Analysis of genomics and immune infiltration patterns of epithelial-mesenchymal transition related to metastatic breast cancer to bone

Abstract

Objective: This study aimed to design a weighted co-expression network and a breast cancer (BC) prognosis evaluation system using a specific whole-genome expression profile combined with epithelial-mesenchymal transition (EMT)-related genes; thus, providing the basis and reference for assessing the prognosis risk of spreading of metastatic breast cancer (MBC) to the bone. Methods: Four gene expression datasets of a large number of samples from GEO were downloaded and combined with the dbEMT database to screen out EMT differentially expressed genes (DEGs). Using the GSE20685 dataset as a training set, we designed a weighted co-expression network for EMT DEGs, and the hub genes most relevant to metastasis were selected. We chose eight hub genes to build prognostic assessment models to estimate the 3-, 5-, and 10-year survival rates. We evaluated the models’ independent predictive abilities using univariable and multivariable Cox regression analyses. Two GEO datasets related to bone metastases from BC were downloaded and used to perform differential genetic analysis. We used CIBERSORT to distinguish 22 immune cell types based on tumor transcripts. Results: Differential expression analysis showed a total of 304 DEGs, which were mainly related to proteoglycans in cancer, and the PI3K/Akt and the TGF-β signaling pathways, as well as mesenchyme development, focal adhesion, and cytokine binding functionally. The 50 hub genes were selected, and a survival-related linear risk assessment model consisting of eight genes (FERMT2, ITGA5, ITGB1, MCAM, CEMIP, HGF, TGFBR1, F2RL2) was constructed. The survival rate of patients in the high-risk group (HRG) was substantially lower than that of the low-risk group (LRG), and the 3-, 5-, and 10-year AUCs were 0.68, 0.687, and 0.672, respectively. In addition, we explored the DEGs of BC bone metastasis, and BMP2, BMPR2, and GREM1 were differentially expressed in both data sets. In GSE20685, memory B cells, resting memory T cell CD4 cells, T regulatory cells (Tregs), γδ T cells, monocytes, M0 macrophages, M2 macrophages, resting dendritic cells (DCs), resting mast cells, and neutrophils exhibited substantially different distribution between HRG and LRG. In GSE45255, there was a considerable difference in abundance of activated NK cells, monocytes, M0 macrophages, M2 macrophages, resting DCs, and neutrophils in HRG and LRG. Conclusions: Based on the weighted co-expression network for breast-cancer-metastasis-related DEGs, we screened hub genes to explore a prognostic model and the immune infiltration patterns of MBC. The results of this study provided a factual basis to bioinformatically explore the molecular mechanisms of the spread of MBC to the bone and the possibility of predicting the survival of patients.

Keywords

• Breast cancer metastases
• Bone metastases
• Differential gene expression
• WGCNA
• Prognostic model
• Immune infiltration pattern