Identification of Prognosis Signature and Analysis of the Immune Microenvironment in Gastric Cancer Based on ALKBH5

doi:10.21203/rs.3.rs-2084871/v1

Download PDF

Research Article

Identification of Prognosis Signature and Analysis of the Immune Microenvironment in Gastric Cancer Based on ALKBH5

https://doi.org/10.21203/rs.3.rs-2084871/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background

N6-methyladenosine(m6A) RNA regulators play important roles in cancers, but the functions and mechanism of them have not been demonstrated clearly in gastric cancer (GC).

Methods

In this study, the GC samples with clinical information and RNA transcriptome were downloaded from TCGA database. The different expression genes were compared by absolute value and median ± standard deviation (sd). Samples with complete information were randomly divided into training dataset and test datasets. The differential expression genes (DEGs) between ALKBH5-low and ALKBH5-high subgroups were identified in training dataset and constructed a risk model by Cox and LASSO regression. The model was testified in test datasets, the overall survival (OS) was compared with Kaplan-Meier method and immune cells infiltration was calculated by CIBERSORT algorithm in the low-risk and high-risk subgroups based on the model.

Results

ALKBH5 was the only one regulator whose expression was lower in tumor samples than that in normal samples, there was the same phenomenon in GEO dataset GSE29998. Low expression of ALKBH5 led to poor overall survival of GC patients and seemed to be an independent protective factor. The model based on ALKBH5 regulated genes was validated in both two datasets (training/test) and it displayed potential capacity to predict clinical prognosis. Gene Ontology (GO) analysis implied that the DEGs were involved in immune response, CIBERSORT results indicated that ALKBH5 and its related genes could alter the immune microenvironment of GC.

Conclusions

In this study we found that ALKBH5 might be a suppressor of GC, ALKBH5 and its related genes were latent biomarkers and immunotherapy targets.

m6A

ALKBH5

Gastric cancer

Immune cell infiltration

Prognosis

Gastric cancer (GC) is one of the most threatening diseases worldwide, which gave 5th incidence and 4th mortality in cancers in 2020(1). Appropriate option of Surgical resection is the only one strategy to treat it. However, due to inconspicuous early symptoms, patients are always diagnosed in advanced stages so that following chemotherapy and targeted therapy are needed. Unfortunately, the efficiency is limiting because of late detection and lack for therapeutic targets(2), therefore, it is urgent to find novel approaches to improve the cure rate.

N6-methyladenosine (m6A) is the most abundant modification on mRNAs of eukaryote(3), it is a highly conserved and dynamic reversible process regulated by the methyltransferases (writers), demethylases (erasers) and m6A-binding proteins (readers)(4, 5). Writers mainly includes METTL3 (Methyltransferase-like 3), METTL14 (Methyltransferase-like 14), WTAP (Wilms Tumor 1-Associating Protein), VIRMA/KIAA1429 (Vir Like M6A Methyltransferase Associated), RBM15/15B (RNA Binding Motif Protein 15/15B) and ZC3H13 (Zinc Finger CCCH-Type Containing 13) (6–9). Erasers mainly includes FTO (Fat Mass and Obesity-associated Protein) and ALKBH5 (AlkB Homolog 5, RNA Demethylase) (4, 10), Readers mainly includes YTHDF1/2/3 (YTH N6-Methyladenosine RNA Binding Protein 1/2/3), YTHDC1/2 (YTH Domain Containing 1/2), IGF2BP1/2/3 (Insulin Like Growth Factor 2 MRNA Binding Protein 1/2/3), HNRNPC (Heterogeneous Nuclear Ribonucleoprotein C), HNRNPA2B1 (Heterogeneous Nuclear Ribonucleoprotein A2/B1)(11–16). m6A regulators participate in various physiological and pathological processes in tumor occurrence and development, acting as promoters or inhibitors. For instance, in bladder cancer, METTL3 accelerated the maturation of pri-miR221/222, resulting in the reduction of PTEN, which ultimately leads to the proliferation of bladder cancer (17). In breast cancer, FTO degraded BNIP3 through demethylation of m6A in the 3’UTR, leading to tumorigenesis and poor prognosis(18). In renal cell carcinoma, METTL14 attenuated the proliferation and migration ability of RCC cells through decreasing the expression of long noncoding RNA nuclear enriched abundant transcript 1_1 (NEAT1_1)(19). In glioblastoma, YTHDF2 tended to be a therapeutic target, it could stabilize the transcripts of MYC and therefore regulate glucose metabolism in glioblastoma stem cells (GSCs)(20).

In GC, there are a lot of important functions induced by m6A regulators and they played dual roles in various oncogenic signaling pathways too. Wang et al. demonstrated that the level of METTL3 was significantly elevated in GC tissues and associated with poor prognosis. It promoted the tumorigenesis and metastasis of GC by stimulating m6A modification of HDGF mRNA and then activated GLUT4 and ENO2 expression (21). Yue et al. also revealed METTL3 facilitated GC by regulating the m6A level of ZMYM1, which could increase the expression of E-cadherin and promote the epithelial-mesenchymal transition(EMT) process(22). Huo et al. found that METTL3 accelerated the development of GC by METTL3-SPHK2-KLF2 axis(23). What’s more, Pi et al. illustrated YTHDF1 promoted the translation of a key Wnt receptor frizzled7 (FZD7), leading to hyperactivation of the Wnt/β-catenin pathway and promotion of gastric carcinogenesis (24). Chen et al. found YTHDF1 also facilitated tumorigenesis and metastasis of GC by promoting USP14 protein translation in an m6A-dependent manner (25). FTO was verified led to metastasis of GC by increasing the m6A level and expression of ITGB1 (26), meanwhile Yang et al. demonstrated it promoted the development of GC by FTO-m6A-MYC axis (27). These regulators also executed functions through shaping long non-coding RNA (lncRNA). Lv et al. have revealed that the m6A levels of lncRNAs were changed in GC by bioinformatic manners and the difference might have predictive value(28). What’s more, Hu et al. found that LINC01320 could be elevated by METTL14 and promotes the proliferation, migration, and invasion of gastric cancer via miR495-5p/RAB19 axis(29).

ALKBH5 is short of alkylation repair homolog protein 5, a demethylase of m6A. Differential expression and regulatory functions of ALKBH5 in multiple cancers have been reported. In acute myeloid leukemia (AML), ALKBH5 was regulated by KDM4C and high level of ALKBH5 increased the stability of AXL, thus promoted leukemia stem cells(30). In glioblastoma, ALKBH5 promoted the tumorigenesis and development by demethylating FOXM1 nascent transcripts which led to enhanced expression of FOXM1(31). In pancreatic cancer, Guo et al. elucidated ALKBH5 reduced tumor proliferation, migration, invasion by activating PER1 (Period Circadian Regulator 1) in an m6A dependent manner(32), Tang et al. declared ALKBH5 also increased the expression of WIF-1 (Wnt inhibitory factor 1) and inhibited Wnt pathway to suppress these tumor features(33). In non-small cell lung cancer (NSCLC), ALKBH5 inhibited tumor growth and metastasis by decreasing YAP activity and regulating miR-107/LATS2 axis in an HuR-dependent manner(34). However, the role of ALKBH5 in GC was conflicted and obscure. Ge et al. demonstrated that there was lower expression of ALKBH5 in GC peripheral blood compared with healthy controls and it might be a protective gene for GC patients(35). Meanwhile, Zhang et al. elucidated that ALKBH5 promoted GC invasion and metastasis by demethylation of lncRNA NEAT1(36). According to present situation, the expression, function and mechanism of ALKBH5 in GC are still worthy to investigate and validate.

In this study, as showed in the workflow chart (Fig. 1), we downloaded and comprehensively analyzed stomach adenocarcinoma (STAD) samples from The Cancer Genome Atlas database (TCGA, https://portal.gdc.cancer.gov/). Comparing the m6A regulator genes’ expression levels in tumor group, normal group and investigating their relationship, we found that the expression of ALKBH5 was lower in tumor samples than normal ones and high level of ALKBH5 led to better prognosis. Then, the whole cohort was randomly divided into training dataset and test dataset. A risk model and a nomogram model for predicting the prognosis of GC patients were constructed in training dataset and validated in test dataset. Furthermore, Gene Ontology (GO) analysis suggested that the differential expression genes (DEGs) might enriched in immune defense, so that we computed the immune cell infiltration in all of the samples by CIBERSORT algorithm and the proportion of these cells were compared in two risk subgroups. The findings in this study indicate that ALKBH5 may play a suppressor in GC and its related genes act as latent predictive biomarkers.

Data Acquisition and Arrangement

The major clinical information and RNA transcriptome dataset (FPKM-UQ and Counts) of GC samples and normal samples were downloaded from The Cancer Genome Atlas database (TCGA, https://portal.gdc.cancer.gov/). Stomach adenocarcinoma (STAD) is the major type of gastric cancer and the number of other types of gastric cancer is too small to analyze, so we chose STAD samples for our study. A total of 407 samples were downloaded, including 375 primary tumor samples and 32 normal samples. When analyzing the prognosis of target genes or of different clusters, samples without complete clinical information were wiped out and the remain ones were randomly divided into training dataset and test dataset. Transcriptome data for validating the expression levels of ALKBH5 was downloaded from Gene Expression Omnibus (GEO, https://www.ncbi.nlm.nih.gov/geo/)

Bioinformatic Analysis

Comparison of the expression levels of m6A regulator genes

The different expression levels of m6A regulator genes were compared by absolute value and median ± standard deviation (sd), the results were displayed by R software using “pheatmap” and “ggplot2” packages. We set overall survival (OS) as the major criterion for evaluation of different cohorts’ death risk. it means from randomization to the time of death from any cause. For subjects who were lost to follow-up prior to death, the time of last follow-up is usually calculated as the time of death. The OS between ALKBH5-low/ALKBH5-high and low-risk/high-risk subgroups, as well as the univariate Cox regression and multivariate Cox regression were conducted by “survival” package.

Construction of the risk model

The “edgeR” package was chosen to identify the differential expression genes (DEGs) between ALKBH5-low and ALKBH5-high subgroups of ALKBH5, the relationship of these genes and the samples’ OS was evaluated with univariate Cox regression. After screening out genes with significant difference, the “glmnet” package was used to conduct the least absolute shrinkage and selection operator (LASSO)-penalized Cox regression. According to the result of LASSO regression, the risk model was constructed by the prognosis formula: risk score =\(\sum \text{e}\text{x}\text{p}\text{i} \text{*} \text{c}\text{o}\text{e}\text{f}\text{i}\). Where expi represented log2 (gene expression + 1) and coefi represented the coefficient of each gene. The model was applied in each sample in training/test dataset and the median value was set as the cut-off, thus samples in both datasets were divided into low-risk subgroup and high-risk subgroup respectively. Furthermore, sequence-based RNA adenosine methylation site predictor (SRAMP) database was used to search for the prediction scores of possible adenosine methylation sites on ALKBH5 related genes in risk model.

Validation of the risk model

To validate the model and roughly get its predictive efficacy in clinical GC patients, in training dataset and in test dataset, we compared the OS curves with Kaplan-Meier method and explored the distribution of samples’ vital status and survival time according to risk score with dot plot. Furthermore, univariate/multivariate Cox regression and receiver operating characteristic (ROC) curves were conducted. The ROC curves were accomplished by “pROC” package. Thereafter, to further clarify the prognosis value of the model, the OS in various clinical features cohort (including gender, age, stage, grade and T/N stages) between the low-risk and high-risk subgroups were compared in the two datasets.

Nomogram

A nomogram for predicting the prognosis of GC was built with the “rms” package in training dataset and the calibration curves were conducted subsequently. Then, the nomogram was checked in test dataset and the Decision Curve Analysis (DCA) was performed by the “rmda” package.

Analysis of immune cells infiltration

To seek the potential function of the DEGs, Gene Ontology (GO) analysis was operated with the “clusterProfiler” package (37, 38). Meanwhile, CIBERSORT algorithm was used to calculate 22 immune cells infiltration proportion of each sample. The different infiltration ratios among ALKBH5-low/ALKBH5-high and low-risk/high-risk subgroups were compared in the whole dataset.

Statistical Analysis

R software version R-4.1.2 for windows (The R Foundation for Statistical Computing, Vienna, Austria), SPSS version 22.0 (SPSS Inc., Chicago, IL, USA) and GraphPad Prism version 8.0 for Windows (GraphPad Software, San Diego, USA, www.graphpad.com) were used for data analysis. t test was used to compare continuous variables, chi-square test was used to compare categorical variables. Univariate and multivariate Cox regression were performed to identify independent prognostic factors for OS. Kaplan-Meier method was used to conduct the OS in different groups, and log-rank test was used for comparing the survival curves. When P-value < 0.05, it was considered as statistically significant.

The Expression and Relationship of m6A Related Genes

We first compared the expression levels of 19 m6A regulator genes between 375 primary tumor samples and 32 solid tissue samples, most of these genes showed higher expression in tumor samples than normal samples, including HNRNPA2B1, HNRNPC, IGF2BP1, IGF2BP2, IGF2BP2, KIAA1429, METTL3, RBM15, YTHDF1, YTHDF2, ZC3H13, but the expression level of ALKBH5 was opposite (Fig. 2. A, C), then the difference of ALKBH5’s expression was checked and verified in GSE29998 (Figure S1). The correlation coefficients of these regulators were detected with Pearson method, results demonstrated that among the 19 regulator genes, KIAA1429 and YTHDF3 had the strongest correlation (0.63), the top 3 genes related to ALKBH5 were RBM15B (0.30), FTO (0.30) and METTL14 (0.25) (Fig. 2. B). Furthermore, the normal samples were wiped out, the expression landscape in tumor cohort was conducted to further see the relationship of them. In different clinical and pathologic cohorts including gender, age, stage, grade, T, N and M, the distribution of these genes’ expression levels had difference (Fig. 2. D).

ALKBH5 is a Protective Gene in GC and A Risk Model Based on ALKBH5 Related Genes was Conducted

To investigate the relationship between the regulators and GC patients’ overall survival (OS), 58 tumor cases without complete clinical information were omitted, then the univariate and multivariate Cox regression were performed in remain 317 samples. Both univariate and multivariate Cox model indicated that ALKBH5 might be an independent protect factor for GC patients. RBM15, YTHDF2 also showed same hazard ratio as ALKBH5 but, ALKBH5 showed significance for both univariate and multivariate Cox regression (Figure S2. A, B). Furthermore, we drew survival curves with the vital status data in ALKBH5-low expression and ALKBH5-high expression subgroups and compared them, the cut-off of groups were median values. Result indicated that the ALKBH5-lowe subgroup had a shorter OS than the ALKBH5-highe subgroup with statistical significance (Fig. 3. A).

To better decode the roles of ALKBH5 in GC, remained 317 samples were randomized into training dataset (159) and test dataset (158), baseline of the two datasets was displayed in Table 1. In training dataset, the “edgeR” package was used to obtain DEGs between ALKBH5-low and ALKBH5-high subgroups with the conditions: log Foldchange > 2, adjust-P-value < 0.05 (Fig. 3. B, Table S1). Then univariate Cox regression was conducted to investigate the relationship between the DEGs and the samples’ OS (Table S2). For those significant ones (P-value < 0.05), we performed the LASSO-penalized Cox regression (Fig. 3. C, D). Constructing the risk model with prognosis formula in method section, the Risk score = 0.006188505 * CA10 + 0.050823131 * SLC7A2–0.008562421 * LINC02303 + 0.050382245 * CGB3 + 0.042501742 * C1QL2 + 0.00927148 *CGB8. All samples’ risk scores in both datasets were calculated with the formula, then samples in each dataset were divided into low-risk and high-risk subgroups by median value.

Table 1

The baseline of patients in training dataset and test dataset
Variables	Training Dataset(n = 159)	Test Dataset(n = 158)	P value
Age (n%)			0.1956
≤ 65	64 (40.25%)	76 (48.10%)
> 65	95 (59.75%)	82 (51.90%)
Gender (n%)			0.6118
male	102 (64.15%)	96 (60.76%)
female	57 (35.85%)	62 (39.24%)
Grade (n%)			0.7827
G1 & G2	57 (35.85%)	60 (38.00%)
G3	102 (64.15%)	98 (62.00%)
T (n%)			0.1457
T1-2	34 (21.38%)	46 (29.11%)
T3-4	125 (78.62%)	112 (70.89%)
N (n%)			0.8737
N0	49 (30.82%)	51 (32.28%)
N+	110 (69.18%)	107 (67.72%)
M (n%)			0.4975
M0	150(94.34%)	145 (91.77%)
M1	9 (5.66%)	13 (8.23%)

In addition, prediction scores in SRAMP(39) suggested that there were considerable adenosine methylation sites with very high confidence or high confidence, especially in CA10, LINC02303 and C1QL2 (Figure S3. A-F).

The Risk Model Has Strong Association with Clinical Prognosis in Gastric Cancer

Survival curves, distribution of patients’ status/survival time, univariate Cox regression and multivariate Cox regression were conducted in training dataset and test dataset to testify if the risk model was capable for predicting GC patients’ prognosis. Consistently, both in the two datasets, OS of high-risk subgroups were shorter than that of low-risk subgroups (Fig. 4. A, B). The distribution of GC patients’ vital status and survival time according to risk score were displayed in Fig. 4. C, D, result demonstrated that samples in low-risk subgroup had longer survival time than that in high-risk subgroup, and there were more dead samples in high-risk sub group than in low-risk subgroup. Meanwhile, whenever in univariate Cox regression model or in multivariate Cox regression model, the risk score could be recognized as an independent risk factor of GC patients in training dataset and test dataset (Fig. 4. E, F). What’s more, the ROC curves showed that the risk model had a promising capability to predict of GC patients with the Areas Under the Curve (AUC) of 3-year OS in training and test datasets were 0.633 and 0.668 (Figure S4. A, B), AUC of 5-year OS were 0.562, 0.607 respectively (Figure S4. C, D).

To further validate the model, samples in the two datasets were grouped by clinical prognosis features including age, gender, stage, grade, T and N. OS curves between the low-risk subgroups and high-risk subgroups in above cohorts were compared in Fig. 5. The results indicated that in most of these cohorts, low-risk subgroups had longer OS than high-risk subgroups, however, in some cohorts, there were no statistical significance, these cohorts included stage 1–2 in test dataset, grade 1–2, T 1–2 and N0 in both datasets.

Construction and Validation of Nomogram Model

To obtain a quantitative tool for predicting the OS of GC patients, a nomogram model was built using age, gender, stage, grade, and risk score in training dataset, and was verified in test dataset (Fig. 6). In training cohort, the calibration curves showed strong and acceptable consistency of observed and predicted ratios in 3-year and 5-year OS respectively (Figures S5 A, B). The Decision Curve Analysis (DCA) curves of nomogram indicated that if the threshold probability of 3-year OS was 0.16–0.39, and 5-year OS was 0.1–0.44, the nomogram could offer a higher net benefit than predicting for all patients or no patients (Figures S5, C, D). These results of validation suggested that our nomogram had a strong ability and accuracy in predicting the OS of GC patients.

Relationship of the Risk Model and Immune Cell Infiltration

The DEGs might be participated in various pathways to execute their functions. GO enrichment analysis, a method mainly used to perform enrichment analysis on gene sets (38), here was carried out to investigate the potential biological processes of these DEGs, the result indicated that some of DEGs were enriched in biological processes like epidermis and regulation of peptidase, what’s more, part of them were involved in immunity activities such as defense response (Figure S6 A, B). Thereafter, the training dataset and test dataset were combined, the risk model was checked again in the whole dataset used Kaplan-Meier method, the high-risk subgroup still led to poor OS (Figure S4. E). Subsequently, CIBERSORT algorithm (an analytical tool from the Alizadeh Lab developed by Newman et al. to provide an estimation of the abundances of member cell types in a mixed cell population, using gene expression data) was used to calculate 22 immune cells infiltration proportion of each sample in the whole dataset (Fig. 7. A), results demonstrated that ALKBH5-high subgroup were infiltrated with more naïve B cells, Neutrophils, Plasma cells and follicular helper T cells (Fig. 7. B). Furthermore, high-risk subgroup had more infiltration of naive B cells and resting CD4 T cells, but low-risk subgroup was infiltrated with more activated memory CD4 T cells, CD8 T cells, M1 Macrophages, and follicular helper T cells (Fig. 7. C). The results indicated that the expression of ALKBH5 shaped immune conditions of tumor samples and compared to high-risk subgroup, low-risk subgroup had a better immune microenvironment.

Recent years, as the booming development of diagnosis technology and increased awareness of prevention, GC incidence has decreased rapidly, but the mortality has been staying at high level(1). For patients in advanced stages, surgery often fails to cure, what’s more, the functions of radiotherapy and chemotherapy in GC is limiting, so that safe and effective therapeutic targets are thirsty need of each patient. After the birth of sequencing technology especially the Next-generation sequencing (NGS) (40), the biology processes and modifications to regulate DNA, RNA and proteins were increasingly discovered. As the most prevalent modification of RNA, m6A RNA methylation played important roles in various tumors and regulators of m6A have potency to be therapeutic target of GC.

In GC, a lot of researched have revealed the important roles of m6A regulators, but previous studies mainly focused on the m6A methyltransferase complex at first, especially METTL3(21–23). Though the relationship of “erasers” and patients’ prognosis or pathological features has been discovered more and more(26, 27), there is less reports about what roles does ALKBH5 play in GC. Moreover, some researchers tried to identify biomarkers for predicting the outcome of GC patients using bioinformatics methods. In m6A field, the limit was that they integrated all of the m6A regulator genes or “erasers” together for analysis(41, 42), which made it hard to tell which regulator is responsible and confused readers.

In present study, a new approach to predict GC patients’ prognosis based on ALKBH5 was established and testified. We discovered that ALKBH5 was lowly expressed in GC tumor samples and it significantly decreased the OS of GC. In training dataset, a risk model based on 6 ALKBH5 related genes was constructed. Multiple authentications in training/test datasets indicated us the high-risk subgroup led to poorer OS and risk score was seemed to be an independent risk factor of GC prognosis in Cox regression models. Then, a nomogram model base on risk score and other clinical features was built to predict the 3-year and 5-year OS of GC. GO and CIBERSORT analysis suggested that ALKBH5 related genes might be involved in immune response and shaped the immune cells infiltration of GC samples.

After realized the different expression levels of ALKBH5 between tumor samples and normal samples, we firstly paid attention to it. The risk model and nomogram model were all derived from the DEGs between ALKBH5-low and ALKBH5-high subgroups, thus in GC, not only ALKBH5 itself was an independent protective factor, but also its related genes had probability to predict patients’ outcomes. This is the first model came from a single gene, if the mechanism can be further illustrated, potential targets or therapeutic medicine might come true faster than those biomarkers derived from complex analysis. What’s more, the 6 genes used for constructing risk model were screened from DEGs by LASSO Cox regression, which could provide high prediction accuracy and prevent overfitting. In addition, the adenosine methylation sites in the 6 genes were predicted in SRAMP database, results suggested that there were a lot of sites with very high confidence or high confidence, these sites were the potential targets of ALKBH5.

In our study, there was an absorbing result. When validating the risk model in different clinical cohorts, the OS between low-risk and high-risk subgroups might have no significant difference in early stages, including stage 1–2 in test dataset, grade 1–2, T 1–2 and N0 in both datasets. These results indicated that the risk model was more effective in advanced stages, so the 6 genes screened out might be participated in critical biological process of GC development and had important value in therapy of advanced stage for GC patients. What’s more, this might be the possible reason of higher accuracy in predicting 3-year OS than 5-year OS, because patients in advanced stages had shorter survival time.

Furthermore, GO enrichment analysis in our study revealed that the DEGs related to ALKBH5 were likely involved in immune response. Immunotherapy and immune biomarkers exhibit outstanding value in diagnosis and therapy of tumors, which brought new hope to cancer patients including GC (43, 44). Immune activities are mainly executed by immune cells, results in our study suggested that ALKBH5 changed GC samples immune microenvironment by altering immune cells. Several reports had demonstrated that CD8 T cells, M1 Macrophages and NK cells played an anti-tumor role in GC(45), resting memory CD4 T cells were closely associated with the pathogenesis of GC(46), moreover, the regulation of follicular helper T cells was critical to prevent autoimmunity in cancer(47). In present research, compared with high-risk subgroup, samples in low-risk subgroup were collected with more activated memory CD4 T cells, CD8 T cells, M1 macrophages and follicular helper T cells, indicating better microenvironment in low-risk subgroup. Oppositely, there were more resting memory CD4 T cells in high-risk subgroup which led to poorer prognosis. In addition, the proportions of Neutrophils and Plasma cells were higher in ALKBH5-high subgroup, they emerged as significant but opposite predictors of survival for breast and lung adenocarcinomas(48), so the relationship between GC and them needed further investigating. What’s more, CD4 T cells and Macrophages infiltrated in GC samples could decrease the tumor purity and low tumor purity in GC was associated with unfavorable prognosis and immune-evasion phenotype(49). Conclusively, while our results suggested that ALKBH5 and its related genes might play an important role in GC immune microenvironment and could provide potential targets for immunotherapy in GC, careful consideration should be given before making immunotherapy decisions.

Though present study gave prospective new signatures in GC, there remains some limitations. Firstly, this is a bioinformatic analysis, while the tumor and cells are not in computer and silico, it is need to validate the function of these genes and the immune cells infiltration in vitro and in vivo. Secondly, although we divided samples into training dataset and test dataset to analyze and validate the model, more clinical and sequencing data from other databased and hospitals should be adopted for further analysis. Thirdly, as a retrospective study, there must be selection bias in this study, subsequent prospective studies may further help to verify the signature, especially when gene detection in clinical practice could become convenient and efficient.

In summary, in this study our results suggested that ALKBH5 was lowly expressed in GC and played a role like repressor. We screened the DEGs in ALKBH5-low subgroup/ALKBH5-high subgroup and got 6 genes (CA10, SLC7A2, LINC02303, CGB3, C1QL2, CGB8) to constructed a risk model by LASSO regression. Furthermore, a nomogram was established for predicting survival probability of GC patients. The risk model and nomogram model were validated and showed promising ability for predicting the prognosis. In addition, ALKBH5 and its related genes could alter the proportion of immune cells infiltration and provided potential targets for immunotherapy of GC.

If any of the sections are not relevant to your manuscript, please include the heading and write 'Not applicable' for that section.

Ethics approval and consent to participate

TCGA and GEO belong to public databases. The patients involved in the database have obtained ethical approval. Users can download relevant data for free for research and publish relevant articles. Our study is based on open source data, so there are no ethical issues and other conflicts of interest.

Consent for publication

Not applicable.

Availability of data and materials

The datasets generated and analyzed during the current study are available in the Cancer Genome Atlas database (TCGA) repository (https://portal.gdc.cancer.gov/) and the Gene Expression Omnibus (GEO, https://www.ncbi.nlm.nih.gov/geo/).

Competing interests

The authors declare that they have no competing interests.

Funding

This study is supported by the Founding for Cultivation of Outstanding Youth of the PLA general hospital (2018-JQPY-001) and the National Key Research and Development Program of China (2016YFC0105700).

Authors' contributions

TJ, XG and JL conceived and designed the study. XG and DL refined the research design idea. TJ and XG downloaded the data. TJ and YC analyzed the data. DL, SH and XA further supplemented and validated the data and assisted with the interpretation of the results. TJ, XG and DL drafted the manuscript. SH prepared figure 1. YC and XA revised the logic of the manuscript and polished the language. WJ and SY checked and corrected spelling mistakes and sorted out the references. JL revised and proofread the manuscript. All authors reviewed and approved the final manuscript.

Acknowledgements

This study benefited from the Cancer Genome Atlas (TCGA) database and Gene Expression Omnibus (GEO) database. We appreciate the data platform and the authors uploaded their data.

Footnotes

Not applicable.

Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin. 2021;71(3):209–49.
Cutsem EV, Sagaert X, Topal B, Haustermans K, Prenen H. Gastric cancer. Lancet. 2016:2654–64.
Dubin DT, Taylor RH. The methylation state of poly A-containing-messenger RNA from cultured hamster cells. Nucleic Acids Research. 1975(10):10.
Jia G, Fu Y, Zhao X, Dai Q, Zheng G, Yang Y, et al. N6-methyladenosine in nuclear RNA is a major substrate of the obesity-associated FTO. Nature Chemical Biology. 2011;7(12):885–7.
Yang Y, Hsu PJ, Chen YS, Yang YG. Dynamic transcriptomic m 6 A decoration: writers, erasers, readers and functions in RNA metabolism. Cell Research. 2018.
Ping XL, Sun BF, Lu W, Wen X, Yang X, Wang WJ, et al. Mammalian WTAP is a regulatory subunit of the RNA N6-methyladenosine methyltransferase. Cell Research. 2014;24(002):177–89.
Patil DP, Chen CK, Pickering BF, Chow A, Jackson C, Guttman M, et al. m6A RNA methylation promotes XIST-mediated transcriptional repression. Nature. 2016;537(7620).
Yue Y, Liu J, Cui X, Jie C, Liu J. VIRMA mediates preferential m6A mRNA methylation in 3′UTR and near stop codon and associates with alternative polyadenylation. Cell Discovery. 2018;4(1):10.
Liu J, Yue Y, Han D, Xiao W, Ye F, Liang Z, et al. A METTL3–METTL14 complex mediates mammalian nuclear RNA N6-adenosine methylation. Nature Chemical Biology.
Tang C, Klukovich R, Peng H, Wang Z, Yu T, Zhang Y, et al. ALKBH5-dependent m6A demethylation controls splicing and stability of long 3′-UTR mRNAs in male germ cells. Proc Natl Acad Sci U S A. 2017:201717794.
Dominissini D, Moshitch-Moshkovitz S, Schwartz S, Salmon-Divon M, Ungar L, Osenberg S, et al. Topology of the human and mouse m6A RNA methylomes revealed by m6A-seq. Nature. 2012;485(7397):201.
Xu C, Wang X, Liu K, Roundtree IA, Tempel W, Li Y, et al. Structural basis for selective binding of m(6)A RNA by the YTHDC1 YTH domain. Nature Chemical Biology. 2014;10(11):927–9.
Hsu PJ, Zhu Y, Ma H, Guo Y, Shi X, Liu Y, et al. Ythdc2 is an N6-methyladenosine binding protein that regulates mammalian spermatogenesis. Cell Research. 2017;27(9):1115.
Liu M, Dai Q, Zheng G, He C, Parisien M, Pan T. N6-methyladenosine-dependent RNA structural switches regulate RNA-protein interactions. Nature. 2015;518(7540):560–4.
Alarcón C, Goodarzi H, Lee H, Liu X, Tavazoie S, Tavazoie S. HNRNPA2B1 Is a Mediator of m(6)A-Dependent Nuclear RNA Processing Events. Cell. 2015:S0092867415010247.
Huang H, Weng H, Sun W, Qin X, Shi H, Wu H, et al. Recognition of RNA N6-methyladenosine by IGF2BP proteins enhances mRNA stability and translation. Nature Cell Biology. 2018;20(3):285.
Han J, Wang JZ, Yang X, Yu H, Yang H. METTL3 promote tumor proliferation of bladder cancer by accelerating pri-miR221/222 maturation in m6A-dependent manner. Molecular Cancer. 2019;18(1).
Niu Y, Lin Z, Wan A, Chen H, Liang H, Sun L, et al. RNA N6-methyladenosine demethylase FTO promotes breast tumor progression through inhibiting BNIP3. Molecular cancer. 2019;18(1):46.
Liu T, Wang H, Fu Z, Wang Z, Wang J, Gan X, et al. Methyltransferase-like 14 suppresses growth and metastasis of renal cell carcinoma by decreasing long noncoding RNA NEAT1. Cancer science. 2022;113(2):446–58.
Dixit D, Prager BC, Gimple RC, Poh HX, Rich JN. The RNA m6A Reader YTHDF2 Maintains Oncogene Expression and Is a Targetable Dependency in Glioblastoma Stem Cells. Cancer Discovery. 2020;11(2).
Wang Q, Chen C, Ding Q, Zhao Y, Wang Z, Chen J, et al. METTL3-mediated mA modification of HDGF mRNA promotes gastric cancer progression and has prognostic significance. Gut. 2020;69(7):1193–205.
Yue B, Song C, Yang L, Cui R, Cheng X, Zhang Z, et al. METTL3-mediated N6-methyladenosine modification is critical for epithelial-mesenchymal transition and metastasis of gastric cancer. Molecular cancer. 2019;18(1):142.
Huo F, Zhu Z, Zhu W, Du Q, Liang J, Mou J. METTL3-mediated mA methylation of SPHK2 promotes gastric cancer progression by targeting KLF2. Oncogene. 2021;40(16):2968–81.
Pi J, Wang W, Ji M, Wang X, Wei X, Jin J, et al. FZD7YTHDF1 Promotes Gastric Carcinogenesis by Controlling Translation of. Cancer research. 2021;81(10):2651–65.
Chen X, Liang R, Yi Y, Fan H, Chen M, Zhang J, et al. The mA Reader YTHDF1 Facilitates the Tumorigenesis and Metastasis of Gastric Cancer via USP14 Translation in an mA-Dependent Manner. Frontiers in cell and developmental biology. 2021;9:647702.
Wang D, Qu X, Lu W, Wang Y, Jin Y, Hou K, et al. N-Methyladenosine RNA Demethylase FTO Promotes Gastric Cancer Metastasis by Down-Regulating the m6A Methylation of ITGB1. Frontiers in oncology. 2021;11:681280.
Yang Z, Jiang X, Zhang Z, Zhao Z, Xing W, Liu Y, et al. HDAC3-dependent transcriptional repression of FOXA2 regulates FTO/m6A/MYC signaling to contribute to the development of gastric cancer. Cancer gene therapy. 2021;28:141–55.
Lv Z, Sun L, Xu Q, Xing C, Yuan Y. Joint analysis of lncRNA m(6)A methylome and lncRNA/mRNA expression profiles in gastric cancer. Cancer cell international. 2020;20:464.
Hu N, Ji H. N6-methyladenosine (m6A)-mediated up-regulation of long noncoding RNA LINC01320 promotes the proliferation, migration, and invasion of gastric cancer via miR495-5p/RAB19 axis. Bioengineered. 2021;12(1):4081–91.
Wang J, Li Y, Wang P, Han G, Zhang T, Chang J, et al. Leukemogenic Chromatin Alterations Promote AML Leukemia Stem Cells via a KDM4C-ALKBH5-AXL Signaling Axis. Cell Stem Cell. 2020;27(1).
Zhang S, Zhao B, Zhou A, Lin K, Zheng S, Lu Z, et al. mA Demethylase ALKBH5 Maintains Tumorigenicity of Glioblastoma Stem-like Cells by Sustaining FOXM1 Expression and Cell Proliferation Program. Cancer cell. 2017;31(4):591–606.e6.
Guo X, Li K, Jiang W, Hu Y, Xiao W, Huang Y, et al. RNA demethylase ALKBH5 prevents pancreatic cancer progression by posttranscriptional activation of PER1 in an m6A-YTHDF2-dependent manner. Molecular Cancer. 2020;19(1).
Tang B, Yang Y, Kang M, Wang Y, Wang Y, Bi Y, et al. m6A demethylase ALKBH5 inhibits pancreatic cancer tumorigenesis by decreasing WIF-1 RNA methylation and mediating Wnt signaling. Molecular Cancer. 2020;19.
Jin D, Guo J, Wu Y, Yang L, Wang X, Du J, et al. m6A demethylase ALKBH5 inhibits tumor growth and metastasis by reducing YTHDFs-mediated YAP expression and inhibiting miR-107/LATS2–mediated YAP activity in NSCLC. Molecular Cancer. 2020;19.
Ge L, Zhang N, Chen Z, Song J, Wu Y, Li Z, et al. Level of N6-Methyladenosine in Peripheral Blood RNA: A Novel Predictive Biomarker for Gastric Cancer. Clinical chemistry. 2020;66(2):342–51.
Zhang J, Guo S, Piao H, Wang Y, Wu Y, Meng X, et al. ALKBH5 promotes invasion and metastasis of gastric cancer by decreasing methylation of the lncRNA NEAT1. Journal of physiology and biochemistry. 2019;75(3):379–89.
Yu G, Wang LG, Han Y, He QY. clusterProfiler: an R package for comparing biological themes among gene clusters. Omics: a journal of integrative biology. 2012;16(5):284–7.
Xing Z, Chu C, Chen L, Kong X. The use of Gene Ontology terms and KEGG pathways for analysis and prediction of oncogenes. Biochimica et biophysica acta. 2016;1860(11 Pt B):2725–34.
Zhou Y, Zeng P, Li YH, Zhang Z, Cui Q. SRAMP: prediction of mammalian N6-methyladenosine (m6A) sites based on sequence-derived features. Nucleic Acids Res. 2016;44(10):e91.
Morganti S, Tarantino P, Ferraro E, D'Amico P, Duso BA, Curigliano G. Next Generation Sequencing (NGS): A Revolutionary Technology in Pharmacogenomics and Personalized Medicine in Cancer. Adv Exp Med Biol. 2019;1168:9–30.
Guan K, Liu X, Li J, Ding Y, Li J, Cui G, et al. Expression Status And Prognostic Value Of M6A-associated Genes in Gastric Cancer. Journal of Cancer. 2020;11(10):3027–40.
Xu X, Zhou E, Zheng J, Zhang C, Zou Y, Lin J, et al. Prognostic and Predictive Value of m6A "Eraser" Related Gene Signature in Gastric Cancer. Frontiers in oncology. 2021;11:631803.
Goel G, Sun W. Cancer immunotherapy in clinical practice -- the past, present, and future. Chinese journal of cancer. 2014;33(9):445–57.
Li K, Zhang A, Li X, Zhang H, Zhao L. Advances in clinical immunotherapy for gastric cancer. Biochimica et biophysica acta Reviews on cancer. 2021:188615.
Sammarco G, Varricchi G, Ferraro V, Ammendola M, De Fazio M, Altomare D, et al. Mast Cells, Angiogenesis and Lymphangiogenesis in Human Gastric Cancer. International journal of molecular sciences. 2019;20(9).
Chen J, Chen J, Sun B, Wu J, Du C. Integrative analysis of immune microenvironment-related CeRNA regulatory axis in gastric cancer. Mathematical biosciences and engineering: MBE. 2020;17(4):3953–71.
Vinuesa CG, Linterman MA, Yu D, MacLennan IC. Follicular Helper T Cells. Annual review of immunology. 2016;34:335–68.
Gentles AJ, Newman AM, Liu CL, Bratman SV, Feng W, Kim D, et al. The prognostic landscape of genes and infiltrating immune cells across human cancers. Nat Med. 2015;21(8):938–45.
Gong Z, Zhang J, Guo W. Tumor purity as a prognosis and immunotherapy relevant feature in gastric cancer. Cancer medicine. 2020;9(23):9052–63.

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

Identification of Prognosis Signature and Analysis of the Immune Microenvironment in Gastric Cancer Based on ALKBH5

Status:

Version 1

Abstract

Background

Methods

Results

Conclusions

Figures

Background

Methods

Results

Discussion

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1