Breast cancer (BRCA) currently has the highest incidence rate among female tumours (1). In most patients, the disease is diagnosed in the early stages, and they show favorable prognosis following surgical resection of the primary tumours. However, once patients experience metastases, it will lead to the majority of BRCA-related deaths. As a type of regional metastasis, lymph node metastasis is less lethal than distant metastasis. However, it is the most common form of metastasis in BRCA patients, an important indicator affecting the efficacy of BRCA treatment, and a definite risk factor affecting long-term prognosis of individuals with BRCA (2).

The immune system is considered a decisive factor in cancer formation and progression (3). As the most important component of the human immune system, the lymphatic system performs the tasks of immunological surveillance and immune regulation (4). Moreover, the lymphatic circulatory system can regulate and maintain homeostasis and mediate lymphatic metastasis of tumours. Tumour-infiltrating lymphocytes (TILs) cross the boundary of oncology and immunology. TILs refer to immune cell populations infiltrating tumour tissues with high immune-related gene (IRG) expression, and a higher proportion of TILs is strongly associated with a higher survival rate of specific BRCA subtypes (5, 6). Several studies have reported TILs in tumours, including BRCA (7, 8), and increased proportions of TILs are associated with HER2 amplification, indicating prolonged survival (9). Increased proportions of TILs in BRCA tissue may suggest favorable responses to neoadjuvant therapy and have considerable predictive significance for adjuvant chemotherapy as well (10). With the development of bioinformatics (11, 12), researchers have started to quantify TILs and uncover personalised immune-related characteristics for the prognosis of different cancers by utilising the expression of IRGs (13–15). In particular, there have been several studies in BRCA that have developed prognostic models based on IRGs characteristics (16–19).

Accumulating evidence demonstrates that the immune system and IRGs might perform an important role in regulating BRCA patients’ treatment responses and long-term survival (7). Because IRGs are potentially correlated with lymphatic metastasis and the correlation between IRGs and lymph node metastatic prognosis in BRCA patients is yet to be systematically evaluated, we specifically selected patients with stage N+M0 BRCA from The Cancer Genome Atlas (TCGA) database, studied the expression profiles of IRGs and their predictive value using systematic bioinformatics analysis, and validated the prognostic model using a validation cohort from GSE20685 and Shandong Cancer Hospital and Institute (SCHI). Subsequently, the relevant regulatory axes in BRCA were explored. Our findings could add to the body of knowledge supporting prognostic biomarkers and treatment strategies for stage N+M0 BRCA.

The occurrence and development of many types of malignancies are associated with disorders of the immune system. IRG activation has been shown in some studies to diminish BRCA recurrence in patients with BRCA (22). Previous studies have shown that immune-related prognostic markers are mostly associated with TILs (23–25). With the development of bioinformatics in recent years, we can now quantify TILs by determining gene expression. In BRCA, models based on TILs have been developed to predict patient prognosis and drug efficacy (26–29). However, the role of IRGs in N+M0 BRCA before this study was unclear; hence, we investigated the role of IRGs in N+M0 BRCA. We first elucidated the expression of IRGs in N+M0 BRCA. We identified 423 differentially expressed IRGs. We constructed an IRG prognostic risk scoring model, according to RNA-seq and clinical data of patients with stage N+M0 BRCA in TCGA, and it enabled us to accurately differentiate patient prognosis based on these criteria. Immunotherapy may be more beneficial to those with a low-risk score. Immune cell infiltration and tumour microenvironment differed significantly between the two risk score subgroups. Age and IRG-based risk score were independent factors affecting N+M0 BRCA patient prognosis. These factors were included in a prognosis nomogram. We also constructed a regulatory axis of the ceRNA network to explore the mechanism of the progression of BRCA.

We first identified the differentially expressed IRGs from breast cancer tissues versus paired normal tissues in the TCGA database, and we considered that screening for DEGs between such paired samples might reduce the bias from individual differences. The protein-protein interaction analysis suggested that CDH1 is a hub gene. Apart from PI3K-Akt and MARK, differentially expressed IRGs were mainly enriched in pathways associated with inflammatory responses and immune responses, including IL-17, T cell receptor, NF-kappa B, and PD-L1 expression. These results indicate that cellular immunity may have a significant impact on the occurrence and progression of BRCA. As a cancer suppressor gene coding for E-cadherin and a type of calcium-dependent cell adhesion protein, CDH1 plays a role in regulating cell adhesion, proliferation, and migration. Hence, CDH1 dysfunction can promote certain biological behaviours of cells, such as invasion and metastasis. Mutation of this gene is closely associated with BRCA, rectal cancer, and other cancer types, and the expression of CDH1 is downregulated in various tumour tissues (30, 31). Jian et al. (32) confirmed that methylation in the CDH1 gene promoter region might cause a reduction in CDH1 expression, and BRCA patients with lower CDH1 expression were more prone to lymph node metastasis and experienced lower OS rate and shorter disease-free survival. The protein encoded by the FGFR3 gene corresponds to an important component of the fibroblast growth factor receptor family that binds to fibroblast growth factor and plays a key role during overall skeletal maintenance as well as development. Aberrant expression of FGFR3 is also commonly observed in various cancers and can directly or indirectly activate various downstream signalling pathways, such as the FGFR3 signalling pathway (33), PI3K-AKT signalling pathway (34), and RAS/RAF/MEK/MAPK pathway (35), which are key mediators of malignant tumorigenesis and progression. According to TCGA and GTEx data, FGFR3 is significantly upregulated in breast tumours. In addition, FGFR3 activation can make BRCA cells less sensitive to drugs, such as fulvestrant and tamoxifen (36), and thus FGFR3 can be a candidate therapeutic target (37). S100B encodes S100 calcium-binding protein B, a molecule associated with tumour metastasis and progression, which regulates the proliferation and metabolism of cancer cells through physical interactions with other molecules. For example, overexpression of S100B leads to enhanced migration and invasion of lung cancer cell lines, thereby promoting brain metastasis (38). Serum S100B and S100B autoantibodies are biomarkers of lung cancer brain metastasis (39). In melanoma, serum S100B concentration was positively correlated with tumour stage and negatively correlated with survival rate (40). Furthermore, elevated serum S100B levels were associated with melanoma metastasis, and lower serum S100B levels were associated with improved survival (41). In the present study, S100B was downregulated in BRCA tissues, which was also corroborated by Yen et al. (42), who also noted that a higher S100B expression predicted good OS in ER-negative BRCA patients and longer metastasis-free survival in all BRCA patients. The expression levels of INHBA and SCG2 in this model were negatively correlated with OS in BRCA patients, and the INHBA gene encodes a member of the transforming growth factor beta superfamily. This gene is significantly overexpressed in BRCA tissues and its high expression status in a variety of tumours, such as colon cancer (43), oesophageal squamous cell carcinoma (44), lung adenocarcinoma (45), and bladder and uroepithelial carcinoma (46), suggests a poor prognostic outcome. SCG2 is a member of neuroendocrine proteins, whose important functions include promoting neointima formation and enhancing endothelial angiogenesis (47). Cury et al. (48) found that SCG2 could be an important indicator to differentiate the progression and prognosis of NSCLC patients.

The expression of five IRGs was subsequently used to build a prognostic risk score model. Similar attempts have been made in studies on tumours, such as BRCA, cervical cancer, osteosarcoma, and Ewing sarcoma (16, 18, 19, 49–54). For example, Zhu et al. (16) and Zhao et al. (18) constructed 12- and 27-gene models of IRGs to predict the prognosis of BRCA patients, respectively. Wang et al. (49) constructed a 5-gene IRG prognostic model for triple-negative breast cancer (TNBC) patients, while Yang et al. (50) combined both hypoxia and immune genes to construct a 6-gene composite prognostic model for the triple-negative subgroup. Tian et al. (19) used an 8-gene IRG model to predict BRCA recurrence, and Tan et al. (51) used a 9-gene IRG model to predict preoperative axillary lymph node metastasis in TNBC. Compared to the above studies, we made the first attempt in N+M0 stage BRCA, and the number of included IRGs was minimal, while ensuring the predictive efficacy of the model, which allowed for a more simplified model. S100B and SCG2 have been reported in the above models, while CDH1, FGFR3, and INHBA are specific to this cohort, which may be related to the characteristics of patients with stage N+M0 BRCA and to the fact that we filtered out some low-expressed genes to facilitate model application and promotion. We also validated the performance of this model using the training and validation cohorts. The results indicated considerable statistical differences in OS between the different risk score subgroups of stage N+M0 BRCA patients and that high risk scores resulted in a higher incidence of patient fatalities. Although the SCHI validation cohort was followed up for 3329 days at most (markedly less than the maximum 8556 days of follow-up for TCGA training cohort), the prognostic model satisfactorily predicted the 1-year, 3-year, and 6-year OS. We also intend to validate the long-term prognostic performance of the model with 10-year or even 20-year follow-up data. Risk score was found to be an independent prognostic factor for individuals with stage N+M0 BRCA in multivariate COX regression analyses. Subgroup analyses suggested that regardless of clinical staging (AJCC stage I-II vs. AJCC stage III), tumour size (pT1-2 vs. pT3-4) or the number of lymph node metastases (pN1 vs. pN2-3), risk score was always an effective tool for forecasting the OS of patients with stage N+M0 BRCA. To further improve the predictive performance, we used mutation data of TCGA cohort to calculate the TMB and MATH of each patient with stage N+M0 BRCA. Previous research has found that cancer patients with high TMB have a better survival (55–57), and MATH is a new approach to characterize intra-tumour heterogeneity. Multiple studies have suggested that high MATH values are correlated with poor cancer prognosis (58–60). Unfortunately, TMB and MATH were not shown to be independent predictive factors in the study; therefore, we integrated age and risk score into the plotted nomogram. The model performed satisfactorily in predicting the 5-year, 10-year, and even 20-year OS, according to the validation results.

Another major finding is that the proportion of TILs in tumour tissue correlated significantly with the expression levels of the five IRGs in the model. In addition, the relationship between risk score and tumour microenvironment further confirmed the role of IRGs in the tumour microenvironment. Immune checkpoint inhibitors may benefit individuals with malignant tumours, including metastatic BRCA, and immunotherapy is becoming a novel treatment option. The level of immune checkpoint molecules (e.g. CTLA4, PD-L1, and PD-1) has been used as a biomarker for predicting immunotherapy efficacy (61, 62). TIDE is a new algorithm that efficiently predicts the efficiency of immune checkpoint inhibitors by combining the two mechanisms of tumour immune escape (rejection reaction and immune dysfunction), outperforming single markers (21). Given the above, we evaluated the efficacy of immunotherapies, demonstrating that immunotherapies may be more beneficial to patients with low risk scores.

When analysing the relationship between IRGs and clinical staging, we noticed that CDH1, as a hub gene, has a correlation with both T_stage and N_stage; its coefficient in the model was up to 1.928, suggesting that CDH1 contributed most to the risk score and might be closely associated with the progression of BRCA. To explore the mechanism by which CDH1 regulates BRCA progression, we constructed a lncRNA-miRNA-mRNA network and established a lncRNA KCNQ1OT1-LINC00665-TUG1/miR-9-5p/CDH1 regulatory axis after strict filtering. Bandini et al. (63) suggested that miR-9-5p can inhibit BRCA cell proliferation by negatively regulating the androgen receptor, whereas TUG1 can regulate EIF5A2 expression by endogenous competition with miR-9-5P, thereby regulating the sensitivity of BRCA cells to doxorubicin (64). Moreover, via modulating miR-379-5p/LIN28B, LINC00665 can promote BRCA cell proliferation, invasion, and migration (65). miR-107 can be regulated by lowering the expression of KCNQ1OT1 to inhibit BRCA cell proliferation and migration (66). Furthermore, we discovered that miR-9-5p correlated with the prognosis of BRCA patients. These lines of evidence suggest that the lncRNA KCNQ1OT1-LINC00665-TUG1/miR-9-5p/CDH1 regulatory axis might play a crucial role in BRCA development. Therefore, additional research is required to validate this finding.

This study has several limitations. First, given that BRCA is a typical multigenic disease, the constructed prognostic model based only on IRGs has an inherent bias. Second, because the clinical data of TCGA training cohort did not include data on molecular subtyping, further studies need to be conducted to confirm whether the predictive performance of the model changes with different molecular subtypes. Finally, the regulatory axis of the ceRNA network needs to be validated in vitro and in vivo, and we are currently conducting studies on the mechanisms involved in the effects of miR-9-5p expression levels on the development and prognosis of breast cancer.

In summary, we developed a five-IRG prognostic scoring model via comprehensive and systemic bioinformatics investigation. The results of the SCHI validation cohort suggested that regardless of clinical staging, tumour size, or the number of lymph node metastases, this model demonstrated good predictive performance in forecasting the prognosis of patients with stage N+M0 BRCA and can determine the sensitivity of patients to immunotherapies, which will be conducive for developing personalised therapeutic strategies and inspiring the research and development of new medications. Our study results also confirmed that the lncRNA KCNQ1OT1-LINC00665-TUG1/miR-9-5p/CDH1 regulatory axis might play an essential role in BRCA progression. Further studies are required to verify these results.

The original contributions presented in the study are included in the article/ Supplementary Material . Further inquiries can be directed to the corresponding authors.

The studies involving human participants were reviewed and approved by The institutional review board of Shandong Cancer Hospital and Institute. The ethics committee waived the requirement of written informed consent for participation.

CT collected the patient’s clinicopathological data and downloaded external data sets. CT followed up on the patient’s survival information, performed qRT-PCR experiments on the SCHI cohort, and prepared the manuscript. XS and YW supervised the research and revised the manuscript. All authors contributed to the article and approved the submitted version.

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.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.