Genomic instability-derived plasma extracellular vesicle-microRNA signature as a minimally invasive predictor of risk and unfavorable prognosis in breast cancer

Abstract

Background: Breast cancer (BC) is the most frequently diagnosed cancer and the leading cause of cancer-associated deaths in women. Recent studies have indicated that microRNA (miRNA) regulation in genomic instability (GI) is associated with disease risk and clinical outcome. Herein, we aimed to identify the GI-derived miRNA signature in extracellular vesicles (EVs) as a minimally invasive biomarker for early diagnosis and prognostic risk stratification.

Experimental design: Integrative analysis of miRNA expression and somatic mutation profiles was performed to identify GI-associated miRNAs. Then, we constructed a discovery and validation study with multicenter prospective cohorts. The GI-derived miRNA signature (miGISig) was developed in the TCGA discovery cohort (n = 261), and was subsequently independently validated in internal TCGA validation (n = 261) and GSE22220 (n = 210) cohorts for prognosis prediction, and in GSE73002 (n = 3966), GSE41922 (n = 54), and in-house clinical exosome (n = 30) cohorts for diagnostic performance.

Results: We identified a GI-derived three miRNA signature (MIR421, MIR128-1 and MIR128-2) in the serum extracellular vesicles of BC patients, which was significantly associated with poor prognosis in all the cohorts tested and remained as an independent prognostic factor using multivariate analyses. When integrated with the clinical characteristics, the composite miRNA-clinical prognostic indicator showed improved prognostic performance. The miGISig also showed high accuracy in differentiating BC from healthy controls with the area under the receiver operating characteristics curve (ROC) with 0.915, 0.794 and 0.772 in GSE73002, GSE41922 and TCGA cohorts, respectively. Furthermore, circulating EVs from BC patients in the in-house cohort harbored elevated levels of miGISig, with effective diagnostic accuracy.

Conclusions: We report a novel GI-derived three miRNA signature in EVs, as an excellent minimally invasive biomarker for the early diagnosis and unfavorable prognosis in BC.

Keywords: Breast cancer; Exosomes; Extracellular vesicle; Genomic instability; microRNA.

Fig. 1

Identification and functional characterization of GI-associated miRNAs in BC patients. a Unsupervised clustering of 522 BC patients based on the expression pattern of 18 DEmiRNAs. The left purple cluster is GS-like group, and the right khaki cluster is GU-like group. Violin diagram of CSPM burden (b), UBQLN4 expression level (c) and aneuploidy scores (d) in the GU-like group and GS-like group. Horizontal lines: median values. Statistical analysis was performed using the Mann–Whitney U test. *P-value < 0.05, **P-value < 0.01, ***P-value < 0.001. e, f. Functional enrichment analysis of KEGG and GO for target genes of miRNA. G. Kaplan–Meier estimates of overall survival of patients in the GU-like group and GS-like group. HRs and 95% CIs for high vs. low miGISig score was estimated using the univariate Cox analysis. P values comparing risk groups were calculated with the log-rank test. BC breast cancer, CIs confidence intervals, GI Genome instability, CSPM cumulative somatic point mutation, GO Gene Ontology, GS genomically stable, GU genomically unstable, HRs Hazard ratios, KEGG Kyoto Encyclopedia of Genes and Genomes

Fig. 2

Development and validation of the miGISig in prognostic risk stratification. Kaplan–Meier estimates of OS or DRFS of patients with low or high miGISig score in the discovery cohort (a), internal testing cohort (b) and GSE22220 cohort (c). HRs and 95% CIs for high vs. low miGISig score were estimated using the univariate Cox analysis. P values comparing risk groups were calculated with the log-rank test. Violin diagram of CSPM burden and aneuploidy scores in the low risk group and high risk group in TCGA-BC cohort (d) and in the GS-like group and GU-like group in TCGA-OV cohort (e). Statistical analysis was performed using the Mann–Whitney U test. *P-value < 0.05, **P-value < 0.01, ***P-value < 0.001. BC breast cancer, CIs confidence intervals, CSPM cumulative somatic point mutation, DRFS Distant relapse-free survival, GI Genome instability, GS genomically stable, GU genomically unstable, HRs Hazard ratios, OS overall survival

Fig. 3

Establishment and performance comparison of CMCPI. Kaplan–Meier estimates of OS or DRFS of patients with low or high CMCPI scores in the discovery cohort (a), internal testing cohort (b) and GSE22220 cohort (c). HRs and 95% CIs for high vs. low miGISig score was estimated using the univariate Cox analysis. P values comparing risk groups were calculated with the log-rank test. The ROC analysis at five years of OS or DRFS for the CMCPI, miGISig, stage and age in the discovery cohort (d), internal testing cohort (e) and GSE22220 cohort (f). BC breast cancer, CIs confidence intervals, CMCPI composite miRNA-clinical prognostic indicator, DRFS Distant relapse-free survival, HRs Hazard ratios, OS overall survival

Fig. 4

Value of miGISig for the diagnosis of BC. Violin diagram of the miGISig miRNAs expression level in healthy controls and BC patients in TCGA-BC cohort (a), GSE73002 cohort (b) and GSE41922 cohort (c). Statistical analysis was performed using the Mann–Whitney U test. ROC curve for the performance of the miGISig in the diagnosis in the TCGA-BC cohort (d), GSE73002 cohort (e) and GSE41922 cohort (f). g Violin diagram of the miGISig miRNAs expression level in healthy controls and BC patients in the in-house clinical exosome cohort. Statistical analysis was performed using the Mann–Whitney U test. h ROC curve for the performance of miR-128 and miR-421 in the clinical exosome cohort. AUC the area under the curve, BC breast cancer, ROC area under the receiver operating characteristic curve. *P-value < 0.05, **P-value < 0.01, ***P-value < 0.001

Fig. 5

Overexpression of miR128-1, miR128-2 and miR421 promotes genomic instability, induces an S-phase arrest and promotes the growth of breast cancer cells. QRT-PCR detection of miR128-1, miR128-2 and miR421 expression in seven breast cancer cells and normal human mammary epithelial cells 184A1 and MCF-10A (a). QRT-PCR detection of miR128-1, miR128-2 and miR421 expression transfected with miRNAs mimics in MDA-MB-231 cells (b). The percentage of multinuclei and micronuclei in MDA-MB-231 cells after overexpression of miR128-1, miR128-2 and miR421 (c). Cell cycle distribution of MDA-MB-231 cells after overexpression of miR128-1, miR128-2 and miR421 (d). Cell growth curve of MDA-MB-231 cells with miR128-1, miR128-2 and miR421 overexpression (e)

Fig. 6

Study flowchart. The study was performed in multicenter cohorts, including TCGA-BC, GSE22220, GSE73002, GSE41922, and in-house clinical exosome cohorts. Genome instability-derived miRNAs signature (miGISig) was identified in the discovery cohort. The miGISig was applied to an internal validation cohort and multiple external validation cohorts to verify its value in prognosis and diagnosis of BC. The effects of the miGISig on BC growth were investigated using the in vitro functional assays. BC breast cancer, GI Genome instability