Exploration of prognosis and immunometabolism landscapes in ER+ breast cancer based on a novel lipid metabolism-related signature

Abstract

Introduction: Lipid metabolic reprogramming is gaining attention as a hallmark of cancers. Recent mounting evidence indicates that the malignant behavior of breast cancer (BC) is closely related to lipid metabolism. Here, we focus on the estrogen receptor-positive (ER+) subtype, the most common subgroup of BC, to explore immunometabolism landscapes and prognostic significance according to lipid metabolism-related genes (LMRGs).

Methods: Samples from The Cancer Genome Atlas (TCGA) database were used as training cohort, and samples from the Molecular Taxonomy of Breast Cancer International Consortium (METABRIC), Gene Expression Omnibus (GEO) datasets and our cohort were applied for external validation. The survival-related LMRG molecular pattern and signature were constructed by unsupervised consensus clustering and least absolute shrinkage and selection operator (LASSO) analysis. A lipid metabolism-related clinicopathologic nomogram was established. Gene enrichment and pathway analysis were performed to explore the underlying mechanism. Immune landscapes, immunotherapy and chemotherapy response were further explored. Moreover, the relationship between gene expression and clinicopathological features was assessed by immunohistochemistry.

Results: Two LMRG molecular patterns were identified and associated with distinct prognoses and immune cell infiltration. Next, a prognostic signature based on nine survival-related LMRGs was established and validated. The signature was confirmed to be an independent prognostic factor and an optimal nomogram incorporating age and T stage (AUC of 5-year overall survival: 0.778). Pathway enrichment analysis revealed differences in immune activities, lipid biosynthesis and drug metabolism by comparing groups with low- and high-risk scores. Further exploration verified different immune microenvironment profiles, immune checkpoint expression, and sensitivity to immunotherapy and chemotherapy between the two groups. Finally, arachidonate 15-lipoxygenase (ALOX15) was selected as the most prominent differentially expressed gene between the two groups. Its expression was positively related to larger tumor size, more advanced tumor stage and vascular invasion in our cohort (n = 149).

Discussion: This is the first lipid metabolism-based signature with value for prognosis prediction and immunotherapy or chemotherapy guidance for ER+ BC.

Keywords: estrogen receptor-positive breast cancer; lipid metabolism; prognostic signature; therapy response; tumor immune microenvironment.

Figure 1

Identification of LMRG expression patterns associated with different prognoses and levels of immune infiltration in ER+ BC in the TCGA cohort. (A, B) GO and KEGG analyses of the 130 identified survival-associated LMRGs showing enrichment of lipid metabolic pathways. The top 20 pathways are presented. (C) Unsupervised consensus clustering showing two lipid metabolism-related clusters in ER+ BC. (D) Heatmap showing distinct expression patterns of these 130 LMRGs in two cluster subtypes. Some important genes are labeled. (E) K-M survival curve of patients stratified by cluster subtype. (F) Immune cell infiltration landscapes of the two cluster subtypes according to CIBERSORT. The abundances of some cell subpopulations significantly differed between the two clusters. **p < 0.01, ***p < 0.001.

Figure 2

Construction of a survival-associated LMRG-based signature for ER+ BC. (A, B) LASSO coefficient profiles and cross-validation via minimum criteria to select significant prognostic LMRGs. (C) Forest plot of univariate Cox regression analysis results showing that the nine lipid metabolism genes used for signature construction were related to poor prognosis. (D) Distributions of risk scores, survival status and gene expression in individual patients from TCGA. As the risk score increases, the number of deaths and gene expression levels also increase. (E) The risk scores of patients who died were higher than those of patients who lived. (F) K-M curve of OS in ER+ BC patients from the TCGA cohort classified based on the risk score. (G) Forest plot showing survival differences between the high- and low-risk groups in subgroups stratified by age, tumor size and lymphatic metastasis. The superior prognosis of the low-risk group was maintained in all subgroups. ***p < 0.001.

Figure 3

A risk-stratification-based clinicopathologic nomogram for OS prediction of patients with ER+ BC. (A, B) Univariate and multivariate Cox analyses of clinicopathologic factors and the risk score in ER+ BC patients in the TCGA cohort. Age, T stage and risk score were independent prognostic indicators. (C) Development of a prognostic nomogram considering the risk score, age and T stage to predict 1-, 3-, and 5-year OS in ER+ BC patients. (D) Calibration curve of the predicted and actual OS values, showing the stable performance of the nomogram. (E) ROC curves of clinicopathologic factors, the risk score, and the nomogram in predicting 5-year OS. The AUC values of each factor are shown.

Figure 4

Analyses of biological processes and pathways related to the LMRG-based signature. (A) Heatmap of the expression levels of all 1034 LMRGs showing distinct expression patterns between the high- and low-risk groups from the TCGA cohort. (B) Correlations between the two clusters and the two risk groups. The majority of patients in Cluster 2 were categorized into the low-risk group. (C) KEGG pathway analysis of the DEGs between the two risk groups, revealing differentially activated pathways. (D, E) Representative KEGG pathways and GO biological processes enriched in the high- and low-risk patients, as determined by GSEA.

Figure 5

Immune microenvironment patterns and immune checkpoint profiles related to the LMRG-based signature in ER+ BC patients in the TCGA cohort. (A) Correlation between the risk score and pantumor immune subtypes. The C4 subtype (lymphocyte depletion) displayed the highest risk score. (B) Comparison of immune cell infiltration levels calculated according to CIBERSORT analysis between the two risk groups. (C) Correlation heatmap showing the correlations between immune cell infiltration levels and the LMRG-based risk score. The risk score was negatively correlated with the immune score estimated by the ESTIMATE algorithm. (D) Correlogram showing the correlations between the risk score and the six metagenes (STAT1, MHC-I, MHC-II, LCK, interferon, and HCK), which reflect inflammatory responses. (E) The expression levels of most immune checkpoints were higher in the low-risk group. *p < 0.05, **p < 0.01, ***p < 0.001.

Figure 6

Association between the risk score and chemotherapy response and drug screening for high-risk tumors. The estimated IC50 values of docetaxel (A), doxorubicin (B), and cisplatin (C) in the low- and high-risk groups. High-risk tumors were more likely to be resistant to chemotherapy. (D) The pCR rate after receipt of neoadjuvant chemotherapy in the two risk groups from the GSE25066 and GSE4779 cohorts. ER+ BC patients with low risk were more likely to achieve pCR. (E, F) K-M survival curves of ER+ BC patients who received chemotherapy stratified by risk score in the TCGA and METABRIC cohorts. (G) CMap analysis of high-risk versus low-risk patients. The DEGs between the two risk groups were uploaded into the CMap database to predict potential drug targets. The top 15 drugs (with negative correlations) for treating high-risk tumors are listed.