Background
Pancreatic cancer is an aggressive cancer of the digestive system with high morbidity and mortality. It is the seventh leading cause of cancer death in both males and females worldwide because of its poor prognosis. The high death rate of pancreatic cancer has been a continuous challenge to the medical field [
1]. Data from the National Centre for Health Statistics (NCSH) show a reduction in the occurrence of breast cancer, colorectal cancer, and prostate cancer in the past decade, but the incidence of pancreatic cancer continues to increase, and it has become the fourth leading cause of cancer death [
2]. Studies have predicted that it could be the second leading cause of cancer death by 2030. The estimated incidence of pancreatic cancer in the United States in 2020 is around 57,600 with a projected 47,050 deaths, indicating a mortality above 80% [
1]. Furthermore, the 5-year survival rate of pancreatic cancer is 9% and is the lowest across all types of cancer. Pancreatic adenocarcinoma (PAAD) is the most common type of pancreatic cancer, which encompasses 85% of pancreatic cancer [
3]. While the precise cause and underlying mechanisms of PAAD are still unclear, genetics is an essential factor related to tumour development and mutations in DNA or RNA may also drive the initiation of PAAD [
4].
RNA modification is a crucial part of epigenetics, which together with gene and protein modification, plays an important role in regulating cellular processes. According to the data from the MODOMICS update in 2017, about 163 different modifications have been identified in RNA molecules so far, including N1-methyladenosine, N7-methyladenosine, 5-methylcytosine, pseudouridine, N6,2′-O-dimethyladenosine (m6A) and 2′-O-methylation. In eukaryotes, m6A is the most prevalent and abundant form of mRNA modification, which was identified in the 1970s [
5,
6]. m6A is a dynamic and reversible process of mRNA modification and has been widely studied, although the exact mechanism of m6A modification remains unelucidated due to the technical limitations. The formation of m6A is composed of three individual proteins, and it is possible to detect these m6A proteins as a reflection of m6A modification. These proteins are classified into three categories: writers, erasers, and readers, which function as modulators in the process of m6A modification. Writer proteins mainly include METTL3, METTL14, WTAP, CBLL1 and METTL16 [
7‐
10]. RBM15, ZC3H13, and KIAA1429 have been newly identified and play a role in mediating the formation of the writer protein complex [
11‐
14]. These writer proteins transfer the methyl group from S-adenosyl methionine to the RNA nucleotides. Erasers, on the other hand, are demethylases and remove the methyl group from the RNA molecule. Members of this group include FTO and ALKBH5 [
15,
16]. Finally, reader proteins members that recognise m6A modification, including the family of YT521-B homology (YTH) domain-containing proteins (YTHDF1, YTHDF2, YTHDF3, YTHDC1, and YTHDC2) and the IGF-2 mRNA-binding proteins (IGF2BP1, IGF2BP2, and IGF2BP3) [
17,
18], heterogeneous nuclear ribonucleoproteins hnRNPA2B1 and hnRNPC also have been identified as m6A RNA readers [
19].
m6A plays a vital role in the splicing, translation, and stability of RNA and regulates chromatin state and transcription [
20]. This has significant consequences in various human diseases where m6A alterations may enhance heart failure [
21] influence brain development and function [
22], immune response to viral infection [
23], and bone metabolism [
24]. An increasing amount of evidence indicates that m6A RNA methylation modulators are associated with the development and progression of several malignant tumours, and play a dual role in cancer development. On the one hand, m6A regulates the expression of oncogenes or tumor suppressor genes, thereby affecting cancer progression. On the other hand, it can regulate the expression and activity of the m6A enzyme, thereby affecting the role of m6A in cancer. In bladder cancer, m6A modulator METTL3 promotes oncogenes’ expression [
25]. Furthermore, YTHDF1 has been linked with EIF3C, leading to the occurrence and metastasis of ovarian cancer [
26]. Studies have also found that m6A RNA methylation modulators cause genetic alterations such as mutations and copy number variations across diverse cancer types [
27]. Moreover, m6A RNA methylation modulators have been shown to correlate with clinical parameters and have been utilized as tumour biomarkers to determine the diagnosis and prognosis of several cancer types and monitor tumor development [
27]. m6A and its modulators are widely studied in hepatocellular cancer [
28], colorectal cancer, and genital and nervous system tumours [
26,
29‐
31]. Recently, He et al. demonstrated that ALKBH5 inhibits pancreatic cancer motility by demethylating long non-coding RNA KCNK15-AS1 [
32]. Other studies indicate that ALKBH5 decreases WIF-1 RNA methylation and inhibits pancreatic cancer tumorigenesis through the Wnt signaling pathway [
33]. However, therapies targeting signaling pathways in pancreatic cancer lack efficacy due to the heterogeneity of tumor microenvironment (TME) in patients, TME prevents antitumor drug resistance, immune escape and tumor cell’s infiltration during cancer progression. Increasing evidence demonstrated that m6A modification is associated with TME of cancer, and the polarization of tumor-associated macrophages-M2 enabled the oxaliplatin resistance via the elevation of METTL3-mediated m6A modification [
34]. Dali et al. found a risen level of CD8+ cytotoxic T cells and natural killer (NK) cells while a reduced infiltration of myeloid-derived suppressor cells (MDSC) from Ythdf1−/− tumor mice compared to WT mice, besides, the therapeutic efficacy of PD-L1 checkpoint blockade is enhanced in Ythdf1−/− mice [
35].
Since the tumor is a result of the coordinated action of multiple factors and genes, we integrated the gene expression profiles information from The Cancer Genome Atlas (TCGA) and GTEx database, extracted expression of 28 widely reported m6A RNA methylation modulators from 140 patients with PAAD, assessed the influence of m6A RNA methylation modulators in the clinical and pathological characteristics of PAAD comprehensively, explored its relationship with the TME cell-infiltrating characteristics and predicted the efficacy of immune checkpoint inhibitors.
Materials and methods
Data collection and pre-processing
RNA-seq transcriptome data (FPKM value) and corresponding clinical information of pancreatic cancer patients were obtained from TCGA database (
https://cancergenome.nih.gov/). The gene expression profile was measured experimentally using the Illumina HiSeq2000 RNA Sequencing platform by the University of North Carolina TCGA genome characterization center. We obtained expression profiles of 178 patients with PAAD; however, 34 cases should not have been included in the PAAD study according to the quality annotation from TCGA, and one patient was a repeated measure, furthermore, PAAD patients with incomplete data on age, gender, survival time (futime), survival status (fustat), pathological grade of the tumor, and pathologic stage were excluded. A total of 140 patients with PAAD were eventually enrolled in our study from TCGA database in the end. The excluded and enrolled patients were listed in Table
S1. RNA-seq transcriptome data of 167 normal human pancreatic tissues was download from the Genotype-Tissue Expression (GTEx) project (
https://xenabrowser.net/). As to GTEx, FPKM values were extracted, and log2(x + 0.001) transformed into FPKM values, then the FPKM values from TCGA and GTEx database were transformed into transcripts per kilobase million (TPM) values. Data of gene expression quantification was integrated from the TCGA database and GTEx project; 140 patients with PAAD and 167 healthy populations were incorporated into our study in the end.
We obtained the list of immune-related genes from import database (
https://immport.niaid.nih.gov/), a long-term, sustainable data warehouse developed to promote re-use of immunological data generated by NIAID DAIT and DMID funded investigators. Besides, the copy number of the genes in 140 PAAD patients was download in the UCSC database (
https://xenabrowser.net/). Expression of prognosis m6A modulators mRNA levels was analyzed to compare normal and tumor tissues in various cancer types using the Oncomine database (
https://www.oncomine.org). The GSE62452 dataset from the GEO database (
https://www.ncbi.nlm.nih.gov/geo/) was used for validation.
We selected m6A RNA methylation modulators which were widely reported in a variety of cancers. Twenty-eight m6A RNA methylation modulators including 10 writers, 16 readers, and 2 erasers were included in this study (Table
1). mRNA expression data of 28 m6A RNA methylation modulators from TCGA and GTEx projects were extracted for further bioinformatics analysis.
Table 1
the lists of the 28 m6A RNA methylation regulative factors
METTL3 | Methyltransferase like 3 | “writers” |
METTL14 | Methyltransferase like 14 | “writers” |
METTL16 | Methyltransferase like 16 | “writers” |
WTAP | WT1 associated protein | “writers” |
RBM15 | RNA binding motif protein 15 | “writers” |
RBM15B | RNA binding motif protein 15B | “writers” |
ZC3H13 | Zinc finger CCCH-type containing 13 | “writers” |
VIRMA | vir like m6A methyltransferase associated | “writers” |
CBLL1 | Cbl proto-oncogene like 1 | “writers” |
ZCCHC4 | zinc finger CCHC-type containing 4 | “writers” |
LRPPRC | leucine rich pentatricopeptide repeat containing | “readers” |
ELAVL1 | ELAV like RNA binding protein 1 | “readers” |
YTHDC1 | YTH domain containing 1 | “readers” |
YTHDC2 | YTH domain containing 2 | “readers” |
YTHDF1 | YTH N6-methyladenosine RNA binding protein 1 | “readers” |
YTHDF2 | YTH N6-methyladenosine RNA binding protein 2 | “readers” |
YTHDF3 | YTH N6-methyladenosine RNA binding protein 3 | “readers” |
HNRNPC | Heterogeneous nuclear ribonucleoprotein C | “readers” |
HNRNPA2B1 | Heterogeneous nuclear ribonucleoprotein A2B1 | “readers” |
EIF3A | Eukaryotic translation initiation factor 3 subunit A | “readers” |
EIF3H | Eukaryotic translation initiation factor 3 subunit H | “readers” |
IGF2BP1 | IGF2 mRNA-binding protein 1 | “readers” |
IGF2BP2 | IGF2 mRNA-binding protein 2 | “readers” |
IGF2BP3 | IGF2 mRNA-binding protein 3 | “readers” |
CBLL1 | Casitas B-lineage lymphoma-transforming sequence-like protein 1 | “readers” |
PRRC2A | The proline-rich coiled-coil 2A | “readers” |
FTO | Fat mass and obesity-associated protein | “erasers” |
ALKBH5 | α-ketoglutarate-dependent dioxygenase AlkB homolog 5 | “erasers” |
Constructing of m6A RNA methylation modulators signature in pancreatic cancer
Twenty-eight m6A RNA methylation modulators were incorporated into univariate Cox proportional hazard regression, then modulators with a p-value < 0.1 were involved in the subsequent LASSO regression for further screening, and the selected modulators were applied to conducted prognosis model. Here m6Ascore = Σni = 1Coef × xi where Coef represents the coefficient and xi is the expression value of each selected modulator, and nomograph was portrayed to predict the risk of death and the survival rate.
Pathways and function enrichment analysis
PAAD samples were categorised into high-risk and low-risk groups according to the median of m6Ascore, “limma” package of R was used to analyse the differentially expressed genes (DEGs) with stated threshold values where false discovery rate (FDR) < 0.05 and |Log2 fold change (FC)| > =1.0, all the DEGs was performed in volcan plot by “ggplot2” package in R, intersection of DEGs and immune-related genes from import database was displayed in venn plot by “VennDiagram” package of R software and exhibited in heatmap partly. Then the function of differentially expressed immune-related gens between two group was performed by Gene ontology (GO) enrichment and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses, visualised by “ggplot2” package in R. KEGG pathways of m6A RNA methylation modulator genes were identified by GESA-3.0.jar software. Gene sets (c2.cp.kegg. v7.4 symbols.gmt) with nominal p-value < 0.05 and false discovery rate (FDR) q-value < 0.25 were considered significantly enriched.
Estimation of tumor-infiltrating immune cell
Estimation proportions of the 22 tumor-infiltrating immune cells (B cells naive, B cells memory, Plasma cells, T cells CD8, T cells CD4 naive, T cells CD4 memory resting, T cells CD4 memory activated, T cells follicular helper, T cells regulatory (Tregs), T cells gamma delta, NK cells resting, NK cells activated, Monocytes, Macrophages M0, Macrophages M1, Macrophages M2, Dendritic cells resting, Dendritic cells activated, Mast cells resting, Mast cells activated, Eosinophils, and Neutrophils) from each sample were calculated by using the “CIBERSORT” (R package) according to the gene expression profiles. The tumor cellularity was inferred through the ESTIMATE algorithm using the “estimate” package in R with default parameters. We acquired ImmuneScore, StromalScore, and ESTIMATEScore of each sample in pancreatic cancer to predict the level of infiltrating stromal and immune cells based on specific gene expression signatures of stromal and immune cells.
Immune response analysis
The immunophenoscore (IPS) of 140 PAAD patients was obtained from The Cancer Immunome Database (TCIA) (
https://www.tcia.at/home), which provides results of comprehensive immunogenomic analyses of next-generation sequencing data (NGS) data for 20 solid cancers from The Cancer Genome Atlas (TCGA) and other data sources. Furthermore, three immunotherapeutic cohorts, including IMvigor210 cohort, TCGA-SKCM cohort, and GSE78220 cohort downloaded from GEO were used to predict the m6Ascore for curative effect of immunotherapy, expression data and detailed clinical information of IMvigor210 cohort were obtained from
http://research-pub.gene.com/imvigor210corebiologies, and normalized by the DEseq2R package in R and then the count value was transformed into the TPM value. As to GSE78220 and TCGA-SKCM cohort, the FPKM data of gene expression profiles were also converted to the more comparable TPM value.
Clinical tissue specimens
A total of six paired pancreatic cancer and adjacent non-tumorous tissue used for immunohistochemistry (IHC) staining were collected from the First Affiliated Hospital of Anhui Medical University. All patients have already provided written informed consent. This work was approved by the Academic Committee of The First Affiliated Hospital of Anhui Medical University and was conducted following the Declaration of Helsinki.
Immunohistochemistry staining analysis
Collected tissue specimens were formalin-fixed and embedded with paraffin. Tissue sections (4 μm thickness) were used in IHC staining analyses, and the slices were treated with methanol containing 3% hydrogen peroxide to inactivate the endogenous peroxidase and treated with citric acid buffer (pH = 6.0) to obtain optimal antigen recovery. The slices were incubated in 1% bovine serum albumin in phosphate buffer for 30 min to block non-specific binding. In addition, the slices were stained with primary antibody overnight at 4 °C. Then, these sections were subjected to three 5-min mild washing in phosphate buffer saline, followed by staining with secondary antibody (HRP polymer) at 1:200 for 50 min. Diaminobenzidine was applied to the slices before being counterstained with hematoxylin. Finally, the samples were sealed, observed, and photographed by a light microscope.
Statistical analysis
R 4.0.3 (
https://www.r-project.org/) was used for our statistical analyses. Wilcoxon test was used to perform a difference comparison between two group, including differential expression of m6A RNA methylation modulators between cancerous and healthy tissues and the DEGs between high and low m6Ascore group. The relationship between m6Ascore and age, gender, pathological stage and histologic grade were analysed by Chi-square test. Association between m6Ascore and tumor-infiltrating immune cells was computed by Spearman’s correlation. Multiple regression analysis was used to analyse the association between m6A RNA methylation modulators and clinical features. Cox regression analysis was used to examine the prognosis of pancreatic cancer. Kaplan-Meier method was used to analyses the difference in overall survival (OS) between the high-risk and low-risk group.
P-values less than 0.05 on both sides were statistically significant. Graphics were visualized by R and Graphad prism 8.
Discussion
The high death rate of pancreatic cancer has been a continuous challenge in the field, it can progress rapidly with only mild symptoms, and there may be a multitude of causes, including a combination of lifestyle, environment, and genetic mutation. Surgery is the preferred treatment for resectable pancreatic cancer with a poor prognosis. Pancreatic cancer is generally insensitive to both chemotherapy and radiotherapy. Therefore, new treatment strategies focusing on the immunotherapy and targeted therapy are urgently needed for patients with pancreatic cancer. m6A is involved in the development of many cancers and some studies have shown its potential in the treatment of diverse cancers. In this study, we explored the function of multiple m6A methylation modulators in pancreatic cancer, identified the role of m6A methylation modulators signature in the TME cell infiltration and estimated the therapeutic effects of immunotherapy in pancreatic cancer.
We analysed 28 widely reported m6A RNA methylation modulators in pancreatic cancer according to the TCGA and GTEx database, a higher expression of these modulators except METTL3 was found in patients with PAAD compared to the normal tissues, suggesting an alterant level of m6A RNA methylation may play a carcinogenic role in pancreatic cancer. The prognosis model was conducted based on the expression of the selected methylation modulators, including KIAA1429, IGF2BP2, IGF2BP3, LRPPRC, METTL3, and EIF3H in our study. A similar study published in 2020 conducted a prognosis model based on the expression of KIAA1429, IGF2BP2, IGF2BP3, HNPNPC, METTL3, and YTHDF1. The discrepancy could be due to the difference of enrolled samples, but the similar finding was that METTL3 acts as a favourable factor while KIAA1429, IGF2BP2 and IGF2BP3 were risk factors in both pieces of research [
36]. An overexpression of METTL3 in the MIA-PaCa-2 and BxPC-3 pancreatic cancer cell lines was involved in pancreatic carcinogenesis and promoted proliferation and invasion of pancreatic cancer cell lines in Xia T’s study [
37]. Taketo K et al. pointed out that METTL3-depleted cells showed higher sensitivity to chemo- and radioresistance in pancreatic cancer cells [
38], which was adverse with our analysis. IGF2BP2 and IGF2BP3, members of the mammalian IGF2 mRNA-binding protein family, were found correlated with an overall poor prognosis and metastasis of various cancer [
39]. We found a higher pathologic stage and histologic grade associated with the overexpression of IGF2BP2 and IGF2BP3 in our study. The results from GSEA showed that overexpression of IGF2BP2 and IGF2BP3 was related to the progression of cancer, IGF2BP3 may involve rearrangement of peripheral actin, which led to the formation of additional membrane protrusions, promoting cell invasiveness and tumor metastasis of pancreatic cancer [
40], IGF2BP2 was confirmed to be an oncogenic factor; miR-141 was a downstream regulatory gene in the progression of pancreatic cancer in a previous study [
41]. EIF3H showed a function on cell cycle from GSEA in our study; however, the specific mechanism in PAAD need to be confirmed. Further, LRPPRC was a multifunctional protein involved in energy metabolism, and we found high LRPPRC enriched in pancreatic cancer, and it may play an essential role in pancreatic tumorigenesis by connecting an oncogenic gene expression with energy production. KIAA1429 overexpression is seen in many types of cancer that drive tumor growth and liver cancer cells’ metastasis by targeting GATA3 [
42].
Tumorigenesis is a process that relies on the coordination of multiple factors, results showed that risk score was an independent prognosis indicator, and the combined expression of the selected modulators had a good prediction of 1-, 2-, and 3-year OS of PAAD, patients with high-risk score related with an unfavourable OS. Previous reports described m6A regulator-mediated methylation modification patterns played a non-negligible role in tumor microenvironment infiltration in gastric cancer [
43]. Comparing the immune-related genes between high and low m6Ascore group, we found
IL36B and IL36RN exhibited a higher level in the high m6Ascore group where
IL36B was a gene involved TH17 pathway and played a role in polarizing T-helper responses.
IL36RN encoded an interleukin-36-receptor antagonist [
44] in renal cell carcinoma cells, and its overexpression inhibited proliferation, migration, invasion, and colony formation of tumor cells by suppressing β-catenin [
45].
CD19, a target of chimeric antigen receptor (CAR) T-cell immunotherapy, was downregulated in PAAD patients with low m6Ascore. The results of GO and KEGG pointed out that different immune-related genes participated in the activation and differentiation of T cell and other immune-related pathways, revealing a significant role of m6Ascore signature in tumor’s immune environment. Besides, patients with high m6Ascore had a lower immune score and estimated score than patients with low m6Ascore, while a recent study declared that gliomas patients with high m6Ascore had a more abundance of immune infiltration with higher immune score and stromal score, which was conversed in PAAD [
46]. As to the discrepancy of immune cell infiltration, patients with low m6Ascore had an elevation of Tregs and CD8
+T cells but a low level of resting memory CD4
+T cells, indicating an association between m6A modification and tumor immunity of PAAD. Moreover, a high level of CD8
+T cell presented a high survival rate, where exhaustion of CD8
+T cell may lead to an unfavourable OS in patients with high m6Ascore. Similar results in Han D’s study exhibited a deficiency of YTHDF1 that promoted the antigen-specific activation of CD8+ T cells and attributed to the enhanced capability of DCs to present tumor neoantigens [
35]. Increasing evidence had shown that m6A RNA methylation modification plays a role in tumor immunity. Previous research reported that conduction of m6A signature was beneficial to reflect tumor microenvironment features of lung adenocarcinoma in different cohorts [
47].
Immunotherapy had been shown been as an effective therapeutic strategy in diverse cancers, however, some patients faced an unsatisfactory effect due to individual variation. TME is composed of tumor vessels and different immune microenvironments, patients have different therapeutic responses due to the discrepancy of their structure. Checkpoint inhibitors have been the most mature and sufficient clinical research and the most widely used of tumor immunotherapy. Immune checkpoint therapy is a therapeutic method to inhibit tumor cells by regulating the activity of T cells through a series of pathways such as co-inhibition or co-stimulation. Various immune checkpoints have been used to treat PAAD, however, PD-1/PD-L1 and CTLA-4 blockade observed limited effect in PAAD treatment [
48,
49]. We further predicted the efficacy of PD-1/PD-L1, CTLA-4, LAG-3, TIM-3 and TIGIT, and found a higher expression of PD-1, CTLA-4, and LAG-3 in the low m6Ascore group. Nevertheless, there was no difference in IPS between the two groups, and the estimated results from three immunotherapy cohorts indicated no association between m6Ascore and the curative effect of immunotherapy. Immunotherapy in pancreatic cancer is conflicted, and it needs more clinical trials to confirm the curative effect. For example, a phase I trial included one pancreatic patient who failed to show any efficacy with a treatment of anti-PD-L1 (MPDL-3280A) therapy [
50]. In contrast, another clinical trial that enrolled 32 patients with pancreatic cancer showed the best response to MK-3475 and disease-free progression for 20 weeks after anti-PD-1 therapy [
51]. Therapy targets immune checkpoints in pancreatic cancer showed a poor effect, and more efforts are required to overcome resistance to PD-1/PD-L1 targeted immunotherapy.
In general, our research found a discrepancy of m6A RNA modification modulators in pancreatic cancer based on TCGA and GTEx databased, and found an association between the expression of m6A RNA modification modulators and clinical characteristic. We conducted an m6A risk model and observed that high m6Ascore presented an unfavourable prognosis in PAAD patients. Besides, patients with high m6Ascore had a lower level of infiltration of CD8+ T cells, suggesting that m6A RNA modification may related to tumor microenvironment. However, the inadequate difference in IPS and m6Ascore showed no function in evaluating the therapeutic effects of checkpoint inhibitors treatment in PAAD. Furthermore, the exact mechanism of m6A modulators needs to be elucidated with future research.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.