Evaluation of microRNA-10b prognostic significance in a prospective cohort of breast cancer patients

Table 1 summarizes descriptive statistics for the 101 cases selected for our analysis. The median age of the study population is 59 years (range, 36 to 82), median tumor size is 2.5 cm (range, 1.0 to 10.0). Metastases at diagnosis (synchronous metastases) were present in 11 cases whereas, among non-metastatic patients, 34 experienced disease progression and 30 of them developed distant metastases (metachronous metastases).

All patients received adequate local treatment (breast conserving surgery or total mastectomy) plus sentinel node biopsy or complete axillary dissection. Post-surgery treatments were performed according to the following guidelines: San Gallen, NCCN and ASCO. Adjuvant therapy in association with postoperative breast irradiation (RT) was performed in 89 patients because one subject refused treatment.

miR-10b expression was evaluated in paired normal and tumor tissues obtained from 101 patients. As expected from previous studies[7, 21‐23] overall miR-10b expression levels (miR-10b/RNU48x1000) were lower in tumor tissues as compared with normal breast with median values of 28.33 (IQR 10.68-62.71) and 254.95 (IQR 110.28-495.09). Thus, to determine tumor specific changes we evaluated for each patient the ratio between the levels of miR-10b expression in cancer specimen to the levels of miR-10b expression in paired normal tissue (RER). RERs ranged from 0.05 to 1.7 with a median value of 0.10 (IQR 0.05-0.32).

The only significant association with clinicopathological characteristics was found between miR-10b RERs and the presence of distant metastases at diagnosis (Additional file1: Table S1). miR-10b RERs were significantly higher in the subgroup of patients with metastases (Median 0.25; IQR 0.11-1.02) as compared with patients without metastases (Median 0.09; IQR 0.04-0.29) (p = 0.028). No statistically significant difference was found in miR-10b RERs between patients with synchronous (N = 11) and metachronous metastasis (N = 30) (t-test p = 0.096).

In the 41 patients with synchronous or metachronous distant metastases, the group of patients with brain metastases (n = 6) had significantly higher miR-10b RERs (Median 0.47; IQR 0.20-1.62) as compared with patients (n = 35) showing metastases in other organ sites (Median 0.10; IQR 0.03-0.61) (t-test p = 0.043) (Additional file1: Table S1).

We evaluated by IHC the expression of miR-10b target gene HOXD10 in three normal breast tissues from reductive mammoplasty and 10 paired normal and tumor tissues (Additional file1: Table S1). HOXD10 was constitutively expressed in normal ductal and lobular epithelium (Figure 1a). In tumor tissues HOXD10 was variably expressed with tumors showing a diffuse immunostaining (Figure 1c-d) and tumors with a low percentage of stained cancer cells (Figure 1b). A statistically significant inverse correlation was found among miR-10b expression levels and percentage of HOXD10 expressing cells (Spearman Rho −0.713 p < 0.001).

The association with survival was evaluated by using miR-10b RERs values as a continuous variable in the group of patients without metastases at diagnosis (n = 90). In univariable Cox regression model, patients with higher miR-10b RERs showed an increased risk of disease progression (HR1.16, p = 0.021), metastases development (HR1.17, p = 0.019), and cancer related death (HR1.20; p = 0.015) (Table 2). Other factors associated with outcome in univariable analysis are shown in Additional file1: Table S2. A multivariable Cox regression model adjusted for tumor size, lymph node metastases, Grade, ER and PgR status, and KI67 labeling index (n = 79), confirmed that increased miR-10b RERs were associated with higher risk of disease progression (HR1.30; p < 0.001), distant metastases (HR1.34; p < 0.001) and worse overall survival (HR1.31; p = 0.003) (Table 2). The assumption of proportional hazards was satisfied.

The association of miR-10b RERs with response to therapy in terms of DFS and MFS was also evaluated. In univariable Cox regression analysis, a statistically significant association between higher RERs and risk of metastases development was found in the subgroup of patients treated with hormone therapy in association with chemotherapy (HR1.22, p = 0.039). A trend toward an association was found for the same subgroup with DFS (HR1.18, p = 0.063) (Additional file1: Table S3).

We evaluated whether the addition of miR-10b RERs to the model with NPI index alone was able to provide improvements in discriminatory power and risk reclassification abilities for the clinical outcomes at 36 months. As shown in Table 3, the survival C-indices significantly increased from 0.849 to 0.889 (p = 0.009) for OS and from 0.735 to 0.767 (p = 0.050) for DFS with the inclusion of miR-10b RERs, along with a very good calibration (all HL p-values were greater than 0.94) for the OS and DFS outcomes, respectively. Furthermore, the addition of miR-10b RERs to the NPI for OS allowed to correctly reclassify 31 out of 89 patients, where 1 of 11 were events (10.4%) and 30 of 70 were non-events (43.4%), providing a cNRI of 0.538 (p = 0.061).

This study is part of a single institution project initiated in 2006, aimed to the identification of novel biomarkers predicting disease progression and metastases development in breast cancer patients. The study is conducted according to the REporting of tumor MARKer Studies (REMARK) guideline[26] and a prospectively written research, pathologic evaluation, and statistical analysis plan. Paired breast cancer and normal mammary tissues are collected at the Breast-Unit, IRCCS “Casa Sollievo della Sofferenza”. Upon receipt from surgery, tissue from the bulk of the tumor, and normal breast tissue at least 2 cm distant from cancer are sampled by a pathologist (MC), immediately frozen in liquid nitrogen and stored at −80°C until used. For legal reason only one normal and one tumor specimen (approximately 50–100 mg of frozen tissue in weight) can be collected from each patient. Prior written and informed consent is obtained from each patient in accordance with Institutional Guidelines. In order to be included in the study, patients must be female, aged more than 18 years, and tumor must be more than 1.0 cm in diameter due to legal reasons. We selected among the 257 breast cancer cases collected from January 2006 to December 2011, 150 consecutive cases with at least 36 months follow-up (Figure 2). For each case a 5 μm eosin/ematoxylin stained section was prepared to ensure that each tumor sample contained more than 70% of cancer cells and to confirm the absence of tumor cells in the normal specimen. After this analysis, 113 samples were suitable for RNA extraction. Additional 12 cases were excluded because RNA showed a RNA Integrity Number (RIN) <7.0 (n = 101).

Pathological assessment includes evaluation of histological type, grade and stage. Estrogen Receptor (ER), Progesterone Receptor (PgR), KI-67 labelling index and HER2 expression were evaluated by immunohistochemistry[27, 28]. The Nottingham Prognostic Index (NPI) score was calculated according to the following formula: NPI = 0.2xT(cm) + N(1–3) + G(1–3), where T is the maximum diameter in cm, N the number and the level of node metastases (1 = no positive axillary lymph nodes; 2 = 1–3 positive axillary lymph nodes or involvement of a node in the internal mammary chain; 3 = more than three positive axillary lymph nodes or involvement of both axillary and internal mammary lymph nodes) and G the Elston and Ellis grade. Patients are classified at low risk for NPI less or equal to 3.4, at intermediate risk for NPI between 3.4 and 5.4, and at high risk for NPI over 5.4[29].

According to Trizol reagent protocol (Life Technologies) 80 mg of frozen specimen were carefully and mechanically homogenized and the mixture was transferred into a clean 1.5 ml tube using a sterile scraper. Total RNA was extracted from samples using the TRIzol reagent according to the manufacturer’s instructions. RNA was eluted in RNAse free-water and stored at −80°C until used. RNA quality was measured by using 2100 Expert Analyzer (Agilent Technology) and only RNAs with RNA Integrity Number (RIN) ≥7.0 were processed. RNA concentration was quantified by the absorbance measurement at 260 and 280 nm using the NanoDropTM.1000 spectrophotometer (NanoDrop Technologies).

Single-stranded cDNA was synthesized from 5.5 ng of total RNA using 50 nM specific stem-loop RT primers (miR-10b P/N 4373152 and RNU48 P/N 4373383), 1X RT buffer, dNTPs (each at 0.25 mM), 0,25 U/μl RNase inhibitor and 3.33 U/μl MultiScribe reverse transcriptase. 15 μl reactions were incubated in a GeneAmp PCR System 9700 Thermocycler at 16°C for 30 min, 42°C for 30 min and 85°C for 5 min. RT positive and negative controls were included in each batch of reactions. All reagents were purchased from Life Technologies.

A relative quantification method with standard curve was developed to determine miR-10b expression in tissues[30]. PCR fragments for the miR-10b and for RNU48 endogenous control were generated using TaqMan miRNA assay (miR-10b P/N 4373152 and RNU48 P/N 4373383, Life Technologies), cloned in the StrataClone™ PCR Cloning Vector pSC-A (Stratagene®) and introduced in StrataClone™SoloPack® Competent Cells. Plasmid DNA from the selected transformant cells was isolated by using the QIAprep® Spin Miniprep Kit (Qiagen) and linearised with Not I (Amersham) Concentration value of plasmid DNA was measured by spectrophotometry and the plasmid copy number was calculated using the following formula: (X μg/μl plasmid DNA/(plasmid and insert length) × 660 g/mole) × 6.023 × 10²³ = Y molecular number/μl. X represents the concentration of recombinant plasmid DNA, 660 g/molecule the average MW of a double-stranded DNA molecule and Y represents copy number. Five plasmid dilutions of pSC-A_miR-10b and pSC-A_RNU48 (in the range of 1 × 10⁶ copies to 1 × 10² copies) were used to construct the five points calibration curves for real-time PCR.

Real-time PCR reactions were performed in 384-well plates on ABI PRISM 7900HT Sequence Detection System (Life Technologies). 10 μl of reaction mix contained 0.5 μl of TaqMan microRNA assay mix, 5 μl of TaqMan Universal PCR Master Mix, No AmpErase® UNG (Life Technologies) and 1 μl of template. PCR conditions were as follows: at 95°C for 10 min, following by 40 cycles (95°C for 15 s, 60°C for 1 min). Each plate included the miR-10b and RNU48 calibration curves, paired normal and tumour cDNA samples from patients, positive and negative controls of reverse transcription and multiple water blanks; all samples were run in triplicates. The analysis was performed by using SDS 2.4 software (Life Technologie). Standard curves were constructed by plotting the threshold cycle (Ct) values against logarithm10 of the copy number and fitting by linear least square regression. The level of miR-10b expression in each sample was determined as the ratio of the miR-10b copy number to the RNU48 copy number and then multiplied by 1000 for easier tabulation ((miR-10b/RNU48) × 1000).

For each patient, Relative Expression Ratio (RER) was determined as the ratio of miR-10b expression level in the tumor sample to its expression level in the paired normal tissue as previously described[31].

Efficiency of amplification was calculated for each real-time PCR run for both miR-10b and RNU48 as follows: E = (10^(−1/slope)-1) using the slope of the standard curve plots of Ct versus log input of cDNA. The average slope (s) of the standard curves was −3.481 ± 0.160 for miR-10b and −3.510 ± 0.172 for RNU48, indicating efficiencies of 0.941 ± 0.051 and 0.927 ± 0.058, respectively (Additional file2: Figure S1a).

One breast cancer case showing high RER and one breast cancer case showing low RER were used to estimate the precision performance of RT-qPCR assay according to the Clinical and Laboratory Standards Institute (CLSI) recommendation. Intra-run variability was assessed by running real-time PCR reactions in triplicate for each paired tumor and normal sample on a 384 well plate. Inter-run variability was assessed by repeating the assay in five independent real-time PCR runs. Intra- and Inter-run variability among RER values were evaluated from the standard deviation and coefficient of variation (CV) (Additional file2: Figure S1b).

Representative tumour blocks were sectioned at 3 μm thickness. Immunohistochemical staining was performed by the streptoavidin-biotin method. Endogenous peroxidase in the section was blocked by incubation with 3% hydrogen peroxide. A rabbit polyclonal antibody against the human HOXD10 (H-80: sc-66926; Santa Cruz Biotechnology) was used as primary antibody at a 1/100 dilution. Sections were incubated with LSAB2 (Dakocytomation, Carpinteria, CA). 3-30-diaminobenzidine was used for colour development and hematoxylin was used for counterstaining. Negative controls were obtained by omitting primary antibody.

Patients’ baseline characteristics were reported as median along with Inter Quartile Range (IQR) or frequencies and percentages for continuous and categorical variables, respectively. Time-to-event analysis was performed for patients without metastases at diagnosis by univariable and multivariable proportional hazards Cox regression models. Models included: miR-10b RERs, T, N, Grade, ER, PgR, HER2 and KI67. Risks were reported as Hazard Ratios (HR) along with their 95% Confidence Interval (CI 95%). Overall Survival (OS) was defined as the time between the enrollment date and cancer related death. Disease Free Survival (DFS) was defined as the time between the enrollment date and the tumor progression. Metastasis Free Survival (MFS) was defined as the time between the enrollment date and the development of distant metastases.

The assumption of proportionality of the hazards was tested by using scaled Schoenfeld residuals[32]. For the miR-10b RERs only, HR were reported for each unitary increment of 0.1 expression level (Additional file3: Table S4).

Predicted risk probabilities were derived from the estimated Cox regression models.

Models’ calibration, i.e. the agreement between observed outcomes and predictions, was assessed using the survival-based Hosmer-Lemeshow (HL) goodness-of-fit test, a chi-squared test based on grouping observations into deciles of predicted risk and testing associations with observed outcomes.

Models’ discrimination, i.e. the ability to distinguish subjects who will develop an event from those who will not, was assessed by computing the modified C-statistic for censored survival data. Comparison between C-statistics was carried out according to Pencina and colleagues[33].

Reclassification improvement for the prediction of the different endpoints offered by mir-10b RER over the NPI was quantified using the survival-based net reclassification index (NRI), following the Kaplan-Meier approach with one-sided bootstrap-based p-values[34, 35]. Since no established risk cut-offs were available for our high risk population, the continuous NRI (cNRI) was used.

Improvements in model discriminatory power and risk reclassification were assessed at a time horizon of 36 months, since it is well established that the peak hazard for both breast cancer recurrence and development of distant metastases falls within 24–36 months from surgery[36].