3.1. Identification of the miRNAs between lymph node non-metastasis tissues and lymph node metastasis tissues. R software was used to research the gene expression profiles from the GSE100453 and GSE38167. According to the cut-off criteria (P<0.05 and |log2FC|≥1), 30 differentially expressed miRNAs were identified. After analyzing 569 samples of the TCGA database, 80 differentially expressed miRNAs were identified and 6 of them exist in both filter results which were consisted of 3 downregulated and 3 upregulated miRNAs. The result is shown in Table 1.
3.2. Target prediction and GO analysis. The target mRNAs of those 6 differently expressed miRNAs were downloaded from two miRNA target prediction websites (targetscan and miRDB). 499 mRNAs were identified after filtering. The Network between miRNAs (2 upregulated miRNAs and 2 downregulated miRNAs) and target mRNAs was shown in Figure 2. To learn more about the function of these mRNAs, we uploaded them into DAVID to perform GO and KEGG analysis. In the CC ontology, upregulated mRNAs were enriched in ‘adherens junction’ (29 mRNAs), ‘focal adhesion’ (24 mRNAs) and ‘cell-substrate adherens junction’ (24 mRNAs), downregulated mRNAs were enriched in ‘nuclear envelope’ (27 mRNAs), ‘focal adhesion’ (27 mRNAs) and ‘nuclear speck’ (26 mRNAs). In the BP ontology, upregulated mRNAs were enriched in ‘regulation of cytoskeleton organization’ (27 mRNAs), ‘myeloid cell differentiation’ (23 mRNAs), and ‘regulation of microtubule-based process’ (16 mRNAs), downregulated mRNAs were enriched in ‘regulation of gene silencing by miRNA’ (18 mRNAs), ‘regulation of gene silencing’ (17 mRNAs) and ‘regulation of posttranscriptional gene silencing’ (17 mRNAs). In the MF ontology, upregulated mRNAs were enriched in ‘proximal promoter sequence-specific DNA binding’ (22 mRNAs), ‘protein serine/threonine kinase activity’ (20 mRNAs) and ‘enzyme activator activity’ (20 mRNAs), downregulated mRNAs were enriched in ‘protein heterodimerization activity’ (32 mRNAs), ‘transcription coregulator activity’ (31 mRNAs) and ‘protein serine/threonine kinase activity’ (31 mRNAs) (Figure 3).
3.3. KEGG pathways of differently expressed mRNAs. The KEGG pathway analysis identified many significant enriched pathways. Upregulated mRNAs were enriched in ‘Hepatitis B’, ‘Epstein-Barr virus infection’, ‘Kaposi sarcoma-associated herpesvirus infection’, ‘Thyroid cancer’, and ‘Human T-cell leukemia virus 1 infection’, downregulated mRNAs revealed involvement in ‘Non-small cell lung cancer’, ‘Central carbon metabolism in cancer’, ‘Cellular senescence’, ‘Chronic myeloid leukemia’, and ‘Longevity regulating pathway’. (Figure 4).
3.4. Construction of a PPI network. 499 mRNAs were entered into the Metascape website to get interactive data. Then, if the combined score was ≥0.9, we would choose the selected mRNAs to build a PPI network. (Figure 5). In the PPI network, 11 modules, including RBAK, TUBA1C, CCR4, HK1, FBXO17, INTS7, F2RL2, RPS29, FOXJ2, OBSL1, and GPI were identified. The outcomes of the KEGG pathway between modules were related to ‘Oxidative Stress Induced Senescence’, ‘Class A/1’, ‘Antigen processing: Ubiquitination & Proteasome degradation’, ‘RNA polymerase II transcribes snRNA genes’, ‘G alpha (q) signaling events’, and ‘Formation of the ternary complex, and subsequently, the 43S complex’ (Table 2). Based on the key mRNAs and related miRNAs, we constructed a miRNA-mRNA network (Figure 6), and It may become potential therapeutic targets and new biomarkers for lymph node metastasis breast cancer.
3.5. Analysis of the key mRNAs expression in normal tissues and cancer tissues. The human protein atlas was utilized to research the expression of human proteins in different tissues. RBAK, TUBA1C, and HK1 were selected from 11 key mRNAs. After entering them into the database, we found that three mRNAs have a positive strong expression in breast cancer tissues and negative weak expression in normal tissues (Figure 7a). To verify our conclusion, TCGAportal was then used. It is a website that downloads statistics from the TCGA database and contains 1102 breast cancer sample tissues. After inputting relevant mRNAs, we also found that the three mRNAs level was higher in cancer tissues than that in normal tissues (Figure 7b).
3.6. Analysis of the miRNAs and their relationship with breast cancer prognosis. To research the prognosis of patients with breast cancer, we used the Kaplan-Meier Plotter. After uploaded 6 miRNAs, we got 6 survival graphs. The results indicated that overexpression of hsa-miR-4274, hsa-miR-6880-3p, and hsa-miR-670-5p (Figure 8) were related to worse overall survival in patients with breast cancer. However, the expression level of has-miR-149, has-miR-1-3p, and has-miR-30b-3p may have no significant relationship with the overall survival. This suggested that the selected miRNAs may be potential targets.
3.7. Evaluation of the 6-miRNA Signature for overall survival. The AUC of 3 years survival for the 6-miRNA signature achieved 0.809 and the AUC of 5 years survival achieved 0.981, which proved that the model has good performance in predicting the survival risk of breast cancer patients (Figure 9a). Besides, the box diagram also proved our conclusion (Figure 9b).
3.8. Verification of potential biomarker expression by qRT-PCR. 3 miRNAs were verified to have relationship with cancer prognosis, and it was found that miR-223-3p and miR-448 had high reliability, which all target STAT1. Then, the selected biomarkers including hsa-miR-4274, hsa-miR-6880-3p, and hsa-miR-670-5p were validated in breast cancer plasma samples using qRT-PCR analysis. Consistent with the prediction, the results showed that the expression levels of hsa-miR-4274 (P-value = 0.015) and hsa-miR-670-5p (P-value = 0.013) in plasma of lymph node non-metastasis breast cancer patients were obviously lower than that of lymph node metastasis breast cancer patients and hsa-miR-670-5p (P-value = 0.013) were higher in lymph node non-metastasis breast cancer patients plasma (Figure 10). And we also found that hsa-miR-4274 and hsa-miR-670-5p had high reliability, which all target STXBP5L.