SEARCH

SEARCH BY CITATION

Keywords:

  • microRNA;
  • circulating microRNA;
  • breast cancer;
  • early detection;
  • molecular marker

Abstract

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

In recent years, circulating miRNAs have attracted a great deal of attention as promising novel markers for various diseases. Here, we investigated their potential to serve as minimally invasive, early detection markers for breast cancer in blood plasma. We profiled miRNAs extracted from the plasma of early stage breast cancer patients (taken at the time-point of diagnosis) and healthy control individuals using TaqMan low-density arrays (TLDA). Selected candidates identified in the initial screen were further validated in an extended study cohort of 207 individuals including 127 sporadic breast cancer cases and 80 healthy controls via RT-qPCR. Four miRNAs (miR-148b, miR-376c, miR-409-3p and miR-801) were shown to be significantly upregulated in the plasma of breast cancer patients. ROC curve analysis showed that the combination of only three miRNAs (miR-148b, miR-409-3p and miR-801) had an equal discriminatory power between breast cancer cases and healthy controls as all four miRNAs together (AUC = 0.69). In conclusion, the identified miRNAs might be of potential use in the development of a multimarker blood-based test to complement and improve early detection of breast cancer. Such a multimarker blood test might for instance provide a prescreening tool, especially for younger women, to facilitate decisions about which individuals to recommend for further diagnostic tests.

Breast cancer is the most common type of cancer and cause of cancer-related death among women in industrialized countries. Worldwide ∼ 1.3 million women develop breast cancer each year.1 Mortality rates have continued to decrease over the years due to the advances made in early diagnosis and treatment.1 Nevertheless, thousands of women die from this disease each year. In the United States, the overall 5-year survival is 98% when diagnosed at an early stage as opposed to 23% when the disease has already spread to distant organs.2 Thus, early breast cancer detection is one of the major challenges in the struggle against this disease. Mammographic screening is currently applied as the screening standard. However, it has limitations due to its use of ionizing radiation and a decreased detection sensitivity as the mammographic breast density increases (generally associated with younger age).3

Protein-based circulating tumor markers such as carcinoembryonic antigen (CEA) and carbohydrate antigen 15-3 (CA 15-3) are occasionally used as prognostic markers, as well as for monitoring breast cancer treatment success and follow-up.4, 5 However, the sensitivity of these markers is low. Therefore, new sensitive and specific as well as minimally invasive markers are needed.

miRNAs are small, noncoding RNAs (∼ 18–25 nucleotides in length) that regulate gene expression on a post-transcriptional level by degrading mRNA molecules or blocking their translation.6 Hence, they play an essential role in the regulation of a large number of biological processes, including cancer.7 Lawrie et al. were among the first to demonstrate the presence of circulating miRNAs in cell-free bodily fluids such as plasma and serum.8 Since then, circulating miRNAs have been reported as being aberrantly expressed in blood plasma or serum in different types of cancer, e.g., prostate, colorectal and esophageal carcinoma.9–11 Their most important advantages include the possibility of their repeated measurement in a noninvasive manner as well as their remarkable stability in plasma/serum, where they circulate even outside of exosomes and are stable due to their binding to Argonaute proteins.12–14

Here, we investigated and compared blood plasma miRNA profiles of breast cancer patients and healthy individuals. To our knowledge, this study is among the first few miRNA profiling studies of circulating miRNAs in breast cancer and the one with the largest plasma sample size analyzed for breast cancer to date. We show that miRNAs miR-148b, miR-376c, miR-409-3p and miR-801 are upregulated in the plasma of breast cancer patients in comparison to control individuals and present a combination of three miRNAs (miR-148b, miR-409-3p and miR-801) as our least redundant and most informative diagnostic panel to discriminate between breast cancer patients and healthy controls.

Material and Methods

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Breast cancer patients and healthy controls

This study was approved by the Ethical Committee of the Medical Faculty in Heidelberg. Blood samples were collected from 127 female sporadic breast cancer patients and 80 healthy female volunteers who served as controls. All cases and controls were Caucasian. Patient blood samples were collected in 2010 and 2011 at the time-point of diagnosis before they underwent any therapeutic procedures, such as surgery, radiation or systemic therapy. Patient histopathology results were confirmed by surgical resection of the tumors and clinicopathological features defined by operative findings. For neoadjuvant patients (n = 26), histopathological characteristics and tumor stage were assessed based on histobiopsy results and imaging techniques. Control blood samples were collected between 2010 and 2011 from healthy women with no history of malignant diseases, no blood donations received in the previous 3 years and no current inflammatory condition. Table 1 and Supporting Information Table 1 summarize the clinical features of the patients and lifestyle data of the healthy controls, respectively.

Table 1. Clinicopathological features of breast cancer patients used for validation
inline image

Malignant tissue samples were collected from non-neoadjuvant breast cancer patients during surgery. They were snap-frozen in liquid nitrogen and stored at −80°C within 15 min of harvesting. Supporting Information Table 2 gives an overview of the clinicopathological characteristics of these patients. For women with benign findings tissue was collected during the diagnostic histobiopsy. The mean and median ages of women with benign findings were 53.3 and 50.0 years, respectively.

Table 2. Circulating miRNAs differentially expressed in the plasma of early stage breast cancer cases compared to healthy controls in TLDA array analysis
inline image

Blood processing and miRNA isolation from plasma

EDTA blood samples were collected from cases and control individuals and processed for plasma within 2 h of collection. To avoid contamination with epithelial cells from the initial skin puncture the first blood tube collected during phlebotomy was not processed for plasma. Blood was centrifuged at 1,300 g for 20 min at 10°C. The supernatant (plasma) was transferred into microcentrifuge tubes followed by a second high-speed centrifugation step at 15,500 g for 10 min at 10°C to remove cell debris and fragments. The plasma was aliquoted into cryo vials, snap-frozen in liquid nitrogen and stored at −80°C until use.

Total RNA (including miRNAs) was extracted from 400 μL of plasma. Denaturation and phase separation were conducted using TRIzol LS (Invitrogen, Germany) according to manufacturer's protocol, with a minor modification: 10 fmol of a C. elegans miR-39/miR-238 mixture was spiked-in. The aqueous phase was transferred into another tube, 1.5 volumes of absolute ethanol were added and the mixture was applied to miRNeasy Mini kit columns (Qiagen, Germany). After washing miRNAs were eluted in 30 μL of RNase-free water.

miRNA profiling of plasma with TaqMan low-density arrays

Profiling was carried out using TaqMan low-density arrays (TLDA; human MicroRNA Cards A v2.1 & B v2.0) from Applied Biosystems according to manufacturer's protocol. These arrays measured the expression of 667 human miRNAs from miRBase version v.10. In brief, a fixed volume of miRNAs (3 μL) was reverse transcribed using the TaqMan MicroRNA Reverse Transcription Kit and TaqMan MicroRNA Megaplex RT Human Pool Sets A & B. cDNA was pre-amplified for 12 cycles with Megaplex PreAmp Human primer Pools A & B, respectively, and loaded into the TLDA array card ports. Real-time PCR was carried out with an Applied Biosystems 7900HT thermocycler under the following conditions: 50°C for 2 min, 94.5°C for 10 min, followed by 40 cycles of 97°C for 30 sec and 59.7°C for 1 min. Raw data were exported using SDS Relative Quantification Software version 2.2.2 (Applied Biosystems) with automatic baseline and threshold settings.

Raw Ct values of the initial plasma screening step with TLDA arrays were analyzed using the statistical computational environment R version 2.11 (http://www.r-project.org/) and the R package HTqPCR version 1.2.0.15, 16 miRNAs with Ct > 35 across all the samples were filtered out from further analysis. Following quantile normalization, we performed quality control plots: (i) the matrix of Pearson's correlations across samples was clustered using hierarchical complete linkage clustering according to the Euclidean distance metric method and the results displayed as a heat map, and (ii) principal components analysis (PCA) plots were created to identify potential outliers. For differential expression analysis invariant miRNAs (as determined by low inter-quartile ranges) were removed from the data set and duplicates on the TLDA arrays averaged. After limma analysis to identify miRNAs that were differentially expressed between cases and controls, the results were adjusted for multiple testing by controlling the false discovery rate (FDR) according to the method of Benjamini-Hochberg.17, 18 All statistical analyses were performed using normalized data.

Validation of selected marker candidates

Reverse transcription (RT) reactions were performed using TaqMan MicroRNA Reverse Transcription Kit (Applied Biosystems, Germany) and miRNA-specific RT primers for hsa-miR-139-3p, hsa-miR-148b, hsa-miR-206, hsa-miR-376c, hsa-miR-409-3p, hsa-miR-571, hsa-miR-801 and cel-miR-39 (Applied Biosystems, Germany). Singleplex reactions were carried out in a volume of 7.5 μL. Each reaction comprised 0.75 μL 10× RT buffer, 0.075 μL dNTPs (100 mM), 0.5 μL 5× miRNA-specific RT primers, 0.095 μL RNase inhibitor, 0.5 μL Multiscribe Reverse Transcriptase and a fixed volume of miRNA template (2 μL). Blinding of samples and a randomized, simultaneous investigation of cases and controls on reaction plates was intended to minimize bias and batch effects during validation. RT was carried out in a G-STORM GS2 PCR cycler (Alphametrix, Germany) under the following conditions: 16°C for 30 min, 42°C for 30 min and 85°C for 5 min, followed by a hold at 4°C.

TaqMan real-time PCR reactions were performed in triplicates in scaled-down reactions comprising 2.5 μL TaqMan 2× Universal PCR Master Mix with No AmpErase UNG (Applied Biosystems, Germany), 0.25 μL 20× miRNA-specific primer/probe mix (Applied Biosystems, Germany) and 2.25 μL of the reverse transcription product (diluted 1:4). Real-time PCR was carried out in a LightCycler 480 thermocycler (Roche, Germany) under the following conditions: 95°C for 10 min, then 40 cycles of 95°C for 15 s, 60°C for 30 s and 72°C for 30 s, followed by a hold at 4°C.

Raw data from validation studies were normalized to spiked-in cel-miR-39 by the principle described by Kroh et al., but with the use of a single miRNA, as normalizing to cel-miR-39 alone seems to be equally as good as normalizing to a combination of three different spike-in miRNAs.19, 20 Power simulations for two-group comparisons were conducted to find the minimal sample size necessary to find a true two-fold change with at least 90% statistical power. Wilcoxon rank sum tests with continuity correction were used to identify miRNAs that were differentially expressed between cases and controls in the validation set. To detect associations between miRNA expression levels and clinicopathological (breast cancer cases) or lifestyle data (healthy controls), the following nonparametric tests were used: Wilcoxon rank sum test (for relating miRNA expression to binary categorical variables), Spearman's rank correlation test (relating to continuous variables) and Jonckheere–Terpstra test (for assessing associations with ordinal variables). A two-tailed p < 0.05 was considered statistically significant. To evaluate the breast cancer detection potential of individual miRNAs and miRNA combinations, logistic regression models were fitted, receiver operating characteristic (ROC) curves constructed and areas under the curves (AUC) calculated as well as specificities for fixed sensitivity values with corresponding 95% confidence intervals (CI).21, 22 Interrelationships between miRNA expressions were investigated by computing Spearman rank correlation coefficients (ρ).

Expression levels of selected markers in primary breast cancer and benign tissue

Total RNA containing small RNAs was isolated using miRNeasy Mini Kit (Qiagen, Germany) according to manufacturer's protocol. In brief, slides of fresh-frozen tumor tissue (breast cancer patients) or biopsy-tissue (women with benign findings) were lyzed with 700 μL QIAzol Lysis Reagent (Qiagen, Germany) and homogenized using a syringe and needle. After denaturation and phase separation, RNA was bound to the column membrane, washed and eluted in 50 μL of RNase-free water.

RT reactions were performed using TaqMan MicroRNA Reverse Transcription Kit (Applied Biosystems, Germany) as described for plasma samples, but with a few modifications. Each miRNA was multiplexed with specific RT primers for the endogenous control RNU6B (Applied Biosystems, Germany). Reactions were carried out in 15 μL and comprised the following: 1.5 μL 10× RT Buffer, 0.15 μL dNTPs (100 mM), 2 μL specific RT primers (1 μL for the target miRNA and RNU6B, each), 0.19 μL RNase-Inhibitor, 1 μL of Multiscribe Reverse Transcriptase and 5 ng RNA. TaqMan real-time PCR reactions were performed in triplicates. Ct values were normalized to RNU6B as described in User Bulletin #2: ABI PRISM 7700 Sequence Detection System (Applied Biosystems). Benign and malignant primary breast cancer tissue samples were compared using a Wilcoxon rank sum test and a two-tailed p < 0.05 was considered significant.

Results

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

miRNA profiling revealed putative marker candidates for breast cancer detection in plasma

In an initial screening step using TLDA arrays we analyzed plasma miRNA profiles of 10 early stage breast cancer patients as well as 10 healthy controls. The patients all had an invasive ductal carcinoma, which was ER/PR+ and HER2 with an AJCC TNM stage I or II.23 Patients were age-matched to healthy controls. The mean and median ages of patients were 54.0 and 51.0 years respectively, while they were 53.0 and 54.5 years for controls.

After normalization of raw array data and filtering a total of 169 variant miRNAs remained for statistical analysis. A heat map illustrating the results of hierarchical cluster analysis of the correlations across samples and PCA plots identified one control sample (B024) as an outlier, which was then removed from further statistical analysis (data not shown). Limma analysis revealed 16 circulating miRNAs with statistically significant differences in expression between cases and controls. Six miRNAs were downregulated (miR-139-3p, miR-193a-3p, miR-206, miR-519a, miR-526b* and miR-571) and ten upregulated (miR-127-3p, miR-148b, miR-148b*, miR-184, miR-190, miR-376a, miR-376c, miR-409-3p, miR-424 and miR-801) in the plasma of early stage breast cancer patients. A list of these miRNAs can be found in Table 2 together with their (i) p-values, (ii) p-values adjusted for multiple testing according to the method of Benjamini–Hochberg (indicating false discovery rates, FDRs), (iii) mean Ct values for both investigated groups and (iv) differences in mean Ct values (ΔCt) between the control and cases group. Based on the expression of these 16 deregulated circulating miRNAs, the samples clustered into two distinct groups (Fig. 1).

thumbnail image

Figure 1. Circulating miRNAs deregulated in the plasma of early stage breast cancer patients. Heat map of Euclidean hierarchical complete linkage clustering based on the expression of 16 deregulated circulating miRNAs.

Download figure to PowerPoint

miR-148b, miR-376c, miR-409-3p and miR-801 are upregulated in plasma of breast cancer patients

The following criteria were applied to choose the best candidates for marker validation studies in plasma: p < 0.05, FDR < 33.3% and mean Ct < 33 in at least one investigated group (as miRNA expression should be stably detectable in at least one group) (Table 2). The application of these criteria resulted in seven candidates for validation: miR-139-3p, miR-148b, miR-206, miR-376c, miR-409-3p, miR-571 and miR-801.

To find the appropriate sample size necessary to detect fold changes as small as two-fold, power simulations were carried out. Statistical power was estimated based on observed standard deviations in the preliminary small-scale validation experiments. In all tested scenarios in which total sample size was ≥200 and included at least 80 controls statistical power was very high (>93%).

A total of 127 breast cancer and 80 healthy control plasma samples were analyzed for their expression of the aforementioned seven marker candidates. After a preliminary small-scale validation on 50 samples, miR-139-3p showed a clearly nonsignificant p-value (p = 0.60) and miR-571 could not be detected in either group (data not shown). For these reasons, validation was continued only with the five remaining miRNAs. A comparison of case and control groups using a Wilcoxon rank sum test resulted in four circulating miRNAs validated as being upregulated in the plasma of breast cancer patients. These miRNAs were miR-148b (p < 0.001), miR-376c (p < 0.0001), miR-409-3p (p < 0.0001) and miR-801 (p < 0.001), whereas miR-206 (p = 0.26) did not reach statistical significance (Figs. 2a2d).

thumbnail image

Figure 2. Circulating miRNAs validated as being upregulated in the plasma of breast cancer patients compared to healthy controls. Box and whisker plots of cel-miR-39 normalized Ct values for miR-148b (a), miR-376c (b), miR-409-3p (c) and miR-801 (d) with their corresponding p-values.

Download figure to PowerPoint

Correlation of miRNA expression to clinicopathological and lifestyle data

A summary of clinicopathological features considered for correlation with miRNA expression levels can be found in Table 1. In brief, they included all clinically relevant characteristics such as hormone receptor status, grading, tumor size and lymph node affliction. Both miR-148b and miR-801 displayed a borderline significant correlation of miRNA expression levels with age and menopausal status in the cases group (p = 0.04 and p = 0.03 for miR-148b and p = 0.03 and p = 0.04 for miR-801, respectively). miR-148b has a negative and miR-801 a positive correlation with age, indicating that the discriminatory potential of miR-148b seems to be better for younger patients and that of miR-801 for older patients (Supporting Information Figure 1). Additionally, Supporting Information Table 1 provides an overview of lifestyle characteristics that were analyzed for possible associations with miRNA expression in the control group. Available information included some features considered to be breast cancer risk factors: BMI, age of menarche, hormone intake, smoking and alcohol consumption. It is noteworthy that neither miR-148b nor miR-801 expressions showed any correlation with age in the control group (p-values were p = 0.79 and p = 0.58, respectively). miR-376c and miR-409-3p expressions were not correlated with any of the analyzed features.

Diagnostic potential of miR-148b, miR-376c, miR-409-3p and miR-801 in plasma

ROC curve analysis was performed to evaluate the diagnostic potential of miR-148b, miR-376c, miR-409-3p and miR-801 for breast cancer detection in blood plasma. The discriminatory power between tumor and control samples is depicted by the areas under the curves (AUC). Individually, miR-148b had an AUC of 0.65 (95% CI: 0.58–0.73), miR-376c of 0.66 (95% CI: 0.59–0.74), miR-409-3p of 0.66 (95% CI: 0.59–0.74) and miR-801 of 0.64 (95% CI: 0.56–0.72) (Figs. 3a3d).

thumbnail image

Figure 3. Evaluation of the diagnostic potential of miR-148b, miR-376c, miR-409-3p and miR-801 in the plasma of breast cancer patients. ROC curves for individual miRNAs miR-148b (a), miR-376c (b), miR-409-3p (c) and miR-801 (d). A combined ROC curve showing the least redundant and most informative diagnostic panel consisting of miR-148b, miR-409-3p and miR-801 (e).

Download figure to PowerPoint

We found that miR-148b, miR-376c and miR-409-3p expressions correlate to each other with Spearman rank correlation coefficients as follows (all p < 0.00001): (i) ρ = 0.64 between miR-148b and miR-409-3p, (ii) ρ = 0.66 between miR-148b and miR-376c and (iii) ρ = 0.91 between miR-376c and miR-409-3p. The correlation coefficient between miR-148b and miR-801 expressions is also considerable (ρ = 0.35), but other correlations were not substantial. By investigating different combinations of miR-148b, miR-376c, miR-409-3p and miR-801 we found that a combined ROC curve with miR-148b, miR-409-3p and miR-801 gave the most informative and least redundant miRNA panel with an AUC of 0.69 (95% CI: 0.61–0.76) (Fig. 3e). At 70% sensitivity the median specificity was 55% (95% CI: 0.39–0.70). The discriminatory power of all four significantly upregulated miRNAs (AUC = 0.69) did not outperform this panel and other miRNA combinations performed only slightly poorer (not shown).

miR-148b, miR-376c and miR-409-3p are downregulated in malignant primary breast cancer tissue

A total of 24 primary breast cancer surgery tissue samples and eight benign breast biopsies were analyzed for their miR-148b, miR-376c, miR-409-3p and miR-801 expression levels. A comparison of these two sample groups showed that, in contrast to plasma, miR-148b (p = 0.007), miR-376c (p < 0.0001) and miR-409-3p (p = 0.002) were downregulated in malignant breast cancer tissue in comparison to benign breast tissue samples (Figs. 4a4c). In the case of miR-801 (p = 0.80) no significant differences in expression levels were detected (Fig. 4d).

thumbnail image

Figure 4. Expression of miR-148b, miR-376c, miR-409-3p and miR-801 in benign versus malignant breast tissue. Box and whisker plots with RNU6B normalized relative expression levels of miR-148b (a), miR-376c (b), miR-409-3p (c) and miR-801 (d) in benign and malignant breast tissue.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Early cancer detection remains a major challenge in breast cancer research as it holds promise to result in a more favorable disease outcome. However, the current standard screening method, mammography, uses potentially harmful ionizing radiation and has false positive rates of 8–10%. Thus, early detection marker development is turning towards novel and minimally invasive techniques. Blood-based tests, such as measuring circulating miRNAs in the plasma, could offer an important complement of existing diagnostic tools allowing improved performance in breast cancer screening and detection. Here we investigated the potential of circulating miRNAs, to serve as an early detection marker by analyzing plasma miRNA profiles of breast cancer patients and healthy individuals.

Pilot studies regarding circulating miRNAs as potential novel breast cancer markers have shown promising results, but were focused on a handful of predefined target miRNAs and/or analyzed only a small sample size. In their pilot study in serum, Zhu et al. could not detect significant differences between healthy women and breast cancer patients, but found that progesterone receptor positive patients had higher miR-155 expression than patients with tumors negative for this receptor.24 Roth et al., who also analyzed serum samples, found that miR-155 can discriminate between nonmetastatic breast cancer patients and healthy individuals.25 Zhao et al., who investigated plasma samples from Caucasian and African American breast cancer patients and healthy women, found in particular, besides deregulated miRNAs, differences in miRNA expression between different ethnic groups.26 Guo and Zhang investigated the levels of miR-181a in serum of breast cancer patients and healthy controls and found that miR-181a has better diagnostic value than conventional tumor markers CA-15-3 and CEA.27 van Schooneveld et al. profiled cancerous and normal breast tissue and found that miR-299-5p and miR-411 are not just differentially expressed in tissue, but also in the serum of patients with metastatic breast cancer compared to healthy volunteers.28 The lack of reproducibility of published studies on circulating miRNAs in breast cancer might have several reasons: (i) differences in miRNA expression due to sample type used (plasma or serum), (ii) differences in blood processing protocols (e.g., possibility of blood cell contamination in plasma/serum samples in some studies), (iii) differences in study populations (age, histopathological characteristics, proportion of breast cancer stages, ethnicity etc.) and (iv) differences in time-points of sample collection (i.e., collection of blood prior, during or after therapeutic treatment).

The strengths of our study are (i) the standardized processing of blood samples to generate plasma within 2 h of collection with a two-step centrifugation protocol, (ii) the use of array-based screening for putative miRNA candidates, (iii) carrying out validation studies in a blinded manner, (iv) analysis of the largest plasma sample cohort for breast cancer to date, which consisted of at least 75% stage I and II cases and (v) investigating plasma samples which have been taken at the time-point of breast cancer diagnosis before the patients underwent any therapeutic procedures (exclusion of possible effects of therapeutic treatments on miRNA expression levels). To minimize (pre)analytical variability, which could lead to bias in miRNA quantification, we standardized our blood handling, i.e., we processed all blood within 2 h of collection and used a two-step centrifugation protocol. In this protocol the second step is an additional high-speed plasma centrifugation step prior to freezing, which allows the preparation of cell-free plasma devoid of cellular components and debris. This is especially important, since recent publications demonstrated that blood cells can contribute significantly to circulating miRNA levels in plasma if the latter is not processed properly.20, 29, 30 Additionally, we matched our healthy controls to the investigated patients by gender and ethnicity (Caucasian) because of the reported differences in miRNA expression between different gender and ethnic groups.26, 29

Although more and more reports on circulating miRNAs as potential disease/cancer markers are being published, data normalization still remains an important and unresolved issue. In tissue analysis RNU6B is commonly used for normalization, but in contrast to miRNAs, RNU6B does not have a remarkable stability in plasma/serum, i.e., it is not protected from RNases in plasma/serum by being bound to Argonaute proteins. So far no circulating miRNA(s) have been established or validated as suitable endogenous control(s) for normalization in blood plasma or serum. Some researchers use miR-16, despite existing reports which describe its high variability or altered expression in the circulation of cancer patients, as well as in hemolyzed samples (due to its very high expression in red blood cells).30–32 At present there are some approaches which can be used to minimize experimental variation, e.g., using spike-in miRNAs and processing the same, constant volumes of samples at each step of the analysis. In our validation study we have utilized both of these approaches by spiking-in a fixed amount of cel-miR-39 during plasma miRNA isolation and by using constant volumes of samples throughout the analysis. Nevertheless, a consensus endogenous control and/or normalization procedure need to be established.

We identified four miRNAs (miR-148b, miR-376c, miR-409-3p and miR-801) as being significantly upregulated in the plasma of breast cancer patients (Fig. 2). To our knowledge none of these miRNAs has been reported to play a role in breast cancer before. Zhang et al. identified miR-148b as being significantly upregulated in the sera of esophageal squamous cell carcinoma patients compared to control individuals.11 Additionally they investigated the expression of miR-148b in colorectal, ovarian and pancreatic carcinoma, but found no alteration in expression in the circulation of these cancer types. Here, we show that circulating miR-148b is also deregulated in breast cancer (Fig. 2a). So far, miR-376c has been reported as being upregulated in a subset of acute myeloid leukemias and also in the sera of gastric carcinoma patients compared to individuals with superficial or mild chronic atrophic gastritis (Fig. 2b).33, 34 Regarding miR-409-3p, there are no reports mentioning its aberrant expression in the plasma or serum of any carcinoma patients. Recently, Zhou et al. reported that miR-801 had significantly higher expression levels in plasma of hepatocellular carcinoma patients compared to healthy controls.35 This finding further strengthens our observation of circulating miR-801 playing a role in cancer and having potential to serve as a plasma marker (Fig. 2d).

Further on we evaluated the diagnostic potential of miR-148b, miR-376c, miR-409-3p and miR-801 by ROC curve analysis. For individual miRNAs the AUC values ranged from 0.64 to 0.66 (Figs. 3a3d). Breast cancer detection efficiency was enhanced when we combined miRNAs into diagnostic panels. Our most informative and least redundant model comprises only three miRNAs: miR-148b, miR-409-3p and miR-801 (Fig. 3e), which gave the same AUC value (AUC = 0.69) as the combination of all four significantly deregulated miRNAs. The discriminatory power of other combinations was only marginally lower (data not shown) and is most likely due to the strong correlation we observed between our miRNA expressions, especially between miR-376c and miR-409-3p.

We also investigated the expression of miR-148b, miR-376c, miR-409-3p and miR-801 in benign versus malignant breast tissue. As a result we found miR-148b, miR-376c and miR-409-3p to be significantly downregulated in malignant primary breast cancer tissue, whereas miR-801 expression did not differ between the two groups (Fig. 4). To our knowledge there are no reports on miR-376c or miR-801 expression in other types of cancers comparing benign and malignant tissue. However, concordant with our findings, miR-148b has been reported as being downregulated in a few other types of cancer tissue, such as colon, gastric and colorectal carcinomas.36–38 MiR-409-3p has also been reported as being downregulated in gastric carcinomas compared to matched nontumorous tissue.39, 40

Although recent plasma/serum-based miRNA reports including the here presented study show that miRNA profiles differ between cancer cases and healthy controls, the origin of the circulating miRNAs remains unknown, along with how they are released into the blood stream. When studies investigating circulating miRNAs as potential cancer markers first began, many researchers focused on already known cancer-associated miRNAs. They expected to find similar expression patterns in the tissue and circulation of cancer patients. Concerning breast cancer, Roth et al. and van Schooneveld et al. were able to show that this is only true for some cancer-associated miRNAs.25, 28 It seems that the miRNAs that are upregulated in the tumor tissue of patients are not necessarily the same miRNAs that are upregulated in their respective plasma. Here, our investigated miRNAs generally displayed the opposite expression pattern in tissue and plasma. This might give reason to speculate that malignant cancer cells can selectively release specific miRNAs, as indicated by Pigati et al.41 This selective miRNA release into the blood stream could theoretically cause levels of particular miRNAs to go up in blood derivatives such as plasma and go down in the tumor tissue/cells from which they originate. They might be released from cells protected in microvesicles or exosomes as suggested by Valadi et al. or Gallo et al.42, 43 On the other hand, it was postulated that circulating miRNAs might mainly represent the by-products of dying/dead cells and are predominantly circulating outside of exosomes/microvesicles in plasma.13, 14 Some of the circulating miRNAs could also be derived from blood cells or be some kind of an immune response to the tumor.30, 44 Furthermore, given the importance of the stromal compartment and tumor microenvironment in cancer initiation and progression it is also possible that circulating miRNAs or a fraction of the circulating miRNAs deregulated in the plasma of cancer patients originate from this source.45 In conclusion, the observations presented herein might be pointing to important biological processes and mechanisms that need to be resolved in the future.

In summary, we identified four miRNAs as being upregulated in the plasma of breast cancer patients compared to healthy control individuals. Also, we defined a panel of plasma miRNAs that might contribute to the discrimination of breast cancer from healthy state. However, additional molecular markers are needed for the development of a multimarker blood-based test that could complement existing screening methods and improve (early) detection, for instance by providing a prescreening tool, especially for younger women, to make decisions about which individuals to recommend for further diagnostic tests.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

We thank our colleagues and technicians for their great assistance and support, especially Ms. R. Yang for the helpful discussions and Ms. C. Dinkelacker, who helped establishing the control cohort. We also thank the medical staff (especially Prof. Dr. P. Sinn) and study nurses of the Women's clinic and NCT in Heidelberg. Due to a US patent application relating to this manuscript K. Cuk, M. Zucknick, D. Madhavan and B. Burwinkel declare a potential conflict of interest.

References

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
IJC_27799_sm_SuppInfo.doc510KSupporting Information

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.