Oncogenic signaling pathway-related long non-coding RNAs for predicting prognosis and immunotherapy response in breast cancer.

Background: The clinical outcomes of breast cancer (BC) are unpredictable due to the high level of heterogeneity and complex immune status of the tumor microenvironment (TME). When set up, multiple long non-coding RNA (lncRNA) signatures tended to be employed to appraise the prognosis of BC. Nevertheless, predicting immunotherapy responses in BC is still essential. LncRNAs play pivotal roles in cancer development through diverse oncogenic signal pathways. Hence, we attempted to construct an oncogenic signal pathway-based lncRNA signature for forecasting prognosis and immunotherapy response by providing reliable signatures. Methods: We preliminarily retrieved RNA sequencing (RNA-seq) data from The Cancer Genome Atlas (TCGA) database and extracted lncRNA profiles by matching them with GENCODE. Following this, Gene Set Variation Analysis (GSVA) was used to identify the lncRNAs closely associated with 10 oncogenic signaling pathways from the TCGA-BRCA (breast-invasive carcinoma) cohort and was further screened by the least absolute shrinkage and selection operator Cox regression model. Next, an lncRNA signature (OncoSig) was established through the expression level of the final 29 selected lncRNAs. To examine survival differences in the stratification described by the OncoSig, the Kaplan-Meier (KM) survival curve with the log-rank test was operated on four independent cohorts (n = 936). Subsequently, multiple Cox regression was used to investigate the independence of the OncoSig as a prognostic factor. With the concordance index (C-index), the time-dependent receiver operating characteristic was employed to assess the performance of the OncoSig compared to other publicly available lncRNA signatures for BC. In addition, biological differences between the high- and low-risk groups, as portrayed by the OncoSig, were analyzed on the basis of statistical tests. Immune cell infiltration was investigated using gene set enrichment analysis (GSEA) and deconvolution tools (including CIBERSORT and ESTIMATE). The combined effect of the Oncosig and immune checkpoint genes on prognosis and immunotherapy was elucidated through the KM survival curve. Ultimately, a pan-cancer analysis was conducted to attest to the prevalence of the OncoSig. Results: The OncoSig score stratified BC patients into high- and low-risk groups, where the latter manifested a significantly higher survival rate and immune cell infiltration when compared to the former. A multivariate analysis suggested that OncoSig is an independent prognosis predictor for BC patients. In addition, compared to the other four publicly available lncRNA signatures, OncoSig exhibited superior predictive performance (AUC = 0.787, mean C-index = 0.714). The analyses of the OncoSig and immune checkpoint genes clarified that a lower OncoSig score meant significantly longer survival and improved response to immunotherapy. In addition to BC, a high OncoSig score in several other cancers was negatively correlated with survival and immune cell infiltration. Conclusions: Our study established a trustworthy and discriminable prognostic signature for BC patients with similar clinical profiles, thus providing a new perspective in the evaluation of immunotherapy responses. More importantly, this finding can be generalized to be applicable to the vast majority of human cancers.