whereas K63-linked chains, the next most abundant linkage, are primarily associated with non-proteolytic regulatory processes [49]. In this study, several lines of evidence indicate that METTL3 is regulated through the ubiquitin–proteasome system downstream of LINC01963. Specifically, proteasome inhibition abolished the effect of LINC01963 on METTL3 protein abundance and stability, and MG132 treatment consistently resulted in the accumulation of ubiquitinated METTL3 in Co-IP-based assays, supporting a proteasome-dependent turnover mechanism. Guided by this proteasome dependence, we next focused on K48-linked ubiquitination as the most functionally relevant modification underlying METTL3 destabilization. Indeed, silencing LINC01963 selectively enhanced K48-linked ubiquitination of METTL3. In contrast, proteasome blockade stabilized ubiquitinated METTL3 in both control and LINC01963-depleted cells, thereby attenuating the apparent difference between groups. While other ubiquitin linkages, such as K63-linked chains, may contribute to non-proteolytic regulation of METTL3, their potential involvement was not addressed here and warrants future investigation.

AlphaFold 3.0 structural modeling predicted that LINC01963 and KDM1B may competitively engage METTL3 through partially overlapping, multi-residue interfaces, with Tyr328 located at a predicted interface hotspot. Consistent with these predictions, Co-IP and RIP-qPCR assays precisely mapped the METTL3 interaction regions for KDM1B and LINC01963 and revealed substantial overlap within the ZnF2 and MTD domains. Furthermore, mutation of multiple interface residues within the ZnF2 domain (Mut-KL) markedly impaired both KDM1B and LINC01963 binding far more strongly than mutation of Tyr328 alone (Mut-328), indicating that these interactions are governed by a distributed interface rather than a single dominant residue; notably, this effect was more pronounced than deletion of the entire ZnF2 domain, suggesting a reliance on specific interface residues rather than the intact domain structure per se. To directly validate the competitive binding model, Co-IP assays showed that the Mut-KL construct exhibited the most pronounced loss of interaction with both KDM1B and LINC01963. Importantly, this mutation abolished the LINC01963 knockdown-induced increase in KDM1B-METTL3 association, providing functional evidence for a competitive, multi-site binding mechanism. Together, these findings demonstrate that LINC01963 stabilizes METTL3 by antagonizing KDM1B binding through shared, multi-site interaction surfaces, rather than by shielding a single residue.

Accumulating evidence has demonstrated that lncRNAs may serve as molecular scaffolds, stabilizing target genes by facilitating the formation of functional complexes between m⁶A reader and m⁶A-modified mRNAs [19, 50–52]. To identify the key m⁶A reader in the LINC01963/METTL3/c-Myc axis, RNA pull-down assays coupled with mass spectrometry revealed IGF2BP2 as a critical component. LINC01963 acted as a scaffold to form a ternary complex between IGF2BP2 and the m⁶A-modified c-Myc mRNA, thereby stabilizing c-Myc in an IGF2BP2-dependent manner. This finding was further validated by dual-luciferase reporter assays, RIP-qPCR, RNA-FISH-IF, and AlphaFold3.0 structural modeling, all of which mapped the interaction sites within the LINC01963/IGF2BP2/c-Myc complex. Notably, our RIP-qPCR experiments revealed that IGF2BP2 binding to m⁶A-modified c-Myc was significantly reduced upon LINC01963 depletion and could not be rescued by METTL3 overexpression. These results demonstrated that LINC01963 is indispensable for IGF2BP2-mediated recognition and stabilization of m⁶A-modified c-Myc. In addition, our study reveals a novel mechanism by which LINC01963 stabilizes c-Myc mRNA by simultaneously modulating both the “m⁶A writer” (METTL3) and “m⁶A reader” (IGF2BP2). This finding not only highlights LINC01963 as a crucial regulator of c-Myc stability, but also extends our understanding of c-Myc regulation beyond the canonical ceRNA mechanism [39]. Up to now, this study provides the first demonstration that a single lncRNA can coordinate multiple m⁶A modulators to maintain oncogenic mRNA stability. Notably, although METTL3 and IGF2BP2 act at distinct stages of the m⁶A regulatory process, the stabilizing effect on c-Myc mRNA ultimately depends on IGF2BP2-mediated recognition of m⁶A modifications, representing a convergence point of these regulatory mechanisms. However, the relative quantitative contribution of each mechanism cannot be precisely resolved in the current experimental framework. In the future, the relative quantitative contribution and hierarchical order of these regulatory layers remain to be further explored.

To explore the mechanisms underlying LINC01963 overexpression in PDAC, we initially analyzed potential genomic alterations through cBioPortal database, while no mutation was found. We also investigated the copy number alterations (CNAs) data and found only three cases with CNAs (data not shown). These results indicate that genomic alterations are unlikely to account for LINC01963 overexpression in PDAC. We further predicted potential transcription factors (TFs) binding to the LINC01963 promoter via the CHIPBase database, and identified several TFs such as P300, HSF1, and SIP4, that potentially bound to the LINC01963 promoter (data not shown). These suggest that upregulation of LINC01963 may be caused by dysregulation of transcription, which needs further exploration.

Conclusions

In summary, our study has identified LINC01963 as a novel oncogenic lncRNA in PDAC that stabilized c-Myc mRNA through METTL3/IGF2BP2 axis-mediated m⁶A modification. LINC01963 enhanced c-Myc m⁶A modification by preventing METTL3 from KDM1B-mediated K48-linked ubiquitination and proteasomal degradation and subsequently promoted IGF2BP2-dependent stabilization of m⁶A-modified c-Myc mRNA. Silencing LINC01963 inhibited PDAC cell proliferation in vitro and tumor growth in vivo by disrupting c-Myc/p21-mediated cell cycle regulation. These findings not only elucidate a novel LINC01963/METTL3/IGF2BP2/c-Myc regulatory axis in PDAC, but also provide a potential therapeutic strategy for indirectly targeting the “undruggable” c-Myc.

Abbreviations

PDAC

Pancreatic ductal adenocarcinoma

LncRNAs

Long noncoding RNAs

K48

Lysine 48

m⁶A

N6-methyladenosine

METTL3

Methyltransferase-like 3

IGF2BP2

Insulin-like growth factor 2 mRNA-binding protein 2

KDM1B

Lysine Demethylase 1B

HPNE

Human normal pancreatic ductal cells

qRT-PCR

RNA extraction and quantitative real-time PCR analysis

LC–MS/MS

Liquid Chromatography-Tandem Mass Spectrometry

MeRIP-seq

Methylated RNA Immunoprecipitation Sequencing

RIP-qPCR

RNA Immunoprecipitation Followed by qPCR

RNA-FISH-IF

RNA FISH and immunofluorescence staining

H&E

Hematoxylin and eosin

Authors’ contributions

Z. Liang and H. Wu conceived the study and supervised the project. Q. Liu, B. Liu and R. Li performed most experiments. Q. Liu and X. Shi carried out statistical and bioinformatics analyses. Q. Liu and B. Liu conducted the animal experiments. H. Liu, X. Yin, X. Ju, S. Zhang and J. Wang collected clinical samples and performed data preprocessing. X. Liu, D. Li and L. Chen contributed to histopathological analyses. Y. Niu provided animal facility support and scientific advice. All authors have read and approved the final manuscript.

Funding

This work was supported by Capital’s Funds for Health Improvement and Research (CFH) (2024–2-4012), Chinese Academy of Medical Sciences (CAMS) Innovation Fund for Medical Sciences (CIFMS) (2024-I2M-C&T-C-001), the Fundamental Research Funds for the Central Universities (3332024005), National Natural Science Foundation of China (82403474), National High Level Hospital Clinical Research Funding (2025-PUMCH-D-002), the National Key Clinical Specialty Construction Project (U114000) and Beijing Municipal Natural Science Foundation (L252174).

Data availability

MeRIP-seq raw data generated in this study are publicly available in the Gene Expression Omnibus at GSE295171. The data generated in this study are available upon reasonable request from the corresponding author.

Declarations

Ethics approval and consent to participate

All animal experiments were approved by the Institutional Animal Care Use and Welfare Committee at the Center for Experimental Animal Research of the Institute of Basic Medical Sciences, Chinese Academy of Medical Sciences (IACUC-A02-2023–059). Animals were handled in accordance with institutional guidelines.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.