A co-culture genome-wide RNAi screen with mammary epithelial cells reveals transmembrane signals required for growth and differentiation
© Burleigh et al.; licensee BioMed Central. 2015
Received: 31 October 2013
Accepted: 18 December 2014
Published: 9 January 2015
The extracellular signals regulating mammary epithelial cell growth are of relevance to understanding the pathophysiology of mammary epithelia, yet they remain poorly characterized. In this study, we applied an unbiased approach to understanding the functional role of signalling molecules in several models of normal physiological growth and translated these results to the biological understanding of breast cancer subtypes.
We developed and utilized a cytogenetically normal clonal line of hTERT immortalized human mammary epithelial cells in a fibroblast-enhanced co-culture assay to conduct a genome-wide small interfering RNA (siRNA) screen for evaluation of the functional effect of silencing each gene. Our selected endpoint was inhibition of growth. In rigorous postscreen validation processes, including quantitative RT-PCR, to ensure on-target silencing, deconvolution of pooled siRNAs and independent confirmation of effects with lentiviral short-hairpin RNA constructs, we identified a subset of genes required for mammary epithelial cell growth. Using three-dimensional Matrigel growth and differentiation assays and primary human mammary epithelial cell colony assays, we confirmed that these growth effects were not limited to the 184-hTERT cell line. We utilized the METABRIC dataset of 1,998 breast cancer patients to evaluate both the differential expression of these genes across breast cancer subtypes and their prognostic significance.
We identified 47 genes that are critically important for fibroblast-enhanced mammary epithelial cell growth. This group was enriched for several axonal guidance molecules and G protein–coupled receptors, as well as for the endothelin receptor PROCR. The majority of genes (43 of 47) identified in two dimensions were also required for three-dimensional growth, with HSD17B2, SNN and PROCR showing greater than tenfold reductions in acinar formation. Several genes, including PROCR and the neuronal pathfinding molecules EFNA4 and NTN1, were also required for proper differentiation and polarization in three-dimensional cultures. The 47 genes identified showed a significant nonrandom enrichment for differential expression among 10 molecular subtypes of breast cancer sampled from 1,998 patients. CD79A, SERPINH1, KCNJ5 and TMEM14C exhibited breast cancer subtype–independent overall survival differences.
Diverse transmembrane signals are required for mammary epithelial cell growth in two-dimensional and three-dimensional conditions. Strikingly, we define novel roles for axonal pathfinding receptors and ligands and the endothelin receptor in both growth and differentiation.
The identification of distinct cell types that appear to be hierarchically organized in the mammary epithelial glands of healthy women is now well established . This hierarchy is defined largely by two prospectively separable subsets of cells that generate colonies containing only one or both lineages (myoepithelial and/or luminal) of cells that make up the bulk of the normal mammary gland structure. The bipotent, clonogenic, progenitor-enriched basal cell fraction also contains putative human mammary stem cells identified in xenotransplantation assays [2,3]. The ability of human mammary cells to be propagated both in vitro and in vivo at limited densities is known to be markedly enhanced by the presence of fibroblast ‘feeders’ [2,4,5]. These and many other studies have shown that fibroblast interactions are important to the growth of mammary epithelial cells [6-12]. However, a comprehensive characterization of the mechanisms by which fibroblasts regulate the growth and functional organization of normal mammary epithelial cells has been lacking.
Genome-wide RNA interference (RNAi, small interfering RNA (siRNA)) screens offer an attractive strategy by which to investigate such questions. They have previously been used with success to identify mediators of Ras oncogene-induced senescence, suppressors of p16 gene expression, genes that regulate cell migration and cell survival genes in mammary cells [13-16]. This type of investigation is nevertheless dependent on a source of cells that can be obtained in large numbers and readily transfected. Because primary normal mammary epithelial cells, even those derived from human mammoplasties, do not satisfy either of these requirements, we sought an alternative in a clonal diploid isolate of hTERT-immortalized cells  that we found remains dependent on fibroblast stimulation for its rapid growth when cultured at low density. By combining automated imaging with siRNA screening of these cells, we identified 43 signal-transducing receptors and secreted factors with functionally validated roles in mediating the in vitro growth of primary normal human mammary epithelial cells.
Passage 6 184-hTERT polyclonal infection pool mammary epithelial cells (obtained from ) were contributed to the study by CB and LA. As described previously , these pools were generated from anonymised primary mammary epithelial sample 184 (see ) and not subject to specific institutional review board approval. We generated the monoclonal cell lines (184-hTERT-L9 or 184-hTERT-E11) and used the 184-hTERT-L9 cell line to generate subsequent polyclonal cell lines (stably infected with lentiviruses or NucLight Red (Essen BioScience, Ann Arbor, MI, USA), for example). The experiments were conducted under University of British Columbia Research Ethics Board protocols H06-0289, H06-0210 and B13-0126.
Passage 6 184-hTERT cells  were cloned in 96-well plates and subcultured in serum-free mammary epithelial cell basal media (MEBM; Lonza, Walkersville, MD, USA) supplemented with the mammary epithelial cell growth media in the SingleQuots kit (Lonza), 5 μg/ml transferrin (Sigma-Aldrich, St Louis, MO, USA) and 10−5 M isoproterenol (Sigma-Aldrich), referred to as mammary epithelial cell growth medium (MEGM).
Multicolour fluorescence in situ hybridization (FISH) was performed as previously described . Immunofluorescence cell staining in three-dimensional Matrigel cultures was performed as previously described  with primary antibodies to GM130 (BD Biosciences, San Jose, CA, USA), CD49f and MUC1 (STEMCELL Technologies, Vancouver, BC, Canada), as well as Alexa Fluor 680–conjugated secondary antibodies (Invitrogen, Carlsbad, CA, USA). Cells were counterstained with Oregon Green 488 or Alexa Fluor 546 phalloidin (Invitrogen) and DRAQ5 nuclear staining prior to imaging on a confocal laser scanning microscope (Nikon Instruments, Melville, NY, USA). For calcein acetoxymethyl ester (calcein AM) and ethidium homodimer 1, 21-day Matrigel cultures were stained unfixed for 20 minutes and counterstained with Hoeschst 33342 (Invitrogen). Immunofluorescence staining of cells in three-dimensional Matrigel cultures cultured for 3 weeks was performed with primary antibodies to E-cadherin (E-cad; Calbiochem, San Diego, CA, USA), GM130 (BD Biosciences), CD49f (STEMCELL Technologies) and Alexa Fluor 680–conjugated secondary antibodies (Invitrogen) and imaged on a Nikon confocal laser scanning microscope. Colonies were counted at five discrete, randomly chosen positions per well using a Nikon confocal laser scanning microscope. Only discrete, well-separated structures were counted. In the cases where two colonies touched or merged, both colonies were ignored for counting purposes. For caspase 3 staining, three-dimensional Matrigel cultures were formalin-fixed and paraffin-embedded, and sections were stained with caspase 3 antibody (Cell Signaling Technology, Danvers, MA, USA). To quantify the proportion of structures with wild-type epithelial organization and polarization, RNAi-treated samples were scored and compared to wild-type localization (see Additional file 1: Figure S7 and Additional file 2: Figure S8A) for examples of each category. For CD49f, the presence of a single basement membrane type of immunoreactive structure was considered the wild type. For GM130 immunoreactivity, wild-type polarization was deemed to be signal-localized between the edge of the colony and an unstained lumen. Lack of polarization would be reflected in the indistinguishable staining patterns between the outer and inner cell layers.
Genome-wide siRNA screen protocol
Black-walled clear-bottom 96-well plates (Greiner Bio-One, Monroe, NC, USA) were seeded with 3,000 freshly irradiated NIH 3T3 cells per well in Dulbecco’s modified Eagle’s medium with 5% foetal bovine serum. After 24 hours, the medium was aspirated at a low flow rate with a multichannel vacuum aspirator and replaced with 550 184-hTERT-L9 cells per well in MEGM without bovine pituitary extract (BPE) added. After an additional 24 hours, Lipofectamine 2000 reagent–siRNA complexes were generated.
siRNAs were purchased from the siGENOME library (Dharmacon, Lafayette, CO, USA) as deconvolved sets of four individual siRNAs and resuspended at 10 μM in 1× siRNA buffer (Dharmacon) as described elsewhere . Lipofectamine 2000 transfection reagent (Invitrogen) was diluted in MEBM and incubated for 5 minutes prior to mixture with an equal volume of prediluted siRNA in MEBM. Complexes were allowed to form for 20 minutes before they were added directly to the cells at a final concentration of 30 nM siRNA and 0.3 μl of transfection reagent per well. The control wells were static in position and were composed of Lipofectamine alone, siControl-3 and a siRNA pool targeting PLK1. The entire 96-well plates of these controls were staggered throughout the duration of this screen to allow for statistical correction of plate positional effects. After an additional 4 days of growth, cells were fixed with 4% paraformaldehyde and stained with 1 μg/ml 4′,6-diamidino-2-phenylindole (DAPI) prior to being imaged on an IN Cell Analyzer (GE Healthcare Bio-Sciences, Pittsburgh, PA, USA) using the 10× lens objective (numerical aperture, 0.45) with charge-coupled device pixel binning. Twenty-one fields per well were collected using HQ360/40 excitation and HQ535/50 emission filters with a multi-bandpass dichroic filter (Q505lp; Chroma Technology, Bellows Falls, VT, USA). For each image field, a single focal plane was captured using a hardware (laser/photodetector) autofocusing algorithm, which estimated the surface area on which the cells were lying. An image segmentation and postprocessing software workflow was developed using CellProfiler . Cell counts per well produced in the Cell Profiler image analysis were further processed using quantile normalization to correct for data distributional differences induced by factors such as time-dependent stain degradation and other plate-handling artefacts. Quantile-normalized data were then analysed using statistical linear mixed effects regression models to compare cell counts under the siRNA condition with counts under the control condition, adjusted for technical artefacts of well position and plate effects, thereby yielding plate-normalized growth effect estimates for each siRNA. Well position effects were assessed by screening additional plates containing the same control condition in all wells. Plate effects were assessed by using multiple plates for each condition, thereby allowing for adjustment of plate-to-plate variability in the statistical model. The model-estimated cell count under the siRNA condition was then divided by the model-estimated cell count under the control condition to produce an overall measure of relative effect. Measured effects were ranked from smallest to largest, and the latter (those reducing cell growth by 75% or more) were selected for further study. Enrichment network analysis was performed as previously described  using the Reactome Functional Interactome plugin in Cytoscape v2.8.1 [23,24].
Lentiviral transduction procedure
184-hTERT-L9 cells were transduced at an estimated multiplicity of infection (MOI) of 5:1 with 8 μg/ml polybrene. After 18 hours at 37°C, cells were washed, and, after 24 hours in MEGM, they were selected with 2 μg/ml puromycin in MEGM (replaced every 24 to 48 hours). Dissociated primary human mammary epithelial cells were infected with an estimated MOI of 10. Transduction was conducted in suspension with 8 μg/ml polybrene at a density of 5 × 105 cells in 100 μl of serum-free 7 medium (DMEM/F12 (STEMCell Technologies) supplemented with 5% FBS, 2 mM glutamine (Gibco), 0.1% w/v BSA, 10 ng/ml EGF (Sigma), 10 ng/ml cholera toxin (Sigma), 1 μg/ml insulin (Sigma), 0.5 μg/ml hydrocortisone (Sigma)) with 50 μg/ml GA-1000 (Lonza). After 18 hours at 37°C, cells were washed and counted and equal numbers plated into two-dimensional colony-forming cell assays.
Primary mammary tissue
Discarded tissue was collected from premenopausal women (ages 19 to 40 years) who provided informed consent as approved by the University of British Columbia Research Ethics Board, as previously described . Suspensions selectively enriched in bipotent progenitor cells were obtained by fluorescence-activated cell sorting (FACS) of cells positively costained with an allophycocyanin-conjugated rat antibody to human CD49f (eBioscience, San Diego, CA, USA) and a phycoerythrin-conjugated mouse antibody to human epithelial cell adhesion molecule (EpCAM; eBioscience) . Hematopoietic stem cells and endothelial cells were eliminated using antibodies to human CD45 and human CD31 (eBioscience), respectively.
Proliferation measurement by 5-ethynyl-2′-deoxyuridine incorporation
To determine the proliferation rate, 184-hTERT-L9 cells stably infected with NucLight Red were seeded into wells of 24-well plates (16,750 cells per well). Twenty-four hours after plating, the cells were transfected with siRNAs (30 nM concentration) and cultured in the IncuCyte ZOOM incubator (Essen BioScience). Twenty hours after transfection, the siRNA-containing medium was removed and replaced with standard 184-hTERT culture medium. Sixty-eight hours after transfection, 5-ethynyl-2′-deoxyuridine (EdU) was added to each well to a final concentration 40 μM. After 1 hour, wells were washed with phosphate-buffered saline, then the cells were harvested with a trypsin/ethylenediaminetetraacetic acid mixture. To ensure sufficient cell numbers for FACS, 1 × 105 nonfluorescent 184-hTERT cells were added to the cells harvested from each well. EdU incorporation was then detected by staining with the Click-iT EdU Alexa Fluor 488 Flow Cytometry Assay Kit (Life Technologies, Grand Island, NY, USA) according to the manufacturer’s instructions, followed by FACS acquisition on a BD FACSAria III cell sorter (BD Biosciences). DAPI was used to detect DNA content and to gate out cell fragments. The percentage of test cells that incorporated EdU was determined by gating on the NucLight Red–positive cell population and applying an EdU-positive gate set with reference to negative control (carrier) cells processed with the Click-iT EdU assay but not previously exposed to an EdU pulse.
Determination of proliferation and apoptosis by live cell imaging
Where appropriate, 184-hTERT-L9 cells were stably infected with NucLight Red, which marks the nuclei red to distinguish them from unlabelled irradiated feeders used the in co-culture experiments. To control for feeder effects, we conducted the experiment using both feeder-free and feeder-containing conditions, with BPE medium supplement (see details in methods) used for the nonfeeder condition. For the feeder-present condition, 3,000 irradiated feeders per well were plated in 96-well plates 24 hours prior to plating 550 184-hTERTs stably expressing NucLight Red. Transfection of 30 nM siRNA (complexing as described in methods) was performed 24 hours later, and the complexes were washed out after 24 hours, with the addition of CellPlayer Caspase-3/7 reagent (Essen BioScience), which labels apoptotic cells green, detecting caspases 3 and 7. The plate was then imaged using an IncuCyte ZOOM live cell microscope (Essen BioScience), and images were taken every 4 hours for an additional 84 hours. The data were analysed as follows. Triplicate red cell objects (representing hTERT with NucLight Red) and green objects (apoptotic cells with activated caspase 3, as the percentage of red objects) were counted (counts per square millimetre) using the GraphPad Prism statistical software suite (GraphPad Software, La Jolla, CA, USA), and the respective areas under the curve (AUCs) for serial measurements were calculated. Where comparisons were made, the AUC values were subjected to a one-way analysis of variance test to compare the mean of each siRNA to the control condition, in which only Lipofectamine 2000 transfection reagent was used to determine significance.
Gene association analyses
Expression and patient outcome data for 1,996 breast cancer patients were obtained from the Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) study (European Genotype-phenome Archive study accession number: EGAS00000000083) . Relative hazard estimates for cases with high versus low gene expression were obtained using a Cox proportional hazards model with the binarized expression variable and stratified by integrative cluster (IntClust) breast cancer subtype groups to mitigate for the nonproportional hazards exhibited between various IntClust groups. Expression variables were binarized at the 15%, 25%, 50% and 75% quantiles, and Akaike information criteria (AIC) were calculated using a Cox model containing the binarized expression variable. The binarization cut point was chosen as the quantile yielding the minimum AIC value. In the case that the minimum and maximum AIC values differed by less than 3.0, the median expression level was used as the cut point to ensure adequate case counts within the breast cancer subtype groups. To identify potential instances of interaction between expression level and breast cancer subtype (high versus low expression showing improved survival in one subgroup and poorer survival in another subgroup), a Cox model with binarized expression, IntClust subgroups and their interaction terms was fitted and compared to a Cox model containing IntClust subgroups only, yielding an omnibus test of survival difference due to the biomarker. P-values across all 46 fitted models were adjusted for multiple comparisons using the method of Benjamini and Hochberg , and significant findings after adjustment for multiple comparisons were identified. To guard against the issue of changing hazards over time and nonproportional hazards between IntClust subgroups, each binarized biomarker was tested using the G-rho rank test procedure stratified by IntClust subgroups and setting ρ = 1 to place heavier weight on earlier time point observations . P-values from all 46 G-rho tests were adjusted for multiple comparisons, and genes with low false discovery rates were identified.
For analysis of transcripts in flow-sorted mammary epithelial cell lineages, we made use of the NIH and Canadian Roadmap Epigenomics mammary epithelial cell RNA-seq libraries. The consortium data generation protocols and data can be accessed online [28,29]. The flow sorting of mammary epithelial subsets was performed using CD10, MUC1 and CD73 as described previously .
where e ijk is the normalized expression value for subject i, gene target j, in data centre k; med ik is the median expression value for gene target j in data centre k; and igs jk is the interquartile spread for gene target j in data centre k. (Interquartile spread is the difference between the 75th-percentile value and the 25th-percentile value, a robust measure of standard deviation.) Heat maps of this robustly centred and scaled data were generated with complete linkage clustering of gene targets indicated on the vertical axis. To yield comparable colour schemes across all genes in the heat map, values less than −3.0 were set to −3.0 and values greater than 3.0 were set to 3.0. Expression value differences among breast cancer subtype groups (IntClust and PAM50) were assessed using the Kruskal-Wallis test. Beanplots  of expression values within breast cancer subtype groups for each of the 47 final identified gene targets were ordered in Kruskal-Wallis test P-value order.
Isolation and characterization of a cytogenetically normal clone of 184-hTERT cells
We next compared the three-dimensional growth characteristics of 184-hTERT-L9 with flow-sorted primary mammary epithelial cells and with MCF10A cells . When placed into three-dimensional Matrigel culture, 184-hTERT-L9 cells formed spherical, multilayered acini with apicobasal polarity, as determined by staining with antibodies raised against apical marker GM130 and basal marker CD49f (Figure 1B). The immunoreactivity of inner cells with antibodies to the luminal marker protein MUC1 suggests that differentiation is multilineage. A key element of three-dimensional mammary differentiation is the death of cells forming the lumen of spherical acini. This was observed using calcein AM and ethidium homodimer 1 staining of unfixed three-dimensional cultures (Figure 1B). As with primary bipotent mammary epithelial progenitors , 184-hTERT-L9 acini exhibit squamous differentiation of the inner cells in three-dimensional Matrigel cultures, which is appreciable by the morphology of the inner cells stained with phalloidin (Figure 1B). The similarities and differences between 184-hTERT-L9 and MCF10A cells are shown in Additional file 4: Table S1.
Defining the genes required for mammary epithelial cell growth under co-culture conditions
We therefore focused on further analysis of genes with subcellular location annotations  (Additional file 7: Table S2B) that indicated the presence of transmembrane, extracellular, and secreted proteins (388 genes with greater than 75% primary growth inhibition). To control for clonal cell line effects, we performed a secondary screen with these 388 genes (using a co-culture assay as primary screen) with both 184-hTERT-L9 and 184-hTERT-E11, a sister clonal cell line to 184-hTERT-L9 with identical growth characteristics, but with a different hTERT integration site. Of these 388 genes, 140 (36% of those retested) were considered reproducible in that they produced the same magnitude of growth reduction upon reassessment in the secondary screen (Figure 3C) in one or both clonal cell lines. We determined which of the 140 screen-reproducible siRNA pools were likely acting on target initially by quantitative RT-PCR (qRT-PCR) to determine which siRNA pools resulted in significant target transcript knockdown 48 hours after transfection into 184-hTERT-L9 cells. For this purpose, cells were grown in BPE as opposed to co-culture with irradiated feeder cells to reduce interfering mRNA signals. We found that 47 (33.6%) of 140 of the siRNA pools achieved statistically significant mRNA silencing (adjusted P < 0.05 by Benjamini-Hochberg analysis) (Additional file 7: Table S2C), and these were designated as the target genes for further study.
Additional validation was conducted as follows. (1) We deconvolved each siRNA pool to individual siRNAs and noted that in 20 (42.5%) of 47 of the siRNA sets, at least three of the individual siRNAs produced a statistically significant decrease (adjusted P < 0.05 by Benjamini-Hochberg multiple comparisons method) (Additional file 7: Table S2C) in growth within the two-dimensional co-culture assays. (2) We tested growth inhibition using independently designed lentiviral short-hairpin RNA (shRNA) constructs for each of the 47 target genes. One to three shRNA constructs from the GIPZ human lentiviral shRNA library (Dharmacon) were selected for analysis , and shRNAs were transduced in BPE-supplemented medium to maximize the potential of rescuing growth and formation of stable clones. Stably growing clones could not be isolated from 29 shRNAs covering 24 genes, suggesting the inability of these cells to grow upon gene silencing of these genes (verified by qRT-PCR in short-term culture; see Additional file 7: Table S2D), even with BPE present in the medium. Even among the stable clones derived, fibroblast-dependent growth was reduced. In total, 34 (72.3%) of 47 of the transmembrane and/or extracellular space protein encoding transcripts appeared to be required for epithelial cell growth, as determined by siRNA pool deconvolution and/or by knockdown with a second gene RNAi method (shRNA) (Additional file 6: Figure S2B and Additional file 7: Table S2C).
The increase and decrease in cell numbers observed in the primary and secondary screens likely resulted from a mix of proliferation (cell division) and apoptosis. To determine the relationship and relative contribution of these two factors, we quantified caspase 3/7 activity and proliferation simultaneously by using high-content live cell imaging (see IncuCyte ZOOM description in the Methods section). We observed that there was an inverse correlation (Additional file 8: Figure S4B (rank correlation −0.954) and Additional file 9: Figure S5 (rank correlation, −0.959)) between caspase activity and the rate of cell increase (determined as AUC over time). We verified that cell proliferation measured by high-content live cell imaging as AUC over time was positively correlated with the fraction of cells in S-phase, which we determined by EdU incorporation (Additional file 8: Figure S4A (rank correlation, 0.735)). The relationship between caspase activation and proliferation was similar, regardless of whether the epithelial cells were grown in feeder-free conditions (Additional file 8: Figure S4B) or with a feeder monolayer (Additional file 9: Figure S5) (feeder to no-feeder rank correlation, 0.892; P < 1 × 10−5 in randomization test). Taken together, these data show that the genes of greatest effect on cell growth tend to affect both proliferation (cell cycle) and caspase activation (apoptosis). Some exceptions were noted, such as the EDG (LPAR3) receptor, where the effect on proliferation was higher in rank than the degree of apoptosis.
Functional annotation clustering was performed using the DAVID Bioinformatics Database, with enrichment based upon the Gene Ontology biological process for the 47 genes required for 184-hTERT-L9 growth . Enrichment scores were generated based upon the functional classification of these genes and denote the relatedness of a seemingly heterogeneous group of genes (Additional file 7: Table S2F). Notably, a number of guidance factors required for neuronal development are necessary for fibroblast-driven mammary epithelial cell growth [40,41]. Also enriched are clusters of genes involved in axon guidance, signal transduction through protein kinase cascades, intracellular ion homeostasis, GPCR signalling and cell migration (Additional file 10: Figure S6B). Additional file 10: Figure S6A displays the axon guidance pathway in which SEMA3C, and part of its receptor complex, PLXNA2, ROBO3, EFNA4, NTN1 and NTN2L from our target gene list, are all highlighted. Axon guidance molecules are increasingly recognized to play a role in mammary gland development and breast tumourigenesis , and our data significantly strengthen this association. We considered the possibility that some of the 47 genes identified may have differential expression in the luminal and basal developmental lineages, and therefore we inspected flow-sorted RNA-seq libraries of mammary lumen and myoepithelium (three independent pairs from the RNA-seq library; see Additional file 11: Table S3 and discussion of NIH Canadian Roadmap Epigenomics mammary epithelial cell RNA-seq libraries in the Methods section). However, no statistically significant differences were observed (Holm-Bonferroni-adjusted P-values >0.05).
Distinct requirements for transmembrane and/or extracellular genes in three-dimensional epithelial cell growth and differentiation
The requirements for three-dimensional growth are distinct from monolayer cultures, and the process of acinar formation can be disrupted for many reasons. In the second approach, we sought to determine if three-dimensional effects were limited solely to growth by looking at acinar formation after 21 days when polarization and lumen formation have occurred. For a 21-day culture, long-term RNAi is required, and we therefore examined shRNA stable lines in which growth could be rescued (for initial clone generation) by BPE media supplements. Among the three genes showing the greatest effects on acinar formation, SNN, HSD17B2 and PROCR (Additional file 7: Table S2C), the shRNAs inhibited growth to a degree precluding derivation of stable clones for all but PROCR. The PROCR stable clone exhibited a 14.1-fold (99% CI, 7.16 to 28.57; RQ Manager software; Applied Biosystems, Foster City, CA, USA) compared to nontargeting controls (Additional file 7: Table S2D). When placed into a quantitative two-dimensional co-culture assay, growth was decreased 3.73-fold (95% CI, 3.38 to 4.16; Student’s t-test) in comparison to a stable cell line generated with a nontargeting shRNA construct.
Requirement of screen-identified transmembrane and/or secreted genes for primary mammary progenitor cell growth
Expression of screen-identified genes among breast cancer subtypes
We examined each gene individually for subtype-specific expression. Significant differences in expression distribution were seen for 40 of the 47 genes across the 10 METABRIC datasets (Figure 9B and Additional file 12: Figure S9) (Kruskal-Wallis-adjusted P-value <0.05, Benjamini-Hochberg multiple-comparisons method) and for 39 of the 47 genes in the PAM50 groups (Additional file 13: Figure S10) . Several genes (for example, RIPK2, EFNA4 and TMEM9B) (Figure 9B, Additional file 12: Figure S9 and Additional file 13: Figure S10) showed differential expression in subtypes associated with high proliferation, such as IntClust 5 (predominantly HER2/ERBB2-amplified cancers) and IntClust 10 (predominantly basal expression type cancers). However, other breast cancer groups (IntClust 4, 6, 7 and 8, predominantly ER+ subtypes) (Figure 9A and B, Additional file 7: Table S2B) also showed significant enrichment for over- and underexpression of several genes. We noted that 21 of 47 genes required for 184-hTERT cells lie within chromosomal hotspots for copy number amplification (Additional file 7: Table S2C). All 47 genes have previously identified mutations in human tumours, and 18 of these genes have known mutations in breast tumour tissues (Additional file 7: Table S2C) [56,57].
To test the group as a whole, we also performed a randomization simulation study to assess the strength of association demonstrated by this set of 47 genes within the 10 IntClust subtypes. In 10,000 simulation runs, selecting 47 genes at random from among the 4,103 transmembrane/extracellular gene set, we found that 211 (0.21%) of the 10,000 of the random sets showed 40 or more genes to be significantly associated with the 10 METABRIC groups (null hypothesis simulation Kruskal-Wallis P = 0.021). For the PAM50 subtypes, 637 (0.637%) of 10,000 of the random sets of 47 genes showed 39 or more genes with an adjusted P-value <0.05 (null hypothesis simulation Kruskal-Wallis P = 0.064). Thus, the identified set of 47 genes is unlikely to be simply a randomly assembled set, suggesting that this gene set is enriched with regard to association with breast cancer subtype.
Finally, we asked whether expression differences in any of the 47 genes exhibit independent disease outcome associations (overall survival) in the 1,998-patient dataset. This was tested using two methods of multivariable analysis to determine if the prognostic significance of the gene expression was independent of that already carried by the breast cancer subtype (see the Methods section for details). The ranked proportional hazards with respect to overall survival (Figure 9C) for each gene showed a range of effect sizes. Two gene targets showed significant differences in survival within IntClust breast cancer subtype groups as assessed by Kaplan-Meier G-rho-stratified analysis and by Cox proportional hazards analysis, after adjustment for multiple comparisons using the method of Benjamini and Hochberg . Elevated CD79A expression showed improved survival for IntClusts 8 and 10 (Additional file 14: Figure S11). Elevated SERPINH1 expression showed poorer survival for IntClusts 6, 9 and 10. An additional two gene targets showed significant differences as assessed by the Kaplan-Meier G-rho-stratified analysis, after adjustment for multiple comparisons. Elevated KCNJ5 expression showed poorer survival for IntClusts 4 and 8. Elevated TMEM14C expression showed improved survival for IntClust 4.
The extracellular factors and transmembrane signals that regulate mammary epithelial growth and differentiation remain poorly understood. In part, this is due to a lack of methods for systematic, genome-wide, genetic and functional interrogation of genes in relation to mammary epithelial growth and differentiation. Although genome-wide functional screens using RNAi methods have proven successful in many similar instances (for example, see [58,59]), these have mostly been undertaken with transformed, somatically mutated epithelial cell types, where key extracellular interactions that modulate growth cannot be recapitulated. Important features of the endogenous milieu, such as growth stimulation by fibroblast stroma have been undersampled as a consequence. To overcome this, we isolated and characterized a cloned, diploid, nontransformed mammary epithelial cell line, 184-hTERT-L9, which retains both fibroblast growth dependence and the capacity to differentiate in three-dimensional growth conditions. We used this cell line in a genome-wide RNAi screen to identify, in a systematic manner, genes required for mammary progenitor cell growth and differentiation.
The 184-hTERT-L9 clone described here is derived from primary mammary epithelium immortalized by hTERT transfection after limited initial passages. The clonal line retains important properties, such as fibroblast-dependent growth and the ability to differentiate in three-dimensional cultures, and is diploid and nontransformed.
Fibroblasts are known to possess an instructive role in regulating mammary epithelial cells in normal development and oncogenesis [6,60,61]. More specifically, the in vitro growth of bipotent progenitor cells is reliant upon the presence of fibroblasts as feeder cells. The 184-hTERT-L9 cells mimic the growth of bipotent progenitor cells when plated with increasing densities of irradiated NIH 3T3 feeder cells in a well-defined colony-forming assay. This is not observed in epithelial cell lines with genomic aberrations and/or additional adaptation events, such as MCF10A cells (Figure 2) and transformed epithelial cells derived from malignancies. Thus, 184-hTERT-L9 cells provided a genomically characterized model system, also amenable to RNAi transfection and image-based, high-content screening, whereby we could replicate in vitro the fibroblast-dependent growth environment of mammary progenitor cells in order to investigate the signalling pathways involved in regulating mammary epithelial and progenitor cell growth.
Among the 21,121 siRNA pools tested in the primary screen, 2,337 demonstrated statistically significant abrogation of growth to less than 25% of the control condition. This was a stringent selection criterion, given that knockdown of receptors for two of the four defined medium components (insulin, EGF, transferrin and isoproterenol) did not achieve this level of growth inhibition. Surprisingly, GPCR and associated signalling proteins were also found amongst this list, suggesting an underappreciated role of this class of receptors within mammary gland biology.
Relative effects of the 47 target genes in two-dimensional and three-dimensional cultures a
Relative rank, primary screen ( N = 47)
Median relative rank, secondary screen ( N = 47)
Relative rank, three-dimensional growth assay ( N = 44)
Three-dimensional acinar formation, fold decrease (siRNA)
shRNA knockdown effect on clone
Although these 47 genes have diverse functions, they are strikingly enriched for both GPCRs (LPAR3, FZD2, ADMR, BDKRB2, GPR39, GPR80 and GPR182) (noted in the primary screen) and axonal guidance molecules (SEMA3C, PLXNA2, ROBO3, EFNA4, NTN1 and NTN2L). For many of these genes, we provide the first description of a role in growth regulation or mammary biology.
To better understand the roles of these genes in growth and differentiation (reviewed in ), we assessed the requirement of the 47 validated target genes for growth in three-dimensional culture. The silencing of all but four of the genes (ADCY4, PCDHB13, KCNJ5 and FLOT2) decreased three-dimensional acinar formation to a level comparable to that seen with PLK1 silencing (which is essential for mitosis). Intriguingly, we have shown that LPAR3 is required for two-dimensional and three-dimensional mammary growth. Given the established importance of LPAR1, LPAR2 and LPAR3 in mammary tumourigenesis, we wanted to confirm that LPAR1 and LPAR2 are indeed irrelevant for growth of normal epithelium (as determined in the primary genetic screen). Colony formation in two-dimensional assays and three-dimensional acinar formation within Matrigel still occur upon silencing of LPAR1 and LPAR2, suggesting that they are not required for the growth of normal, nontransformed epithelial cells. LPAR3 has a greater binding affinity for 2-acyl-LPA with unsaturated fatty acids, whereas LPAR1 and LPAR2 are more responsive to saturated acyl chains . With responsiveness to a similar ligand, it is possible that compensatory redundancy exists between LPAR1 and LPAR2.
Some of the genes identified as regulating growth in two dimensions also affected differentiation when epithelial cells were grown in three-dimensions, with SNN, HSD17B2 and PROCR showing greater than tenfold reduction in acinar formation. The requirement for long-term cultures (21 days) and the lethality of the shRNAs for SNN and HSD17B2 precluded analysis of these two genes; however, reduction of PROCR in long-term cultures was associated with absence of lumen formation and disorganized epithelial growth. We quantified the relative effects on epithelial organization and lumen formation and observed that disordered differentiation was also present for two axonal pathfinding associated genes, EFNA4 and NTN1, and for LGALS1 (Additional file 1: Figure S7, Figure 8), with differential effects on polarization and epithelial organization. PROCR has been implicated as a receptor for protease-cleaved substrates in breast cancer migration  and as a marker of colony-forming cells in malignant cell lines . Here we show for the first time, to our knowledge, a role in growth and differentiation of primary breast epithelium. Loss of NTN1 causes disorganization in the terminal end buds, an effect that is proposed to occur through the loss of cellular adhesions . It has also been shown that implantation of NTN1-secreting pellets into mammary glands during pregnancy increases the number of alveolar structures that develop . In the present study, we show a role for NTN1 in both luminal and bipotent progenitor cell growth in that it was required for colony formation in our in vitro assays.
Finally, genes required for growth and differentiation are often implicated in tumourigenesis. In this study, we identified a subset of genes that have not previously been implicated in mammary gland growth or development. We sought to determine if expression of these genes correlated with any of the breast cancer subtypes. Significant, nonrandom differences in expression distribution across the 10 METABRIC datasets was seen for 40 of the 47 genes, with several genes (for example, RIPK2, EFNA4 and TMEM9B) differentially expressed in breast cancer subtypes with high proliferation. Furthermore, we were able to demonstrate independent prognostic significance for CD79A (with elevated expression improving survival in two of the ten METABRIC subtypes) and SERPINH1 (with elevated expression decreasing survival in three of the ten METABRIC subtypes). The possible roles of these genes in the tumour subtypes studied requires future functional studies in representative models; however, it is notable that all of the genes studied here are accessible by virtue of solubility or membrane location, making them a practical choice for intervention.
This work shows for the first time, to our knowledge, the diversity of transmembrane signals and/or proteins required for the growth of nontransformed mammary epithelial cells in the physiological state where extracellular signals from fibroblasts are required. In doing so, we demonstrate the functional requirement of several transmembrane and extracellular proteins in normal mammary growth in multiple well-accepted models of in vitro mammary physiology. Taken further, these proteins were differentially associated with breast cancer subtypes, which were examined in 1,998 patients, indicating that these proteins may be associated with the biology of breast cancer subtypes. This is of particular note, as the location of these proteins makes them amenable to therapeutic intervention.
Akaike information criterion
Area under the curve
Bovine pituitary extract
Oestrogen receptor α
Epithelial cell adhesion molecule
Fluorescence-activated cell sorting
False discovery rate
Fluorescence in situ hybridization
Green fluorescent protein
G protein–coupled receptor
Human epidermal growth factor receptor 2
Mammary epithelial cell basal medium
Mammary epithelial cell growth medium
Multiplicity of infection
Small interfering RNA
The authors acknowledge excellent technical contributions from Ken Fong, John Fee, Darcy Wilkinson, Gavin Ha and the staff of the Flow Cytometry Facility of the Terry Fox Laboratory and the staff at the Centre for Translational and Applied Genomics at the BC Cancer Agency. Mammoplasty tissue was obtained with the assistance of Drs. Jane Sproul, Peter Lennox, Nancy Van Laeken and Richard Warren. We are grateful to Martin Hirst and members of the NIH and Canadian Roadmap Epigenomics programs for access to RNA-seq libraries. We thank Dr. Sarah Mullaly for editing the manuscript.
- Visvader JE. Keeping abreast of the mammary epithelial hierarchy and breast tumorigenesis. Genes Dev. 2009;23:2563–77.View ArticlePubMedPubMed CentralGoogle Scholar
- Eirew P, Stingl J, Raouf A, Turashvili G, Aparicio S, Emerman JT, et al. A method for quantifying normal human mammary epithelial stem cells with in vivo regenerative ability. Nat Med. 2008;14:1384–9.View ArticlePubMedGoogle Scholar
- Lim E, Vaillant F, Wu D, Forrest NC, Pal B, Hart AH, et al. Aberrant luminal progenitors as the candidate target population for basal tumor development in BRCA1 mutation carriers. Nat Med. 2009;15:907–13.View ArticlePubMedGoogle Scholar
- Smith HS, Lan S, Ceriani R, Hackett AJ, Stampfer MR. Clonal proliferation of cultured nonmalignant and malignant human breast epithelia. Cancer Res. 1981;41:4637–43.PubMedGoogle Scholar
- Kuperwasser C, Chavarria T, Wu M, Magrane G, Gray JW, Carey L, et al. Reconstruction of functionally normal and malignant human breast tissues in mice. Proc Natl Acad Sci U S A. 2004;101:4966–71.View ArticlePubMedPubMed CentralGoogle Scholar
- Parmar H, Young P, Emerman JT, Neve RM, Dairkee S, Cunha GR. A novel method for growing human breast epithelium in vivo using mouse and human mammary fibroblasts. Endocrinology. 2002;143:4886–96.View ArticlePubMedGoogle Scholar
- Haslam SZ. Mammary fibroblast influence on normal mouse mammary epithelial cell responses to estrogen in vitro. Cancer Res. 1986;46:310–6.PubMedGoogle Scholar
- Lühr I, Friedl A, Overath T, Tholey A, Kunze T, Hilpert F, et al. Mammary fibroblasts regulate morphogenesis of normal and tumorigenic breast epithelial cells by mechanical and paracrine signals. Cancer Lett. 2012;325:175–88.View ArticlePubMedPubMed CentralGoogle Scholar
- Darcy KM, Zangani D, Shea-Eaton W, Shoemaker SF, Lee PP, Mead LH, et al. Mammary fibroblasts stimulate growth, alveolar morphogenesis, and functional differentiation of normal rat mammary epithelial cells. In Vitro Cell Dev Biol Anim. 2000;36:578–92.View ArticlePubMedGoogle Scholar
- Stingl J, Eaves CJ, Zandieh I, Emerman JT. Characterization of bipotent mammary epithelial progenitor cells in normal adult human breast tissue. Breast Cancer Res Treat. 2001;67:93–109.View ArticlePubMedGoogle Scholar
- van Amerongen R, Bowman AN, Nusse R. Developmental stage and time dictate the fate of Wnt/β-catenin-responsive stem cells in the mammary gland. Cell Stem Cell. 2012;11:387–400.View ArticlePubMedGoogle Scholar
- Makarem M, Spike BT, Dravis C, Kannan N, Wahl GM, Eaves CJ. Stem cells and the developing mammary gland. J Mammary Gland Biol Neoplasia. 2013;18:209–19.View ArticlePubMedPubMed CentralGoogle Scholar
- Silva JM, Marran K, Parker JS, Silva J, Golding M, Schlabach MR, et al. Profiling essential genes in human mammary cells by multiplex RNAi screening. Science. 2008;319:617–20.View ArticlePubMedPubMed CentralGoogle Scholar
- Bishop CL, Bergin AM, Fessart D, Borgdorff V, Hatzimasoura E, Garbe JC, et al. Primary cilium-dependent and -independent Hedgehog signaling inhibits p16INK4A. Mol Cell. 2010;40:533–47.View ArticlePubMedGoogle Scholar
- Borgdorff V, Lleonart ME, Bishop CL, Fessart D, Bergin AH, Overhoff MG, et al. Multiple microRNAs rescue from Ras-induced senescence by inhibiting p21Waf1/Cip1. Oncogene. 2010;29:2262–71.View ArticlePubMedGoogle Scholar
- Zhang Y, Smolen P, Baxter DA, Byrne JH. The sensitivity of memory consolidation and reconsolidation to inhibitors of protein synthesis and kinases: computational analysis. Learn Mem. 2010;17:428–39.View ArticlePubMedPubMed CentralGoogle Scholar
- Annab LA, Terry L, Cable PL, Brady J, Stampfer MR, Barrett JC, et al. Establishment and characterization of a breast cell strain containing a BRCA1 185delAG mutation. Gynecol Oncol. 2000;77:121–8.View ArticlePubMedGoogle Scholar
- Horikawa I, Cable PL, Mazur SJ, Appella E, Afshari CA, Barrett JC. Downstream E-box-mediated regulation of the human telomerase reverse transcriptase (hTERT) gene transcription: evidence for an endogenous mechanism of transcriptional repression. Mol Biol Cell. 2002;13:2585–97.View ArticlePubMedPubMed CentralGoogle Scholar
- Bayani J, Squire J. Multi-color FISH techniques. Curr Protoc Cell Biol. 2004;24:Chapter 22:Unit 22.5.Google Scholar
- Debnath J, Muthuswamy SK, Brugge JS. Morphogenesis and oncogenesis of MCF-10A mammary epithelial acini grown in three-dimensional basement membrane cultures. Methods. 2003;30:256–68.View ArticlePubMedGoogle Scholar
- Poon SS, Wong JT, Saunders DN, Ma QC, McKinney S, Fee J, et al. Intensity calibration and automated cell cycle gating for high-throughput image-based siRNA screens of mammalian cells. Cytometry A. 2008;73:904–17.View ArticlePubMedGoogle Scholar
- Shah SP, Roth A, Goya R, Oloumi A, Ha G, Zhao Y, et al. The clonal and mutational evolution spectrum of primary triple-negative breast cancers. Nature. 2012;486:395–9.PubMedGoogle Scholar
- Smoot ME, Ono K, Ruscheinski J, Wang PL, Ideker T. Cytoscape 2.8: new features for data integration and network visualization. Bioinformatics. 2011;27:431–2.View ArticlePubMedGoogle Scholar
- Wu G, Feng X, Stein L. A human functional protein interaction network and its application to cancer data analysis. Genome Biol. 2010;11:R53.View ArticlePubMedPubMed CentralGoogle Scholar
- Curtis C, Shah SP, Chin SF, Turashvili G, Rueda OM, Dunning MJ, et al. The genomic and transcriptomic architecture of 2,000 breast tumours reveals novel subgroups. Nature. 2012;486:346–52.PubMedPubMed CentralGoogle Scholar
- Benjamini HY. Controlling the false discovery rate: a practical and powerful approach to multiple hypothesis testing. J Roy Stat Soc Stat Soc. 1995;57:289–300.Google Scholar
- Harrington DP, Fleming TR. A class of rank test procedures for censored survival-data. Biometrika. 1982;69:553–66.View ArticleGoogle Scholar
- Canadian Epigenetics, Environment and Health Research Consortium (CEEHRC) Platform. Data downloads. http://www.epigenomes.ca/downloads.html. Accessed 13 Jan 2015.
- Canadian Epigenetics, Environment and Health Research Consortium (CEEHRC). Protocols and standards. http://www.epigenomes.ca/protocols.html. Accessed 13 January 2015.
- Roy S, Gascard P, Dumont N, Zhao J, Pan D, Petrie S, et al. Rare somatic cells from human breast tissue exhibit extensive lineage plasticity. Proc Natl Acad Sci U S A. 2013;110:4598–603.View ArticlePubMedPubMed CentralGoogle Scholar
- Kampstra P. Beanplot: a boxplot alternative for visual comparison of distributions. J Stat Softw. 2008;28:Code Snippet 1.View ArticleGoogle Scholar
- Raouf A, Brown L, Vrcelj N, To K, Kwok W, Huntsman D, et al. Genomic instability of human mammary epithelial cells overexpressing a truncated form of EMSY. J Natl Cancer Inst. 2005;97:1302–6.View ArticlePubMedGoogle Scholar
- Hammond SL, Ham RG, Stampfer MR. Serum-free growth of human mammary epithelial cells: rapid clonal growth in defined medium and extended serial passage with pituitary extract. Proc Natl Acad Sci U S A. 1984;81:5435–9.View ArticlePubMedPubMed CentralGoogle Scholar
- Lasfargues EY, Moore DH. A method for the continuous cultivation of mammary epithelium. In Vitro. 1971;7:21–5.View ArticlePubMedGoogle Scholar
- White MT, Hu AS, Hamamoto ST, Nandi S. In vitro analysis of proliferating epithelial cell populations from the mouse mammary gland: fibroblast-free growth and serial passage. In Vitro. 1978;14:271–81.View ArticlePubMedGoogle Scholar
- Petersen OW, Rønnov-Jessen L, Howlett AR, Bissell MJ. Interaction with basement membrane serves to rapidly distinguish growth and differentiation pattern of normal and malignant human breast epithelial cells. Proc Natl Acad Sci U S A. 1992;89:9064–8.View ArticlePubMedPubMed CentralGoogle Scholar
- Ashburner M, Ball CA, Blake JA, Botstein D, Butler H, Cherry JM, et al. Gene Ontology: tool for the unification of biology. Nat Genet. 2000;25:25–9.View ArticlePubMedPubMed CentralGoogle Scholar
- Paddison PJ, Silva JM, Conklin DS, Schlabach M, Li M, Aruleba S, et al. A resource for large-scale RNA-interference-based screens in mammals. Nature. 2004;428:427–31.View ArticlePubMedGoogle Scholar
- Huang DW, Sherman BT, Lempicki RA. Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat Protoc. 2008;4:44–57.View ArticleGoogle Scholar
- Kanehisa M, Goto S. KEGG: Kyoto Encyclopedia of Genes and Genomes. Nucleic Acids Res. 2000;28:27–30.View ArticlePubMedPubMed CentralGoogle Scholar
- Kanehisa M, Goto S, Sato Y, Furumichi M, Tanabe M. KEGG for integration and interpretation of large-scale molecular data sets. Nucleic Acids Res. 2012;40:D109–14.View ArticlePubMedGoogle Scholar
- Harburg GC, Hinck L. Navigating breast cancer: axon guidance molecules as breast cancer tumor suppressors and oncogenes. J Mammary Gland Biol Neoplasia. 2011;16:257–70.View ArticlePubMedPubMed CentralGoogle Scholar
- Yang J, Balakrishnan A, Hamamoto S, Beattie CW, Das Gupta TK, Wellings SR, et al. Different mitogenic and phenotypic responses of human breast epithelial cells grown in two versus three dimensions. Exp Cell Res. 1986;167:563–9.View ArticlePubMedGoogle Scholar
- Wang F, Weaver VM, Petersen OW, Larabell CA, Dedhar S, Briand P, et al. Reciprocal interactions between β1-integrin and epidermal growth factor receptor in three-dimensional basement membrane breast cultures: a different perspective in epithelial biology. Proc Natl Acad Sci U S A. 1998;95:14821–6.View ArticlePubMedPubMed CentralGoogle Scholar
- Koch A, Poirier F, Jacob R, Delacour D. Galectin-3, a novel centrosome-associated protein, required for epithelial morphogenesis. Mol Biol Cell. 2010;21:219–31.View ArticlePubMedPubMed CentralGoogle Scholar
- Liu S, Umezu-Goto M, Murph M, Lu Y, Liu W, Zhang F, et al. Expression of autotaxin and lysophosphatidic acid receptors increases mammary tumorigenesis, invasion, and metastases. Cancer Cell. 2009;15:539–50. A published erratum appears in. Cancer Cell. 2009;16:172.View ArticleGoogle Scholar
- Stoker M, Perryman M, Eeles R. Clonal analysis of morphological phenotype in cultured mammary epithelial cells from human milk. Proc R Soc Lond B Biol Sci. 1982;215:231–40.View ArticlePubMedGoogle Scholar
- Edwards PA, Brooks IM, Monaghan P. Expression of epithelial antigens in primary cultures of normal human breast analysed with monoclonal antibodies. Differentiation. 1984;25:247–58.View ArticlePubMedGoogle Scholar
- O’Hare MJ, Ormerod MG, Monaghan P, Lane EB, Gusterson BA. Characterization in vitro of luminal and myoepithelial cells isolated from the human mammary gland by cell sorting. Differentiation. 1991;46:209–21.View ArticlePubMedGoogle Scholar
- Stingl J, Eaves CJ, Kuusk U, Emerman JT. Phenotypic and functional characterization in vitro of a multipotent epithelial cell present in the normal adult human breast. Differentiation. 1998;63:201–13.View ArticlePubMedGoogle Scholar
- Banerji S, Cibulskis K, Rangel-Escareno C, Brown KK, Carter SL, Frederick AM, et al. Sequence analysis of mutations and translocations across breast cancer subtypes. Nature. 2012;486:405–9.View ArticlePubMedPubMed CentralGoogle Scholar
- Ellis MJ, Ding L, Shen D, Luo JQ, Suman VJ, Wallis JW, et al. Whole-genome analysis informs breast cancer response to aromatase inhibition. Nature. 2012;486:353–60.PubMedPubMed CentralGoogle Scholar
- Stephens PJ, Tarpey PS, Davies H, Van Loo P, Greenman C, Wedge DC, et al. The landscape of cancer genes and mutational processes in breast cancer. Nature. 2012;486:400–4.PubMedPubMed CentralGoogle Scholar
- Koboldt DC, Fulton RS, McLellan MD, Schmidt H, Kalicki-Veizer J, McMichael JF, et al. Comprehensive molecular portraits of human breast tumours. Nature. 2012;490:61–70.View ArticleGoogle Scholar
- Dawson SJ, Rueda OM, Aparicio S, Caldas C. A new genome-driven integrated classification of breast cancer and its implications. EMBO J. 2013;32:617–28.View ArticlePubMedPubMed CentralGoogle Scholar
- Forbes SA, Bhamra G, Bamford S, Dawson E, Kok C, Clements J, et al. The Catalogue of Somatic Mutations in Cancer (COSMIC). Curr Protoc Hum Genet. 2008;57:Chapter 10:Unit 10.11.Google Scholar
- Bamford S, Dawson E, Forbes S, Clements J, Pettett R, Dogan A, et al. The COSMIC (Catalogue of Somatic Mutations in Cancer) database and website. Br J Cancer. 2004;91:355–8.PubMedPubMed CentralGoogle Scholar
- Whitehurst AW, Bodemann BO, Cardenas J, Ferguson D, Girard L, Peyton M, et al. Synthetic lethal screen identification of chemosensitizer loci in cancer cells. Nature. 2007;446:815–9.View ArticlePubMedGoogle Scholar
- Simpson KJ, Selfors LM, Bui J, Reynolds A, Leake D, Khvorova A, et al. Identification of genes that regulate epithelial cell migration using an siRNA screening approach. Nat Cell Biol. 2008;10:1027–38.View ArticlePubMedGoogle Scholar
- Bhowmick NA, Chytil A, Plieth D, Gorska AE, Dumont N, Shappell S, et al. TGF-β signaling in fibroblasts modulates the oncogenic potential of adjacent epithelia. Science. 2004;303:848–51.View ArticlePubMedGoogle Scholar
- Bhowmick NA, Neilson EG, Moses HL. Stromal fibroblasts in cancer initiation and progression. Nature. 2004;432:332–7.View ArticlePubMedPubMed CentralGoogle Scholar
- Schmeichel KL, Bissell MJ. Modeling tissue-specific signaling and organ function in three dimensions. J Cell Sci. 2003;116:2377–88.View ArticlePubMedPubMed CentralGoogle Scholar
- Choi JW, Herr DR, Noguchi K, Yung YC, Lee CW, Mutoh T, et al. LPA receptors: subtypes and biological actions. Annu Rev Pharmacol Toxicol. 2010;50:157–86.View ArticlePubMedGoogle Scholar
- Beaulieu LM, Church FC. Activated protein C promotes breast cancer cell migration through interactions with EPCR and PAR-1. Exp Cell Res. 2007;313:677–87.View ArticlePubMedGoogle Scholar
- Hwang-Verslues WW, Kuo WH, Chang PH, Pan CC, Wang HH, Tsai ST, et al. Multiple lineages of human breast cancer stem/progenitor cells identified by profiling with stem cell markers. PLoS One. 2009;4:e8377.View ArticlePubMedPubMed CentralGoogle Scholar
- Srinivasan K, Strickland P, Valdes A, Shin GC, Hinck L. Netrin-1/neogenin interaction stabilizes multipotent progenitor cap cells during mammary gland morphogenesis. Dev Cell. 2003;4:371–82.View ArticlePubMedGoogle Scholar
- Strizzi L, Mancino M, Bianco C, Raafat A, Gonzales M, Booth BW, et al. Netrin-1 can affect morphogenesis and differentiation of the mouse mammary gland. J Cell Physiol. 2008;216:824–34.View ArticlePubMedPubMed CentralGoogle Scholar
- Underwood JM, Imbalzano KM, Weaver VM, Fischer AH, Imbalzano AN, Nickerson JA. The ultrastructure of MCF-10A acini. J Cell Physiol. 2006;208:141–8.View ArticlePubMedGoogle Scholar
- Chua HL, Bhat-Nakshatri P, Clare SE, Morimiya A, Badve S, Nakshatri H. NF-κB represses E-cadherin expression and enhances epithelial to mesenchymal transition of mammary epithelial cells: potential involvement of ZEB-1 and ZEB-2. Oncogene. 2007;26:711–24.View ArticlePubMedGoogle Scholar
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.