Identification of differentially expressed microRNAs in human male breast cancer

For the comprehensive determination of the microRNA expression profile of human male breast cancer 9 specimens were retrieved from the archives. The primary selection criteria were tumor size and tumor cell content, because several microgram of total RNA are required for the full profiling with five different bead pools (see Materials and Methods). All tumors were grade 2 or 3, with a mean proliferation rate of 20% (as determined by MIB1 labeling, median: 20%, range: 10 - 40%). They were all estrogen and androgen receptor positive, all but one were also progesterone receptor positive, but negative for Her2 overexpression, p63 expression and expression of basal cytokeratines 5 and 14, thus representing typical male breast cancer cases [1].

The 12 additional cases for independent validation were comprised by 6 primary invasive carcinomas, 3 local recurrences, and 3 metastases. The distribution of tumor grade, proliferation rate and the immunohistochemically determined tumor marker profiles were very similar to the first nine cases under study. A detailed description of all 21 specimens is provided in Table 1.

For comparison a series with 15 male gynecomastia specimens was analyzed representing a benign proliferative condition of the ductal epithelium.

The expression of 319 microRNAs was analyzed in 9 primary human male breast cancer specimens using fluorescence-labeled bead technology. As a control the expression of these 319 microRNAs was measured in 15 epithelial cell fractions from male gynecomastia specimens, which were pooled in 4 mixes due to low RNA yield.

Unsupervised clustering of all microRNA expression data separated very well the carcinoma specimens from the pools of control samples, clearly indicating cancer-specific alterations of microRNA expression in male breast cancer (Figure 1).

T-test based analysis for comparison of carcinoma and normal samples identified several differentially expressed microRNAs (Figure 2), among them miR-21 reported to be overexpressed in female breast cancer and miR-145 reported to be down-regulated in female breast cancer [7, 8]. Up-regulation of miR-519d, miR-183, miR197, and miR-493-5p (more than twofold in comparison to gynecomastia, see Table 2) has not been reported so far in female breast cancer.

A detailed analysis revealed 33 statistically significantly up-regulated and 21 statistically significantly down-regulated microRNAs in human male breast cancer in comparison to male gynecomastia specimens (Table 2). Strongest up-regulation was found for miR-21, already described to be significantly up-regulated in female breast cancer ([4] and references therein) and other solid tumors. ([3] and references therein)

In order to validate the results of the profiling experiments, the expression level of the two most up-regulated (miR-21, miR-519d) and the two most down-regulated (miR-497, miR-145) microRNAs were validated in the carcinoma training cohort (n = 9), an independent carcinoma test cohort (n = 12) and the four control samples using real-time PCR methodology (stem-loop primer-based kits from Applied Biosystem). Correlation analysis revealed a good correlation between the real-time PCR data (stem loop primer PCR) and the Luminex™ data (fluorescence labeled beads) for miR-21 (r = 0.93), miR-145 (r = 0.513), and miR-497 (0.71).

Welch-based t-test demonstrates a statistically significant difference between carcinoma samples and the controls for miR-21, miR-145, and miR-497, but only a trend for miR-519d (Figure 3).

To further explore our findings, we analyzed the expression of several target genes of miR-21. We concentrated on miR-21, because it is the most significantly up-regulated miR-21 in our study and several validated target genes are described in the literature (see [9] and references therein, see also the new database miRecords [10]). Quantitative real-time RT-PCR was employed to measure mRNA levels of target genes, because quantification of protein expression levels in archival clinical specimens is difficult and depends on the availability of suitable antibodies. Mean expression in carcinoma specimens compared to the control group was measured for MASPIN, PDCD4, GLCCl1, BMPR-II, and APAF-1. Whereas no clear differences were discernible for BMPR-II and APAF-1, respectively (data not shown), differential expression, albeit to a different extent, was found for MASPIN, PDCD4 and GLCCl1 (Figure 4). GLCCl1 showed a highly significant reduction in the tumor specimens in comparison to the control group. MASPIN showed only a weak reduction whereas PDCD4 didn't show any difference in this presentation of the data (Figure 4A). However, if the mRNA expression level is correlated to the miR-21 expression level case-by case, a clear inverse correlation is discernible for MASPIN (Figure 4B), and much more pronounced for PDCD4 (Figure 4C), one of the best studied target genes of miR-21 [9, 11]. Nearly all measurements for PDCD4 are within the 95% confidence interval (data not shown).

21 specimens from 20 male patients with the clinically and histopathologically confirmed diagnosis of breast carcinoma in the period from 1995 to 2008 were selected from the archive of the Institute of Pathology, Medizinische Hochschule Hannover, Germany and analyzed anonymously following the guidelines of the local Ethics committee ("Ethik-Kommission der Medizinischen Hochschule Hannover", head: Prof. Dr. Tröger). For the comprehensive profiling of 319 microRNAs (trainings set, n = 9) only tumors with a size greater than 0.5 cm and a total RNA yield of at least 10 μg were selected. If tumor cell content was below 75%, tumor cells were manually microdissected from unstained histological sections using an H&E stained serial section as guidance. For a detailed description of all carcinoma cases see Table 1.

As a reference for normal expression level the RNA isolated from 15 male age-matched gynecomastia specimens was pooled into four control samples ("mix 1 - 4", made up of three or four individual RNA preparations, respectively) because the average RNA yield was quite low due to the limited number of epithelial cells present.

Multi tissue arrays were constructed as described in order to ensure homogenous staining quality for all specimens [28]. Two tissue cores per specimen were analyzed and scored in a blinded fashion. Mouse monoclonal anti-ER (clone: 6F11), anti-PgR (clone: 16), and anti-HER2 (clone: CB11) antibodies were purchased from Ventana (Illkirch, France) and mouse monoclonal anti-CK5/6 (clone: D5/16 B4), anti-p63 (clone: 4A4), anti-MIB-1/Ki 67 (clone: MIB-1), and anti-androgen receptor (anti-AR; clone: AR441) antibodies were purchased from Dako (Glostrup, Denmark). The antibodies used for immunohistochemistry formalin-fixed, paraffin-embedded sections (3 μm) were deparaffinized in xylene and dehydrated through graded alcohols. Slides were rinsed thoroughly with Protein Blocking Agent (UltraTech HRP Streptavidin-Biotin Detection System, Beckman Coulter, Fullerton, CA, USA). Endogenous peroxidase was blocked in 3% hydrogen peroxide (H₂O₂) in methanol for 15 min. Antigen retrieval for ER, PR, HER2, CK5/6, p63, MIB-1, and AR was done by autoclaving at 121°C for 10 min in 0.1 mol/L citrate buffer (pH 6). The sections were incubated with one of the following antibodies: mouse monoclonal anti-ER antibody (predilution), anti-PgR antibody (predilution), anti-HER2 antibody (predilution), anti-CK5/6 antibody (1:100 dilution), anti-p63 antibody (1:50 dilution), anti-MIB-1 antibody (1:50 dilution), and anti-AR antibody (1:50 dilution) for 60 min at room temperature. Then, the secondary biotinylated goat polyvalent antibody (Beckman Coulter) was applied for 10 min at room temperature. The sections were incubated with peroxidase-conjugated streptavidin for 10 min and the reaction products were visualized using 3,3'-diaminobenzidine tetrahydrochloride (DAB) and H₂O₂. Counterstaining was performed using hematoxylin. The slides were rinsed with PBS after each stage of the procedure. As a negative control, all staining was also performed without the first antibody. When >5% of the tumor cells were stained with the antibody for p63 and CK 5/14, it was categorized as positive for antigen expression. HER2 staining was scored following the HER2 testing guidelines for breast cancer. ER, PR and AR expression was evaluated by the Allred score.

Isolation of total RNA from FFPE specimens was performed as described previously [29]. Following the recommendations from Luminex Inc. no microRNA enrichment was performed.

Quantification of microRNAs hybridized to fluorescence labeled beads was performed with a BioPlex 200™ from Biorad (Bio-Rad Laboratories GmbH, München, Germany) using the software Luminex IS™, version 2.3 from Luminex (Austin, Texas, USA). All beads coated with LNA probes complementary to mature microRNAs were purchased from Luminex Inc. (Austin, Texas, USA).

Prior to hybridization total RNA is labeled with biotin which is later bound to a streptavidin-phycoerythrin conjugate. For this purpose the FlexmiR™ MicroRNA Labeling Kit from Luminex Inc. was used. For the comprehensive profiling of 319 microRNAs (FlexmiR™ human microRNA pool, version 8) 2.5 μg total RNA were labeled following the protocol provided by the manufacturer. The 319 different microRNA beads are divided into 5 pools because there are not enough different fluorescence labels available to distinguish more than 70 - 80 beads in a single tube. For every pool 0.5 μg of total RNA are required. Subsequent washing, hybridization and analysis of samples were performed exactly as described in the protocols provided by Luminex Inc.

The system was calibrated using the xMAP™ calibration control reagents from Luminex Inc. following the recommendations of the manufacturer. Every run contained a negative control for background subtraction (water instead of RNA) and a positive control (total human brain RNA from Ambion, provided by Luminex Inc.).

For the measurement of the expression levels of selected microRNAs (mir-21, -145, -497, -519d) the real time-RT-PCR-based detection methodology from ABI (Darmstadt, Germany) was used. Relative expression levels were calculated using the constitutively expressed small RNAs RNU24, U54, and Z-30 as endogenous controls. All primers and probes were purchased from ABI. cDNA synthesis and real-time PCR were performed following the protocols supplied by the manufacturer.

For the measurement of the expression level of selected mir-21 target genes total RNA was transcribed to cDNA as described in ref. [30]. For this purpose an aliquot from the RNA preparation used for microRNA profiling was used. Subsequent real-time RT-PCR using SybrGreen as detection reagent was performed as described in ref. [31] with the same reference genes (βGUS and TBP). The primer sequences for the miR-21 target genes were taken from ref. [9].

The raw intensity values were processed according to Blenkiron et al. [26]. Values smaller than 1 were equalized to 1. All values were log transformed to a basis of 2. The background values derived from the negative control were subtracted from all sample signals. A pool to pool probe effect normalization was performed as described by Blenkiron et al. [26]. Samples and miRNAs were clustered unsupervised hierarchically with the Eisen cluster software (EisenLab Cluster version 2.11) using average linkage and uncentered correlation. Visualization was performed with TreeView (version 1.60).

To compare the two groups (normal vs. carcinoma) Welch based t-tests were carried out using R statistical platform (version 2.80) and visualized using TIGR software (MeV version 4.3, http://​www.​tm4.​org).

The correlation between microRNA expression and relative PCR C_T-values was determined by Pearson product moment correlation.