2.1 Downloading of the PC dataset in the cancer genome atlas (TCGA) and gene expression omnibus (GEO)

We used the R package TCGAbiolinks to download the PC transcriptome data from TCGA database and combined them with healthy human tissue transcriptome data in GTEx to explore differential gene expression in PC. Also, RNA sequencing data was downloaded in GEO to measure the expression of regulators in PC. Patients with rational survival data and clinicopathological characteristics were included in the analysis. For group analysis, the median value was used as the cut-off site.

2.2 Clinical bioinformatic analysis

2.3 Prediction of IGF2BP2 target genes

To predict and verify the possible downstream targets regulated by IGF2BP2, we analyzed and compared the differential genes between the two subgroups (PA1 and PA2), and the differential genes in PC. In addition, the CLIP dataset in m6A2Target database was used to filter the downstream mRNAs of IGF2BP2, the possible targets of IGF2BP2 were then obtained by intersection, and the follow-up experiments were carried out to verify it.

2.4 Clinical PC specimens

PC together with counterpart adjacent normal tissue samples were obtained from department of General Surgery, the First Affiliated Hospital of Soochow University. The research was ratified by the Ethical Committee of the first affiliated hospital of Soochow University and written informed consent were successively obtained from all participants before the study.

2.5 Cell culture

Human CFPAC1 and PANC1 PC cells were purchased from GenePharma Technology Co., Ltd. (Suzhou, China) and used in our research. These cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% foetal bovine serum (FBS) in an incubator with a 5% CO₂ atmosphere at 37°C. All cells were free of bacterial and mycoplasma contamination.

2.6 Transfection

The target mRNA interference fragments and the NC construct si-NC were synthesized (Synbio Technologies, China) and co-transfected with Lipofectamine 2000 reagent (Invitrogen, USA) to knock down gene expression in PC cells; their sequences are shown in Table S1.

2.7 Quantification of overall m6A RNA methylation

We used an EpiQuik m6A RNA Methylation Quantification Kit (colorimetric) (Epigentek, USA) to quantify the overall m6A methylation level in PC cells. First, we extracted total RNA and approximately 200 ng RNA was separated as an initial input. Then, the sample RNA, standard positive/negative control and binding solution were bound to the wells respectively for 1 h to assay and capture RNA according to the manufacturer's instruction. After washing, the detection antibody and enhancer solution were added, and color developer solution and stop solution was added prior to measurement of the absorbance. The values were calculated using linear regression equations.

2.8 RNA extraction and quantitative real-time polymerase chain reaction (qRT-PCR)

Total RNA was extracted with TRIzol from PC cells and tissues. All primers used for qRT-PCR were designed and blasted in the National Center for Biotechnology Information database. RNA was reverse transcribed to cDNA and PCR was then performed using SYBR Green qPCR Master Mix (GenePharma, China) in triplicate. Human β-actin was used as the internal control for mRNA expression. The primer sequences are shown in Table S1.

2.9 CCK-8 proliferation assay

CCK-8 kit (Dojindo Laboratories, Japan) was used to detect the proliferation ability of PC cells. Transfected cells were seeded in 96-well plates at a concentration of 5 × 10³ cells per well. After cultured for 0, 24, 48, 72 and 96 h under the same conditions, PC cells were incubated with diluted CCK-8 reagent following the instructions. Then, the absorbance was measured in a microplate reader at a wavelength of 450 nm.

2.10 EdU assay

An EdU kit (RiboBio, China) was used to detect the cell proliferation capability. In brief, after stable transfection for 48 h, cells were incubated with 5 μM EdU reagent diluted with DMEM for 3 h. Then, cells were permeabilized with 0.5% Triton X-100 for 20 minutes prior to fixation with 4% paraformaldehyde for 30 minutes. Apollo and DAPI dyes were used to stain DNA and then nuclei. Images of EdU- and DAPI-positive cells were acquired under a fluorescence microscope.

2.11 Colony formation assay

To verify colony formation ability, transfected cells were seeded into 6-well plates at a density of 5000 cells per well and distributed evenly by pipetting. After 14 days, the cells were washed with PBS twice, fixed with 4% paraformaldehyde for 30 min and stained with 1% crystal violet for 30 min. Eventually, the cells were imaged and the colonies were counted.

2.12 Transwell invasion assay

To evaluate the cell migration ability, cells were cultured in serum-free medium in the upper chamber of a Transwell plate (Corning, USA), and 700 μl of DMEM containing 10% FBS was added to the lower chamber. After 48 h, the cells were fixed and stained with crystal violet for 20 min, and the cells migrating through the upper chamber membrane were counted. Images were acquired by microscopy.

2.13 Luciferase reporter assay

The B3GNT6 fragment containing the possible m6A modification site and the corresponding mutated fragment were synthesized and cloned into the pmirGLO vector (Promega, USA) named B3GNT6-WT and B3GNT6-Mut. The synthesized vectors were separately co-transfected with IGF2BP2-si or si-NC into 293 T cells. After co-incubation for 24 h, cells were collected, and luciferase activities were measured with a Dual Luciferase Reporter Assay kit (GenePharma, China).

2.14 m6A IP and IGF2BP2 IP assays

Immunoprecipitation assays of m6A and IG2BP2 were performed with an RIP kit (BersinBio Biotech, China). In brief, approximately 1 × 10⁷ treated cells were collected and lysed for 30 min. DNA impurities were removed with DNase, the supernatant was collected after centrifugation, and the supernatant was then incubated with anti-m6A and anti-IGF2BP2 primary antibodies overnight at 4°C in a vertical orientation. Then, magnetic beads were added to incubate for 1 h. Finally, RNA was extracted, and target gene expression was detected by qRT-PCR.

2.15 3-DAA demethylation treatment

The overall methylation inhibitor 3-DAA (Cayman, USA) was used to inhibit RNA methylation levels. After inoculation, cells were treated with 10 ug/ml 3-DAA for 48 hours, then cells were collected for protein analysis.

2.16 IHC analysis

Paraffin sections of 5 paired clinical pancreatic specimens were prepared and baked in an oven at 65°C for 30 min. After routine dewaxing, hydration and antigen repair, sections were sealed with 5% goat serum for 30 min, placed in a wet box and incubated overnight with the IGF2BP2 primary antibody (ID:OTI3C8, OriGene, American) at 4°C. The next day, sections were incubated with the secondary antibody for 30 min, incubated with DAB for color development for 3 min and stained with hematoxylin for 3 min. Sections were sealed with neutral gum, and images were acquired with a microscope.

2.17 IF detection assay

After dewaxing, hydration and antigen repair, paraffin sections were blocked with immunostaining blocking solution for 1 h, incubated with IGF2BP2 and B3GNT6 (ID:21291-1-AP, Proteintech, American) primary antibody at 4°C overnight and secondary antibody for 1 h. After counterstaining with DAPI, sections were sealed with an anti-fluorescence quenching agent and imaged under a fluorescence microscope. Treated PC cells were prepared as cell slides and subsequent experiments were performed as described above.

2.18 Western blot analysis

We used a RIPA mixture containing 1% PMSF lysate to lyse and obtain total protein from transfected cells. A BCA protein detection kit (Thermo Scientific Pierce, USA) was used to determine the protein concentration. Protein separation under 80 V, 30 min and 120 V, 1 h electrophoresis conditions, followed by transferring to polyvinylidene difluoride membranes at 200 mA for 1.5 h. Then, membranes were blocked with 5% skim milk and incubated with IGF2BP2 primary antibody (OriGene, American) and B3GNT6 (Proteintech, American) primary antibody overnight at 4°C. After washing, immunoreactions were visualized with an electroluminescence detection system.

2.19 Animal experiment

To evaluate the effect of IGF2BP2 on the tumor formation ability in vivo, 2 × 10⁶ cells were injected into the axilla of nude mice for subcutaneous tumor formation. After 9 days, modified IGF2BP2-si or si-NC was injected into the tumors every 3 days for 3 weeks. Tumor size was measured every 3 days. Finally, the subcutaneous tumors were excised and weighed to compare the tumor sizes, and protein expression was detected by immunohistochemistry and immunofluorescence. The care of laboratory animals was in accordance with the guidelines and ethical requirements of the Laboratory Animal Centre of Soochow University.

2.20 Statistical analysis

We used GraphPad 8.0 for charting. For statistical analysis, two-tailed Student's t-test between two groups were performed by SPSS Statistic 26 for windows. The Kaplan–Meier method was used for survival analysis. Data was reported as mean ± SD from three duplicate experiments. Results was considered statistically significant when p < 0.05 (*p < 0.05).

3.1 Overall expression pattern of m6A regulators

Considering that methylation regulators perform different biological roles, we first analyzed the expression of 15 regulators in PC and found that compared with normal pancreatic expression profiles in the GTEx database, the regulators were generally upregulated in TCGA, it was also observed in the GEO database (GSE15471) (Figure 1A,B). Spearman correlation analysis showed a close relationship between the factors (Figure 1C). Analysis of the STRING database also revealed a positive correlation between multiple regulators (Figure 1D). Our CNV analysis showed tumor size with regulators CNV was bigger than that without CNV (Figure 1E), suggesting that CNV is an important element leading to the upregulation of regulators and the progression of PC.

3.2 Identification of two subgroups of PC by consensus clustering

To visualize the specific functions of m6A methylation regulators, we removed normal pancreatic samples and PC samples with no suitable clinical data, and used the ConsensusClusterPlus R package to perform consensus clustering. According to the expression similarity of the regulators, the cumulative distribution function (CDF) of our data seemed to be ideal when K = 2, and the PA1 (96 samples) and PA2 (80 samples) subgroups were then defined (Figure 1F,G). We compared the clinicopathological characteristics of PA1 and PA2 and found them to be consistent with the above expectations. K-M survival analysis showed significant differences between the two subgroups (Figure 1H).

The edgeR software package was used to analyze the difference between the expression profiles of PA1 and PA2, and the DEGs were selected based on the criteria |logFC| ≥ 1 and p ≤ 0.05. Finally, 2681 genes were identified, of which 724 genes were upregulated in PA1, and 1957 genes were upregulated in PA2. KEGG and GO enrichment analyses showed that the DEGs were mainly enriched in homeostasis or regulation of membrane potential (Figure 2A,B), such as Cytokine−cytokine receptor interaction and pancreatic secretion, indicating that mRNA methylation affects multiple biological processes.

3.3 Analysis of associations between m6A methylation regulators and clinicopathological characteristics

To further clarify the role of the regulators in survival, we conducted univariate Cox regression analysis. Among the 15 regulators, 5 were related to prognosis: IGF2BP2 (HR = 2.19), IGF2BP3 (HR = 1.53), METTL3 (HR = 0.53), ALKBH5 (HR = 0.6) and KIAA1429 (HR = 1.55) (Figure 2C). Survival models based on high reliability and LASSO regression are widely used to screen prognostic genes from high-dimensional data; thus, we established risk characteristics and performed LASSO-Cox regression analysis to calculate the risk score, and found that KIAA1429, METTL3, IGF2BP2 and ALKBH5 were the main contributors (Figure 2D,E), where the coefficient of KIAA1429, METTL3, IGF2BP2 and ALKBH5 were 0.00027, −0.00056, 0.00015 and −0.00019, respectively. According to the median risk score, low-risk patients had higher survival status (Figure 2F). The respective survival rate analysis showed higher IGF2BP2 contributed to lower survival rate (Figure 2G). Heatmap showed that ALKBH5 and METTL3 were upregulated in the low-risk group, while IGF2BP2 and KIAA1429 were upregulated in the high-risk group (Figure 2H). We evaluated the correlations between the risk subgroups and the clinicopathological features. Univariate and multivariate Cox regression analyses showed that the risk score (p < 0.05) and lymph node metastasis (p < 0.05) were highly significant (Figure 2I,J). In summary, we believe that the accuracy of m6A methylation regulators for predicting the prognosis of PC was further corroborated. Considering that IGF2BP2 is highly expressed in PC, with a high HR, and is strongly independently associated with prognosis, we selected it as the main regulator for subsequent studies.

3.4 IGF2BP2 is a credible molecular prognostic marker in PC

TCGA and IHC analysis revealed high expression of IGF2BP2 in PC (Figure 3A,B). What's more, quantitative analysis of RNA methylation indicates an overall upregulation of m6A modification in PC tissues (Figure 3C). Combining positive experiments and bioinformatics results, we then sought to investigate some specific reasons for high IGF2BP2 and m6A expression. TCGA data analysis showed that increasing CNV contributed to a higher mRNA level (Figure 3D), while the DNA methylation level was negatively correlated with the mRNA expression (Figure 3E), indicating that DNA methylation and CNV were indispensable cooperative factors for high IGF2BP2 expression. These results indicated that IGF2BP2 was a marker of unfavorable prognosis in PC patients.

3.5 IGF2BP2 promotes PC cell proliferation and migration

For further investigation of the biological function of IGF2BP2 in PC, we knocked down its expression in CFPAC1 and PANC1 cells (Figure 3F,H) and quantitative analysis of m6A revealed no changes (Figure S1). CCK-8 and EdU assays indicated that silencing IGF2BP2 obviously suppressed the proliferation of PC cells (Figure 4A–D). Also IGF2BP2 KO decreased cell migration ability with Transwell assay (Figure 4E). The colony formation assay showed that inhibition of IGF2BP2 decreased cell viability (Figure 4F). These results illustrate that IGF2BP2 exerts oncogenic effects in PC.

3.6 IGF2BP2 regulates B3GNT6 expression

As an RNA binding protein, IGF2BP2 has been proven to bind to a variety of RNAs to regulate their expression. To identify potential target genes of IGF2BP2, we analyzed IGF2BP2 CLIP data and identified 6249 targets with relatively high associativity. Subsequently, PC patients were divided into high IGF2BP2 expression and low IGF2BP2 expression group, and differential gene expression analysis was conducted between the groups. In TCGA, we identified 8335 DEGs between PC and normal tissue, among which nine were common highly reliable genes (Figure 5A). qRT-PCR confirmed that after knockdown of IGF2BP2, some genes were downregulated to varying degrees, of which, B3GNT6, DHRS9 and ALPP had the largest decreases (Figure 5B). We performed RIP with IGF2BP2 and m6A antibody (Figure S2) and results showed that B3GNT6 was more highly enriched than DHRS9 and ALPP (Figure 5C). Also, TCGA data analysis indicated that B3GNT6, ALPP and DHRS9 was upregulated in PC (Figure 5D; Figure S3). B3GNT6 is a beta-1,3-N-acetylglucosaminyl transferase that adds an N-acetylglucosamine moiety to N-acetylgalactosamine-modified serine or threonine residues to affect biosynthesis and metabolism. Studies have found that B3GNT6 is closely related to a variety of disease processes, such as immunity deficiency,²¹ colorectal cancer and m6A modification,²² Together, these results indicate that B3GNT6 may be the downstream target of IGF2BP2. Survival analysis with TCGA data showed high B3GNT6 was related to lower survival status, and Spearman correlation analysis also showed the co-expression of IGF2BP2 and B3GNT6 (Figure 5E). Given that B3GNT6 is clinically relevant, closely related to m6A and understudied in PC, we mainly focused our research on it.

3.7 Inhibition of B3GNT6 decreased the proliferative ability of PC cells

We first knocked down B3GNT6 in cells, CCK-8 and EdU incorporation assays indicated that silencing B3GNT6 overtly inhibited the proliferation of PC cells (Figure 5F–H). In addition, Transwell and colony formation assay showed that B3GNT6-si inhibited cell migration and viability (Figure 5I,J). These results indicate that B3GNT6 can exert oncogenic effects in PC.

3.8 IGF2BP2 recognizes m6A methylation on B3GNT6 and increases its stability

To further identify the underlying mechanism of IGF2BP2-mediated B3GNT6 expression, we sought to determine whether and how IGF2BP2 interacts with B3GNT6 to affect its expression. We knockdown IGF2BP2 and RIP showed lower B3GNT6 was enriched (Figure 6A). Dual luciferase reporter assay was conducted and the results revealed that B3GNT6 was possibly recognized and regulated by m6A modification (Figure 6B,C).

Considering that IGF2BPs could regulate RNA stability, we sought to evaluate B3GNT6 mRNA stability. PANC1 cells were transfected with IGF2BP2-si or the si-NC, then treated with actinomycin D (Fdbio science, China) at 10 ng/ml at different time point (0, 4, 8 and 12 h). As shown, B3GNT6 mRNA decay rate in the IGF2BP2-siRNA group was faster than that in the si-NC group (Figure 6D), IGF2BP2-KO and 3-DAA could decreased B3GNT6 protein expression (Figure 6E), suggesting IGF2BP2 can increase B3GNT6 mRNA stability and contribute to the high B3GNT6 protein level in an m6A manner. IF staining revealed that IGF2BP2 and B3GNT6 had similar expression trends in PC tissue (Figure 6F), and knockdown of IGF2BP2 led to an obvious decrease in the B3GNT6 protein level (Figure 6G). Together, these results suggest that B3GNT6 could undergo m6A methylation, and IGF2BP2 functions as a reader of methylated B3GNT6 to increase its stability.

3.9 IGF2BP2 promotes tumorigenesis in vivo

Subcutaneous tumorigenesis experiments in mice showed that knockdown of IGF2BP2 contributed to a decreased tumor size (Figure 6H,I) and tumor weight (Figure 6J), showing its significant ability in tumor formation. We evaluated the B3GNT6 expression level by IHC and IF, finding that knockdown of IGF2BP2 contributed to a lower B3GNT6 expression level (Figure 6K,L). Together, these data indicate that IGF2BP2 regulates B3GNT6 in an m6A manner and serves as an oncogene in PC (Figure 7).