It is expected that approximately 232,670 new cases of invasive breast cancer and 40,000 breast cancer deaths occurred among U.S. women in 2014.1 One in eight women in the United States will develop breast cancer in her lifetime.2 In addition to a breast mass and breast pain, nipple discharge is also a reason that women seek medical advice. Of the patients presenting with nipple discharge, 10–20 % have an underlying malignancy.3,4 Galactography and exfoliative cytology are common diagnostic methods for patients with nipple discharge. Approximately 60 % of patients with nipple discharge show abnormality in galactography.5 The sensitivity of exfoliative cytology was 7 %.6 However, neither diagnostic method has reliable detection sensitivity for atypia or cancer. Thus, novel approaches are urgently needed to improve breast cancer detection via nipple discharge. These methods will complement current diagnostic strategies.

microRNAs (miRNAs) are small, noncoding, single-stranded RNAs ranging between 18 and 22 nucleotides in size. They are typically excised from longer, 60- to 110-nucleotide stem–loop precursors.7,8 Numerous studies have shown that aberrant miRNA expression is associated with the development and progression of various types of cancer.9–13 miRNAs play a conserved and crucial role in fundamental biological processes, including development, differentiation, apoptosis, and proliferation, and they also act predominately as posttranscriptional regulators that either degrade target mRNAs or repress their translation.14 The advancement of the knowledge on miRNAs and the fast-growing number of identified miRNAs have sparked interest in using miRNAs as potential biomarkers. Because they have relatively simple structures without postprocessing modifications, miRNAs can be detected using polymerase chain reaction (PCR).15

Several studies have shown that circulating miRNAs are detectable and stable in serum and plasma.16–23 miRNAs are also detected in body fluids, including blood, saliva, pleural fluid, and urine, as well as human breast milk.17,24–26 However, there have been no reported studies on the presence of miRNAs in nipple discharge or the correlation between breast cancer and specific miRNA levels in nipple discharge. The purpose of this study was to identify and validate miRNAs in nipple discharge and to investigate the role of these miRNAs as novel breast cancer biomarkers.

Materials and Methods

Ethics Statement

Written consent was obtained from each of the patients enrolled onto this study. The study protocol and informed consent were approved by the Ethical Committee of Qilu Hospital of Shandong University.

Sample Collection and Preparation

Twenty-one human nipple discharge samples were collected before surgery in the Department of Breast Surgery of Qilu Hospital of Shandong University between January and June 2014. After surgery, all of the tumors were subjected to a histopathologic diagnosis following the Pathology and Genetics of Tumors of Breast in World Health Organization Classification of Tumors. Nipple discharge samples were collected by a trained surgeon using Eppendorf tubes. The nipple was cleaned with alcohol swabs to remove cellular debris. Nipple discharge was expressed by manual compression. No serious complications occurred. A droplet of nipple discharge was collected by pipette. Usually we were able to collect 20–200 μL from each patient. All of the collected nipple discharge samples were stored in liquid nitrogen until analysis. Cells and large debris from the nipple discharge samples were removed by centrifuging twice at 2000×g for 10 min. The supernatant was then centrifuged at 12,000×g for 30 min to remove cellular debris.

miRNA Microarray and Data Analysis

KangChen Biosciences (China) performed the miRNA microarray analysis. Briefly, total RNA was isolated from nipple discharge using TRIzol LS (Invitrogen, USA) and miRNeasy mini kits (Qiagen, USA) according to the manufacturers’ instructions. After isolating the total RNA, a miRCURY Hy3/Hy5 Power labeling kit (Exiqon, Denmark) was used according to the manufacturer’s guidelines to label miRNA. One microgram of RNA from each sample was 3′ end labeled with a Hy3 fluorescent label using a T4 RNA ligase as follows. RNA in 6.0 μL of water was combined with 3.0 μL of CIP buffer and CIP (Exiqon). The mixture was incubated for 30 min at 37 °C, and the reaction was terminated by incubating the mixture at 95 °C for 5 min. Then 9.0 μL of labeling buffer, 4.5 μL of fluorescent label (Hy3), 6.0 μL of dimethyl sulfoxide, and 6.0 μL of labeling enzyme were added to the mixture. The labeling reaction was incubated for 1 h at 16 °C, and the reaction was terminated with a 15 min at 65 °C incubation. A seventh-generation miRCURY LNA Array (v.18.0) (Exiqon) was used to screen the miRNAs in nipple discharge. The Hy3-labeled samples were hybridized on the miRCURY LNA Array according to the array manual. Briefly, a 25 μL total volume mixture of the Hy3-labeled samples and 25 μL of the hybridization buffer were first denatured for 2 min at 95 °C, incubated on ice for 2 min, and then hybridized to the microarray for 16–20 h at 56 °C in a 12-bay hybridization system (Nimblegen Systems, USA). After hybridization, the array slides were washed several times using a wash buffer kit (Exiqon) and dried by centrifugation for 5 min at 400 rpm. The slides were then scanned using an Axon GenePix 4000B microarray scanner (Axon Instruments, USA). The scanned images were imported using GenePix Pro 6.0 software (Axon Instruments) for grid alignment and data extraction. Replicated miRNAs were averaged, and miRNAs with intensities of ≥30 were chosen for calculating the normalization factor in all samples. The data were normalized using a median normalization. After normalization, significantly differently expressed miRNAs were identified through Volcano Plot filtering. Hierarchical clustering was performed by MEV software (v4.6, TIGR).

Quantitative Reverse-Transcription PCR (qRT-PCR) Analysis

For qRT-PCR, 200 ng of total RNA was used in reverse transcription as previously described.27 The qRT-PCR was performed using Gene Amp PCR System 9700 (Applied Biosystems, USA). The samples were loaded in triplicate, and the results of each sample were normalized to RNU6. All of the primers for reverse transcription and PCR are listed in Supplementary Tables S1 and S2, respectively.

Sequencing Analysis of qRT-PCR Products

All qRT-PCR products were purified by a Universal DNA Purification Kit (Tiangen, China) according to the manufacturer’s instructions. Five sets of purified products were ligated to pMD19-T vectors (Takara, China) according to the user’s manual. The constructed plasmids were verified by sequencing. The results were validated to sequences in microRNA database (http://www.mirbase.org/).

Statistical Analysis

The results were analyzed by SPSS 18.0 software (IBM, USA). A 2-tailed Student’s t test was used to determine statistical significance. Receiver operating characteristic curves were drawn for miRNAs based on sensitivity and specificity. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated. p < 0.05 was considered statistically significant.

Results

An overview of the experimental design and analyses of our study is shown in Fig. 1. Samples from three intraductal carcinoma breast cancer and three intraductal papilloma benign lesions were used for microarray screening. The dot map is shown in Fig. 2a. Of the 3100 human miRNAs, 51 miRNAs were detected to be up-regulated more than 1.5-fold. A heat map analysis showed the levels of differentially expressed miRNAs in different samples (Fig. 2b). Meanwhile, ten miRNAs were found to be at least 1.5-fold down-regulated (Fig. 3a, b). Among these differentially expressed miRNAs, the expression of six miRNAs were identified as significantly different between the carcinoma patients and benign controls (p < 0.05)  (Table 1). A further analysis with a 2-tailed Student’s t test confirmed that miR-4484, miR-K12-5-5p, and miR-3646 were significantly up-regulated and that miR-4732-5p, miR-4687-3p, and miR-S1-5p were significantly down-regulated in the carcinoma group compared to the benign group (Fig. 3c). miRNA expression was normalized to small nuclear RNA (RNU6), and a mean expression level was calculated. To validate the microarray result, qRT-PCR was performed using the remaining RNA from six patient samples. All of the miRNAs, except miR-S1-5p, were detected by qRT-PCR. As shown in Fig. 3d, the levels of four miRNAs (miR-4484, miR-K12-5-5p, miR-3646, and miR-4732-5p) were significantly different (p < 0.05) between the breast cancer and benign groups, which was consistent with the microarray data.

Fig. 1
figure 1

Study design flowchart. miRCURY was used for screening candidate miRNAs, and qRT-PCR was used for validation

Full size image
Fig. 2
figure 2

a microRNA microarray graphs of the breast cancer (BCG) and control group (CG). b Heat map diagram showing hierarchical clustering of differentially expressed miRNAs in the breast cancer group compared to the control group. The color scale shown at the top illustrates the relative expression of miRNAs, with red for relatively high and green for relatively low expression

Full size image
Fig. 3
figure 3

a Fifty-one miRNAs were detected to be up-regulated more than 1.5-fold (the ratio of normalized intensities between the 2 groups). b Ten miRNAs were down-regulated by at least 1.5-fold in the breast cancer group compared to the control group. c A total of 6 miRNAs were differentially expressed between the breast cancer group and the benign control group (p ≤ 0.05). d Validation of selected miRNAs by qRT-PCR of the 3 breast cancer samples and 3 benign control samples (*p < 0.05). e Average levels of nipple discharge RNU6 were not significantly different (p > 0.05) between the breast cancer group (BCG) and control group (CG)

Full size image
Table 1 List of miRNA candidates
Full size table

Because the expression of miR-S1-5p was not detectable with qRT-PCR, we used the five remaining miRNAs (miR-4484, miR-K12-5-5p, miR-3646, miR-4687, and miR-4732-5p) as potential tumor biomarkers in the subsequent experiments. The validation sample set has 21 patient samples from 8 carcinomas and 13 benign tumors, including the 6 samples used for microarray analysis. Supplementary Table S3 shows the baseline characteristics of the patients and validation samples. The average ages of the breast cancer patients were 44.4 years (range 30–58 years) and 47.3 years (range 37–64 years), respectively. Other clinical information about the patients and samples, including age at diagnosis, unilateral side, nipple discharge color, and nipple discharge traits are shown in Supplementary Table S3.

The expression levels of miR-4484, miR-K12-5-5p, miR-3646, miR-4732-5p, and miR-4687-3p were examined in the 21 nipple discharge samples (the validation set) using qRT-PCR. The distributions of the ∆Ct values for each candidate miRNA in both the cancer and control groups are shown in Fig. 4. The sensitivity, specificity, PPV, and NPV of each miRNA are shown in Supplementary Table S4. They were calculated on the basis of the receiver operating characteristic curve in Supplementary Fig. S1. miR-3646 and miR-4732-5p showed the best sensitivity (100 %), specificity (100 %), PPV (100 %), and NPV (100 %) compared to other miRNAs. The results indicate that all three up-regulated miRNAs identified in the microarray profiling were significantly higher in the carcinomas than the benign tumors (Fig. 4a–c; p < 0.05). miR-4732-5p was the only miRNA candidate with significantly lower expression in the carcinoma group than the benign group (Fig. 4d; p < 0.001). In the screening set, miR-4687-3p expression was significantly lower in carcinomas compared to benign tumors. However, the expression level of miR-4687-3p in the validation set was scattered over a wide range, and no significant difference between the malignant and benign groups was observed (p = 0.519; Fig. 4e).

Fig. 4
figure 4

Validation of selected miRNAs in a different set of nipple discharge samples. A–E The level of selected miRNAs in each sample was normalized to the internal control, RNU6, and shown as the relative expression. Compared to the control group, A miR-3646, B miR-4484, and C miR-K12-5-5p expressions were elevated, while D miR-4732-5p and E miR-4687-3p levels were decreased in the breast cancer patients. a–e Two-tailed t tests were performed to examine the differences of the selected miRNAs in nipple discharge between the breast cancer patients and benign controls. p < 0.05 was considered significant

Full size image

Ultimately, all of the qRT-PCR products were purified and ligated to the pMD19-T vector. The constructed plasmids were verified using sequencing. Supplementary Fig. S2 shows that the sequencing results of miR-3646, miR-4484, miR-K12-5-5p, miR-4732-5p, and miR-4687-3p matched the sequences in the microRNA database.

Discussion

Calin et al. showed that half of the known miRNAs are in cancer-associated genomic regions or fragile sites, suggesting that miRNAs might be involved in the initiation and progression of human malignancies.28 Because of their significant biological roles, miRNAs can be used as potential tumor biomarkers for detecting different cancers. Thus, miRNAs are considered to be potential diagnostic markers for patients with breast cancer and may provide new strategies for diagnosis. The broad range of clinical applications for miRNAs is impressive, as is the fast pace of evolution in miRNA-based technologies. Identifying and applying tumor biomarkers in detecting cancers as a noninvasive diagnostic method has been a focus in the field of early diagnosis. Recent studies suggest that the profile of plasma miRNAs, as well as tissue miRNAs, could be considered diagnostic markers for cancer.25,29 miRNAs are found in many body fluids and show distinct compositions in different fluid types.25 These findings indicate that miRNAs can serve as biomarkers as long as there is a correlation between specific miRNA levels and certain types of cancer.

Identifying breast cancer-specific miRNA profiles in nipple discharge is gaining interest as a potential diagnostic marker for breast cancer. In this study, we utilized microRNA microarray and qRT-PCR to identify six miRNAs as potential tumor biomarkers. Although it has been reported that miRNAs are frequently down-regulated in cancer patients, our study found that some miRNAs in nipple discharge are up-regulated and others are down-regulated in the breast cancer patients in this study.30 In the screening set, we found three significantly up-regulated miRNAs (miR-4484, miR-K12-5-5p, and miR-3646) and three considerably down-regulated miRNAs (miR-4732-5p, miR-4687-3p, and miR-S1-5p) among the 3100 miRNAs in our microarray panel. We then validated the results in an independent cohort using qRT-PCR.

This is the first clinical report describing four miRNAs—miR-4484, miR-K12-5-5p, miR-3646, and miR-4687-3p—with expression levels related to breast cancer. Of the five miRNAs that are differentially expressed in nipple discharge between carcinomas and benign tumors, miR-4732-5p is the only miRNA that has been previously reported to be associated with breast cancer.31 The involvement of these miRNAs in other diseases has also been studied. Previous reports have shown that miR-4484 is up-regulated in cervical squamous cell carcinomas and macrophages infected with virulent and avirulent Mycobacterium tuberculosis. 32,33 It has been reported that miR-3646 showed strong expression in bladder cancer cell lines and tissue and is increased in human patients with acetaminophen hepatotoxicity or ischemic hepatitis.34,35 Expression of miR-K12-5-5p is considered to be related to the transcriptional silencing and posttranscriptional silencing of ORF50 and antisense transcription throughout the Kaposi sarcoma-associated herpesvirus genome.36,37 There are no reports to date regarding the function of miR-4687-3p. In this study, we found significant increases of miR-4484, miR-K12-5-5p, and miR-3646 and a considerable reduction of miR-4732-5p in nipple discharge samples from breast cancer patients compared to the benign group. Consistent with the results from the analysis of screening sample set, the levels of aforementioned miRNAs were, in general, substantially deregulated in the validation set.

Our study is the first attempt to screen miRNAs from nipple discharge and identify miRNAs specific to breast cancer. The model of miRNAs had a satisfactory sensitivity, specificity, PPV, and NPV. Because nipple discharge samples can be easily obtained, examining tumor biomarkers in lactiferous duct-acquired fluid has a great application potential in breast cancer diagnosis. Instead of screening a large number of miRNA expression patterns, specific miRNAs can be used as biomarkers to differentiate malignant breast tumors from benign ones. Our finding that there is differential expression of four miRNAs in the nipple discharge of breast cancer patients and the control group will facilitate the detection of breast cancer by studying nipple discharge.

To establish miRNAs as breast cancer biomarkers in the clinical setting, several issues need to be addressed. First, an optimal combination of miRNAs for differentiating breast cancer from benign tumors needs to be determined. Second, our study detected one significantly down-regulated miRNA, miR-4732-5p, in breast cancer. The underlying mechanism of this miRNA in breast cancer development and progression needs to be further investigated. Third, more samples are needed to validate the feasibility of using miRNAs as a noninvasive diagnostic method for breast cancer. Finally, long-term follow-up studies are required to confirm the relationship between miRNA levels and breast cancer progression.

In conclusion, we found an elevation of miR-4484, miR-K12-5-5p, and miR-3646 and a reduction of miR-4732-5p in the nipple discharge of breast cancer patients in this study. The expression levels of miRNAs may have a potential predictive value as novel tumor biomarkers for diagnosing breast cancer via nipple discharge.