Introduction
Cervical cancer is the fourth most common cancer in women [
1,
2]. More than 570,000 new cases of cervical cancer are diagnosed worldwide each year, with approximately 311,000 deaths [
3]. Presently, at least 170 RNA modifications have been discovered in cervical cancer, for instance, N
6-dimethyladenosine (m
6A), inosine, pseudouridine, 5-methylcytidine (m5C), 5-hydroxymethylcytidine (hm5C), and N1-methyladenosine (m1A). Furthermore, most RNA species contain at least one chemical modification and are widely associated with physiopathological processes [
4‐
7]. m
6A is one of the most abundant mRNA modifications [
8,
9]. m
6A modifies approximately 0.1–0.4% of all adenosines in RNA, accounting for ~ 50% of the total methylated ribonucleotides [
10]. m
6A is a dynamic and reversible epigenetic modification. The cellular m
6A status is regulated by groups of proteins called m
6A methyltransferases (“writers”), m
6A demethylases (“erasers”), and m
6A-binding proteins (“readers”), which add, remove, or recognize m
6A-modified sites, respectively, modulating the stability, splicing, intracellular distribution, and translational changes of mRNA while affecting certain biological processes [
11‐
14]. As a potential tumor biomarker, m
6A plays a key role in various biological processes and malignancies [
15]. Changes in m
6A-modifying enzyme levels affect the expression of downstream oncogenes or tumor suppressor genes by altering mRNA methylation [
16]. Furthermore, epigenetic markers potentially serve as diagnostic, prognostic, and predictive biomarkers and might be used as novel targets for cancer precision medicine [
17‐
20].
Here, we studied the expression and mutation of m6A-related genes in patients with cervical squamous cell carcinoma and endocervical adenocarcinoma (CESC), obtained the gene characteristics of patients with CESC based on an m6A-risk model, and constructed a prognostic m6A feature model based on m6A signature genes. The effects of the m6A risk score on biological function, immune characteristics, and genomic changes of patients with CESC were analyzed. The sensitivity of patients with CESC to different small molecule drugs was evaluated based on the m6A risk score, and a clinical prediction model was constructed. Subsequently, we studied the expression and function of ZC3H13 in cervical cancer tissues according to the screening results. We selected rapamycin to study the effects of the screened drug on m6A and cervical cancer phenotype. Through this research, we attempted to study the gene signatures, immune infiltration, and drug sensitivity based on a comprehensive analysis of m6A RNA methylation regulators in cervical cancer.
Materials and methods
Data
Data were downloaded from the TCGA GDC website (
https://portal.gdc.cancer.gov/), and the cervical squamous carcinoma and adenocarcinoma (cervical squamous cell carcinoma and endocervical adenocarcinoma, CESC) expression profile data (FPKM results) of gene expression sequencing in patients were obtained. The FPKM values were converted into the TPM values as follows:
$${\mathrm{TPM}}_{i}=\left(\frac{{\mathrm{FPKM}}_{i}}{{\sum }_{j}{\mathrm{FPKM}}_{\mathrm{j}}}\right)\bullet {10}^{6}$$
Clinical data of corresponding patients, including age, TNM stage, and survival prognosis, were downloaded. After excluding patients with CESC with missing clinical information, data regarding 279 tumor tissues and three paracancer tissues were obtained. The copy number variation (CNV) of somatic cells of patients with CESC was downloaded. RCircos package was used to map the gene CNV in 23 pairs of chromosomes [
21]. “Masked mutation” was selected as the somatic mutation data, and R’s MAfTools package was used to visualize somatic mutations. The tumor mutation burden (TMB) of each patient was obtained [
22]. The sequencing results of 19 cervical tissues were downloaded from the GTEx database and converted into TPM values.
In addition, gene expression data of the GSE52903 chip, clinicopathological features, and prognostic information of patients were downloaded from samples in the GEO database, including 55 tumor tissue and 17 normal cervical tissue samples [
23]. Among them, all chip data samples were from
Homo Sapiens, and the platform was mainly based on the GPL6244 [Hugene-1_0-ST] Affymetrix Human Gene 1.0 ST Array [Transcript (Gene) version]. Next, TCGA data, GTEx data, and GEO chip data were combined, and the limma and sva packages of R were used for homogenization and batch removal [
24,
25]. The total process is composed of two steps. The first is data consolidation; only common genes and corresponding expression values are retained in the three data sets. The second step involves removing the batch effects. The first step adopts the merge function in R to merge the three data sets. The second step uses the combat function in the SVA package to remove the batches. The demographic and clinical characteristics of the cervical cancer patients are shown in Additional file
1 materials.
Risk model construction based on m6A-related genes
To analyze the expression of m
6A-related genes in cervical cancer, we first analyzed the differential expression of m
6A-related genes in cervical cancer and normal tissues, the correlation of gene expression, and its impact on the prognosis of patients with CESC. Subsequently, expression data of patients with CESC were pooled using TCGA-CESC and GEO data. m
6A-related genes were included in the model, and the least absolute shrinkage and selection operator (LASSO) algorithm was used to analyze the dimension reduction and obtain characteristic genes associated with prognosis. A risk score formula was established using the penalty coefficient weighted normalized gene expression values obtained by LASSO Cox analysis. The patients were divided into high- and low-risk groups based on the median risk score.
$$\mathrm{risk}Score = \sum_{i}Coefficient \left({hub gene}_{i}\right)*mRNA Expression (hub {gene}_{i})$$
Gene set enrichment analysis (GSEA)
Gene Ontology (GO) analysis is a common method for large-scale functional enrichment studies that includes three categories: biological process (BP), molecular function (MF), and cellular component (CC). KEGG is a widely used database that stores information regarding genomes, biological pathways, diseases, and drugs. R's clusterProfiler package was used for GO annotation analysis and KEGG pathway enrichment analysis of signature genes. A value of FDR < 0.05 was considered statistically significant [
26].
To study the differences in BPs between different subgroups, we performed GSEA based on the gene expression profile dataset of patients with CESC. GSEA is a computational method to analyze whether a specific gene set has statistically significant differences between two biological states and is usually used to estimate changes in pathways and BP activity among samples [
27]. The “h.all.v7.2.symbols.gmt” gene set was downloaded from MSigDB for GSEA analysis [
27]. An adjusted P-value of less than 0.05 was considered statistically significant.
Assessment of patients’ biological characteristics between risk groups
We further analyzed the correlation between different subgroups and some biologically related processes using the GSVA method [
27]. Mariathasan et al. constructed a gene set for storing genes related to some BPs, including (1) immune checkpoint; (2) antigen processing; (3) characteristics of CD8
+ T cells; (4) epithelial-mesenchymal transformation (EMT) markers, including EMT1, EMT2, and EMT3; (5) angiogenesis; (7) characteristics of TGF-β response in pan-FTBRs; (8) WNT characteristics; (9) DNA damage repair; (10) mismatch repair; (11) nucleotide excision repair; (12) DNA replication; and (13) antigen handling and presentation [
28‐
30]. Gene sets corresponding to different biological characteristics were downloaded to calculate the corresponding enrichment scores of patients and to compare the differences between two groups.
To identify genes associated with the m
6A risk model, THE limma package of R was used to analyze DEGs between high and low subgroups in patients with CESC, and the DEGs with significant differences were defined as genes with absolute log (Fold change) > 0.5 and FDR < 0.05 [
24]. Hierarchical clustering was used to divide the tumors into different gene groups based on the Euclidean distance of differential gene expression and named them Geneclusters A, B, and C. Among them, R's ConsensusClusterPlus package was used to determine the number of clusters in the dataset and was repeated 1000 times to ensure the stability of classification [
31]. Meanwhile, based on the expression changes of specific genes, they were divided into signature-A and -B gene groups.
Calculation of dimensionality reduction and m6A score
First, unsupervised clustering was used to classify TCGA data of patients according to DEG values. m
6A signature-A and -B gene groups were reduced in dimension according to the Boruta algorithm. The principal component PC1 was extracted by the principal component analysis (PCA) algorithm as the A score. Finally, we applied a method similar to the gene expression grade index to define each patient’s immune checkpoint inhibitor (ICI) score:
$$m6A score = \sum_{i}PC1A- \sum_{i}PC1B$$
Identification and correlation analysis of tumor immune infiltrating cells
To further quantify the proportion of different immune cells in CESC samples, the CIBERSORT algorithm, and LM22 gene set were used to investigate the phenotypes of 22 human immune cells (including B cells, T cells, and natural killer cells) in the tumor microenvironment (TME) [
32]. Macrophages are highly sensitive and specific. CIBERSORT is a deconvolution algorithm that uses a set of reference gene expression values (with 547 characteristic genes) considered to be the minimum representative of each cell type. Based on these values, the proportion of cell types can be inferred in data from a large tumor sample with mixed cell types.
At the same time, R's ESTIMATE package was used to assess tumor immune activity [
33]. ESTIMATE an immune score for each tumor sample obtained by quantifying the immune activity (level of immune invasion) in the tumor sample based on its gene expression profile. Then, differences in the characteristics of immune infiltration in patients with CESC between the high- and low-risk groups were obtained.
Analysis of CNV
To analyze the changes in the copy number in different risk score groups of TCGA-patients with CESC, the TCGAbiolinks package of R was used to download the masked copy number segment data of patients. GISTIC 2.0 analysis of the downloaded CNV fragments was performed by GenePattern5. During analysis, except for a few parameters (e.g., the confidence is 0.99; Excluding X chromosomes prior to analysis), GISTIC 2.0 analysis was used with the default settings. Finally, R's MAfTools package was used to visually display the results of the GISTIC 2.0 analysis.
Sensitivity analysis of anticancer drugs
Genomics of Drug Sensitivity in Cancer (GDSC;
https://www.cancerrxgene.org/), which is used for molecular cancer therapy and mutation, was used to explore public databases [
34]. R's pRRophetic package was used to download cell line gene mutation data and IC
50 values of different anticancer drugs to analyze the correlation between patients with high- and low-risk scores and sensitivity to different anticancer drugs [
35].
Building a clinical prediction model based on the m6A risk model
To prove that the risk score combined with the clinicopathological features can help in the personalized assessment of patient prognosis, univariate and multivariate Cox analyses were conducted to analyze the ability of the risk score combined with clinicopathological features to predict the overall survival (OS). Then, a nomogram was constructed by incorporating the risk scoring model and clinicopathological features into the model. To quantify the differentiated performance, Harrell’s Consistency index (C-index) was measured. A calibration curve was generated to evaluate the performance of the rosette by comparing the predicted values of the rosette with the actual observed survival rates.
Human tissue samples
Ten cervical cancer tissues and ten normal cervical tissues were obtained from patients with cervical cancer who underwent ovariectomy prior to chemotherapy and radiotherapy. The resected cervical cancer tissues and normal tissues were immediately stored at − 80 °C for further study. All patients signed informed consent forms. This study was performed in accordance with the Helsinki Declaration and approved by the ethics committee of the Second Affiliated Hospital of Zhengzhou University.
Cell lines and culture conditions
The human cervical cancer cell lines HeLa and SiHa were purchased from Procell Life Science & Technology Co., Ltd. (Wuhan, P.R. China). To prepare the complete growth medium, the cell culture media were supplemented with fetal bovine serum (Gibco, Grand Island, USA) at a final concentration of 10%.
Total RNAs of human cells were extracted using the Trizol reagent (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s instructions and treated with RQ1 DNase (Promega, Madison, WI, USA) to remove DNA. The quality and quantity of purified RNA were determined by measuring the absorbance at 260 nm and 280 nm (A260 and A280, respectively) using a SmartSpec Plus Spectrophotometer (Bio-Rad Laboratories, Inc., Hercules, CA, USA). RNA integrity was further verified by electrophoresis using a 1.5% agarose gel. All RNA samples were stored at -80 °C for future analysis. Reverse transcription reactions were carried out using the ReverTra Ace qPCR RT Kit (TOYOBO Life Science, Shanghai, P.R. China), according to the manufacturer’s instructions.
Quantitative real-time PCR (qRT-PCR)
Expression levels of the ZC3H13 gene were detected by qRT-PCR. The human (species) ACTB gene was used as a control. Specific primers were designed based on cDNA sequences. Primer sequences of ZC3H13 were as follows: 5′- ACATTCATTAGGCTCTGGTGC -3′ forward), 5′—TTCTCCTCATCCTGTTGGTCC—3′ (reverse). qRT-PCR was performed on a Bio-Rad S1000 with Bestar SYBR GreenRT-PCR Master Mix (TOYOBO). PCR conditions consisted of denaturing at 95 °C for 1 min, and 40 cycles of denaturing at 95 °C for 15 s followed by annealing and extension at 60 °C for 30 s. The relative gene expression was calculated using the Livak and Schmittgen 2−ΔΔCt method (Livak and Schmittgen 2001), normalized with the reference gene actin. PCR amplifications were performed in triplicate for each sample.
Immunohistochemical analysis
The tissue slides (4-μm-thick sections) were initially treated for deparaffinization, rehydration, and antigen-retrieval using 3% H2O2. The sections were incubated with anti-ZC3H13 (Bioss, P.R. China) and then with horseradish peroxidase (HRP)-labeled IgG secondary antibodies (Beyotime Institute of Biotechnology, Shanghai, P.R. China). Fields from each slide were examined and photographed using a light microscope (Olympus, Japan).
Western blot analysis
Cells were lysed in a radioimmunoprecipitation assay buffer containing 1 nM phenylmethanesulfonyl fluoride, and protein concentrations were determined using a bicinchoninic acid (BCA) protein assay kit (Biosharp, Guangzhou, China) according to the manufacturer’s instructions. Protein samples (40 μg) of each group were boiled with 6 × sodium dodecyl sulfate loading buffer for 10 min before electrophoresis on 12% sodium dodecyl sulfate–polyacrylamide gel electrophoresis gels. The resolved proteins were electrotransferred onto polyvinylidene difluoride membranes (Millipore, Billerica, MA, USA) in a transferring buffer (25 mM Tris, 0.2 M glycine, and 25% methanol). After blocking with 5% skimmed milk, the membranes were incubated with anti-ZC3H13 (Bioss, P.R. China) and anti-β-actin antibodies (Bioss, P.R. China), followed by incubation with appropriate HRP-conjugated secondary antibodies (Abcam, England).
Cell transfection
ZC3H13-siRNA (5′-GAAGACAUCUGCAGUAUCU-3′, antisense 5′-AGAUACUGCAGAUGUCUUC-3′) was synthesized by GeneCreate Bioengineering Co., Ltd. Gene Create (Wuhan, P.R. China). According to the manufacturer’s instructions, Lipofectamine 2000 (Invitrogen, USA) was used for transfecting siRNAs into HeLa and SiHa cells.
CCK8 assay
Cells were digested with trypsin and inoculated into 96-well plates. The culture plate was placed in a 5% CO2 incubator for 0, 24, and 48 h at 37℃. Next, rapamycin (GlpBio, USA) and/or FBS (Gibco, USA) were added. Then, the CCK-8 (Solarbio, Beijing, China) solution was added, and the samples were incubated for 0.5 h. The absorbance of each well was measured at 450 nm.
Transwell chamber assay
The cell invasion ability was evaluated using Transwell chambers precoated with Matrigel. A total of 1 × 104 cells were inoculated into the top chamber and incubated at 37 °C with 5% CO2 for 48 h. After cells on top of the filter were removed, cells on the bottom of the filter were fixed in 4% paraformaldehyde and stained with 1% crystal violet (Beyotime, P.R. China). The invading cells were counted under a microscope.
Wound-healing assay
Approximately 5 × 105 HeLa cells were seeded in 6-well plates. The cells were washed with PBS three times, and fresh medium was added. Cells were cultured in a 37 ℃ 5% CO2 incubator, photographed, and recorded under the microscope at 0, 24, and 48 h.
m6A ELISA
The total RNA extracted was detected by EpiQuik ™ m6A RNA Methylation Quantification Kit (Epigentek, USA), following the manufacturer’s instructions.
Statistical analysis
All data processing and analysis were completed in R (version 3.6.2). To compare two groups of continuous variables, the statistical significance of the normally distributed variables was estimated using the independent student’s t-test, and the differences between non-normally distributed variables were analyzed using the Mann–Whitney U test (i.e., Wilcoxon rank-sum test). The Chi-square test or Fisher’s exact test was used to compare and analyze the statistical significance between two groups of categorical variables. The correlation coefficients of different genes were calculated using Pearson correlation analysis. R’s survival package was used for survival analysis, the Kaplan–Meier survival curve was used to highlight the survival difference, and the log-rank test was used to evaluate the significance of the survival time difference between two groups [
36]. R’s pROC package was used to plot the receiver operating characteristic (ROC) curve and calculate the area under the curve (AUC) to assess the accuracy of the risk score in estimating prognosis. Univariate and multivariate Cox analyses were used to determine independent prognostic factors [
37]. All statistical P-values were bilateral, and P < 0.05 was considered statistically significant.
Discussion
As one of the most common RNA modifications, m
6A mRNA methylation is closely associated with cervical cancer. m
6A mRNA methylation might promote cervical cancer development. The m
6A level was significantly reduced in cervical cancer compared to adjacent normal tissue. The reduction in m
6A levels significantly correlated with the FIGO stage, tumor size, differentiation, lymph invasion, and cancer recurrence [
38]. m
6A methyltransferase methyltransferase-like 3 (METTL3) enhanced the stability of FOXD2-AS1, and its expression was maintained. METTL3/FOXD2-AS1 accelerated cervical cancer progression via an m
6A-dependent modality [
39]. METTL3 enhanced the HK2 stability through YTHDF1-mediated m
6A modification, thereby promoting the Warburg effect in cervical cancer [
40]. GAS5-AS1 interacted with the tumor suppressor GAS5 and increased its stability by interacting with RNA demethylase ALKBH5 and decreasing GAS5 m
6A modification. m
6A-mediated GAS5 RNA degradation relied on the m
6A reader protein YTHDF2-dependent pathway [
41]. FTO can control the m
6A modification of E2F1 and Myc transcripts to regulate the proliferation and migration of cervical cancer cells [
42]. The FTO-mediated stabilization of HOXC13-AS epigenetically upregulated FZD6 and activated Wnt/β-catenin signaling, driving CC proliferation, invasion, and EMT [
43]. FTO and its substrate m
6A may be critical factors for regulating chemo-radiotherapy resistance [
44]. YTHDF1 regulated the translation of RANBP2, which potentiated the growth, migration, and invasion of cervical cancer cells in an m
6A-dependent manner without any effect on its mRNA expression [
45]. circARHGAP12 exerted the oncogenic role in cervical cancer progression through the m
6A-dependent IGF2BP2/FOXM1 pathway [
46]. KCNMB2-AS1 and IGF2BP3 formed a positive regulatory circuit that increased the tumorigenic effect of KCNMB2-AS1 in cervical cancer [
47]. m
6A-associated downregulation of miR-193b promoted cervical cancer aggressiveness by targeting CCND1 [
48]. ZFAS1 and its m
6A modification may be a promising target for cervical cancer treatment [
49].
According to our study, differential analysis results showed differences in the expression of multiple m
6A-related genes between cervical cancer tissues and normal tissues; furthermore, these genes were related to prognosis. The established m
6A risk scoring system is closely related to the biological function of cervical cancer, immune invasion of cervical cancer, and sensitivity to small molecule drugs. The model based on the m
6A signature genes and the model based on the m
6A risk score can accurately predict the prognosis of patients with cervical cancer. Studies showed that m
6A was closely related to epithelial–mesenchymal transition, angiogenesis, Myc targets V1, KRAS signal pathway, and estrogen in the occurrence and development of tumors [
43,
50‐
54]. Our results showed that epithelial-mesenchymal transition, angiogenesis, Myc targets V1, and other pathways are mainly enriched in the high-risk groups, while bile acid metabolism, KRAS signaling DN, late estrogen response, and other pathways are significantly enriched in low-risk patients. The results of the present study may help understand the relationship between these pathways and m
6A, revealing their mechanism in cervical cancer occurrence and development. We also showed that the m
6A regulator correlates with the survival and clinicopathological characteristics of patients with CESC. The m
6A regulator-based prognostic signature may predict the prognosis of CESC [
55].
According to our analysis, the IC
50 of multiple chemotherapy drugs differed significantly between patients with high and low m
6A risk scores. Meanwhile, previous studies have also shown a close relationship between m
6A, mTOR pathway, and rapamycin. mTORC1 could stimulate oncogenic signaling and control anabolic cell growth via m
6A [
56,
57]. m
6A has been reported to play a fundamental role in the function of the mTOR pathway in gastrointestinal cancer [
58]. Reductions in the m
6A levels increased the expression of mTORC2 in endometrial cancer [
59]. The FTO protein was reported to participate in tumorigenesis via interactions with the target of the mammalian target protein Rapamycin (mTOR) [
60]. METTL3 was shown to cause the activation of mTORC1 signaling and colorectal cancer development in an m
6A-dependent manner [
61]. Rapamycin may regulate the m
6A levels and affect the biological characteristics of OSCC cells [
62]. METTL3-high tumors showed more sensitivity to everolimus, a rapamycin analog, in gastric cancer [
63].
Our study showed that knocking down ZC3H13, an m6A methyltransferase, can promote the proliferation, migration, and invasion abilities of cervical cancer cells and reduce the m6A levels. Rapamycin suppresses the proliferation, migration, and invasion abilities of cancer cells and can enhance the m6A levels, demonstrating the role of the m6A methylation modulator in the development of cervical cancer and the effectiveness of the drugs targeting the modulators of m6A methylation in treating cervical cancer.
In conclusion, our study shows that m6A regulatory factors are closely related to the occurrence, development, immune invasion, drug sensitivity, and prognosis of cervical cancer. Drugs targeting the factors regulating m6A offer good prospects for treating cervical cancer.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.