Breast invasive ductal carcinoma diagnosis with a three-miRNA panel in serum

As the second most common cancer worldwide, breast neoplasms are very common [1]. In the USA in 2017, approximately 316,000 females were diagnosed with breast cancer, comprising over 80% of all detected malignancies [2]. The histological classification of breast tumors is complicated with groups including epithelial, mixed connective tissue and epithelial, nonepithelial and other tumors. Invasive ductal carcinoma (IDC), which is the general name of nonspecific invasive carcinoma belonging to epithelial tumors classification group, is an extremely malignant subtype among breast cancers [3]. IDC is the most common type of breast cancer and a leading cause of cancer-related death in females. Over the past 30 years, in spite of advances in screening, diagnostic technologies and treatments and despite behavioral interventions such as controlling alcohol consumption and maintaining an appropriate weight, the prognosis of IDC patients remains poor [4,5]. Furthermore, it has been shown that 11% of the females with IDC will suffer a recurrence within 5 years of surgery [6]. In spite of the presence of advanced imaging techniques and biopsies, current diagnostic measures have severe limitations such as radical risks, invasive injures or false positives, thus largely accounting for the poor prognosis in these cases [5,7]. Therefore, it is necessary to find a noninvasive method with high sensitivity and specificity for IDC detection.

Since the discovery of miRNAs about 20 years ago, a number of studies have linked the miRNA family to cell growth and disease development [8]. Containing 22–25 nucleotides, miRNAs lack coding ability but have the ability to regulate gene expression. Numerous studies have proven the tight correlation between miRNAs and disease, including cancer. In cancer biology, aberrant expression of miRNAs is frequently observed in tumorigenesis, tumor progression and metastasis. Additionally, it has been demonstrated that miRNAs could monitor the efficacy of therapeutic approaches in breast cancer [9]. More importantly, miRNAs can be detected in tumor tissues and circulating fluid [10], which provides a noninvasive alternative to traditional diagnostic approaches. An increasing number of studies have explored the possibility of evaluating the levels of circulating miRNAs as a biomarker for reflecting the diagnosis or prognosis index [11]. Hence, we evaluated the possibility of utilizing miRNAs for screening IDC.

In this study with 260 participants, a three-cohort experiment was carried out to verify the potential of an miRNA panel for screening IDC using quantitative reverse transcription-PCR (qRT-PCR) and spiked-in normalization method. The primary aim of our research was to determine the diagnostic efficacy of an assembled miRNA panel for screening IDC by assessing the profile of miRNA expression in serum. Furthermore, bioinformatics methods were utilized to generate additional biological information related to target genes.

Materials & methods

Subjects & ethics information

With the approval of the Ethics Committee of Shenzhen Hospital, Peking University, it was ensured that all participants fully understood the use of specimens and signed the informed consent form voluntarily before sample collection. A total of 260 samples were collected between November 2017 and August 2019, comprising 135 IDC patients and 125 healthy controls (HCs). All the participants were females. IDC patients were diagnosed by tissue biopsy, and HCs underwent comprehensive health checks in Shenzhen Hospital, Peking University. Any participant with disturbing factors would be excluded from this study. Table 1 exhibits the clinical characteristics of the participants.

Study procedure

First, 28 candidate miRNAs were selected from the PubMed or Gene Expression Omnibus database with the following search strategy: ((‘Breast Neoplasms’ [Mesh] OR [invasive ductal carcinoma] OR [breast cancer]) AND (‘MicroRNAs’ [Mesh] OR miRNA OR microRNA)). All selected miRNAs were connected with IDC or Breast Neoplasms. Next, a three-cohort experiment (shown in Figure 1) was executed. We set up five IDC pools and three HC pools at the screening cohort. Each pool had an average of five corresponding samples with randomly placed IDC patients to IDC pools, HCs to HC pools. By means of high-throughput qRT-PCR, candidate miRNAs were included in the testing cohort according to their change in expression levels under the cut-off criteria p < 0.05 and fold change (FC) >1.5 or ≤1. 5. Then, in the testing cohort, the expression levels of candidate miRNAs were evaluated further one by one with 30 IDC specimens and 30 HC specimens through qRT-PCR. In the validation cohort, 80 IDC serum samples and 80 HC serum samples were utilized to confirm the differential expression of final candidate miRNAs. Finally, we constructed an miRNA compound panel for verifying the diagnostic value.

Sample disposal & RNA extraction

We drew 10 ml blood from all participants for serum preparation. The blood samples were then centrifuged at 1000× g for 10 min and 15,000× g for 5 min at 4°C. All the samples were processed within 2 h after blood collection. Before miRNA extraction and purification, 2 μl of synthetic C. elegans miR-39 (cel-miR-39) (10 nM/l, RiboBio, Guangzhou, China) was added to each serum sample for normalization. As per the manufacturer’s instructions in the TRIzol LS isolation kit (Thermo Fisher Scientific, MA, USA), we extracted total miRNA and resuspended them in 30 μl RNase-free water. The NanoDrop 2000 spectrophotometer (NanoDrop, DE, USA) was used to determine miRNA concentration and purity. The extracted miRNAs were stored in fresh tubes at -80°C for further analysis.

Quantitative RT-PCR

Before beginning the real-time PCR (rtPCR) in 384-well plates, we prepared specific primers of reverse transcription from Bulge-Loop miRNA qRT-PCR Primer Set (RiboBio, Guangzhou, China) and the rtPCR was performed using a SYBR Green qPCR kit (SYBR Pre-mix Ex Taq II, TaKaRa) on LightCycler 480 Real-Time PCR System (Roche Diagnostics, Mannheim, Germany) under the following conditions: 95°C for 30 s, 95°C for 10 s, 60°C for 20 s and 70°C for 10 s, repeating 30 times. Melt curve analysis was performed to examine the specificity of the PCR outcome. The experiments were repeated at least three times to establish the reliability of the results. The relative expression levels of selected miRNAs were analyzed using the 2^-ΔΔCq method [12].

Bioinformatic analysis

The bioinformatic analysis included miRWalk3.0 (http://mirwalk.umm.uni-heidelberg.de/) for targeted genes prediction, DAVID database (version 6.8; http://david.abcc.ncifcrf.gov/) for Gene Ontology (GO) annotation and Kyoto encyclopedia of genes and genomes (KEGG) pathway analysis. GO was divided into three functional groups: biological process (BP), cellular component (CC) and molecular function (MF).

Statistical analysis

The information on clinical manifestations was expressed as a number or percentage, and as the mean value ± standard deviation if there were continuous variables. We adopted the following corresponding tests or analysis tools to deal with data: Kruskal–Wallis rank test for multiple comparisons among separate independent phases, Students’ t-test or Mann–Whitney test for the difference in expression levels of each miRNA between IDC patients and NCs samples, multiple logistic regression analysis for the construction of miRNA signature, receiver-operating characteristic (ROC) curves and the area under the ROC curve (AUC) for the diagnostic capability of miRNAs panel. SPSS software (SPSS 23.0 Inc., IL, USA), GraphPad Prism 8 (GraphPadSoftware Inc., CA, USA) and Medcalc (Version 19) software were used for statistical analysis. p-value <0.05 was deemed to be statistically significant.

Results

Clinical characteristics of participants

There were a total of 260 study subjects enrolled in our study. One hundred and thirty five of them were patients who had been diagnosed with IDC by histology, based on WHO criteria. Details about tumor size, lymphatic metastasis, TNM stage system, histological grade and status of PR, ER and HER2 are shown in Table 1. One hundred and twenty five HCs enrolled in our study were all without any interfering disease history. Moreover, there was no significant difference between the age of HCs and IDC patients in statistics at each phase (p-values >0.05).

Screening miRNA cohort

Twenty eight miRNAs were tested for their expression levels in pools (five IDC pools and three HCs pools) under the cut-off criteria p < 0.05 and FC >1.5 or ≤1.5 (Figure 2). At the end of screening, 11 miRNAs were selected to progress to the next cohort. Five of them (miR-125b-5p, miR-196b-5p, miR-129-3p, miR-20a-5p, miR-9-5p) were highly expressed in IDC patients compared with HCs. The rest of these miRNAs (miR-1-3p, miR-145-5p, miR-146a-5p, miR-34a-5p, miR-34b-3p, miR-143-3p) were expressed at lower levels. Supplementary Table 1 exhibits the detailed screening result.

Testing miRNAs cohort

qRT-PCR analysis was performed on the serum of 30 IDC patients and 30 HCs for confirming whether the expression level of 11 candidate miRNAs was dysregulated in the testing cohort. As shown in Figure 3, only seven miRNAs (miR-9-5p, miR-34b-3p, miR-1-3p, miR-146a-5p, miR-20a-5p, miR-34a-5p, miR-125b-5p) had eminently aberrant expression between IDC patients and HCs. The filtration of the testing cohort was significant, with a p-value <0.05.

Validation miRNA cohort & ability to diagnose IDC

The serum from rest of the participants, including 80 IDC patients and 80 HCs, was involved in the verification of the diagnostic capability of the seven candidate miRNAs. The relative expression levels of miRNAs and their corresponding ROC curves can be found in Figure 4. MiR-34a-5p was excluded from the study after the validation cohort because its value had no statistical significance. The relative expression levels of the other six miRNAs showed the same outcomes as the testing cohort, with miR-9-5p, miR-20a-5p and miR-125b-5p upregulation and miR-34b-3p, miR-1-3p, miR-146a-5p downregulation, in IDC patients.

The diagnostic value of seven candidate miRNAs was also exhibited in Figure 4 through ROC curve analysis. The AUCs and their corresponding 95% CI are listed as follows: miR-125b-5p: 0.669 (95% CI: 0.590–0.741; Figure 4B), miR-20a-5p: 0.685 (95% CI: 0.606–0.756; Figure 4D), miR-9-5p: 0.767 (95% CI: 0.693–0.830; Figure 4F), miR-1-3p: 0.653 (95% CI: 0.574–0.726; Figure 4H), miR-146a-5p: 0.789 (95% CI: 0.718–0.849; Figure 4J), miR-34a-5p: 0.505 (95% CI: 0.425–0.585; Figure 4L) and miR-34b-3p: 0.680 (95% CI: 0.601–0.751; Figure 4N).

The construction of an miRNA panel for better detection of IDC

There was a low sensitivity and specificity with the detection of a single miRNA. Therefore, it was necessary to develop an miRNA panel for enhancing accuracy. We selected the three most effective miRNAs including miR-9-5p, miR-34b-3p, miR-146a-5p. Combining the data of the three selected miRNAs in the validation cohort together, we drew the ROC curve and set up a stepwise logistic regression model (Figure 5). The AUC of the miRNA panel was 0.880 (95% CI: 0.819–0.926, sensitivity = 86.25%, specificity = 81.25%). The model was expressed by a formula: Logit(P) = 0.211 + 2.711 × miR-9-5p - 2.320 × miR-34b-3p - 2.492 × miR-146a-5p.

Relationship between the relative expression of serum miRNAs & clinical manifestation

The Kruskal–Wallis rank test was used to verify whether the expression levels of the three selected miRNAs (miR-9-5p, miR-34b-3p, miR-146a-5p) were statistically significant among clinical manifestations of the disease. The subject tests included training and validation cohorts. When IDC patients were divided into groups according to size, lymphatic metastasis, TNM stage, histological stage and molecular phenotype in the training and validation cohorts, as shown in Table 2, there was no significant correlation between the expression levels of miR-9-5p and the clinical manifestations, except for lymphatic metastasis. The expression levels of miR-34b-3p had no significant correlation with the clinical manifestations. However, there seemed to be a significant association of miR-146a-5p with the histological grade, PR and HER2 status of IDC.

Bioinformatics analysis of candidate miRNAs

Using the miRWalk 3.0 software, a total of 353 genes related to miR-9-5p, miR-34b-3p, miR-146a-5p were adopted if the prediction of these genes appeared in more than two miRNAs or two databases. Then, selected targeted genes were put into the DAVID database for GO annotation and KEGG pathway enrichment. GO functional annotation including BP, CC and MF showed numerous terms. The top five of each components of GO functional annotation are listed in Figure 6A–C. Figure 6A shows the pathways associated with BP, including transcription, DNA-template (GO:0006351), protein phosphorylation (GO:0006468), positive regulation of cell migration (GO:0030335), negative regulation of apoptotic process (GO:0043066) and positive regulation of cell proliferation (GO:0008284). The details of pathways associated with CC, nucleus (GO:0005634), nucleoplasm (GO:0005654), intracellular (GO:0005622), cell surface (GO:0009986) and synapse (GO:0045202) can be found in Figure 6B. The category of MF included metal ion binding (GO:0046872), DNA binding (GO:0003677), sequence-specific DNA binding (GO:0043565), protein kinase activity (GO:0004672) and androgen receptor binding (GO:0050681) in Figure 6C. Enrichment pathways in KEGG pathway analysis were listed in Figure 6D, including pathways in cancer, Rap1 signaling pathway, MAPK signaling pathway, Wnt signaling pathway and cGMP-PKG signaling pathway.

Discussion

Breast neoplasms are the most prevalent cancer in females, with the second-highest mortality rate. Among breast tumors, IDC is a kind of malignant tissue subtype comprising approximately 80% of cases with poor prognosis. Approximately 11% of women experience a recurrence within 5 years after surgery. In spite of the advances in diagnostic techniques, some drawbacks like radiation risks, detection costs and physical injuries still exist. The primary limitation of these diagnostic measures is low specificity [5]. Therefore, it is necessary to seek a noninvasive approach with high sensitivity and specialty. miRNAs have gained increasing relevance because of their functions in the growth of cancer cells such as hyperplasia and migration [8]. In recent years, several studies including the research of Papachristopoulou et al. [13] and Guo et al. [14] illuminated the potential relationship between miRNA and breast tumors, while their researches only clarify the role of one particular miRNA in breast neoplasms. For other correlative studies, they paid more attention on the relationship between clinicopathological characteristics and miRNAs [15,16]. Other studies, such as that by Bitaraf et al. [11], expounded the use of combined miRNA panel screening for detection of breast neoplasms. However, these studies only focused on one miRNA or clinicopathological manifestation, or breast cancer families. Hence, in the current study, we primarily focused on the diagnosing for IDC with a multi-miRNAs panel and finding the relation between candidate miRNAs and clinicopathological characters in IDC. The aim of this study was to set up an miRNA panel for improving the sensitivity and specificity of the diagnosis value through three rigorous cohorts. Using qRT-PCR and other methods, through the screening, testing and validation cohorts, we constructed a three-miRNA serum panel comprised of miR-9-5p, miR-34b-3p, miR-146a-5p (AUC = 0.880; 95% CI: 0.819–0.926; sensitivity = 86.25%, specificity = 81.25%) for enriching IDC diagnosis methods.

Among three identified miRNAs, miR-9-5p is a complex miRNA that plays different roles in a number of tumors. In our study, miR-9-5p is highly upregulated in the serum of IDC patients. As reported by Naorem et al. [17], miR-9-5p was also overexpressed in breast cancer cells. Additionally, for other cancers, like colorectal cancer, miR-9-5p is similarly expressed at high levels [18]. However, in gastric cancer, miR-9-5p was downregulated [19]. In our study, there was a significant correlation between miR-9-5p expression levels and lymphatic metastasis. Other studies have also confirmed that the overexpression of miR-9 was related to lymph node metastasis in patients with breast cancer, which determines the cellular invasiveness by affecting E-cadherin expression, a protein that is involved in epithelial-to-mesenchymal transition [20,21]. Furthermore, it has been reported that the overexpression of miR-9-5p interferes with the efficacy of chemotherapy by targeting ONECUT2 in breast cancer stemness [22]. Together, these findings indicate that miR-9-5p is closely correlated with the invasiveness and malignancy of breast cancer; this strongly supports the observation that miR-9-5p serves as a noninvasive diagnostic biomarker for patients with IDC.

miR-34b-3p is a member of the miR-34 family, with other members including miR-34a and miR-34c. Some studies have proven that the miR-34b family has tumor-suppressive capabilities [23]. In these studies, the researchers paid great attention to the mechanism of action of miR-34b-5p (the opposite site to miR-34b-3p) in breast cancer. For example, miR-34b-5p affected cell proliferation and apoptosis by regulating the expression of NK1R-FL and NK1R-Tr in breast cancer cells [24]. For miR-34b-3p, a few studies have reported its role in some cancers, but none in breast cancer. To our knowledge, our study is the first to reveal the relationship of miR-34b-3p with IDC. Therefore, there is still room for further studies to focus on the mechanism of gene targeting of mir-34b-3p in IDC. In a cervical cancer cell study, high expression levels of miR-34b-3p restrained the proliferation and migration of cervical cells, but not invasion [25]. In non-small-cell lung cancer, the overexpression of miR-34b-3p targeted cyclin-dependent kinase-4 to inhibit the growth of non-small-cell lung cancer cells [26]. Combined with our observation that miR-34b-3p was downregulated in IDC patient serum, we expect that miR-34b-3p could be also considered as a tumor suppressor and a noninvasive biomarker in IDC.

In recent years, several studies have identified that miR-146 takes part in the life cycle of breast cancer cells. Long et al. [27] found that the overexpression of miR-146a-5p markedly restrained the proliferation, migration and invasion of breast cancer cells by targeting IL-1 receptor-associated kinase 1. Moreover, Liang et al. [28] demonstrated that overexpression of miR-146a-inhibited breast cancer stem-like cell division and the ability to self-renew by virtue of miR-146a/LIN28/Wnt signaling pathway leading to upregulation of Let-7. In summary, a low level of miR-146a-5p could promote the degree of malignancy in breast cancer. Our study showed that miR-146a-5p was downregulated in the serum of IDC patients, suggesting that miR-146a-5p played an important role such as controlling proliferation or invasiveness in IDC. The specific mechanisms of miR-146-5p in IDC need further exploration. In addition, our results showed that the relative expression of miR-146a-5p was significantly different from that of Her-2 status. It has been determined that SNPs lay on the stem of the miR-146a-5p precursor, which was associated with Her-2 and breast cancer survival status, and provided the basis for the role of miR-146a-5p in IDC [29].

Recent studies have focused on the relationship between IDC and miRNAs. Ren et al. [30] illustrated that the MALAT1/miR-216b-5p/Pyridoxine 5′-phosphate oxidase axis was involved in the development of IDC. Additionally, Shams et al. [31] demonstrated that miR-100 affected the clinicopathologic feature of IDC by regulating C-X-C Chemokine Receptor Type-7. To the best of our knowledge, the three selected miRNAs in our study, especially miR-34b-3p, have not been reported in relation to IDC. The relationship between miR-34b-3p and IDC was revealed for the first time in our study. How miR-34b-3p works in IDC remains unknown. However, there are some limitations to our study, such as small sample size and the presence of other functional miRNAs. However, our results indicate that miR-9-5p, miR-34b-3p and miR-146a-5p may have potential in the diagnosis of IDC.

Finally, we performed bioinformatics analysis to explore the probable pathways associated with the three selected miRNAs, such as Rap1 signaling pathway, MAPK signaling pathway, Wnt signaling pathway and cGMP-PKG signaling pathway, which might contribute to the mechanism of IDC progression. As reported by Shah et al. [32], the downregulation of Rap1Gap could lead to breast ductal carcinoma in situ to IDC via ERK/MAPK activation. Additionally, miR-146a was involved in Wnt signaling pathway and affected cell division and self-renewal [28]. Moreover, cGMP-PKG, by virtue of interaction with the actin/myosin-associated protein caldesmon, regulates breast cell migration and invasion [33]. Hence, the KEGG pathway enrichment supported the possibility that the three selected miRNA could be cancer-specific biomarkers of IDC.

Conclusion

Our study uncovered the close relationship between miR-9-5p, miR-34b-3p, miR-146a-5p and IDC. We confirmed that the change of three selected miRNA levels plays a role in the development of IDC. Furthermore, according to the diagnostic value of three selected miRNA (miR-9-5p, AUC = 0.767; miR-34b-3p, AUC = 0.680; miR-146a-5p, AUC = 0.789), we built up a serum three-miRNA panel (AUC = 0.880) to enhance the diagnostic efficacy of IDC. According to KEGG pathway analysis, the probable functional pathways of the three selected miRNAs involved in IDC were Rap1, MAPK, Wnt and cGMP-PKG pathways. Therefore, the serum three-miRNA panel including miR-9-5p, miR-34b-3p and miR-146a-5p has great potential to become a noninvasive biomarker of IDC diagnosis.

Future perspective

The study posits the idea that a combination of miRNAs could have high specificity for the diagnosis of noninvasive diagnosing breast IDC, and moreover, miRNAs could be used for monitoring the progress of treatments or prognosis.