Esophageal cancer (EC) is the seventh most common cancer and the sixth leading cause of cancer death worldwide, with a 5-year survival rate of lower than 20%, seriously affecting human health (1, 2). In 2017, the number of global new EC cases and deaths is 473000 and 436000, respectively (3). It is estimated that there will be about 18440 diagnosed EC cases in 2020 causing about 16170 deaths in the United States (4). The treatment of EC mainly focuses on preoperative chemotherapy or postoperative radiotherapy and chemotherapy. However, about 50% EC patients respond unsatisfactorily to the treatment due to resistance of cancer cells to the chemotherapeutic drugs (5). Esophageal squamous cell carcinoma(ESCC) and esophageal adenocarcinoma (EAC) are two pathological types of EC characterized by different distribution, etiology and risk factors (6, 7). In addition, they vary in molecular characteristics and undergo different changes in their specific genes (8).

In recent years, immunotherapy has emerged as a new treatment for various malignant tumors including EC, knowing that it plays a role in immunosuppression in the tumor micro-environment. Several clinical trials have shown that immune checkpoint inhibitors (ICI) nivolumab and Pembrolizumab can be regarded as new standard second-line treatment strategies for EC (9–14). ICIs include anti-PD1, anti-PD-L1, and anti-CTLA-4 (15). However, only a small number of patients can clinically benefit from ICI treatment because of drug resistance, and some patients even deteriorate sharply after immunotherapy. Therefore, it is necessary to identify biomarkers that can accurately predict patients who could benefit from immunotherapy for the sake of providing individualized treatment (9).

At present, microsatellite instability (MSI), PD-L1, tumor mutation burden (TMB), DNA mismatch repair deficiency (dMMR) and TILs factors are the only factors that are confirmed to be able to predict the efficacy of ICI response (16, 17). One study identified that immune-related genes (ABL1, ATF2, ATG5, C6, CD38, HMGB1, ICOSLG, IL12RB2, and PLAU) were significantly associated with overall survival (OS) of ESCC patients (18). With the emergence of open data sets of gene expression, some studies established immune gene prognostic markers for predicting the survival of EC patients (19). A large-scale multicenter retrospective study analyzed and established immune markers based on four genes (SERPINE1, MMP12, PLAUR and Eps8) in predicting pathological complete remission of neoadjuvant radiotherapy and chemotherapy in EC patients, which is believed to lay a foundation for the combination of neoadjuvant radiotherapy and chemotherapy with immunotherapy (20). The study of Lu et al. showed that CD103+CD8+TIL displayed the tissue-resident memory T cell phenotype and showed high expression of immune checkpoint (PD-1, TIM-3) in ESCC. After blocking anti-PD-1, CD103+CD8+TIL induced strong anti-tumor immunity (21). To help find individual immunotherapy targets, some recent studies established comprehensive prognostic indicators of lung squamous cell carcinoma, colorectal cancer and other tumors based on immune genome map analysis (22–29). However, the prognostic value of immune-related genes in ESCC and EAC as two pathological sub-types of EC has not been fully elucidated.

Using TCGA and ImmPort database, we combined the patient’s clinical information with an immune-related genomic map, and found immune-related genes that were significantly related to prognosis with ESCC and EAC respectively. Based on these genes, we constructed individual immune prognostic index models for EC patients, hoping that they could lay a foundation for promoting individualized immunotherapy of EC.

In recent years, immunotherapy has become a research hotspot in cancer therapy. It is necessary to explore the indicators to predict the prognosis of patients on immunotherapy for the sake of screening out immunotherapy beneficiaries. So far, we found no clear consensus biomarker to predict the efficacy of immunotherapy for EC. Few studies have classified the results of immune-related genes in EC according to its histological sub-types. In this study, we analyzed the whole genomes of ESCC and EAC in the TCGA database and established the immune gene prognostic models related to the prognosis of ESCC and EAC based on respective 12 and 11 genes. In addition, we explored the clinical significance of these immune-related genes and their correlations with immune cell infiltration with a view to improve the efficacy of immunotherapy and promote the development of individualized immunotherapy for EC patients.

Genome analysis strongly suggests that ESCC and EAC are different tumor entities, and their genetic change profiles are very different (8, 30). A study on the genomic characteristics of EC found that ESCC and EAC had a set of driving genes that were almost mutually exclusive, indicating that their development was independent (31). Our univariate COX regression analysis showed that 19 and 24 differential immune genes were related to the prognosis of ESCC and EAC respectively. Among them, we mainly analyzed four genes (EGFR and BMP1 in ESCC and IL-1B and MAPT in EAC) that were related to gender, tumor stage, lymph node metastasis, distant metastasis, and other clinical features in EC, supposing that they might play essential roles in the prognosis of EC patients.

The overexpression of epidermal growth factor receptor (EGFR) in ESCC often indicates a poor prognosis (32). It was reported that an ESCC-targeted antibody could improve the radiosensitivity of recurrent ESCC with overexpression of EGFR, suggesting an effective treatment (33). In their in vitro and in vivo model study, Hoi and his team found that chemotherapy upregulated the expression of PD-L1 in ESCC by activating the EGFR/ERK pathway, suggesting that anti-PD-L1 immunotherapy combined with conventional chemotherapy could achieve a better therapeutic effect (5). Another study showed a correlation between the expression of PD-L1 and EGFR in ESCC, as well as a negative correlation between them on tumor cells or tumor-infiltrating immune cells in ESCC (34). However, few studies focused on the relationship between EGFR gene expression in EC and the prognosis of patients undergoing immunotherapy. Our study found that EGFR was an immune differential gene in patients with ESCC, and its down-regulation was associated with poor prognosis. The expression of EGFR in patients without distant metastasis was higher than that in patients with distant metastasis, and its high expression often indicated a better survival outcome. Our study identified that BMP1 was also an immune differential gene related to prognosis in ESCC, and a high expression of BMP1 tended to indicate a lower possibility of distant metastasis. It was reported that the up-regulation of BMP1 may indicate the poor prognosis of gastric cancer (35), osteosarcoma (36), renal clear cell carcinoma (37), and other tumors. Nevertheless, the study of BMP1 in EC has not been reported.

Interleukin1β (IL-1β) is considered an essential regulatory factor that promotes tumor progression, metastasis and immunosuppression (38, 39). In addition, IL could be used as a biomarker for the prognosis of non-small cell lung cancer (40). A team study on the relationship between genetic polymorphisms of IL1A and IL1B and thyroid cancer in the Chinese Han population showed that IL1Ars3783521 was a risk factor for thyroid cancer, and rs3136558 and rs1143623 in the IL1B gene suggested susceptibility to the disease in patients older than 48 years (38). Our pilot study showed that IL-1B was a risk immune-related differential gene for EAC, and its expression was generally higher in females than that in males. It was found in the present study that MAPT expression was associated with poor prognosis in EAC, and its expression in men was higher than that in women. Some studies using the TCGA database found that MAPT gene expression was closely related to survival of patients with low-grade gliomas (41). By analyzing the clinical data of patients with breast cancer, Pan et al. concluded that MAPT-AS1 may be a potential therapeutic target for ER-negative breast cancer related to tumor growth, invasion and drug resistance (42). The prognostic role of MAPT in prostate cancer and childhood blastoma were also proposed in some studies (41).

More importantly, we reported herein two prognostic index models of immune-related genes based on the expression levels and corresponding regression coefficients of specific genes in ESCC and EAC. The risk score of EC patients was calculated to provide reference for predicting the efficacy of immunotherapy in EC patients. According to the score, the patients were divided into a high-risk group and a low-risk group. The results showed that the survival prognosis of the high-risk group of ESCC and EAC was significantly worse than that of the low-risk group, and the higher the risk score, the worse the prognosis. Previous studies explored some markers to predict the efficacy of chemotherapy and radiotherapy in EC patients, but there are few biological indicators to evaluate the efficacy of immunotherapy for the EC subtypes. A study of immune-related genes in EC is similar to ours, but it does not reflect the difference of immune-related genes in different pathological types of EC. In this study, we established a prognostic score model based on nine genes (HSPA6, CACYBP, DKK1, EGF, FGF19, GAST, OSM, ANGPTL3 and NR2F2). We also analyzed the clinical associations of these genes and found that DKK1 was associated with worse T stages of EC. The expression of OSM in patients with tumor stage III and IV was reportedly higher than that in patients with stage I and II (43). Xi et al. developed a prognostic model of EC based on the histological grade, tumor location, baseline PET SUV max and lymph node size and found that it could be used to evaluate whether induction chemotherapy before neoadjuvant radiotherapy and chemotherapy could benefit the patients (44). Wang and his team established a model for predicting the postoperative survival of ESCC by using UBE2C and MGP genes, as well as the clinicopathological factors including the tumor staging and grade (45). A study on the genome of ESCC used data from two clinical centers and found that NFE2L2 may be a tumor suppressor in ESCC. However, its mutation was found to be associated with poor prognosis, and hot spot mutations in the SLC35E2 promoter region also indicated a low survival rate (46).

Besides, We also explored the relationship between our model risk score and immune cell infiltration. The tumor immune microenvironment contains a variety of immune cells including dendritic cell, natural killer cell, macrophage, T and B lymphocyte, all of which can affect the efficacy of ICIs (47). However, we only concluded that the immune differential gene risk score related to prognosis was negatively correlated with B cell infiltration in ESCC. In the study of other immune cells, no correlation was found between the infiltration level and the score of the model. A recent study used tissue microarray of ESCC and EAC to explore changes in immune cell infiltration in the two EC subtypes by the immunohistochemical method. Their results showed that CD45RO+ and CD8+ cells were highly expressed in EC, and the level of invasion in ESCC was higher than that in EAC (48). Guo et al. also created immune-related genes in the prognostic index score in EC. Unlike our results, they found no statistically significant correlation between B cell infiltration and the score, while the positive level of dendritic cell and macrophage neutrophil suggested a higher risk score (43), which is inconsistent with our results. Therefore, we analyzed their research and found that sample differences might cause this inconsistency.

Our work had some potential limitations. First, our research data came from the public databases, which may lack detailed clinical treatment plans and follow-up information in some patients. Second, our study is a retrospective analysis, and it is therefore necessary to verify the prognostic role of these immune-related genes in EC in larger-scale studies. In addition, current work lacked clear evidences that patients with specific scores could or could not benefit from immunotherapy. Finally, we failed to clarify the relationship between immunogenomics, proteomics, and metabonomics and explore the immunobiological mechanism of EC at the molecular level.

We identified the immune differential genes of different EC subtypes by using the public databases for genome analysis and established a model index of immune differential genes related to prognosis. The results of the present study may provide a new idea for individualized immunotherapy for EC patients.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/supplementary material.

YZ and TX conceived and designed the study and assisted in writing the manuscript. ZF and RX performed the data analyses and contributed to the writing of the manuscript. ZC, JX, and YG reviewed the manuscript. All authors contributed to the article and approved the submitted version.

This study was supported by grant from the Natural Science Foundation of Zhejiang Province (No. LY21H280012) and the new-shoot Talents Porgram of Zhejiang province (No. 2021).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.