Large-scale transcriptome analysis identified RNA methylation regulators as novel prognostic signatures for lung adenocarcinoma

Abstract

Background

The abnormal expression of genes is an essential factor affecting the prognosis of cancer. RNA modification is a way of regulating post-transcriptional levels, including m⁶A, m⁵C, m¹A RNA methylation. Studies have found that RNA methylation regulates tumorigenesis development and stem cell regeneration. However, there are few studies on lung adenocarcinoma. This study aims to explore the clinical value of RNA methylation for lung adenocarcinoma.

Methods

We summarized thirty-one RNA methylation regulators. The training set was obtained from The Cancer Genome Atlas (TCGA) database, and the test set was obtained from the Gene Expression Omnibus (GEO) database. The Wilcoxon test was used to analyze the expression of RNA methylation regulators. We constructed tumor subgroup models and risk models based on the expression of those regulators. Principal component analysis (PCA) and the receiver operating characteristic (ROC) confirmed the accuracy of the models. Real-time polymerase chain reaction (PCR) validates the results in vitro.

Results

Most RNA methylation regulators had distinct expressions in tumor tissues and adjacent tissues (P<0.05). All the models showed high predictive performance (AUC: 0.65–0.82), and the five-year survival of patients in each group was statistically different (P<0.05). The patients in the high-risk group were more likely to have a higher stage, more lymph node metastases, and distant metastases, showing a poor clinical outcome. Patients with high expression of NOP2 or HNRNP were more likely to have a poorly differentiated in vitro experiment.

Conclusions

With our study, we found that the expressions of most RNA methylation regulators were significantly different in cancer and para-cancerous tissues. Different molecular phenotypes constructed by RNA methylation regulators can be independent risk factors for the prognosis of lung adenocarcinoma. Our study demonstrates the critical role of RNA methylation in lung adenocarcinoma, and it is expected to supply a reference for the prognostic stratification and treatment strategy development of lung adenocarcinoma.

Introduction

Lung cancer is one of the most common malignant tumors in the world. It is estimated that lung cancer-related deaths account for 23% of the total cancer-related deaths in 2020 (1). Non-small cell lung cancer accounts for 75% to 80% of all lung cancers, including adenocarcinoma and squamous cell carcinoma (2). Surgery is the primary treatment of non-small cell lung cancer. Tumor volume and lymph node metastasis are critical indicators to assess whether patients can undergo surgery and tumor staging (3). Traditional tumor staging is often used to predict the prognosis of lung cancer patients. However, patients with the same tumor stage sometimes have different prognoses due to individual differences (4). Recent studies have found that molecular expression is a key factor influencing lung cancer prognosis. For example, PD-1 expression reflects the response to immunotherapy (5), which suggests the vital value of molecular level studies in lung cancer prognosis.

RNA modification is a way of regulating post-transcriptional levels (6,7). More than 100 types of chemical RNA modifications, which are widely distributed in messenger RNA (mRNA), transfer RNA (tRNA), non-coding small RNA, and long non-coding RNA (lncRNA), have been identified at present (6,8,9). N6-methyladenosine (m⁶A) methylation is the most common type of RNA modification on mRNA (10). Limited by the low sensitivity of early detection techniques, it was not until 2011 that the biological function of the first m⁶A methylation regulators on mass and obesity (FTO) was clarified (11). 5-methylcytosine (m⁵C) and N1-methyladenosine (m¹A) are new types of RNA modifications that have attracted full attention in recent years. Bisulfite-based transcriptome sequencing study finds thousands of m⁵C modification sites on mRNAs in HeLa cells, and research on RIP sequencing technology found that m¹A is a novel transcriptome control with evolutionary conservation (12,13). However, the characteristics and functions of m⁵C and m¹A modification in mRNA are not entirely apparent.

RNA methylation regulates transcriptome expression under the dynamic regulation of methyltransferases (“writers”), binding proteins (“readers”), and demethylases (“erasers”) (14). Through the study of RNA methylation related proteins, thirty-one RNA methylation regulators were found (Table 1) (15-17). Recent studies have demonstrated that RNA methylation dynamically and reversibly regulates critical biological functions such as RNA metabolism and processing, as well as directional differentiation of stem cells (18). In the field of cancer research, it has been found that RNA methylation exists in multiple processes of tumorigenesis, development, and metastasis (19). For example, a study based on The Cancer Genome Atlas (TCGA) database found that m⁶A methylation regulators are involved in malignant progression and can predict the prognosis of liver cancer patients (20). The m⁶A demethylase FTO promotes the growth of lung cancer cells by regulating the expression of USP7 mRNA, and m⁶A methylation, which is also associated with afatinib resistance in lung cancer cells (21,22). However, m⁵C and m¹A methylation are rarely studied in lung cancer, and the overall level of the predictive value of m⁶A methylation in lung cancer is insufficient.

Table 1. RNA methylation regulators of m⁶A, m⁵C, and m¹A methylation.

RNA methylation	Writer	Reader	Eraser
m6A	METTL3, METTL14, RBM1, WTAP, ZC3H13, KIAA1429	HNRNPC, YTHDC1, YTHDC2, YTHDF1, YTHDF2	ALKBH5, FTO
m5C	NOP2, NSUN2, NSUN3, NSUN4, NSUN5, NSUN7, DNMT1, TRDMT1, DNMT3A, DNMT3B	ALYREF	TET2
m1A	TRMT6, TRMT10C, BMT2, RRP8		ALKBH1, ALKBH3

In this study, we systematically analyzed the expression of m⁶A, m⁵C, and m¹A methylation regulators in lung adenocarcinoma, as well as the associations between the clinicopathological characteristics. We constructed different tumor subgroup models and risk models to show the predictive value of RNA methylation for lung adenocarcinoma.

Methods

Datasets acquisition

The RNA-seq transcriptome data and corresponding clinical information of LUAD were downloaded from TCGA (https://cancergenome.nih.gov/) database. RNA-seq transcription of LUAD data included 59 cases of para-cancerous tissues and 535 cases of tumor tissues. Two hundred and ninety-four clinical cases were obtained after removing invalid data. Clinical information included age, gender, stage, T, N, M, overall survival (OS) time, and survival status (Table 2).

Table 2. Clinicopathological features of patients included in TCGA database.

Variable	Number	Percentage
Total cases	294	100.00
Age
<65	124	42.18
≥65	170	57.82
Gender
Male	145	49.32
Female	149	50.68
Stage
Stage I	152	51.70
Stage II	71	24.15
Stage III	51	17.35
Stage IV	20	6.80
T
T1	88	29.94
T2	156	53.06
T3	25	8.50
T4	25	8.50
N
N0	186	63.27
N1	64	21.77
N2	44	14.96
M
M0	274	93.20
M1	20	6.80

TCGA, The Cancer Genome Atlas.

Also, the test set data was downloaded from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/) under the accession number (GSE37745). It is used to test the reliability of the Cox risk models constructed from TCGA. Sixty-seven tumor cases from lung adenocarcinoma and clinical information include age, gender, stage, OS, and survival status (Table 3).

Table 3. Clinicopathological features of patients included in GEO database.

Variable	Number	Percentage
Total cases	67	100.00%
Age
<65	38	56.72%
≥65	29	43.28%
Gender
Male	27	40.30%
Female	40	59.70%
Stage
Stage i	46	68.66%
Stage ii	13	19.40%
Stage iii	5	7.46%
Stage iv	3	4.48%

GEO, Gene Expression Omnibus.

Lung adenocarcinoma and adjacent tissues from 11 patients (average age =48.0±13.4 years; 8 male patients and 3 female patients) were collected from The First Affiliated Hospital of China Medical University. The surgery and samples collection period is October 2019 to present. The study was conducted in accordance with the Declaration of Helsinki. The study was approved by ethics board of The First Affiliated Hospital of China Medical University (No. YB M-05-02) and informed consent was taken from all the patients.

Bioinformatic analysis

After correcting for transcriptome expression on the training and test sets, we separately extracted expression data of the thirteen m⁶A RNA methylation regulators, twelve m⁵C RNA methylation regulators, six m¹A RNA methylation regulators, and all the regulators. According to the classification of tumor tissues and adjacent tissues, we analyze the expression of those RNA methylation regulators. Violin maps and heatmaps were drawn to observe the expression distribution of RNA methylation regulators.

To investigate the role of different RNA methylation regulators in lung adenocarcinoma, we removed adjacent samples and grouped tumor samples using the “Consensus Cluster Plus” package. Principal component analysis (PCA) analysis was used to test the clustering effects. Combined clinical information, we draw survival curves to analyze the survival differences of different subgroups and draw heatmaps to explore the correlation between subgroups and clinical characteristics.

To clarify the prognosis risk of the RNA methylation regulators, we use the “survival” package to constructed Cox models to screen out risk molecules and divide patients into the high-risk group or low-risk group. We calculated the risk score using the following formula, where Coefi is the coefficient, and xi is the expression value of each selected molecule. Receiver operating characteristic (ROC) curve was used to test model effectiveness. Multivariate Cox regression was used to analyze the independent prognostic role of the risk model.

$$\mathrm{Risks}\mathrm{core}={\displaystyle \sum _{i=1}^n\mathrm{Coe}{f}_i*x_i}$$

Lastly, gene set enrichment analysis (GSEA) is used to find the possible mechanism of RNA methylation affecting the prognosis of lung adenocarcinoma. The whole gene expression data of patients in the various risk groups were uploaded to the GSEA v4.0.3 (www.gsea.com) and the Kyoto Encyclopedia of Genes and Genomes (KEGG) software for analysis, which was performed with 1,000 iterations. The pathways that may be enriched in the different risk groups are scored. Screen the KEGG pathway based on the enrichment score.

In vitro experiment

Real-time polymerase chain reaction (PCR) was validated in vitro after screening RNA methylation regulators associated with clinical pathology by bioinformatic analysis. RNA extraction from lung adenocarcinoma and adjacent tissues was performed using TRIzol reagent (TIANGEN Biotech Company, China). RNA was reverse-transcribed into cDNA with the FastKing RT Kit (TIANGEN Biotech Company, China). Real-time PCR was on TL988 Real-Time PCR Detection System (TIANLONG, China), and the levels were normalized to the level of β - actin. The primers were as follows: HNRNPC: forward: 5'-TCCTCCTCCTATTGCTCGGG-3' and reverse: 5'-GTGTTTCCTGATACACGCTGA-3'. NOP2: forward: 5'-AAGGGTGCCGAGACAGAACT-3' and reverse: 5'-GAGCACGACTAGACAGCCTC-3'. β-actin: forward: 5'-ATAGCACAGCCTGGATAGCAACGTAC-3' and reverse: 5'-CACCTTCTACAATGAGCTGCGTGTG-3'.

Statistical analysis

Statistical analysis of all RNA-seq transcriptome data was conducted using R v3.4.1 (https://www.r-project.org/). Wilcoxon test was used to compare the differences in the expression of RNA methylation regulators between different tumor stage or between tumor tissues and adjacent tissues. OS is defined as the interval from the date of diagnosis to the date of death. The Kaplan-Meier method was used to compare the OS of the patients in distinct groups. A chi-square test was used to analyze the correlation between the separate groups and clinical characteristics. P<0.05 was considered statistically significant.

Results

Expression of different RNA methylation regulators in lung adenocarcinoma

According to the classification of cancer tissues and adjacent tissues, we respectively analyzed the expression levels of the distinct types of RNA methylation regulators. The expressions of five m¹A molecules, nine m⁵A molecules, and six m⁶A molecules were higher in tumor tissues than that in the adjacent tissues. One m⁵C molecule and four m⁶A molecules were lower in cancer tissues than in the adjacent tissues (P<0.05) (Figure 1A,B,C,D,E,F). Overall, the expression of most RNA methylation regulators in lung adenocarcinoma tissues is specific (P<0.05), which are shown in Figure 1G,H. These results suggest that RNA methylation may play an essential role in the malignant progression of lung adenocarcinoma.

Figure 1.

Expression of RNA methylation regulators in TCGA database. (A,D) The expression levels of each m¹A RNA methylation regulators. (B,E) The expression levels of each m⁵C RNA methylation regulators. (C,F) The expression levels of each m⁶A RNA methylation regulators. (G,H) The expression levels of all RNA methylation regulators. *, P<0.05; **, P<0.01 and ***, P<0.001. TCGA, The Cancer Genome Atlas.

Consensus clustering of RNA methylation regulators identified cluster subgroups of lung adenocarcinoma with different clinical features

After removed adjacent samples, we grouped cancer samples using the “Consensus Cluster Plus” package. Considering the clustering stability and the number of each group, we divided the patients into two subgroups clustered by k=2. In the cluster models of lung adenocarcinoma, we can see that the distribution of the two subgroups of the m¹A model overlaps (Figure 2A,B,C), while the other models show good dispersion by PCA (Figure 2D,E,F,G,H,I,J,K,L). These results show that our classification by RNA methylation regulators are correct.

Figure 2.

Identification of consensus clusters by RNA methylation regulators in lung adenocarcinoma. (A,B,C) Cluster model and PCA based on m¹A RNA methylation regulators. (D,E,F) Cluster model and PCA based on m5C RNA methylation regulators. (G,H,I) Cluster model and PCA based on m6A RNA methylation regulators. (J,K,L) Cluster model and PCA based on all RNA methylation regulators. PCA, principal component analysis.

To better understand the correlation between clustering results and clinicopathological features, we performed a survival analysis with the OS. The model constructed by m¹A methylation regulators does not show differences in survival or correlation with clinicopathological features (P>0.05) (Figure 3A,B). However, there are significant differences in survival between tumor subgroups of m⁵C and m⁶A models, which is also related to the clinical characteristics of stage, T, N, and M classification (P<0.05) (Figure 3C,D,E,F). The analysis results for all methylation regulators model are consistent (Figure 3G,H). Overall, these results suggest are consistent. Further analysis of the models suggests that cluster 2 patients had a lower five-year survival rate and this population is more likely to have a higher stage, lymph node metastases, and distant metastasis, indicating a poor clinical prognosis.

Figure 3.

Correlation of different subgroups with OS and clinical characteristics in TCGA database. (A,B) Survival curve and clinical characteristics of the clustering model based on m¹A RNA methylation regulators. (C,D) Survival curve and clinical characteristics of the clustering model based on m⁵C RNA methylation regulators. (E,F) Survival curve and clinical characteristics of the clustering model based on m⁶A RNA methylation regulators. (G,H) Survival curve and clinical characteristics of the clustering model based on all RNA methylation regulators. *, P<0.05; **, P<0.01. OS, overall survival; TCGA, The Cancer Genome Atlas.

Prognostic value of RNA methylation regulators, and risk models built using selected RNA methylation regulators

To better understand the prognostic role of RNA methylation regulators in lung adenocarcinoma, we constructed Cox models combining with the expression of RNA methylation regulators and OS from the TCGA database. Based on the risk score, patients were divided into high-risk and low-risk groups. The risk score coefficients of different risk models are shown in Table 4. The survival rate of the high-risk group in m¹A prognostic signature is lower than that of the low-risk group (P<0.05), and the prediction effect is high (AUC=0.73) (Figure 4A,B). The prognostic signatures of m⁵C and m⁶A also show significant differences in survival (P<0.05), of which m⁵C signature has a higher predictive effect (AUC: 0.77) and m⁶A signature has the lowest predictive effect (AUC: 0.71) (Figure 4C,D,E,F). The prognostic signature constructed by all RNA methylation regulators has the highest predictive effect (AUC=0.82) (Figure 4G,H).

Table 4. Risk model constructed by m¹A, m⁵C, m⁶A, or all methylation regulators.

Risk model	Gene	Coefficient	HR	HR.95L	HR.95H	P
m1A RNA methylation regulators	BMT2	−0.062	0.940	0.866	1.021	0.141
	TRMT6	0.052	1.053	1.003	1.106	0.037
	RRP8	−0.316	0.729	0.563	0.945	0.017
m5C RNA methylation regulators	NSUN2	0.012	1.012	0.998	1.026	0.095
	NSUN4	−0.319	0.727	0.598	0.883	0.001
	TET2	0.164	1.178	1.039	1.335	0.010
	ALYREF	0.013	1.013	1.004	1.022	0.003
m6A RNA methylation regulators	METTL14	−0.255	0.775	0.590	1.018	0.067
	METTL3	−0.070	0.932	0.867	1.002	0.057
	ZC3H13	0.115	1.122	1.050	1.198	0.001
	YTHDF1	−0.017	0.983	0.967	1.001	0.060
	HNRNPC	0.018	1.018	1.005	1.032	0.006
	RBM15	0.144	1.155	1.040	1.283	0.007
All RNA methylation regulators	YTHDF2	0.026	1.026	0.992	1.062	0.131
	BMT2	−0.086	0.918	0.840	1.002	0.055
	YTHDC1	−0.071	0.931	0.846	1.024	0.141
	ZC3H13	0.093	1.098	1.018	1.184	0.016
	YTHDF1	−0.028	0.973	0.953	0.993	0.009
	NOP2	0.034	1.035	1.006	1.065	0.020
	HNRNPC	0.023	1.024	1.008	1.040	0.003
	NSUN4	−0.413	0.662	0.540	0.812	<0.001
	RBM15	0.219	1.245	1.101	1.407	<0.001
	TRMT10C	−0.039	0.962	0.921	1.004	0.074
	ALYREF	0.012	1.012	1.003	1.022	0.012

Figure 4.

Survival and ROC curves for different risk models from TCGA database. (A,B) The risk model constructed by m¹A methylation regulators. (C,D) The risk model constructed by m⁵C methylation regulators. (E,F) The risk model constructed by m⁶A methylation regulators. (G,H) The risk model constructed by all methylation regulators. ROC, receiver operating characteristic; TCGA, The Cancer Genome Atlas.

We further used univariate and multivariate Cox regression to explore the predictive value of the signatures, as well as the associations between the clinicopathological characteristics. The analysis results of all prognostic signatures are consistent. Risk score in those signatures can be independent risk factors affecting the prognosis of lung adenocarcinoma (HR>1, P<0.05), and the high-risk patients are more likely to have a higher stage, more lymph node metastases, and distant metastases, indicating a poor clinical prognosis (Figure 5).

Figure 5.

Clinical characteristics of different risk groups from TCGA database. (A,B,C) The risk model constructed by m¹A methylation regulators. (D,E,F) The risk model constructed by m⁵C methylation regulators. (G,H,I) The risk model constructed by m⁶A methylation regulators. (J,K,L) The risk model constructed by all methylation regulators. *, P<0.05; **, P<0.01. TCGA, The Cancer Genome Atlas.

Verify the reliability of the risk model constructed by the TCGA database through the GEO database

To test the reliability of the risk models based on TCGA data, we use the GSE37745 dataset from the GEO for validation. By calculating the patient’s risk score, all the patients are divided into a high-risk group or low-risk group. The survival rate of the high-risk group in m¹A prognostic signature is lower than that of the low-risk group (P<0.05), which is consistent with the training set, but the prediction effect is general (AUC =0.65) (Figure 6A,B). The prognostic signatures of m⁵C and m⁶A also show significant differences in survival (P<0.05), with higher predictive effects (AUC: 0.75–0.76) (Figure 6C,D,E,F). The prognostic signature constructed by all RNA methylation regulators has the highest predictive effect (AUC =0.79) (Figure 6G,H).

Figure 6.

Survival and ROC curves for different risk models in the test set. (A,B) Verification of the risk model constructed by m¹A methylation regulators. (C,D) Verification of the risk model constructed by m⁵C methylation regulators. (E,F) Verification of the risk model constructed by m⁶A methylation regulators. (G,H) Verification of the risk model constructed by all methylation regulators. ROC, receiver operating characteristic.

The test set only followed the patient’s gender, age, and stages, and we also performed univariate and multivariate Cox regression analysis. These results are a bit different. The risk score of the m¹A signature can be a risk factor that affects the prognosis of lung adenocarcinoma (HR >1, P<0.05), but it has no correlation with clinical features (Figure 7A,B,C). The analysis of the m⁵C signature is not statistically significant (P>0.05) (Figure 7D,E,F). The risk score of other signatures also can be independent risk factors for the prognosis of lung adenocarcinoma (HR >1, P<0.05) (Figure 7G,H,I,J,K,L). Generally, the test set successfully verifies the results of the training set, which suggests that RNA methylation is a valid marker for predicting patients with lung adenocarcinoma.

Figure 7.

Clinical characteristics of different risk groups f in the test set. (A,B,C) Verification of the risk model constructed by m¹A methylation regulators. (D,E,F) Verification of the risk model constructed by m⁵C methylation regulators. (G,H,I) Verification of the risk model constructed by m⁶A methylation regulators. (J,K,L) Verification of the risk model constructed by all methylation regulators. *, P<0.05.

Gene set enrichment analysis

We also performed KEGG analysis of gene expression in patients in the different risk groups. The results of different risk models were consistent. Overall, cell cycle, RNA degradation, P53 signaling pathway, homologous recombination, and mismatch repair were significantly enriched in the high-risk group. Fatty acid metabolism, histidine metabolism, and primary bile acid biosynthesis were significantly enriched in the low-risk group (Figure 8). These results supply a reference for further functions and mechanisms of RNA methylation in lung adenocarcinoma.

Figure 8.

KEGG analysis of different risk groups in lung adenocarcinoma. (A) The risk model constructed by m¹A methylation regulators. (B) The risk model constructed by m⁵C methylation regulators. (C) The risk model constructed by m⁶A methylation regulators. (D) The risk model constructed by all methylation regulators. KEGG, Kyoto Encyclopedia of Genes and Genome.

Validation of RNA methylation regulators expression and clinical pathological correlations in lung adenocarcinoma by real-time PCR

Survival analysis of each molecule identified nine RNA methylation regulators (RRP8, HNRNPC, NOP2, NSUN4, DNMT3B, RBM15, TRDMT1, DNMT1, YTHDF2) that were associated with patient survival (P<0.05) (Figure 9A,B). Among these RNA methylation regulators, patients with high expression of HNRNPC or NOP2 were more likely to be in advanced tumor stage (P<0.05) (Figure 9C,D). By performing real-time PCR on 11 lung adenocarcinoma patient tissues, we found that the expression of HNRNPC and NOP2 was higher in tumor tissues than that in the adjacent tissues, which is consistent with the results of bioinformatic analysis (P<0.05) (Figure 9E,F). Furthermore, patients with high expression of HNRNPC or NOP2 were more likely to have a poorly differentiated (P<0.05) (Figure 9G,H), indicating a poor clinical prognosis. These results suggest the impact of RNA methylation regulators in the clinical pathology of different tumors.

Figure 9.

Validation of RNA methylation regulators expression and clinical pathological correlations in lung adenocarcinoma by real-time PCR. (A,B) Correlation of HNRNPC and NOP2 expression with patient survival from TCGA. (C,D) Correlation of HNRNPC and NOP2 expression with tumor staging of the patient from TCGA. (E,F) Expression of HNRNPC and NOP2 in lung adenocarcinoma and adjacent tissues by real-time PCR. (G,H) Correlation of HNRNPC and NOP2 expression with tumor differentiation of the patient from by real-time PCR. PCR, polymerase chain reaction.

Discussion

Over the past decade, people recognized that the occurrence and development of lung cancer is a complex process involving many factors, and different molecular expressions have variable drug responses and clinical prognosis to patients. Traditional tumor staging only depends on tumor volume and metastasis, which cannot take into account the patient’s gene expression level (23). The mutation rate of functional driver genes in lung adenocarcinoma is about 60%, of which KRAS, EGFR mutation, and EML4-ALK fusion is the most common driver gene, accounting for about 35% to 40% (24). Kris’s study found that the median survival time of patients with such type of driver gene mutation receiving targeted therapy was prolonged by 2–4 years compared with other patients (25). The study of genetic changes in lung squamous cell carcinoma was later than lung adenocarcinoma. TP53, KRAS, GRM are the common gene mutations (26). Weiss’s research found that about 20% of lung squamous cell carcinomas have FGFR1 amplification. Inhibition of FGFR1 in cell lines and mouse models can inhibit cell growth and induce apoptosis (27). These studies suggest that the new molecular typing research is expected to predict the clinical prognosis of patients more from the molecular level.

RNA modification is a way of regulating post-transcriptional levels. For example, m⁶A modification refers to a modification in which one hydrogen atom (-H) attached to the sixth nitrogen atom (N6) on the adenine molecule is substituted with a methyl group (-CH4) (28). The m⁶A methyltransferase promotes mRNA methylation of adenine, and demethylase can eliminate it, and the binding proteins play a recognition role in this process (28,29). m⁵C and m¹A RNA methylation also regulate gene expression by acting at different sites. RNA methylation is a dynamic and reversible modification method, and the entire process is regulated by methyltransferases, demethylase, and binding proteins. FTO is the first identified m⁶A demethylase, and its m⁶A-mediated modification can be a novel cis-element to regulate mRNA splicing and adipose precursor cell differentiation (18). Also, it was found that FTO can demethylate m¹A, too (30). Yang discovered the distribution of m⁵C methylation, identified the main methyltransferase NSUN2 and the first binding protein ALYREF (17). David’s study found that Arabidopsis m⁵C modification regulates tissue development, and the methyltransferase is TRM4B (31). There is less research on m¹A methylation. Li has developed a single-base-resolution m¹A sequencing method and found that nuclear and mitochondrial coding transcripts have different types of m¹A methylation (32).

RNA methylation also regulates tumorigenesis and development, drug response, and stem cell renewal. Zhou’s study found that FTO enhances the chemo-radiotherapy resistance both in vitro and in vivo through regulating the expression of β-catenin by reducing m⁶A levels in its mRNA transcripts (33). Cui’s research suggests that m⁶A mRNA modification is critical for glioblastoma stem cell self-renewal and tumorigenesis. Knockdown of METTL3 or METTL14, key components of the RNA methyltransferase complex, promotes human glioblastoma stem cell growth, self-renewal, and tumorigenesis (34). Chen’s research found that m⁵C RNA methylation promotes the pathogenesis of bladder cancer through stabilizing mRNAs, and high expression of NUSN2 predicts poor survival (35). In the study of TCGA, literature reported that m⁶A RNA methylation regulators contribute to malignant progression and have a clinical prognostic impact (15,36). These studies have suggested that RNA methylation not only plays a role in tumor progression but can also be a signature to predict clinical prognosis, but there are few studies on the effect of RNA methylation on lung cancer. We obtained a large number of patient cases and clinicopathological features by downloading lung adenocarcinoma data from the TCG database. We likewise mined the GEO database, but with a lower number of cases, which lends itself to doing studies that validate the TCGA results. The analysis and validation of these data provide the basis for a prognostic study of RNA methylation in lung adenocarcinoma.

Through the study of RNA methylation, we summarized thirty-one RNA methylation regulators. Further analysis showed that the expression of these regulators in tumor tissues differed from that in adjacent tissues, which suggested that RNA methylation plays a vital role in the development of lung adenocarcinoma. To explore the clinical value of RNA methylation, we constructed several subgroups and risk models, PCA and ROC curves show the excellent accuracy of the models. In different models, we found that the expression of RNA methylation regulators is related to the survival of patients, and different molecular phenotypes can be independent risk factors for the prognosis of lung adenocarcinoma. Real-time PCR validated the results of the bioinformatic analysis. Our study explains the critical role of RNA methylation in lung cancer, and it is expected to supply a reference for the prognostic stratification and treatment strategy development of lung adenocarcinoma.

Supplementary

The article’s supplementary files as

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki (as revised in 2013). The study was approved by ethics board of The First Affiliated Hospital of China Medical University (No. YB M-05-02) and informed consent was taken from all the patients.