Introduction
Pancreatic ductal adenocarcinoma (PDAC) is one of the most prevalent gastrointestinal malignancies in the world and a lethal disease with extremely devastating cancer [
1]. It is reported that an overwhelming number of patients with extremely aggressive PDAC in terms of 5-year survival rate only account for 6% in the United State [
2]. However, the traditional treatment fails to effectively prevent the worsening tumor and reduce the mortality rate of patients with PDAC [
3]. Therefore, it is necessary to develop a novel strategy to impede the rapid progression of PDAC.
Long non-coding RNAs (lncRNAs) do not have the ability of coding protein, which account for roughly four fifths in the entire transcriptome [
4]. It has been demonstrated that multiple biological processes are regulated by numerous lncRNAs involved in tumor cell proliferation, apoptosis, invasion [
5]. Wang et al. identified that significantly up-regulated LINC00240 were positively correlated with the poor prognosis, and silencing of the LINC00240 exhibited the decreased ability of proliferation and migration in gastric cancer [
6]. Another study demonstrated that the upregulated lncRNA JPX promoted lung cancer proliferation and metastasis by the JPX/miR-33a-5p/Twist1 axis activating Wnt/β-catenin signaling both in vitro and in vivo [
7]. Zhao et al. reported that the lncRNA CERS6-AS1 facilitates proliferation, migration and invasion in colorectal cancer cells by affecting miR-15b-5p to upregulate SPTBN [
8].
Epigenetic modification, gene sequence is scarcely influenced, is characterized by the changed gene expression derived from various modifications, including histone modification, DNA methylation, the modification of non-coding RNA [
9‐
11]. Particularly, N6-methyladenosine (m6A) commonly participates in the variety of biological processes. The modification of m6A functionally regulates the carcinogenesis by the three main proteins, such as binding protein (reader), demethylase (eraser), and methyltransferase (writer) [
12]. The increasing research have indicated that abnormal gene expression is caused by the modification of m6A in an incorrect manner [
13]. For instance, Lin et al. demonstrated that YTHDF1, the m6A reader, facilitating the immune evasion in skin cutaneous melanoma (SKCM) [
14]. Moreover, Ban et al. identified that the stability of LNCAROD was augmented in a m6A modification manner and the ability of cell proliferation and mobility was significantly enhanced with the increased LNCAROD expression in head and neck squamous cell carcinoma (HNSC) [
15]. Feng et al. focused on constructing an m6A-immune-related lncRNA risk model for predicting prognosis, immune landscape and chemotherapeutic response in bladder cancer [
16]. However, the relationship between m6A and lncRNAs has not been deeply explored in PDAC. Therefore, it is of vital importance to develop a new treatment strategy involving m6A modification associated with lncRNAs for improving the survival time in PDAC.
In the present study, we constructed a risk prognosis model for improving the clinical outcome of patients with PDAC, and validated the rationality of the risk prognosis model and searched for appropriate immunotherapy drugs that can effectively treat PDAC. Our research aims to explore potential molecular mechanisms and treatment strategies for PDAC.
Materials and methods
The data sources of patients
21 related genes were reported from previous literature [
17], including writers (METTL3, METTL14, WTAP, METTL16), readers (YTHDC1, YTHDF2, YTHDC2, RBMX) and erasers (ALKBH5, FTO). The m6A-related lncRNA prognostic signature genes were designed in three steps. (1) Univariate Cox regression analysis identified those m6A-related lncRNAs significantly influencing the overall survival (OS) of PADC patients. (2) To reduce the potential for overfitting, least absolute shrinkage and selection operator (LASSO) Cox regression was applied—where the penalty parameter was calculated by performing tenfold cross-validation on the training set at the minimum partial likelihood deviance. Lastly, a multivariate Cox regression analysis was conducted to finalize the selection of ideal m6A-related lncRNAs for the prognostic signature. (3) lncRNA expression’s multivariate regression coefficients were utilized to create a prognostic signature for patients. The risk score was computed using the formula: Risk score = β1 * Exp1 + β2 * Exp2 + βn * Expn [
18]. Here, ‘β’ signifies the coefficient while ‘Exp’ symbolizes the expression value of the respective m6A-related lncRNA. Patients were categorized into high- and low-risk groups based on the median risk scores. Differences in survival between these groups were illustrated using Kaplan-Meier analysis. Predictive abilities of the signature were assessed using the receiver operating characteristic (ROC) curves via the “SurvivalROC” R package. The optimal cut off point for the risk score, distinguishing patients from high- to low-risk groups, was determined using the previously described formula in the training set. The survival differences between the two groups were depicted using Kaplan-Meier analysis, and the prognostic signature’s predictive efficacy was evaluated by the time-dependent ROC curve.
Clinical correlation analysis
Univariate and multifactorial COX regression analyses were performed to exclude the influence of other clinical factors (gender, age, TNM stage and so on) and determine independent elements about prognosis [
19].
Principal component analysis (PCA) and Kaplan-Meier survival analysis
We used Principal Component Analysis (PCA) for effective dimensionality reduction, model identification, and grouping visualization of high-dimensional data, including 21 m6A genes, 4 m6A-related long non-coding RNAs (lncRNAs), and a risk model based on the expression patterns of these 4 m6A-related lncRNAs. We employed Kaplan-Meier survival analysis to assess the differences in overall survival (OS) between high-risk and low-risk groups. The R packages were used for this process including survMiner and survival [
20].
Assessment of immune function scores and functional analysis
In our study, the single-sample Gene Set Enrichment Analysis (ssGSEA) algorithm, implemented through the R package ‘GSVA’, was utilized to evaluate differences in immune function scores between the high and low risk groups. Additionally, the Tumor Immune Dysfunction and Exclusion (TIDE) algorithm was employed to predict the efficacy of immunotherapy in these groups [
21]. For differential analysis, we analyzed gene expression data from the high and low risk groups using the ‘limma’ package, adhering to specific criteria (log fold change (log FC) > 1 and a false discovery rate (fdr) < 0.05). Based on the findings of this differential analysis, Gene Ontology (GO) enrichment analysis was conducted using the ‘clusterProfiler’ package to identify and characterize significant GO terms associated with the differentially expressed genes, thereby providing insights into the biological processes and pathways that may underlie the risk stratification observed in our study [
19].
Comparison of the tumor mutation burden (TMB)
The “maftools” R package’s waterfall function was employed to illustrate the mutation panorama in both the high-risk and low-risk groups [
22]. This functionality allowed us to quantify the somatic mutation count and TMB, expressed as mutations per million bases, for each individual patient. To discern any statistical differences in the somatic mutation count and TMB levels between the high- and low-risk groups, Wilcoxon test was conducted. The same “maftools” R package was utilized to determine TMB within the high and low-risk groups. Subsequent to this, we employed the Kaplan-Meier methodology to compare survival rates. This allowed us to assess the differences between the high and low mutation groups, and the risk categories. This varied approach offered a comprehensive view of the mutation landscape in correlation with patient risk level and survival rates.
Drug sensitivity
The R package “pRRophetic” was utilized to forecast the IC50 of drugs deriving from the GDSC (
https://www.cancerrxgene.org/) website site summary in patients with PDAC.
Cell culture
Human pancreatic adenocarcinoma cells (AsPC-1, PANC-1, MIA PaCa-2, Patu8988, CFPAC-1, SW1990, BxPC-3), and human normal pancreatic duct epithelial cell (HPDE) were all preserved in Shanghai Cancer Institute, Shanghai Jiao Tong University. AsPC-1, BxPC-3, CFPAC-1 were kept in RPMI 1640 with 10% fetal bovine serum (FBS), and or PANC-1, Patu8988, MIA PaCa-2, SW1990, HPDE were cultured by DMEM with 10% FBS. The ratio of cell passage is subtlety different according to the cell growth rate, typically ranging from 1:2 to 1:3. Cell passaging was performed when the cell density reached 80-90% confluence in the bottom of culture dish. The old medium in the dish was discarded, and the cells were washed by PBS 1–2 times. Subsequently, the trypsin solution containing EDTA was added into the incubator (5% CO2 at 37 °C) with 2–5 min. To prevent further digestion, DMEM or RPMI 1640 supplemented with 10% FBS was added to the dish when the 70%~80% cells appeared round and the enlarged cell space were observed in the inverted microscope. After that, the cell suspension was transferred to a 15mL centrifuge tube and centrifuged at 800 rpm/min with 3–5 min. Next, supernatant was discarded and the new medium was used for resuspended cells. Finally, cells were inoculated into a new dish appeared round and were suspended in the indicated cell culture medium. The cells gradually returned to the original shape and sticked to the bottom of plate after 4 h. All cells were cultured under the same condition (5% CO2 at 37 °C).
RNA extraction and qPCR
Appropriate Trizol (ShareBio, Shanghai, China) was added into the cells in the dish washed by PBS with 1–2 times (10 cm dish with 2mL, 6 cm dish with 1mL). Then, 1 ml cell suspension was transferred to the 1.5 mL EP tubes after placing at ice for 5 min. 200µL chloroform was added into cell suspension, mixed well for 15s and rested on ice for 3-10 min. After that, centrifugation was performed at 4℃, 12,000 rpm/min for 15 min. Next, the upper water phase was taken into a new EP tube (200–500µL), and then added by equal volume isopropyl alcohol and rested on the ice for 10 min. The cell suspension was centrifuged at 4℃, 12,000 rpm/min for 10 min, and cells were kept and washed with 500µL anhydrous ethanol at twice. The cells were dissolved in DEPC water after drying 15 min. The concentration of RNA was measured and adjusted to 500ng/µL.
The first-strand cDNA was reverse-transcribed by All-in-One First-Strand Synthesis Master Mix reagent Kit (ShareBio, Shanghai, China). The reverse transcription system was defined as follows: Template RNA: 50 ng ~ 1 µg (1µL), All-in-One First-Strand Synthesis MasterMix: 4 µL, dsDNase:1 µL, 10x dsDNase Buffer: 2 µL, Nuclease-Free Water: 12 µL. Secondly, Mix gently and spin briefly. Thirdly, incubate under 37℃ for 2 min to remove the gDNA contamination. Fourthly, Incubate under 55℃ for 15 min. Then, Terminate the reaction by incubate under 85℃ for 5 min. Finally, the cDNA was put on ice immediately and diluted by DEPC (1:20).
The 10µL system of qPCR per well was defined as follows: 0.5µL Forward primer, 0.5µL Reverse primer, 5µL SYBR green qPCR Premix (ShareBio, Shanghai, China) and 4µL cDNA. qPCR was performed with on a 7500 Real-time PCR system (Applied Biosystems, Inc. USA). The expression levels of target genes were calculated by compared to the expression of the reference gene 18s, and quantification was performed using the 2
-ΔΔctmethod. The related experiments were performed in triplicate. The primers used in this study were shown in Supplementary Table
1.
IC50
To identify the sensitivities of drugs for different cell lines, the concentration gradient (0, 1, 10, 50, 100, 200, 500, 1000µM) was defined in the Phenformin (Selleck.cn, Shanghai, China) and Pyrimethamine (Selleck.cn, Shanghai, China). The cells were suspended in 100µL DMEM with 10% FBS. 3 × 103 HPDE, MIA PaCa-2 and Patu8988 were seeded into 96-well plates in the different concentration gradient and drugs treatment groups. Then, the 96-well plates were placed into the incubator (5% CO2 at 37 °C) for 24 h. The cells were treated by the Phenformin or Pyrimethamine with different concentration gradient (0, 1, 10, 50, 100, 200, 500, 1000µM) with 36 h. After that, Cell viability was measured by 10% CCK8 reagent (Share-Bio, shanghai, China, SB-CCK8L) at the incubator (5% CO2 at 37 °C) with 1 h. The absorbance of each 96-well plates was measured at 450 and 600 nm by microplate reader, and the half-maximal inhibitory concentration (IC50) was calculated by non-linear regression using GraphPad Prism 9.0.
Cell-counting kit 8
In order to evaluate cell cytotoxicity, 3 × 103 MIA PaCa-2 or Patu8988 per well in 96-well plate were plated in different groups with/without Phenformin or Pyrimethamine treatment. The cells were cultured in the incubator (5% CO2 at 37 °C) with 24 h. Then, the Phenformin (0 µM, 100 µM, 200 µM, 400 µM) or Pyrimethamine (0 µM, 50 µM, 100 µM, 200 µM) of different concentration gradient was added into MIA PaCa-2 and Patu8988 for 36 h, respectively. 100 µL of CCK-8 mixture (CCK-8 reagent: DMEM = 1:9) were added into these cells at the incubator (5% CO2 at 37℃) for 2 h. The absorbance of each well was observed at 450 and 600 nm. Finally, the optical density at 450 nm in different groups was calculated using GraphPad Prism 9.0.
Statistical analysis
Statistical analysis was performed using R 4.1.3 software and GraphPad Prism 9.0. The wilcoxon rank-sum test and Mann-Whitney U test were used to compare the continuous variables and Spearman analysis to calculate correlation coefficients. Kaplan-Meier method was used to draw survival curve, and log-rank test was used to compare survival differences. The IC50 was calculated by Nonlin fit. P < 0.05 was recognized as statistically significant.
Discussion
In order to explore the role of lncRNAs and m6A in the development of PDAC, m6A-related lncRNAs were screened by matching m6A and lncRNAs using co-expression analysis. The patients with PDAC were classified into the high- and low-risk groups by univariate Cox analysis and Lasso regression model. The survival of patients with PDAC was excellently predicted by ROC curves of the clinical parameters and risk scores in the risk prognosis model.
As a kind of commonly chemical modification in eukaryotic mRNAs and lncRNAs, abnormal expression of m6A influence cell self-renewal, differentiation, apoptosis, invasion [
26,
27]. Zhang H et al. found that the highly expressed SNHG17 modified by METTL3 aggravated the malignant phenotypes in Lung adenocarcinoma (LUAD) gefitinib resistance cells [
28]. Li B et al. demonstrated that lncRNA WEE2-AS post-transcriptionally stabilized by IGF2BP3 promotes Glioblastoma (GBM) progression [
29]. However, little research focused on pathogenic mechanism of lncRNAs and m6A in PDAC were identified. Therefore, it is essential to explore the prognostic signature about m6A-related lncRNAs in the PDAC.
In our research, 4 m6A-related lncRNAs (including EMSLR, AL358944.1, ZNF236-DT, AC087501.4) were selected to construct a prognosis model. It is reported that the highly expressed EMSLR facilitates proliferation of lung cancer by interacting with target gene LncPRESS [
30]. AL358944.1 is a prognostic indicator in the PDAC in the cuproptosis-related lncRNAs signature [
31]. Moreover, the numerous evidences showed that tumor cells not only accelerate growth but also inhibit TMB by interacting the infiltered immune cell around tumor cells [
32]. GO analysis revealed that immune function between the high- and the low-risk groups were widely disparate. For example, Type-I-IFN Response, MHC-I, T cell receptor complex (TCR) and so on. IFNs, a kind of cytokines, affect the growth, migration and presentation of neonatal antigens. cytotoxic lymphocyte can be stimulated by IFNs, thus, the disturbed IFN responses leading to the loss of tumor immune surveillance [
33]. Cytotoxic T lymphocytes can recognize and bind the antigenic peptide-MHC-I complex through TCR receptors to initiate anti-tumor immune function. However, the downregulation of MHC-I leads to the inability of CD8 + T cells to recognize tumor antigen peptides [
34], which is closely related to the immunosuppression and poor prognosis of many malignant tumors. Yamamoto K et al. demonstrated that MHC-I is degraded by binding with NBR1 in the autophagy process, promoting immune escape from pancreatic cancer [
35].
TIDE (Tumor Immune Dysfunction and Exclusion) is a computational method used to assess the efficacy of immune checkpoint blockade therapy. In general, a lower TIDE score indicates a higher possibility for the response to immunotherapy. In our study, lower TIDE scores were obtained in patients from the high-risk groups compared to those from the low-risk groups, suggesting that those patients with lower TIDE scores may have reduced immune dysfunction and exclusion, a better response to immunotherapy. In the future research, a novel treatment strategy may be found by the research based on TIDE scores, and more options are offered for personalized treatment.
Besides, 12 immunotherapy drugs about PDAC from GDSC websites were filtered, which paves the way for patients with PDAC in medical treatments. Pyrimethamine (2,4-diamino-5-p-chlorophenyl-6-ethylpyrimidine), a dihydrofolate reductase (DHFR) inhibitor, is used for treating malaria by inhibiting the proliferation of plasmodium and toxoplasma and also applied to anti-tumor field [
36]. Furthermore, Phenformin, a common biguanide drug, can effectively treat diabetes. Currently, it has been widely used to resist the development of cancers [
37]. The proliferation of MIA PaCa-2 treated with Phenformin and Patu8988 treated with Pyrimethamine was inhibited, indicating Phenformin and Pyrimethamine may be potential highly sensitive immunotherapy drugs. In this way, patients with PDAC in the high- and low-risk groups are more likely to be effectively treated when they receive suitable immunotherapy drugs. The established m6A-related lncRNAs model provides a new strategy for clinical outcome in PDAC.
Nevertheless, there are some limitations to our study that need to be addressed. Firstly, although we validated the risk model using separate training and test cohorts, further validation with larger patient cohorts is needed to confirm its reliability and generalizability. Secondly, the underlying molecular mechanisms linking the identified m6A-related lncRNAs to PDAC pathogenesis and progression remain largely unknown. Future studies should focus on elucidating these mechanisms to gain a deeper understanding of the role of m6A-related lncRNAs in PDAC. Lastly, while our study identified potential immunotherapy drugs based on the risk model, further preclinical and clinical investigations are required to evaluate their efficacy and safety in PDAC patients.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.