Immune related genes play a significant role in tumor progression and immunotherapy. An integrative, genome-wide profiling study to establish the immune gene-based signature to predict the clinical prognosis is urgently needed, and the molecular regulatory mechanisms of immune related genes have not been identified. In the present study, we conducted comprehensive analysis and identified 21 survival associated immune gene-based signature with the TCGA dataset as the training set and two independent GEO datasets as the validation set. The large number of ovarian cancer cases for this study facilitated robust and general results. We identified several immune related genes that significantly correlated with the progression of ovarian cancer, and they may be the valuable clinical biomarkers. Moreover, we proposed the prognostic signature based on selected, differentially expressed, and survival-associated immune related genes to predict clinical outcome. The signature was the independent prognostic factor in overall survival. Furthermore, the network analysis showed that immune genes were closely related with transcription factors, and this has been a critical regulatory mechanism for immune response.
Patients with ovarian cancer are at substantial risk for recurrence and chemotherapy resistant. Immunotherapeutic approaches such as personalized antigen-specific immunotherapy have been recognized as curative potential targets[7]. Currently, the immune-based interventions have gained approval in many solid tumors and hematologic malignancies. However, ovarian cancer has the features of extensive malignant and immunologic heterogeneity. New tumor antigens and prediction signature are critically needed to select cases that are at high risk for recurrence and chemotherapy resistant. Previous studies demonstrated that next-generation sequencing or large-scale sequencing analysis is now available to identify the tumor neo-antigens for personalizing cancer immunotherapies[8], but they had the limitation of small sample size and intra-study heterogeneity [9]. Bioinformatics systematic analysis will enable a more in-depth exploration. In this study, we combined gene expression profiling from TCGA dataset, which had relative large samples with 376 cases, and GEO datasets. The 21 immune genes were identified as reliable biomarkers of ovarian cancer prognoses. Besides, exploration of immune gene patterns and survival-associated immune genes with computational biology that are specifically designed to perform analysis across different platforms can minimize the technical or samples bias, providing further and general insights into biomarkers identification. As such, this immune-gene based prognostic signature may serve as a generalized, individualized estimate of survival of ovarian cancer.
Several clinical trials have demonstrated the potential use of cancer immunotherapy in ovarian cancer, but the immunotherapeutic responses are variable in different patients with high cost for treatment. Therefore, predictive biomarkers that can identify treatment response are urgently needed. Significant research on immune relevant prognostic signature proposed by Wen Jiang et al has led to studies to find biomarkers predicting prognoses and immunotherapeutic responses in bladder cancer, but differences exist between Wen’s article and the present study. Firstly for the study cases included, we used the cases including ovarian cancer patients and normal ones, while only patients were included in Wen’s article. Secondly for the identification of immune associated genes, we identified the differentially expressed genes between cancer and normal cases, while Wen was focused on the differentially expressed genes associated with immune infiltration. Thirdly for the application, Wen constructed the tumor immune infiltration–associated gene (TIM) signature that can predict the immunotherapeutic response and reflect the immune cells infiltration, while the gene signature construction based on survival-associated immune genes in this article stratified ovarian cancer patients into two distinct subgroups related with survival outcomes. Therefore, this proposed signature combined the molecular and clinical characteristics, and identified biomarkers to provide a more accurate estimation of overall survival in ovarian cancer. Combinatorial prognostic immune gene-based signature can illuminate how specific genomic aberration types associated with clinical outcome[10]. The correlation between the gene signature and prognostic factors provoked perspectives on the good predictive value of gene signature on distinct grade or stage disease and further on overall survival in OC.
However, immunotherapy can be prevented by tumor immunological function disruption, and the off-target activity of immune-stimulatory factors may result in severe toxicity. Moreover, individual immunotherapy is not efficiency for strong anti-tumor potential, while combinatorial immunotherapy may increase the risk and severity of adverse effects [4]. Thus, finding tumor-mediated immunosuppression or immunostimulation targets is still challenging[11]. Immunomodulatory gene circuit platform is potential for tumor-specific immune-stimulation by de novo cancer-specific promoter synthesis, with RNA-based design and transcription factors encoding. Differentiated TFs were identified and multiple binding motifs for cancer-specific TFs were encoded to generate synthetic OC-specific promoters, resulting in compact and tumor-specific promoters[12]. It is promising that we can identify the specific TFs for promoters encoding. In the present study, we have demonstrated the interaction between transcription factors and immune genes, showing that the majority of poor survival-associated immune genes were positively correlated with the expression of TFs in serous OC. The most critical TFs were CIITA, BATF, VDR, and CBX2. Among them, CIITA has been shown to drive MHC Class II expressing tumor cells as professional antigen presenting cell (APC) performers, thus activating the immune cells and constructing the specific optimal anti-tumor vaccine[13]. BATF can induce the T cell exhaustion during chronic infection, which is characterized by expression of inhibitory receptors and protect cells from excessive immunopathology[14, 15]. Besides, BATF inhibition can ameliorate the pathophysiologic responses in allergic asthma acting as the important transcription factor by regulating T and B-cell differentiation[16]. Vitamin D and the vitamin D receptor (VDR) is important in immunological regulation in disease such as inflammatory bowel diseases (IBD) and human immunodeficiency virus infection by modulating the function of monocytes/macrophages during infection[17, 18]. Furthermore, polycomb chromobox (CBX) proteins, especially CBX2 were down-regulated in macrophage upon viral infection. Cbx2 knockdown or silencing inhibited IFN-β production and played a critical role in antiviral innate immunity[19]. On the basis of the aforementioned findings, the specific TFs for promoters encoding may be readily translated to clinical practice.
Immune cells infiltration is the important features in tumor microenvironment (TME) of ovarian cancer. Early immune response is always presented with multiple types of immune cell infiltration and cell-mediated immunity-associated gene expression alteration. In ovarian cancer, macrophages and T cells are the main infiltrated type, along with neutrophil and NK cells. Previous studies showed that alternatively activated macrophages (M2) and neutrophils possess the pro-tumor roles and T follicular helper cells (Tfh) play an important role in immune cell recruitment. In our study, we showed that the genes of patients with high risk scores were enriched in pro-tumor or anti-inflammatory pathways relating with M2 macrophages and neutrophils, while the genes of patients with low risk scores were enriched in inflammatory pathways relating with M0 macrophages, Tfh cells and NK cells. The association reflected the landscape of immune infiltration in TME of ovarian cancer. The gene signature would be highly related with the TME especially tumor immune microenvironment and could stratify patients into high-infiltrating TME subgroups and low--infiltrating TME subgroups, and it would be an attractive target for prediction of immune infiltration and therapy intervention.
Our study has some advantages. Firstly, we performed analysis with TCGA dataset which showed larger sample size, thus our gene signature was reliable and general. Secondly, the current study was based on the immune-related genes which showed a strong biological background, thus our study has the advantage on other models which screened from RNA-seq or the whole genome profiling, providing the novel biomarkers and targets for early diagnosis and molecule-targeted therapy exploration. Thirdly, our prognostic model had a promising survival prediction ability which was shown in ROC curves, and our signature simplified the complicated effects of immune genes in clinical outcomes and immunotherapy responses, making it easier for prognosis and therapy response prediction.
But our study also showed some limitation. First, we used the datasets from both GEO and TCGA to get more sufficient validation, but it will show some statistic cohort bias and heterogeneity for the difference of platforms and differences in clinical care, clinical setting, and treatment. Second, only overall survival was remained to estimate the association between immune gene signature and clinical outcome to decrease the missing rate. This approach increased statistical power and data integrality, but it is also a limitation insofar as some patients’ information will be lost, and the signature will be more accurate if other survival parameters are included. Third, this study is developed with genes in ImmPort database, and further biological experiments and validation are warranted in ovarian cancer in the future. At the same time, the gene signature was validated in other two independent GEO datasets, but it will be more reliable with prospective cohort study in the future.
In summary, the current study constructed prognostic signature with the immune-related genes, providing a good ability for prognostic prediction. Network analysis revealed the regulatory relationship and the interaction between immune genes and transcription factors, providing the biomarkers for immunomodulators. Prospective and validation studies are necessary for further establishment of prediction accuracy with this gene signature. The network analysis is warranted to be validated to identify the critical role of transcription factors in survival outcome.