A tissue microRNA signature that predicts the prognosis of breast cancer in young women

Abstract

Since breast cancers in young women are generally aggressive, young patients tend to be intensively treated with anti-cancer drugs. To optimize the strategy for treatment, particularly in young women, prognostic biomarkers are urgently required. The objective of this study was to identify a tissue microRNA (miRNA) signature that predicts prognosis in young breast cancer patients. Total RNA from 45 breast cancer patients aged <35 years was extracted from formalin-fixed paraffin-embedded (FFPE) tissues and analyzed using miRNA microarrays. Patients were categorized into two groups according to recurrence status within the 5 year period after surgery: recurrence (n = 11) and non-recurrent (n = 34). Histological parameters of hormone receptors and Ki-67 were statistically compared between the two groups. Differentially expressed miRNAs were identified, and their associations with overall survival (OS) were evaluated by log-rank test. The median observation period was 5.8 years for the recurrent group, and 9.1 years for the non-recurrent group. Nine miRNAs were significantly differentially expressed between the recurrent and non-recurrent groups. Receiver Operating Characteristic curve analysis was performed to evaluate the prediction accuracy of the identified miRNAs, and the resultant area under the curve was >0.7. Five of the miRNAs were validated by qRT-PCR, and the expression levels of three of those five (miR-183-5p, miR-194-5p, and miR-1285-5p), both alone and in combination, were associated with OS. In conclusion, we identified three candidate miRNAs that could be used separately or in combination as prognostic biomarkers in young breast cancer patients. This miRNA signature may enable selection of better treatment choices for young women with this disease.

Fig 1. Schematic of selection of candidate miRNA biomarkers of prognosis in young breast cancer patients.

(A) Flow diagram showing derivation of the analytic cohort of patients enrolled in this study. (B) Kaplan–Meier plot of recurrent (n = 11) and non-recurrent (n = 34) groups of young breast cancer patients.

Fig 2. miRNA microarray analysis for 45 breast cancer samples and 16 normal breast epithelial samples.

(A) PCA map for 45 breast cancer samples and 16 normal breast epithelial samples. (B) PCA map for 45 breast cancer samples, classified by cancer subtypes. (C) Heatmap showing miRNAs differentially expressed between breast cancer and normal epithelial control samples (fold change ≥2 and p < 0.05; 102 miRNAs).

Fig 3. Candidate miRNA selection based on miRNA microarrays.

(A) Box plots of nine miRNA candidates with significant differences in expression between recurrence and non-recurrence groups. P-values were calculated using a t-test comparing levels between groups, divided according to recurrence status within 5 years after surgery. The middle line indicates the median, the box the interquartile range, and the whiskers the most extreme data point less than 1.5× the interquartile range away from the box. (B) Hierarchical clustering analysis with heatmap for nine candidate miRNAs, considering recurrence status within 5 years after surgery.

Fig 4. Correlation with quantitative RT-PCR and miRNA microarray data of nine candidate miRNAs.

(A–F) Candidate miRNA expression levels were validated by TaqMan-based miRNA qRT-PCR. Average duplicate qPCR data for each miRNA were used as qPCR expression values. Expression of miR-16 was used as an internal control for the normalization. Expression levels measured by qRT-PCR and microarray are shown in scatter plots. R² was calculated to evaluate the correlation.

Fig 5. ROC curves and Kaplan–Meier plots for five validated miRNAs (miR-183-5p, miR-194-5p, miR-205-5p, miR-375, and miR-1285-5p).

(A) ROC curves of five miRNAs and the three-miRNA combination, with AUC values. Arrows indicate cutoff points. (B) Kaplan–Meier plots of overall survival for five miRNAs and the three-miRNA combination. Based on AUC values and cutoffs, the 45 samples were divided into two groups (high and low expression) for each miRNA, and statistical analysis was performed to calculate p-values using the log-rank test.

Fig 6. ROC curves and Kaplan–Meier plots for the three validated miRNAs (miR-183-5p, miR-194-5p, and miR-1285-5p) in ER-positive and -negative subgroups.

(A) ROC curves of three miRNAs and their combination with AUC values. (B) Kaplan–Meier plots of overall survival for three miRNAs and their combination. Cutoff values for miRNA levels were the same as in Fig 5B. p-values were calculated by the log-rank test.