Introduction
Ovarian cancer (OC) is the fifth leading cause of cancer death in women which lead to 5% women die of it in 2021. According to statics from American, there are 21,410 new cases and 13,770 deaths in 2021 [
1]. The incidence of OC is considerably lower than the first most common female malignancy breast cancer, but the mortality is three times than breast cancer, even worse, the mortality rate of OC is predicted to be rise significantly in 2040 [
2]. The reason of high mortality rate of OC due mainly to lack of effective screening means and prognosis evaluating tools that result in its diagnosis in the advanced stages and harder to treat [
3].
Over the past decade, precision diagnostics and treatment strategies in ovarian cancer offer opportunity to improve survival [
4]. Meanwhile, advances in precision oncology strategies have increased a need to identify clinically relevant predictive biomarkers within tumours and the best possible candidates for therapies have become more important [
5]. Precision cancer therapies will have more room for improvement as actionable predictive biomarkers are developed.
In clinical, histological cell type is applied as a significant prognostic factor in OC, it is considered significantly related to clinical outcome of OC patients [
6]. Other prognostic factors including clinical factors such as age and parity, biological factors such as multiple gene expressions and pathologic factors such as the presence of ascites and residual disease after surgery [
7]. Since clinical heterogeneity and some subjective reasons, it is still hard to predict the prognosis accurately and objective now. However, as the rapid development of sequencing technologies and bioinformatic algorithms, some researchers attempt to combine multiple molecular biomarkers to established an algorithm to evaluating prognosis accurately and Clinically practically [
8,
9]. In our paper, we hope to construct a risk model based on immune genes for effective prognosis prediction and throw light on targeted therapy.
Epithelial ovarian cancer (EOC) account for more than 95% of OC and was considered as an immunogenic cancer since 55% of patients was found spontaneous anti-tumor immune response [
10]. The strongly link between OC and immune system can be inferred clearly. The immune system played a multifaceted role in OC and was a significant mediator of ovarian carcinogenesis [
11]. Many reports proved that various immune cells and immune gene products played an integrated role in OC progression and associated with prognosis [
12‐
14]. For better evaluating prognosis of OC patients and achieve individually management, besides histological analysis, we attempt to incorporate immune molecular features in the system of prognosis evaluation in this paper. The development of multiple immune-related prognostic markers in OC can benefit for accurate prognosis prediction, new molecular targets identification and personalized immune precision therapy until finally improve the survival rate of OC patients.
Immune-related genes (IRGs) play significant role in immune system. In this study, we aimed to build a novel immune gene signature based on IRGs for risk stratification and provide therapeutic targets in OC patients. The clinical validity and stability of the immune gene-based risk model for prognosis evaluation was validated in OC patients in the TCGA training cohort and ICGC validation cohort respectively. Our study provided an efficient and promising method for prognosis predicting and can give a valuable clue for personalized immunotherapy.
Methods
Data resources
The following 9 ovarian cancer (OC) expression chip datasets: GSE14407, GSE6008, GSE14001, GSE16708, GSE26712, GSE29450, GSE38666, GSE66957, and GSE105437, were downloaded from the Gene Expression Omnibus (GEO) database (
www.ncbi.nlm.nih.gov/geo/) to compare the expression of 2498 immune-related genes (IRGs) in 428 OC and 77 normal ovary tissue specimens. 2498 IRGs were derived from the ImmPort database (
https://www.ImmPort.org/home). The transcriptome RNA-sequencing data and corresponding clinical information of 374 and 93 OC patients were extracted from TCGA database (
https://portal.gdc.cancer.gov/) and ICGC data portal (
https://dcc.icgc.org/), respectively. The TCGA data and ICGC data were applied as a prognostic model training set an external validation set, respectively. Over 20,000 primary cancer samples and matched normal samples were included in the TCGA database, which held over 2.5 petabytes of genomic, epigenomic, transcriptomic, and proteomic data. The ICGC Data Portal is a collaborative effort to describe genetic anomalies in 50 different cancer types.
Data processing and differentially expressed IRGs (DEIRGs) screening
Expression matrix data from GEO database containing 9 datasets from different labs were normalized with Limma R package. We also removed the batch effect between TCGA and ICGC datasets by SVA package in R. The differentially expressed IRGs (DEIRGs) were identified in 428 OC and 77 normal ovary tissue specimens from 9 independent GEO datasets using the limma R package, and the cutoffs were |log2FoldChange| (|log2FC|) > 1 and p < 0.05.
Functional enrichment analysis and protein genes interacting of DEIRGs
The biological function of the DEIRGs was investigated using GO term enrichment analysis and KEGG pathway enrichment analysis through “clusterProfiler” R package. The protein–protein interaction (PPI) network was determined among all DEIRGs using STRING database (
https://string-db.org/).
Construction of immune gene signature related to prognosis by DEIRGs in the TCGA training cohort
First, DEIRGs with prognostic values were screened via univariate cox analysis in the TCGA training cohort. Then, to avoid overfitting, we performed the least absolute shrinkage and selection operator (LASSO) Cox regression analysis with the identified prognostic genes using R package “glmnet” to construct an immune gene signature for OC patients in the TCGA training cohort. The independent variable in the LASSO analysis was the standardized expression matrix of prognostic DEIRGs identified before, and the response variables were survival status and OS of OC patients in the TCGA training cohort. Finally, a prognostic immune gene signature for assessing survival risk of OC patients was finally constructed using the standardized expression levels of independent prognostic DEIRGs and their corresponding regression coefficients. The formula of the risk score for each patient was:
\(\mathrm{The risk score}=\sum_{\mathrm{i}-\mathrm{1,2},\dots ,\mathrm{n}}\mathrm{regression coefficient }(\mathrm{genei})\times \mathrm{expression value of }(\mathrm{genei})\). OC patients were divided into low- and high-risk groups by the median risk score as the threshold.
Evaluation of the immune gene signature in the TCGA training cohort and ICGC validation cohort
The immune gene-based risk model divide OC patients in the TCGA training cohort and the ICGC validation cohort into a high-risk group and low-risk group respectively. The Kaplan–Meier survival curves were plotted by R package “survminer” to compare the survival differences of OC patients in different risk groups. Five-year receiver operating characteristic (ROC) curves of risk score and clinical features were plotted via “survival ROC” package to describe accuracy and performance of the model in the TCGA training cohort and ICGC validation cohorts. The larger of the area under the curve (AUC) of ROC curve, the more accurate of the model. We also exhibited the relationship between survival status of OC patients and risk scores in the training cohort and validation cohort, respectively. Principal component analysis (PCA) and t-SNE analysis were performed by R package “stats” and “Rtsne” to verify the distribution of high-risk and low-risk group patients.
Integrated analysis of the Prognostic Model and Clinical parameters of OC patients
The risk score was compared with the clinical traits to determine whether the risk score was associated with the clinical characteristics of OC patients in both TCGA training cohort and ICGC validation cohorts. Age, grade, pathological stage, and overall survival (OS) time were among the OC clinical data that were obtained from the TCGA database. The age and OS of patients were obtained from ICGC data portal. The relationship between the risk score ang these clinicopathological indexes were evaluated. To identify the independence of our risk score signature, univariate and multivariable Cox regression analyses were performed with R package “survival” in TCGA cohorts to identify independent prognostic indicators among risk score and clinical factors. Age, grade, pathological stage and Risk score were included in TCGA.
Construction of the nomogram for OC patients
A nomogram was generated using the R package “rms” to predict the probability of 1-, 3-, 5- and 10-year OS of OC patients based on the independent prognostic DEIRGs that screened out for building the risk model.
Correlation analysis between immune cells infiltration and immune gene signature
The immune infiltration of OC patients was derived from Tumor Immune Estimation Resource (TIMER) website (
https://cistrome.shinyapps.io/timer/). The association between the abundance of 6 immune infiltrates cells (CD4+ T cells, dendritic cells, CD8+ T cells, B cells, macrophages, and neutrophils) and the immune gene-based risk model were analyzed using R.
Discussion
Ovarian cancer represented 2.5% of all female malignancies, but lead to 5% mortality among all cancer deaths. The high mortality of OC was mainly due to 80% of patients were diagnosed at an advanced stage with extensive peritoneal cavity metastases [
15,
16]. For these patients diagnosed at an advanced stage, surgery and chemotherapy are still the standard of care [
17]. Since the responses of different patients to treatment is diversity, this reminder us to searching for highly reliable prognostic biomarkers. Efficiency prognostic biomarkers are conducive to distinguish patients at different levels of risk, convenient for treatment choice, and facilitate patient counseling [
18].
At present, it had become a hotspot to establish gene signatures based on specific characteristics for prognosis predicting in cancer research [
19,
20]. Immunoediting is a process present in OC, it comprised of cancer cell elimination, equilibrium and escape from immune surveillance, and was a significant element of the immune system [
21]. The immune system plays a significant and complicated role in OC, it has been proved [
22]. Klemi et al confirmed that T cells in colorectal cancer specimens can predicted the outcome more accurately than standard prognostic factors [
23]. Other studies also showed similar results [
24]. These studies proved the significance of the immune response in prognosis. Although there are some researchers want to explore the relationship between OC and immune response from different perspectives, such as using ceRNA that affecting immune infiltration [
25], or using macrophage-related gene [
26] or immune-related gene pairs [
27] to construct a risk model, our study using immune-related genes to expound the relationship between OC and immune response is more immediately and comprehensive. Our risk model was composed of only 5 risk genes, and verified in 2 independent cohort, the novel risk prediction model based on immune-related genes for OC patients was verified the accuracy and clinical validity from several aspects. Our study is a novel research that construct an immune genes signature for prognosis evaluating, and can provide clues to targeted therapy with immune related genes.
With the development of precision genomic medicine, researchers are committed to identify specific and accurate prognostic factors from massive medical data sets with clinical outcomes [
28]. A multigene-based model for prognosis predicting was obviously more precise and robust compared with using a single gene [
29,
30]. To evaluate prognosis by expression of 5 immune genes in OC patients is convenient, efficient, accurate and cost-effective. We constructed an immune genes signature for OC patients for the first time. There are some studies studied the relationship between 5 risk genes (
ANGPT4 [
31],
PLTP [
32]
, A2M [
33]
, CXCR4 [
34] and
MIF [
35]) that composed the immune genes signature and OC. However, our study constructed a risk model using the 5 risk genes firstly, can with the model, we can predict prognosis for OC patients more accurate than only one biomarker. According to our study, risk score may offer correct risk classification as a standalone prognosis factor, according to prognosis analysis on the risk model. Therefore, our nomograms built on DEIRG-based prognostic markers can help OC patients better quantify their risk. Meanwhile, ANGPT4 and PLTP were high risk DEIRGs while A2M, CXCR4 and MIF were low risk DEIRGs, maybe they are potential therapeutic targets, of course more study should be done in future.
Of course, although our risk model based on immune genes can predict the prognosis of OC patients rather good, there are many other factors associated with the prognosis of OC patients, including metabolism, autophagy and so on. Therefore, further prospective studies should be implemented in multicenter clinical trials. All in all, for the first time, this study established and validated a novel immune gene related prognostic model using strict standards. It may contribute to the development of individualized treatments and improve OC patients’ OS.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit
http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (
http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.