A lactate-related LncRNA model for predicting prognosis, immune landscape and therapeutic response in breast cancer

Abstract

Breast cancer (BC) has the highest incidence rate of all cancers globally, with high heterogeneity. Increasing evidence shows that lactate and long non-coding RNA (lncRNA) play a critical role in tumor occurrence, maintenance, therapeutic response, and immune microenvironment. We aimed to construct a lactate-related lncRNAs prognostic signature (LRLPS) for BC patients to predict prognosis, tumor microenvironment, and treatment responses. The BC data download from the Cancer Genome Atlas (TCGA) database was the entire cohort, and it was randomly assigned to the training and test cohorts at a 1:1 ratio. Difference analysis and Pearson correlation analysis identified 196 differentially expressed lactate-related lncRNAs (LRLs). The univariate Cox regression analysis, least absolute shrinkage and selection operator (LASSO), and multivariate Cox regression analysis were used to construct the LRLPS, which consisted of 7 LRLs. Patients could be assigned into high-risk and low-risk groups based on the medium-risk sore in the training cohort. Then, we performed the Kaplan-Meier survival analysis, time-dependent receiver operating characteristic (ROC) curves, and univariate and multivariate analyses. The results indicated that the prognosis prediction ability of the LRLPS was excellent, robust, and independent. Furthermore, a nomogram was constructed based on the LRLPS risk score and clinical factors to predict the 3-, 5-, and 10-year survival probability. The GO/KEGG and GSEA indicated that immune-related pathways differed between the two-risk group. CIBERSORT, ESTIMATE, Tumor Immune Dysfunction and Exclusion (TIDE), and Immunophenoscore (IPS) showed that low-risk patients had higher levels of immune infiltration and better immunotherapeutic response. The pRRophetic and CellMiner databases indicated that many common chemotherapeutic drugs were more effective for low-risk patients. In conclusion, we developed a novel LRLPS for BC that could predict the prognosis, immune landscape, and treatment response.

Keywords: breast cancer; drug sensitivity; lactate; long non-coding RNA; prognostic signature; tumor immune microenvironment.

FIGURE 1

Identification of the differential expressed LRLs. The volcano plots of the differentially expressed lactate-related genes (A) and differentially expressed lncRNAs (B). (C) The interaction between the differentially expressed lactate-related genes and LRLs.

FIGURE 2

Construction and evaluation of the LRLPS. (A) The univariate Cox regression analysis of LRLs in the training cohort. (B) The cross-validation graph shows the optimal parameter selection with minimum criteria in the LASSO model. (C)The LASSO coefficient profiles of the 14 LRLs. (D) The forest graph showed the results of stepwise multivariable cox proportional hazards regression analysis. (E) The OS curve of the two risk groups. (F) The time-dependent ROC curves of the LRLPS. (G) The risk score, clinical event, and the model genes in the two risk groups. (H) The ROC curves of the risk score and other clinicopathological parameters. The univariate (I) and multivariate (J) Cox regression analyses.

FIGURE 3

Validation of the LRLPS. The OS curve of the two risk groups in test (A) and entire (D) cohorts. The time-dependent ROC curves in test (B) and entire (E) cohorts. The risk score, clinical event, and the model genes in the two risk groups in test (C) and entire (F) cohorts. The ROC curves of the risk score and other clinicopathological parameters in test (G) and entire (J) cohorts. The univariate Cox regression analyses in the test (H) and entire (K) cohorts. The multivariate Cox regression analyses in the test (I) and entire (L) cohorts.

FIGURE 4

Stratification analyses of the prognostic signature. Kaplan-Meier curves indicated the OS of the two risk groups stratified by age (>60 years vs. ≤60 years) (A,B), ER stage (negative vs. positive) (C,D), HER2 stage (negative vs. positive) (E,F), PR stage (negative vs. positive) (G,H), stages (stage I–II vs. stage III-IV) (I,J), AJCC T stage (T I–II vs. T III-IV) (K,L), AJCC N stage (N 0-I vs. T II-III) (M,N), AJCC M stage (M 0 vs. M I) (O,P), respectively.

FIGURE 5

Construction and evaluation of the nomogram. (A) The nomogram for predicting BC patients’ survival probability. (B) The nomogram’s 3-, 5-, and 10-year ROC curves. (C,D,E) The 3-, 5-, and 10-year calibration curves. (F) The 3-, 5- and 10-year DCA curves of the nomogram. (G) DCA curves of clinicopathological factors and the nomogram.

FIGURE 6

Functional enrichment analysis. (A) GO enrichment analysis. (B) KEGG enrichment analysis. (C) The results of GSEA in two risk groups.

FIGURE 7

The immune infiltration and immunotherapy response in the two groups. (A,B) The heatmap and box plots of the proportions of tumor-infiltrating cells in the two risk groups. (C) Comparisons of tumor purity stromal, immune, and ESTIMATE scores between the two risk groups. (D) Correlations between the risk score and tumor purity, stromal, immune, and estimate score. (E) Comparisons of the 27 ICPs in the two risk groups. (F) Comparisons of the proportions of non-responders and responders to ICIs between the two risk groups. (G) Comparison of the risk score between the responders and non-responders. (H–K) Comparison of the IPS between the two risk groups.

FIGURE 8

Association between DNA mutation and prognostic model. Waterfall plots of the top 30 mutated genes in the high-risk (A) and low-risk (B) groups. Comparisons of the mutation status of TP53 (C) and PIK3CA (D) in different risk groups. (E) Comparisons of the TMB between the two risk groups. (F) Correlation between TMB and the risk score. (G) Comparisons of the expression of the lncRNAs in different BRCA1 mutation statuses. (H) Comparisons of the expression of the lncRNAs in different BRCA2 mutation statuses.

FIGURE 9

The sensitivity of chemotherapeutic agents and the prediction of potential drugs. The IC50 values of six chemotherapy and targeted agents in the two risk groups, including 5-Fluorouracil (A), Sorafenib (B), Tamoxifen (C), Temozolomide (D), Temsirolimus (E), and Vinblastine (F). (G) Sensitivity correlation analyses of the LRLs and potential drugs according to the CellMiner Database.