A 4-lethal gene prognostic signature correlates with genomic instability and immunotherapy biomarker in stomach adenocarcinoma

doi:10.21203/rs.3.rs-1963103/v1

Download PDF

Research Article

A 4-lethal gene prognostic signature correlates with genomic instability and immunotherapy biomarker in stomach adenocarcinoma

https://doi.org/10.21203/rs.3.rs-1963103/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

The lethal genes manifesting significant prognostic value have not been systematically explored in stomach adenocarcinoma (STAD). The current study aims at evaluating the cell viability-related lethal genes by examining the results of CRISPR-cas9 screening, in an attempt to reveal novel therapeutic targets for STAD. Patient data were gathered from The Cancer Genome Atlas (TCGA), and the lethal genes were identified by conducting genome-scale CRISPR screening of the RNA sequencing data from DepMap. A few genes were selected from STAD lineage preferential dependency candidates derived from DepMap for investigation of genomic alterations, survival outcomes, immune profiles and signaling activities. After narrowing down the processes, four lethal genes were sorted out to construct a risk signature that proves to be a novel and independent prognostic factor. The enrolled patients were subdivided into a low-risk group and a high-risk group in accordance with the risk score. Gene set enrichment analysis demonstrated that certain cancer-related phenotypes were significantly enriched in the high-risk group. Patients in the high-risk group exhibited a low immune score, higher aneuploidy and down-regulated level of PD1 than the low-risk group.

In conclusion, our study of CRISPR-cas9 Screening constructed a robust four-lethal gene prognostic signature, which could be applied as a useful prognostic biomarker and thereby highlighting the potential therapeutic targets for the intervention of STAD.

Stomach adenocarcinoma

TCGA

CRISPR screening

Immune

Biomarker

As the predominant subtype of stomach cancer, Stomach Adenocarcinoma (STAD) is one of the most frequently observed malignancies and continues to be a major cause of cancer-caused mortality on a global scale[1]. In spite of the improvement of early detection and the development of therapies of STAD, patients with STAD are still facing a daunting challenge owing to the dismal prognosis[2]. The development and application of the high-throughput transcriptional sequencing and DNA methylation contribute to the well-established molecular subtyping of STAD[3]. The results of molecular subtyping based on mathematical and statistical methods can help predict the prognosis of patients and their clinical response, nevertheless the substantial genomic heterogeneity between and with each molecular subtyping techniques for STAD should be noted and warrants in-depth exploration[4].

Recently, large scale screening project DepMap based on CRISPR-Cas9 knock-out technique has provided an remarkable thorough landscape of potential essential genes whose existence is critical for the survival of cancer cell. By applying a genome-scale CRISPR-Cas9 technique to deliberately knock out each gene, DepMap could identify the candidate genes that are specific or critical for tumor events[5]. The researchers were able to investigate the target genes responsible for STAD malignancy by employing DepMap and the gene expression signature data derived from STAD patients.

In the present study, we aim to establish a scoring model by subdividing the STAD patients in light of the CRISPR/cas9 screening of essential genes to predict their prognosis and to guide clinical treatment. The subsequent analysis uncovered four lethal genes, which were implicated with clinical outcome, PD1 expression, genomic alteration and immunotherapy response, respectively.

Study design

The data of gene expression, clinical outcomes, and somatic alteration were collected from TCGA and GEO datasets for the subsequent training and validation of the prognosis signature. The cohort information of STAD patients was retrieved form UCSC Xena (https://xenabrowser.net/datapages/?cohort=TCGA%20Stomach%20Cancer%20(STAD)&removeHub=https%3A%2F%2Fxena.treehouse.gi.ucsc.edu%3A443). The other datasets were collected form Gene Expression Omnibus (GEO), we downloaded Genome-wide CRISRP screening of STAD cells from the DepMap portal (https://depmap.org/portal/). The dependency scores for approximately 17000 candidate genes were measured by applying the CERES algorithm. The candidate genes with a CERES score of <-1 in STAD cell lines (n=8) were defined as essential genes. We used the limma package in R to analyze the differentially expressed gene between tumor tissues and the paired normal tissues from TCGA RNAseq data. The candidate genes with FDR of less than 0.05 and a fold change of greater than 1 were considered to meet the criteria of significant upregulation in cancerous tissues.

DepMap

Cancer DepMap (Dependency Map) is an accessible and commonly used website constructed on the basis of large-scale multi-omics screening projects, including Cancer Cell Line Encyclopedia. The top essential genes from STAD cell lines were identified by using DepMap database. The CERES score of each gene derived from the screening experiments was also collected from DepMap.

Definition of Prognostic Gene Signature

Univariate cox regression and multivariate cox regression analyses were conducted to calculate the overall survival in TCGA dataset. Uni-variable Cox regression analysis was performed to categorize the candidate genes. The genes with P values of less than 0.05 were considered as candidate genes, which were then used for multivariate Cox regression analysis to construct the predictive model. The results of multivariate prognostic analysis for essential genes were acquired using Cox package.

A linear combination of the expression profiles of prognostic genes weighted by the estimated regression coefficient was applied for the construction of the prognostic signature, in accordance with the formula below:

Risk score = (expression-valuegene1 × coef-valuegene1 + expression-valuegene2 × coef-valuegene2 + ⋯ + expression-valuegeneN × coef-valuegeneN). The median was used as the cut-off value for the subdivision of subjects into the low-risk group or the high-risk group.

Validating Datasets

We established the predictive performance of the prognostic gene signature in the validation cohort. STAD gene expression datasets deposited at the Gene Expression Omnibus Web site ( http://www.ncbi.nlm.nih.gov/geo/, accession No. GSE15459 and No. GSE119834) were used.

Correlation Analysis of Risk Score and Cancer Markers

Ploidy, homologous recombination deficiency (HRD), loss of heterozygosity (LOH) were downloaded from the UCSC Xena browser. BCR shannon and TCR Shannon were obtained from the previous study[6].

Statiscal Analysis

All data were shown as mean ± standard error of the mean. P value < 0.05 was considered statistically significant between the groups.

Bioinfomatic analysis of TCGA and construction of a risk score model

A total of eight STAD cell lines had dependence scores in the DepMap database. 876 genes in the eight STAD cell lines with an average dependence score less than -1 were defined as oncogenes.

Afterwards, univariate Cox regression analysis was conducted to explore the correlation between gene expression and the OS (overall survival) of STAD dataset from TCGA. 11 prognostic-related lethal genes were identified. To further narrow down the list of lethal genes, multivariate Cox regression was conducted. 4 lethal genes were sorted out from the 11 aforementioned genes as the independent prognostic factors. A risk assessment model based on the four genes was constructed and all the patients were classified into the high-risk group or the low-risk groups by applying the median risk score in the STAD dataset as the cut-off value. The risk scores were calculated, following the formula below:

Risk score = 0.32*COPS8+0.013*PPP1R12A+0.43*SEC61G+0.35*TOMM20

Figure 1A presents the Kaplan-Meier overall survival curves of the high-risk group and low-risk group based on the four-gene signature (log-rank p=0.041). AUC area was 0.667, 0.599 and 0.61 for 1 year, 3 years and 5 years, respectively (Figure 1B). After adjusting the clinicopathological variables, the results of Cox regression demonstrated that risk score was an independent prognostic signature in the TCGA STAD dataset (Figure 1C, D).

Validation of the clinical correlation from GEO datasets

We developed a four-gene signature to indicate the risk score by applying the gene expression data of TCGA in the training dataset. Our results showed that the higher risk score indicates a briefer overall survival. The gene expression and prognosis data of the GEO datasets were analyzed to validate these findings, and confirmed the consistency with the training results of TCGA (Figure 2).

Relationship between tumor grading of STAD and risk genes expression

The four-gene expression level between tumor tissue and the normal adjacent tissues were compared to explore the clinical significance of the four-gene signature in the TCGA STAD dataset. The expression levels of COPS8, SEC61G and TOMM20 were significantly higher in the tumor tissue than the normal tissues (Figure 3A). In terms of tumor grading, the levels of these four genes were significantly elevated in high-grade tumor than in low-grade tumor (Figure 3B).

The association between somatic mutation, immune checkpoints expression, immune cell infiltration, and IRG signature

Genetic mutation exerts a profound impact on STAD prognosis. We found that that certain driver mutations are associated with the status of the four-gene signature. The analysis of MuTect2 mutation annotation files obtained the mutation distribution from the cohort. The patients presenting somatic mutation were classified into a low-risk group or a high-risk group, following the aforementioned criteria of grouping (Figure 4A, 4B). We further explored the effect of genomic instability in STAD. The high-risk patients had a higher aneuploidy score, higher HRD score and higher LOH score than low-risk patients (Figure 5A-C). The patients of high-risk group exhibited large-scale genomic alterations and genome instability, which might be one of the major reasons for their poor prognosis.

PD-1 and immune infiltration are critical clinical factors by virtue of their vital roles in immuno-therapy responses. We compared the differences in these two factors between two STAD subtypes. As shown in Figure 6A, the expression of PD-1 in the low-risk group was significantly reduced than that in the high-risk group. Meanwhile, ESTIMATE algorithm was applied to calculate Immune Score, ESTIMATE Score, and Stromal Score of the two groups. Our analysis indicate that the low-risk group presents significantly higher ESTIMATE and immune score than the high-risk group (Figure 6B). By analyzing the immune cell profiles, we observed that M1 macrophages and T cells follicular helper were higher infiltrated in the high-risk group. Conversely, dendritic cells resting, plasma cells, T cells CD4 memory resting and T cells regulatory Tregs were more highly infiltrated in low-risk group (Figure 6C). Boxplot showed that the expression of most HLA markers and immune checkpoint genes were also higher in the low-risk group than high-risk group (Figure 7A, B), and the violin plot also showed that BCR and TCR score were higher in the low-risk group than high-risk group (Figure 7C, D). The above results suggested that our four-lethal gene prognostic signature might have a positive impact on the immune response of STAD patients.

Next, we analyzed the relationship between risk-score groups and biological functions. The activity of the eukaryotic protein translation, generic transcription pathway increased as the risk score rose, while the activity of the TNF signaling, inflammasomes, IL-3 signaling pathway and toll-like receptor pathway decreased (Figure 8). To further investigate the biological functions and pathways-correlated lethal gene signature, GSEA was carried out between the two groups in the TCGA STAD cohort. We observed that a number of cancer-related gene sets were significantly enriched in the high-risk group, including those related to RNA processing and cell cycle (Figure 9A and 9B). Moreover, the gene sets that are related to immunoregulation, including chemokine signaling pathway, immune system process, immune response and PD1 signaling, were notably enriched in the low-risk group (Figure 9C-F)。

In the present study, we integrate CRISPR-cas9 screening results of STAD, and construct a four-lethal gene signature to divide patient into low-risk group and high-risk group. We confirmed the prognostic value of the prognostic signature consisting of 4 lethal genes in the validation cohort. Among the 4 lethal-related genes in the signature, COPS8 level was markedly increased in cancerous tissue and suggested an unfavorable prognosis[7]. Analogously, the knockdown of COPS8 inhibited the proliferative, migrative and invasive capacities of cutaneous melanoma cells, whereas the overexpression of COPS8 resulted in exactly an opposite biological effect. The high level of COPS8 indicates the shortened survival time of patients, suggesting that COPS8 could contribute to a dismal outcome of cutaneous melanoma patients. COPS8 executes its pro-cancerous effect on cutaneous melanoma via regulating EMT-related genes[7]. Specifically, the COPS8-loss elevates the expression of E-cadherin whereas lowers the levels of N-cadherin, vimentin, and snail. The inhibition of PPP1R12A gene led to suppressed survival and metastasis of CCA cells, and dramatically decreased the proliferation of TFK-1 and HCCC9810 cells[8]. SEC61G, as a subunit of endoplasmic reticulum translocon, is deeply involved in the onset and progression of a myriad of tumors. The aberrantly high expression in SEC61G in breast cancer cells implies the poor prognosis of patients harboring breast cancer. The knockdown of SEC61G stymies the proliferation, migration, invasion of breast cancer cells, yet facilitates the excessive apoptosis of breast cancer cells in vitro. Evidence suggested that E2F1/SEC61G axis regulates glycolysis and chemo-sensitivity of Herceptin in breast cancer cells and that transcription factor E2F1 directly binds to the promoter of SEC61G to mediate its expression in breast cancer cells [9]. The overexpression of Translocase of outer mitochondrial membrane 20 (TOMM20), which play the key role of a receptor for proteins targeted to mitochondria, was reported in various cancers. TOMM20-loss-of-function led to significantly diminished capacities of proliferation, migration, and invasion of colorectal cancer (CRC) cells, whereas TOMM20 overexpression led to an opposite effect [10].

The genetic variants and immune heterogeneity of the prognostic model were also evaluated in the present study. Programmed cell death protein 1 (PD1) is a critical immune checkpoint functioning to attenuate the response of immune checkpoint system in STAD[11]. The enhanced PD1 expression in tumor cells or tumor infiltrating lymphocytes would induce T cell exhaustion, thereby weakening the tumor-specific immune capacity and pressing forward tumor proliferation[12, 13].

The potential benefit of immunotherapies using immune checkpoint inhibitors (ICIs), such as programed delivery of anti-PD-1 antibody, to block immune-inhibitory signals and thereby strengthening antitumor immunity has been acknowledged and even optimized[14]. The use of immune checkpoint PD1/PDL1 blocking agents yields gratifying efficacy with durable response in the treatment of some typically immunogenic tumors, such as melanoma, renal cell carcinoma, bladder cancer and STAD[15]. Therefore, the blockage of PD1 pathway becomes an viable approach for the intervention of advanced and metastatic STAD.

Unfortunately, no greater than 33% of the patients would respond to ICI treatment, and the exploration of predictive and accurate biomarkers is in dire need of effort, the validation of specific biomarkers remains a daunting challenge for the medical community[16]. To find more effective immune-therapy approaches, an advanced understanding of the immune characteristics of STAD is particularly important and necessary. In this study, we retrospectively analyzed the transcriptome data of STAD patients, who were further classified into a low-risk group and a high-risk group. Our results found that the high-risk group exhibited lower ESTIMATE and immune score than the low-risk group. The immune micro-environment is constructed by tumor cells and tumor infiltrating immune cells, which regulate the expression pattern of immune checkpoints[17]. Consistent with the immune score result, PD1 signaling and immune response were enriched in the low-risk group. The above results jointly suggest that the dismal prognosis of the high-risk group could be attributable to the immunosuppressive microenvironment, which acts to press forward the progression of STAD.

In addition, RNA processing, and RNA metabolic process were enriched in the high-risk group. To date, the results of genome-wide analyses have uncovered that ASEs are correlated closely with the progression and prognosis of STAD[18]. The abnormal RNA processing of individual gene is involved in numerous tumorigenic processes, such as proliferation, cell cycle, apoptosis, metastasis, and immune escape[19]. The RNA expression profile and RNA metabolic process heavily depend on these RNA processing factors, which function to mediate temporally and spatially coordinated gene expression[20].

In a nutshell, our study highlights the prognostic value of 4 lethal-related genes in STAD and reveals the prognostic signature to enhance the prognostic prediction of STAD patients. Meanwhile, genetic variants, pathway activation, the clinical outcomes and immune heterogeneity were identified. The current study builds a solid theoretical basis to advance the knowledge the roles of cell viability-related lethal genes and indicates the promising clinical implications of DepMap results.

STAD

stomach adenocarcinoma

TCGA

The Cancer Genome Atlas

GEO

Gene Expression Omnibus

DepMap

Dependency Map

overall survival

ICIs

immune checkpoint inhibitors

KEGG

Kyoto encyclopedia of genes and genomes.

Acknowledgement

Not applicable.

Author contributions

SJF is responsible for the project design. XM collect data from the TCGA and GEO database. XM completed data analysis and manuscript writing. SJF revised the manuscript.

Ethics approval and consent to participate

Not applicable

Availability of Data and Material (ADM)

The RNA sequencing data of human samples were obtained from the UCSC Xena (https://xenabrowser.net/datapages/?cohort=TCGA%20Stomach%20Cancer%20(STAD)&removeHub=https%3A%2F%2Fxena.treehouse.gi.ucsc.edu%3A443). GeneChip data were retrieved from the GEO data repository (http://www.ncbi.nlm.nih.gov/geo/) with the accession numbers GSE15459 and GSE119834. DepMap dataset can be accessed with the https://depmap.org/portal/.

Consent for publication

Not applicable

Competing interests

The authors declare that they have no competing interests.

Funding

Not applicable

Akshatha CR, Bhat S, Sindhu R, Shashank D, Rose Sommano S, Tapingkae W, Cheewangkoon R, Prasad SK: Current therapeutic options for gastric adenocarcinoma. Saudi J Biol Sci 2021, 28(9):5371-5378.
de la Fouchardiere C, Decoster L, Samalin E, Terret C, Kenis C, Droz JP, Coutzac C, Smyth E: Advanced oesophago-gastric adenocarcinoma in older patients in the era of immunotherapy. A review of the literature. Cancer Treat Rev 2021, 100:102289.
Stachler MD, Jin RU: Molecular Pathology of Gastroesophageal Cancer. Surg Pathol Clin 2021, 14(3):443-453.
Mariette C, Carneiro F, Grabsch HI, van der Post RS, Allum W, de Manzoni G, European Chapter of International Gastric Cancer A: Consensus on the pathological definition and classification of poorly cohesive gastric carcinoma. Gastric Cancer 2019, 22(1):1-9.
Zhao W, Li J, Chen MM, Luo Y, Ju Z, Nesser NK, Johnson-Camacho K, Boniface CT, Lawrence Y, Pande NT et al: Large-Scale Characterization of Drug Responses of Clinically Relevant Proteins in Cancer Cell Lines. Cancer Cell 2020, 38(6):829-843 e824.
Thorsson V, Gibbs DL, Brown SD, Wolf D, Bortone DS, Ou Yang TH, Porta-Pardo E, Gao GF, Plaisier CL, Eddy JA et al: The Immune Landscape of Cancer. Immunity 2018, 48(4):812-830 e814.
Sun L: COPS8 in cutaneous melanoma: an oncogene that accelerates the malignant development of tumor cells and predicts poor prognosis. Biosci Biotechnol Biochem 2021, 85(2):242-250.
Deng X, Zuo M, Pei Z, Xie Y, Yang Z, Zhang Z, Jiang M, Kuang D: MicroRNA-455-5p Contributes to Cholangiocarcinoma Growth and Mediates Galangin's Anti-Tumor Effects. J Cancer 2021, 12(15):4710-4721.
Ma J, He Z, Zhang H, Zhang W, Gao S, Ni X: SEC61G promotes breast cancer development and metastasis via modulating glycolysis and is transcriptionally regulated by E2F1. Cell Death Dis 2021, 12(6):550.
Park SH, Lee AR, Choi K, Joung S, Yoon JB, Kim S: TOMM20 as a potential therapeutic target of colorectal cancer. BMB Rep 2019, 52(12):712-717.
Dai L, Huang Z, Li W: Analysis of the PD-1 Ligands Among Gastrointestinal Cancer Patients: Focus on Cancer Immunity. Front Oncol 2021, 11:637015.
Kim W, Chu TH, Nienhuser H, Jiang Z, Del Portillo A, Remotti HE, White RA, Hayakawa Y, Tomita H, Fox JG et al: PD-1 Signaling Promotes Tumor-Infiltrating Myeloid-Derived Suppressor Cells and Gastric Tumorigenesis in Mice. Gastroenterology 2021, 160(3):781-796.
Potzsch M, Berg E, Hummel M, Stein U, von Winterfeld M, Johrens K, Rau B, Daum S, Treese C: Better prognosis of gastric cancer patients with high levels of tumor infiltrating lymphocytes is counteracted by PD-1 expression. Oncoimmunology 2020, 9(1):1824632.
Liu J, Li H, Sun L, Yuan Y, Xing C: Profiles of PD-1, PD-L1, PD-L2 in Gastric Cancer and Their Relation with Mutation, Immune Infiltration, and Survival. Biomed Res Int 2020, 2020:2496582.
Kraehenbuehl L, Weng CH, Eghbali S, Wolchok JD, Merghoub T: Enhancing immunotherapy in cancer by targeting emerging immunomodulatory pathways. Nat Rev Clin Oncol 2021.
Chao X, Liu L, Sun P, Yang X, Li M, Luo R, Huang Y, He J, Yun J: Immune parameters associated with survival in metaplastic breast cancer. Breast Cancer Res 2020, 22(1):92.
Hu F, Wang Y, Wang WD, Gale RP, Wu BY, Liang Y: Improving prediction accuracy in acute myeloid leukaemia: micro-environment, immune and metabolic models. Leukemia 2021, 35(11):3073-3077.
Ye ZS, Zheng M, Liu QY, Zeng Y, Wei SH, Wang Y, Lin ZT, Shu C, Zheng QH, Chen LC: Survival-associated alternative splicing events interact with the immune microenvironment in stomach adenocarcinoma. World J Gastroenterol 2021, 27(21):2871-2894.
Lou S, Zhang J, Zhai Z, Yin X, Wang Y, Fang T, Xue Y: Development and validation of an individual alternative splicing prognostic signature in gastric cancer. Aging (Albany NY) 2021, 13(4):5824-5844.
Lou S, Zhang J, Zhai Z, Yin X, Wang Y, Fang T, Xue Y: The landscape of alternative splicing reveals novel events associated with tumorigenesis and the immune microenvironment in gastric cancer. Aging (Albany NY) 2021, 13(3):4317-4334.

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

A 4-lethal gene prognostic signature correlates with genomic instability and immunotherapy biomarker in stomach adenocarcinoma

Status:

Version 1

Abstract

Figures

Introduction

Method

Result

Discussion

Abbreviations

Declarations

References

Additional Declarations

Status:

Version 1