Prolactin Pro-Differentiation Pathway in Triple Negative Breast Cancer: Impact on Prognosis and Potential Therapy

Triple negative breast cancer (TNBC) is a heterogeneous disease associated with poor clinical outcome and lack of targeted therapy. Here we show that prolactin (PRL) and its signaling pathway serve as a sub-classifier and predictor of pro-differentiation therapy in TNBC. Using immunohistochemistry and various gene expression in silica analyses we observed that prolactin receptor (PRLR) protein and mRNA levels are down regulated in TNBC cases. In addition, examining correlation of PRLR gene expression with metagenes of TNBC subtypes (580 cases), we found that PRLR gene expression sub-classifies TNBC patients into a new subgroup (TNBC-PRLR) characterized by epithelial-luminal differentiation. Importantly, gene expression of PRL signaling pathway components individually (PRL, PRLR, Jak2 and Stat5a), or as a gene signature is able to predict TNBC patients with significantly better survival outcomes. As PRL hormone is a druggable target we determined the biological role of PRL in TNBC biology. Significantly, restoration/activation of PRL pathway in TNBC cells representative of mesenchymal or TNBC-PRLR subgroups led to induction of epithelial phenotype and suppression of tumorigenesis. Altogether, these results offer potential new modalities for TNBC stratification and development of personalized therapy based on PRL pathway activation.


Tissue microarray
Tissue microarray of TNBC cases (BR487) was commercially purchased from US Biomax (Rockville, MD, USA). The TMA includes 37 cases of invasive ductal carcinoma, 2 cases of intraductal carcinoma and 2 cases of medullary carcinoma, 1 case of infiltrating lobular carcinoma and 1 case of lobular ductal mixed carcinoma. The ER, PR and HER-2 status were provided for all the cases. These cases were negative for the three markers. Stage (TNM) and grade were also provided. Biomax assured the accuracy of all the histological and pathological features. For further assurance of the pathological diagnosis all the cases were reviewed by one of the co-authors (anatomical pathologist). For quality assurance, T47D cell line (positive control) and MDA-MB-231 (negative control) were used for IHC of PRLR ( Figure S2) and AR ( Figure S4).

Immunohistochemistry
Slides were baked for 30 minutes at 55C, followed by deparafinization and rehydration. Antigen retrieval was performed in sodium citrate 10mM, pH 6.0 buffers. Slides were incubated with a rabbit polyclonal antibody against PRLR-L (H300) (Santa Cruz #sc-20992) or a rabbit polyclonal antibody against AR (Santa Cruz #CO215), UltraVision LP Detection System HRP Polymer & DAP Plus Chromogen (Thermo Fisher Scientific, Fremont CA) was used for detection. The TMA slides were scanned using Aperio XT slide scanner (Leica Biosystems).
Membranous and/or granular cytoplasmic staining of PRLR in >10% of malignant cell was considered positive. AR immunostaining was assessed using Allred score 39 . Two investigators including one anatomical pathologist, blindly from the clinical data evaluated the slides independently. If there is any discordance, simultaneous examination was performed to solve the differences.

Human breast cancer gene profiling databases and in silica analyses
The Breast Cancer Gene-Expression Miner Version 3.0 (bc-GenExMiner 3.0) database 54 categorized patients according different methods of classification, based on variations in gene expression patterns derived from different cDNA microarrays analysis. Within this database two methods of classification (Hu and Sorlie) were used containing large number of TNBC patients 2,37 as indicated in results section. This database was used to evaluate PRLR gene expression levels (660 patients), prognosis (AEFS) using the two different sub-classification methods mentioned above (patient number for each gene tested is indicated in results and figures) and correlation of expression (Metagene analysis available in 580 patients out of the 660) (as described in results) 13

Cell lysis, immunoprecipitations and western blotting
For whole cell lysates and immunoprecipitations, cells were lysed in lysis buffer as described previously 56 . Immunoprecipitations were done for 3 hr at 4°C using a poly clonal antibody to PRLR and protein A-Sepharose beads. Proteins were run on SDS-PAGE and transferred to nitrocellulose membrane for western blotting analysis using the appropriate antibodies.

Invasion assay
Invasion assays were performed as described previously 26 . Cells were seeded in 24-well plates HTS multiwell insert system format (BD Falcon), 8.0μm pore size. Inserts were coated with Matrigel (BD biosciences) 1/10 dilution. 80 x10 3 cells were plate inside the inserts in 500μl with a 2% FBS growth medium. The bottom of the well was filled with 10% FBS growth medium.
Cells were treated for 24-72 hours with rhPRL 250ng/ml and doxycycline (100 ng/ml), plated and left to invade for 24 hours. Invaded cells were stained with 1ml of 0.2% Crystal Violet (Fisher) for 1 hr. Images collection, processing and automated cell counting was then performed (McGill University Life Sciences Complex Advanced BioImaging Facility (ABIF)). Results are presented as means ± standard errors of the means (SEM) performed as triplicates of three independent experiments. Statistical significance was assessed using student's t test analysis.

RNA extraction and qRT-PCR
MDA-MB-231/PRLR and MDA-MB-231/vector cells grown to confluence then were pre-treated with doxycycline (10-100ng/ml) before stimulation. Before ligand stimulation, cells were starved in DMEM (2% FBS) and stimulated with rhPRL 250ng/ml for a period of 72 hrs. Cells were lysed in 500 l of trizol. Total RNA was isolated as described by the manufacturer (Invitrogen Life Technologies, Burlington, Ontario, Canada). Samples were quantified by absorbance at 260 nm. Aliquots of 300 to 400 ng of total RNA were used for reverse transcription and PCR amplification in one step using Brilliant II SYBR green quantitative real-time PCR (qRT-PCR) Master Mix kit, 1-Step (Stratagene Amsterdam, Zui-doost, The Netherlands) according to the manufacturer's recommendation. RT-qPCR of EMT markers (slug, snail, twist, FN1, vimentin, e-cadherin, zeb1) was performed and the specificity of the primers was then tested by a dissociation program at 95°C for 1 min, with a ramp-down to 65°C and then a ramp-up to 95°C
Statistical significance was assessed using student's t test analysis. 5 x 10 3 cells of MDA-MB-453 were seeded into 96-well plate overnight and grown in 2%FBS in L-15 media. Cells were then starved in 2% FBS starvation media and either treated or not with hPRL 250ng/ml for a period of 24-72hrs following the same procedure as mentioned above. Results are presented as means ± standard errors of the means (SEM) performed as triplicates of five independent experiments. Statistical significance was assessed using student's t test analysis.

Statistical analysis:
Pearson correlation coefficient was used to evaluate the correlation between PRLR and different members of the metagenes used to distinguish molecular heterogeneity of TNBC. Correlation coefficient was denoted as (r). P value is also provided to evaluate the linear dependence between the two genes. Any event free survivals (AEFS), relapse free survival (RFS) curves in different databases were plotted using the Kaplan-Meier method. In vitro assays were all performed in triplicates of at least three independent experiments. Results were shown as means ± SEM. Student's t-test was used to evaluate the statistical significance. Statistical analyses were performed using GraphPad Prism software.
Tumor volume was measured in two dimensions with a vernier caliper (Mitutoyo, Kawasaki, Japan) and calculated using the formula [length x width 2 ] /2. Mice were euthanized by cervical dislocation following 8 weeks of treatments. For measuring serum levels of injected rhPRL, mice were anesthetised and blood was collected by cardiac puncture. Levels of serum rhPRL were determined using radioimmunoassay at 2hrs and 4hrs post injection (Table S1).

MDA-MB-453 xenograft animal models
Female NOD/SCID mice (12 in total) were purchased from Charles River Laboratories (Saint-Constant, QC, Canada) housed and maintained under specific pathogen-free conditions (RI-MUHC animal facility). The mice were randomly assigned into two groups (n=6 mice/group): MDA-MB-453 untreated and MDA-MB-453 rhPRL treated. Cells (5×10 6 /ml) were re-suspended in Matrigel and implanted subcutaneously into the right flank of each mouse. The mice were treated intra-peritoneal with either vehicle or hPRL (0.1g/g) each second day. Tumor growth was monitored up to 8 weeks after implantation. PET/SPECT/CT scan was performed (please see below) on three mice from each group. At the end of the experiment mice were sacrificed by CO2 asphyxiation and subjected to necropsy.

Whole-body imagines of NOD/SCID/xenograft mice with PET/SPECT/CT scan analysis
The animals were kept fasting for approximately 12hrs before undergoing scanning. They were anesthetized using 1 L/min of 2% isoflurane in 100% oxygen. After receiving the anesthesia, they were injected in the tail vein with the radiotracer (fludeoxyglucose [ 18 F]) (FDG). Following an appropriate uptake period of 45 minutes, animals were re-anesthetized (isoflurane, 5% induction, 1.5-2% maintenance throughout the scan), moved to the Mediso nanoScan and placed in the prone position on an animal bed covered with absorbent paper. Following completion of PET and/or SPECT scans; a CT scan was performed for anatomical localization and attenuation correction. Respiration rate and body temperature was continuously monitored and the temperature maintained at ~37°C throughout the study using a feedback-regulated warming system. Following completion of scanning, the animals were euthanized. Data acquired in listmode format for 60min, full 3D sonograms with corrected efficiency, scattering, attenuation, count losses and decay were reconstructed using an iterative 3D dynamic raw-action maximum likehood algorithm (Drama). After PET scanning a CT scan was performed. Analysis of the PET/SPECT and the CT datasets were imported using the nonproprietary AMIDE software,