Open Collections

UBC Faculty Research and Publications

A co-culture genome-wide RNAi screen with mammary epithelial cells reveals transmembrane signals required… Burleigh, Angela; McKinney, Steven; Brimhall, Jazmine; Yap, Damian; Eirew, Peter; Poon, Steven; Ng, Viola; Wan, Adrian; Prentice, Leah; Annab, Lois; Barrett, J C; Caldas, Carlos; Eaves, Connie; Aparicio, Samuel Jan 9, 2015

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
52383-13058_2014_Article_510.pdf [ 3.91MB ]
Metadata
JSON: 52383-1.0223272.json
JSON-LD: 52383-1.0223272-ld.json
RDF/XML (Pretty): 52383-1.0223272-rdf.xml
RDF/JSON: 52383-1.0223272-rdf.json
Turtle: 52383-1.0223272-turtle.txt
N-Triples: 52383-1.0223272-rdf-ntriples.txt
Original Record: 52383-1.0223272-source.json
Full Text
52383-1.0223272-fulltext.txt
Citation
52383-1.0223272.ris

Full Text

RESEARCH ARTICLE Open AccessA co-culture genome-wide RNAi screen withmammary epithelial cells reveals transmembraneth and differentiationConclusion: Diverse transmembrane signals are required for mammary epithelial cell growth in two-dimensional andBurleigh et al. Breast Cancer Research  (2015) 17:4 DOI 10.1186/s13058-014-0510-y1L3, CanadaFull list of author information is available at the end of the articlethree-dimensional conditions. Strikingly, we define novel roles for axonal pathfinding receptors and ligands and theendothelin receptor in both growth and differentiation.* Correspondence: saparicio@bccrc.ca1Department of Pathology and Laboratory Medicine, University of BritishColumbia, and BC Cancer Agency, 675 West 10th Avenue, Vancouver, BC V5ZAngela Burleigh1, Steven McKinney1, Jazmine Brimhall1, Damian Yap1, Peter Eirew1, Steven Poon1, Viola Ng1,Adrian Wan1, Leah Prentice1,2, Lois Annab3, J Carl Barrett4, Carlos Caldas5, Connie Eaves6 and Samuel Aparicio1*AbstractIntroduction: The extracellular signals regulating mammary epithelial cell growth are of relevance tounderstanding the pathophysiology of mammary epithelia, yet they remain poorly characterized. In this study, weapplied an unbiased approach to understanding the functional role of signalling molecules in several models ofnormal physiological growth and translated these results to the biological understanding of breast cancer subtypes.Methods: We developed and utilized a cytogenetically normal clonal line of hTERT immortalized human mammaryepithelial 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 ofgrowth. 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, weidentified a subset of genes required for mammary epithelial cell growth. Using three-dimensional Matrigel growth anddifferentiation assays and primary human mammary epithelial cell colony assays, we confirmed that these growth effectswere not limited to the 184-hTERT cell line. We utilized the METABRIC dataset of 1,998 breast cancer patients to evaluateboth the differential expression of these genes across breast cancer subtypes and their prognostic significance.Results: 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 theendothelin receptor PROCR. The majority of genes (43 of 47) identified in two dimensions were also required forthree-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 properdifferentiation and polarization in three-dimensional cultures. The 47 genes identified showed a significant nonrandomenrichment 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.signals required for grow© 2015 Burleigh et al.; licensee BioMed Central. This is an Open Access article distributed under the terms of the CreativeCommons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, andreproduction in any medium, provided the original work is properly credited. The Creative Commons Public DomainDedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article,unless otherwise stated.Burleigh et al. Breast Cancer Research  (2015) 17:4 Page 2 of 21IntroductionThe identification of distinct cell types that appear to behierarchically organized in the mammary epithelialglands of healthy women is now well established [1].This hierarchy is defined largely by two prospectivelyseparable subsets of cells that generate colonies containingonly one or both lineages (myoepithelial and/or luminal)of cells that make up the bulk of the normal mammarygland structure. The bipotent, clonogenic, progenitor-enriched basal cell fraction also contains putative humanmammary stem cells identified in xenotransplantationassays [2,3]. The ability of human mammary cells tobe propagated both in vitro and in vivo at limiteddensities is known to be markedly enhanced by thepresence of fibroblast ‘feeders’ [2,4,5]. These and manyother studies have shown that fibroblast interactions areimportant to the growth of mammary epithelial cells[6-12]. However, a comprehensive characterization of themechanisms by which fibroblasts regulate the growth andfunctional organization of normal mammary epithelialcells has been lacking.Genome-wide RNA interference (RNAi, small interferingRNA (siRNA)) screens offer an attractive strategy by whichto investigate such questions. They have previously beenused with success to identify mediators of Ras oncogene-induced senescence, suppressors of p16 gene expression,genes that regulate cell migration and cell survival genes inmammary cells [13-16]. This type of investigation isnevertheless dependent on a source of cells that canbe obtained in large numbers and readily transfected.Because primary normal mammary epithelial cells, eventhose derived from human mammoplasties, do not satisfyeither of these requirements, we sought an alternative in aclonal diploid isolate of hTERT-immortalized cells [17] thatwe found remains dependent on fibroblast stimulationfor its rapid growth when cultured at low density. Bycombining automated imaging with siRNA screeningof these cells, we identified 43 signal-transducing receptorsand secreted factors with functionally validated roles inmediating the in vitro growth of primary normal humanmammary epithelial cells.MethodsCell linesPassage 6 184-hTERT polyclonal infection pool mammaryepithelial cells (obtained from [18]) were contributed tothe study by CB and LA. As described previously [18],these pools were generated from anonymised primarymammary epithelial sample 184 (see [18]) and notsubject to specific institutional review board approval.We generated the monoclonal cell lines (184-hTERT-L9 or184-hTERT-E11) and used the 184-hTERT-L9 cell line togenerate subsequent polyclonal cell lines (stably infectedwith lentiviruses or NucLight Red (Essen BioScience,Ann Arbor, MI, USA), for example). The experimentswere conducted under University of British ColumbiaResearch Ethics Board protocols H06-0289, H06-0210and B13-0126.Cell culturePassage 6 184-hTERT cells [18] were cloned in 96-wellplates and subcultured in serum-free mammary epithelialcell basal media (MEBM; Lonza, Walkersville, MD, USA)supplemented with the mammary epithelial cell growthmedia in the SingleQuots kit (Lonza), 5 μg/ml trans-ferrin (Sigma-Aldrich, St Louis, MO, USA) and 10−5 Misoproterenol (Sigma-Aldrich), referred to as mammaryepithelial cell growth medium (MEGM).ImmunofluorescenceMulticolour fluorescence in situ hybridization (FISH) wasperformed as previously described [19]. Immunofluores-cence cell staining in three-dimensional Matrigel cultureswas performed as previously described [20] with pri-mary 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 OregonGreen 488 or Alexa Fluor 546 phalloidin (Invitrogen)and DRAQ5 nuclear staining prior to imaging on aconfocal laser scanning microscope (Nikon Instruments,Melville, NY, USA). For calcein acetoxymethyl ester(calcein AM) and ethidium homodimer 1, 21-dayMatrigel cultures were stained unfixed for 20 minutesand counterstained with Hoeschst 33342 (Invitrogen).Immunofluorescence staining of cells in three-dimensionalMatrigel cultures cultured for 3 weeks was performed withprimary antibodies to E-cadherin (E-cad; Calbiochem,San Diego, CA, USA), GM130 (BD Biosciences), CD49f(STEMCELLTechnologies) and Alexa Fluor 680–conjugatedsecondary antibodies (Invitrogen) and imaged on aNikon confocal laser scanning microscope. Colonies werecounted at five discrete, randomly chosen positions perwell using a Nikon confocal laser scanning microscope.Only discrete, well-separated structures were counted. Inthe cases where two colonies touched or merged, bothcolonies were ignored for counting purposes. For caspase3 staining, three-dimensional Matrigel cultures wereformalin-fixed and paraffin-embedded, and sectionswere stained with caspase 3 antibody (Cell SignalingTechnology, Danvers, MA, USA). To quantify the propor-tion of structures with wild-type epithelial organizationand polarization, RNAi-treated samples were scored andcompared to wild-type localization (see Additional file 1:Figure S7 and Additional file 2: Figure S8A) for examplesof each category. For CD49f, the presence of a single base-ment membrane type of immunoreactive structure wasto largest, and the latter (those reducing cell growth by(Sigma), 1 μg/ml insulin (Sigma), 0.5 μg/ml hydrocorti-Burleigh et al. Breast Cancer Research  (2015) 17:4 Page 3 of 21considered the wild type. For GM130 immunoreactivity,wild-type polarization was deemed to be signal-localizedbetween the edge of the colony and an unstainedlumen. Lack of polarization would be reflected in theindistinguishable staining patterns between the outerand inner cell layers.Genome-wide siRNA screen protocolBlack-walled clear-bottom 96-well plates (Greiner Bio-One,Monroe, NC, USA) were seeded with 3,000 freshly irradi-ated NIH 3T3 cells per well in Dulbecco’s modified Eagle’smedium with 5% foetal bovine serum. After 24 hours, themedium was aspirated at a low flow rate with a multichan-nel vacuum aspirator and replaced with 550 184-hTERT-L9cells per well in MEGM without bovine pituitary extract(BPE) added. After an additional 24 hours, Lipofectamine2000 reagent–siRNA complexes were generated.siRNAs were purchased from the siGENOME library(Dharmacon, Lafayette, CO, USA) as deconvolved setsof four individual siRNAs and resuspended at 10 μM in1× siRNA buffer (Dharmacon) as described elsewhere[21]. Lipofectamine 2000 transfection reagent (Invitrogen)was diluted in MEBM and incubated for 5 minutes priorto mixture with an equal volume of prediluted siRNA inMEBM. Complexes were allowed to form for 20 minutesbefore they were added directly to the cells at a finalconcentration of 30 nM siRNA and 0.3 μl of transfectionreagent per well. The control wells were static in positionand were composed of Lipofectamine alone, siControl-3and a siRNA pool targeting PLK1. The entire 96-wellplates of these controls were staggered throughout theduration of this screen to allow for statistical correction ofplate positional effects. After an additional 4 days ofgrowth, cells were fixed with 4% paraformaldehyde andstained 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-onefields per well were collected using HQ360/40 excitationand HQ535/50 emission filters with a multi-bandpassdichroic filter (Q505lp; Chroma Technology, BellowsFalls, VT, USA). For each image field, a single focal planewas captured using a hardware (laser/photodetector)autofocusing algorithm, which estimated the surface areaon which the cells were lying. An image segmentation andpostprocessing software workflow was developed usingCellProfiler [21]. Cell counts per well produced in theCell Profiler image analysis were further processed usingquantile normalization to correct for data distributionaldifferences induced by factors such as time-dependentstain degradation and other plate-handling artefacts.Quantile-normalized data were then analysed using statis-tical linear mixed effects regression models to comparesone (Sigma)) with 50 μg/ml GA-1000 (Lonza). After 18hours at 37°C, cells were washed and counted and equalnumbers plated into two-dimensional colony-formingcell assays.Primary mammary tissueDiscarded tissue was collected from premenopausalwomen (ages 19 to 40 years) who provided informedconsent as approved by the University of BritishColumbia Research Ethics Board, as previously described[2]. Suspensions selectively enriched in bipotent progeni-tor cells were obtained by fluorescence-activated cellsorting (FACS) of cells positively costained with anallophycocyanin-conjugated rat antibody to human CD49f(eBioscience, San Diego, CA, USA) and a phycoerythrin-conjugated mouse antibody to human epithelial cell adhe-sion molecule (EpCAM; eBioscience) [2]. Hematopoieticstem cells and endothelial cells were eliminated using anti-75% or more) were selected for further study. Enrichmentnetwork analysis was performed as previously described[22] using the Reactome Functional Interactome plugin inCytoscape v2.8.1 [23,24].Lentiviral transduction procedure184-hTERT-L9 cells were transduced at an estimatedmultiplicity of infection (MOI) of 5:1 with 8 μg/ml poly-brene. After 18 hours at 37°C, cells were washed, and,after 24 hours in MEGM, they were selected with 2 μg/mlpuromycin in MEGM (replaced every 24 to 48 hours).Dissociated primary human mammary epithelial cells wereinfected with an estimated MOI of 10. Transduction wasconducted in suspension with 8 μg/ml polybrene at adensity of 5 × 105 cells in 100 μl of serum-free 7medium (DMEM/F12 (STEMCell Technologies) sup-plemented with 5% FBS, 2 mM glutamine (Gibco), 0.1%w/v BSA, 10 ng/ml EGF (Sigma), 10 ng/ml cholera toxincell counts under the siRNA condition with counts underthe control condition, adjusted for technical artefacts of wellposition and plate effects, thereby yielding plate-normalizedgrowth effect estimates for each siRNA. Well positioneffects were assessed by screening additional platescontaining the same control condition in all wells.Plate effects were assessed by using multiple platesfor each condition, thereby allowing for adjustment ofplate-to-plate variability in the statistical model. Themodel-estimated cell count under the siRNA conditionwas then divided by the model-estimated cell count underthe control condition to produce an overall measure ofrelative effect. Measured effects were ranked from smallestbodies to human CD45 and human CD31 (eBioscience),respectively.Burleigh et al. Breast Cancer Research  (2015) 17:4 Page 4 of 21Proliferation measurement by 5-ethynyl-2′-deoxyuridineincorporationTo determine the proliferation rate, 184-hTERT-L9 cellsstably infected with NucLight Red were seeded into wellsof 24-well plates (16,750 cells per well). Twenty-fourhours after plating, the cells were transfected with siRNAs(30 nM concentration) and cultured in the IncuCyteZOOM incubator (Essen BioScience). Twenty hours aftertransfection, the siRNA-containing medium was removedand replaced with standard 184-hTERT culturemedium. Sixty-eight hours after transfection, 5-ethynyl-2′-deoxyuridine (EdU) was added to each well to a finalconcentration 40 μM. After 1 hour, wells were washedwith phosphate-buffered saline, then the cells wereharvested with a trypsin/ethylenediaminetetraacetic acidmixture. To ensure sufficient cell numbers for FACS,1 × 105 nonfluorescent 184-hTERT cells were added tothe cells harvested from each well. EdU incorporationwas then detected by staining with the Click-iT EdUAlexa Fluor 488 Flow Cytometry Assay Kit (LifeTechnologies, Grand Island, NY, USA) according to themanufacturer’s instructions, followed by FACS acquisitionon a BD FACSAria III cell sorter (BD Biosciences). DAPIwas used to detect DNA content and to gate out cellfragments. The percentage of test cells that incorporatedEdU was determined by gating on the NucLightRed–positive cell population and applying an EdU-positivegate set with reference to negative control (carrier) cellsprocessed with the Click-iT EdU assay but not previouslyexposed to an EdU pulse.Determination of proliferation and apoptosis by live cellimagingWhere appropriate, 184-hTERT-L9 cells were stablyinfected with NucLight Red, which marks the nuclei redto distinguish them from unlabelled irradiated feedersused the in co-culture experiments. To control forfeeder effects, we conducted the experiment using bothfeeder-free and feeder-containing conditions, with BPEmedium supplement (see details in methods) used forthe nonfeeder condition. For the feeder-present condi-tion, 3,000 irradiated feeders per well were plated in96-well plates 24 hours prior to plating 550 184-hTERTsstably expressing NucLight Red. Transfection of 30 nMsiRNA (complexing as described in methods) was per-formed 24 hours later, and the complexes were washedout after 24 hours, with the addition of CellPlayerCaspase-3/7 reagent (Essen BioScience), which labelsapoptotic cells green, detecting caspases 3 and 7. The platewas then imaged using an IncuCyte ZOOM live cellmicroscope (Essen BioScience), and images were takenevery 4 hours for an additional 84 hours. The data wereanalysed as follows. Triplicate red cell objects (represent-ing hTERT with NucLight Red) and green objects(apoptotic cells with activated caspase 3, as the percentageof red objects) were counted (counts per square milli-metre) using the GraphPad Prism statistical software suite(GraphPad Software, La Jolla, CA, USA), and the respect-ive areas under the curve (AUCs) for serial measurementswere calculated. Where comparisons were made, the AUCvalues were subjected to a one-way analysis of variance testto compare the mean of each siRNA to the controlcondition, in which only Lipofectamine 2000 transfectionreagent was used to determine significance.Gene association analysesExpression and patient outcome data for 1,996 breastcancer patients were obtained from the MolecularTaxonomy of Breast Cancer International Consortium(METABRIC) study (European Genotype-phenome Archivestudy accession number: EGAS00000000083) [25]. Relativehazard estimates for cases with high versus low geneexpression were obtained using a Cox proportional hazardsmodel with the binarized expression variable and stratifiedby integrative cluster (IntClust) breast cancer subtypegroups to mitigate for the nonproportional hazardsexhibited between various IntClust groups. Expressionvariables were binarized at the 15%, 25%, 50% and75% quantiles, and Akaike information criteria (AIC)were calculated using a Cox model containing thebinarized expression variable. The binarization cutpoint was chosen as the quantile yielding the minimumAIC value. In the case that the minimum and maximumAIC values differed by less than 3.0, the median expres-sion level was used as the cut point to ensure adequatecase counts within the breast cancer subtype groups. Toidentify potential instances of interaction between expres-sion level and breast cancer subtype (high versus lowexpression showing improved survival in one subgroup andpoorer survival in another subgroup), a Cox model withbinarized expression, IntClust subgroups and their inter-action terms was fitted and compared to a Cox model con-taining IntClust subgroups only, yielding an omnibus testof survival difference due to the biomarker. P-values acrossall 46 fitted models were adjusted for multiple comparisonsusing the method of Benjamini and Hochberg [26], and sig-nificant findings after adjustment for multiple comparisonswere identified. To guard against the issue of changinghazards over time and nonproportional hazards betweenIntClust subgroups, each binarized biomarker was testedusing the G-rho rank test procedure stratified by IntClustsubgroups and setting ρ = 1 to place heavier weight onearlier time point observations [27]. P-values from all 46G-rho tests were adjusted for multiple comparisons, andgenes with low false discovery rates were identified.For analysis of transcripts in flow-sorted mammaryepithelial cell lineages, we made use of the NIH andCanadian Roadmap Epigenomics mammary epithelialBurleigh et al. Breast Cancer Research  (2015) 17:4 Page 5 of 21cell RNA-seq libraries. The consortium data generationprotocols and data can be accessed online [28,29].The flow sorting of mammary epithelial subsets wasperformed using CD10, MUC1 and CD73 as describedpreviously [30].Expression dataExpression data processing and normalization are describedin the supplementary materials of the METABRIC study[25]. Expression data were additionally centred and scaledwithin the study data centre as follows:rijk ¼ eijk−medikigsjkwhere eijk is the normalized expression value for subject i,gene target j, in data centre k; medik is the medianexpression value for gene target j in data centre k;and igsjk is the interquartile spread for gene target jin data centre k. (Interquartile spread is the differencebetween the 75th-percentile value and the 25th-percentilevalue, a robust measure of standard deviation.) Heat mapsof this robustly centred and scaled data were generatedwith complete linkage clustering of gene targets indicatedon the vertical axis. To yield comparable colour schemesacross all genes in the heat map, values less than −3.0 wereset to −3.0 and values greater than 3.0 were set to 3.0.Expression value differences among breast cancer sub-type groups (IntClust and PAM50) were assessed usingthe Kruskal-Wallis test. Beanplots [31] of expression valueswithin breast cancer subtype groups for each of the 47 finalidentified gene targets were ordered in Kruskal-Wallis testP-value order.ResultsIsolation and characterization of a cytogenetically normalclone of 184-hTERT cellsWe sought a cell line model of primary human mammaryepithelial cells that would display fibroblast-enhancedgrowth at low cell density and would also be readilyamenable to high-throughput, genome-wide geneticmanipulation. Preliminary data indicated that this wasa retained feature of 184-hTERT cells. However, thosewidely available are karyotypically abnormal and may bechromosomally unstable [32]. Therefore, we obtained anearly passage of the original 184 primary cell pool trans-duced with hTERT [17]. After cloning approximately80 independent lines, we isolated a clone (184-hTERT-L9)in which all cells showed a stable normal (46, XX)karyotype, a copy number diploid genome (Figure 1Aand Additional file 3: Figure S1D) and a basal phenotype(K5+, E-cad+, MUC1− and oestrogen receptor α–negative(ERα−)) (Additional file 3: Figures S1B and S1C) withexpression of epidermal growth factor receptor and E-cad(Additional file 3: Figure S1B), but not of keratin 18,human epidermal growth factor receptor 2 (HER2) or ERα(Additional file 3: Figures S1B and S1C). 184-hTERT-L9 cellsalso express wild-type p53 (Additional file 3: Figure S1B),but they show silenced p16 expression (Additional file 3:Figure S1C) and have a single integration site for hTERT(Additional file 3: Figure S1A).To quantify fibroblast-dependent growth [33-35], wecompared 184-hTERT-L9 cells with primary basal progeni-tor cell-enriched fractions (CD49fhighEpCam−) sorted fromdissociated reduction mammoplasty tissue. When plated ina colony-forming cell assay in co-culture with an increasingdensity of irradiated fibroblasts, a dose-dependent growthresponse was observed (Figure 2) for both 184-hTERT-L9and primary mammary epithelial cells at low platingdensities. This behaviour is distinct from many immortal-ized mammary epithelial cells lines, including MCF10A(Figure 2), in which colony formation at low epithelialcell density occurs with high efficiency independentof a fibroblast feeder layer (growth at 0 fibroblasts).We next compared the three-dimensional growthcharacteristics of 184-hTERT-L9 with flow-sorted pri-mary mammary epithelial cells and with MCF10A cells[36]. When placed into three-dimensional Matrigel cul-ture, 184-hTERT-L9 cells formed spherical, multilayeredacini with apicobasal polarity, as determined by stainingwith antibodies raised against apical marker GM130 andbasal marker CD49f (Figure 1B). The immunoreactivityof inner cells with antibodies to the luminal markerprotein MUC1 suggests that differentiation is multilineage.A key element of three-dimensional mammary differenti-ation is the death of cells forming the lumen of sphericalacini. This was observed using calcein AM and ethidiumhomodimer 1 staining of unfixed three-dimensionalcultures (Figure 1B). As with primary bipotent mammaryepithelial progenitors [3], 184-hTERT-L9 acini exhibitsquamous differentiation of the inner cells in three-dimensional Matrigel cultures, which is appreciable bythe morphology of the inner cells stained with phal-loidin (Figure 1B). The similarities and differencesbetween 184-hTERT-L9 and MCF10A cells are shown inAdditional file 4: Table S1.Defining the genes required for mammary epithelial cellgrowth under co-culture conditionsTo identify genes involved in the regulation of mammaryepithelial cell growth, we used the 184-hTERT-L9 cellsin a co-culture (fibroblasts and epithelial cells, definedmedium without serum) genome-wide siRNA screenusing 21,121 pools of siRNAs designed to the humangenome (Dharmacon). The primary screen was used forinitial hit identification, and multiple secondary screensfocused on transmembrane and extracellular proteinencoding genes were used to reconfirm and explorePhalloidin MUC1 NucleiPhalloidin GM130 Nuclei10umCalcein-AM EthD-1 Nuclei20umE-cadherin CD49f Nuclei25um10umABFigure 1 (See legend on next page.)Burleigh et al. Breast Cancer Research  (2015) 17:4 Page 6 of 21niidicuolain,(EBurleigh et al. Breast Cancer Research  (2015) 17:4 Page 7 of 21the genes implicated (Figure 3A). For the primary andsecondary screening assays (Figure 3A), a layer offreshly irradiated murine NIH 3T3 cells was left toadhere for 24 hours prior to addition of a low densityof 184-hTERT-L9 cells. After an additional 24 hours,these co-cultures were robotically transfected with thesiRNA pools. Although two cell types were present inthe assay, two features of the assay were used to biasthe primary screen RNAi responses to the epithelialcells. First, the transfection conditions used for theepithelial cells were not capable of efficiently inducingRNAi transcript/protein knockdown in irradiated fibroblastcells (Additional file 5: Figures S3B and S3C), even withspecies-specific siRNAs (murine β-actin for NIH 3T3 cells,human GAPDH for IMR-90). The reduced capacity ofirradiated versus nonirradiated fibroblasts to mediate RNAiwas further demonstrated (Additional file 5: Figures S3Aand S3D) using universally cell-lethal positive controlsiRNA (siTOX; Dharmacon) and siRNA to PLK, showingthat irradiation abrogates the functional effects of thesepositive controls (Additional file 5: Figure S3D), althoughirradiated fibroblasts are still capable of executing celldeath programs. Second, the siRNA library used ishuman-specific, whereas the fibroblasts are mouse-derived,so any siRNAs entering the cell would be of reduced effect-iveness. It is important to note that, in order to confirmthat growth effects were related to gene activity in theepithelial cells, all subsequent secondary screen revalidationexperiments were carried out either in feeder-free condi-tions or with transfection into epithelial cells undertakenbefore placement onto a feeder layer. To quantify cellgrowth, we counted the number of nuclei per well 4 daysafter transfection with automated image recognition induplicate assay plates [21]. In the co-culture conditions(See figure on previous page.)Figure 1 184-hTERT-L9 cells have a normal karyotype and form acimetaphase using colcemid prior to multicolour fluorescence in situ hybrcomplement devoid of numerical or structural changes (n = 7). (B) Cellsprobed with antibodies targeting GM130 (apical polarity), CD49f (basal pphalloidin or an antibody targeting E-cadherin was used as a counterstawere incubated for 20 minutes with calcein AM, ethidium homodimer 1Nikon confocal laser scanning microscope.we used, changes in total nuclear count reflected thenumber of epithelial cells present, irrespective of thefixed number of nondividing feeder cells present per well(Additional file 6: Figure S2A).The primary screen (two replicates per condition)resulted in the identification of 2,337 genes (of 21,121(11.1%)) in which knockdown led to a statistically significantdecrease (adjusted P-value <0.05 by Benjamini-Hochberganalysis) in cell growth to less than 25% of the controlcondition (Figure 3B and Additional file 7: Table S2A). Wenoted that several receptors for required components of thedefined growth media were ranked in the top 50%(Figure 3B) of growth effects (insulin, epidermal growthfactor (EGF), isoproterenol and transferrin). Enrichmentmap analysis using the Reactome Functional Interactomeplugin in Cytoscape v2.8.1 [23,24] of the genes ranked byprimary screen growth inhibition showed four statisticallysignificant linked networks (Figure 4) and several singletons.As expected, the largest network implicated cell cyclefunctions; however, we also noticed a multinode networkand several singleton nodes representing G protein–coupledreceptor (GPCR)-mediated transmembrane signalling.We therefore focused on further analysis of genes withsubcellular location annotations [37] (Additional file 7:Table S2B) that indicated the presence of transmembrane,extracellular, and secreted proteins (388 genes withgreater than 75% primary growth inhibition). To con-trol for clonal cell line effects, we performed a sec-ondary screen with these 388 genes (using a co-cultureassay as primary screen) with both 184-hTERT-L9 and184-hTERT-E11, a sister clonal cell line to 184-hTERT-L9with identical growth characteristics, but with a differenthTERT integration site. Of these 388 genes, 140 (36% ofthose retested) were considered reproducible in that theyproduced the same magnitude of growth reduction uponreassessment in the secondary screen (Figure 3C) in oneor both clonal cell lines. We determined which of the 140screen-reproducible siRNA pools were likely acting ontarget initially by quantitative RT-PCR (qRT-PCR) todetermine which siRNA pools resulted in significant targettranscript knockdown 48 hours after transfection into184-hTERT-L9 cells. For this purpose, cells were grown inBPE as opposed to co-culture with irradiated feedercells to reduce interfering mRNA signals. We found that 47(33.6%) of 140 of the siRNA pools achieved statisticallyin three dimensions. (A) 184-hTERT-L9 cells were arrested inzation analysis. Metaphase spreads had a normal diploid chromosomeltured in three-dimensional Matrigel were fixed after 21 days andrity) and MUC1 (luminal marker). Either Oregon Green–labelledwith DRAQ5 nuclear stain. For viability assessment, unfixed structuresthD-1) and Hoechst 33452 immediately prior to being imaged on asignificant mRNA silencing (adjusted P < 0.05 by Benjamini-Hochberg analysis) (Additional file 7: Table S2C), and thesewere designated as the target genes for further study.Additional validation was conducted as follows. (1)We deconvolved each siRNA pool to individualsiRNAs and noted that in 20 (42.5%) of 47 of thesiRNA sets, at least three of the individual siRNAsproduced a statistically significant decrease (adjustedP < 0.05 by Benjamini-Hochberg multiple comparisonsmethod) (Additional file 7: Table S2C) in growth withinthe two-dimensional co-culture assays. (2) We testedFigure 3 High-throughput screen for genes regulating mammary epithelial cell growth. (A) Summary of screen and follow-up assaysleading to the identification of 47 candidate target genes. (B) Ranked distribution of growth effects for small interfering RNA (siRNA) pools utilizedin the siRNA screen, normalized to the Lipofectamine 2000 (LF2K) reagent-alone control condition. Pooled siRNAs with a statistically significanteffect that reduced growth to 25% or less of the control condition are highlighted in blue. (C) Ranked distribution of growth effects for siRNApools utilized in the secondary screen targeting genes on the plasma membrane and in the extracellular space. Pooled siRNAs with a statisticallysignificant effect in 184-hTERT-L9 and/or 184-hTERT-E11 cells that reduced growth to 25% or less of the control condition are highlighted in red.EGF, Epidermal growth factor; RT-QPCR, Quantitative RT-PCR.Number of coloniesIrradiated NIH 3T3s per dish (x1000)Bi-potent progenitor cells184-hTERT-L9 cellsMCF10A cells0 40 80 120 160 0 20 40 80 100 120 160 200 Figure 2 Fibroblasts influence mammary epithelial cell growth. Colony-forming assays were performed with an increasing density ofirradiated NIH 3T3 cells plated alongside a set density of prospectively fractionated human bipotent mammary progenitor cells (black; n = 6 patientsamples), 184-hTERT-L9 cells (blue; n = 4) and MCF10A cells (red; n = 4). The whisker ends on the box plots denote the lowest and highest data within1.5× the interquartile range of the lower and upper quartiles, respectively.Burleigh et al. Breast Cancer Research  (2015) 17:4 Page 8 of 21IBOMEPNANTATI3KBurleigh et al. Breast Cancer Research  (2015) 17:4 Page 9 of 21BETA-ARRESTINSIN GPCR DESENSITIZATION(B)PROTEASOME(K)DNAREPLICATION(R)NEUROACTIVELIGAND-RECEPTORRHOSTINTERACTIONSOF HIV FACTORS(R) METABOLISM OF AMINO ACIDS(R) IL2 SIGEVEMEDIBY PG1/STRANSITION(R)APC/C-MEDIATEDDEGRADATIONOFCELL CYCLE PROTEINS(R)CELL CYCLE CHECKPOINTS(R)S PHASE(R) CELL CYCLE(P) M/G1TRANSITION(R)APOPTOSIS(R)SIGNALING BY WNT(R)growth inhibition using independently designed lentiviralshort-hairpin RNA (shRNA) constructs for each of the 47target genes. One to three shRNA constructs from theGIPZ human lentiviral shRNA library (Dharmacon) wereselected for analysis [38], and shRNAs were transduced inBPE-supplemented medium to maximize the potentialof rescuing growth and formation of stable clones.Stably growing clones could not be isolated from 29shRNAs covering 24 genes, suggesting the inability ofthese cells to grow upon gene silencing of thesegenes (verified by qRT-PCR in short-term culture;see Additional file 7: Table S2D), even with BPEpresent in the medium. Even among the stable clonesderived, fibroblast-dependent growth was reduced. Intotal, 34 (72.3%) of 47 of the transmembrane and/orextracellular space protein encoding transcripts appearedto be required for epithelial cell growth, as determined bysiRNA pool deconvolution and/or by knockdown with asecond gene RNAi method (shRNA) (Additional file 6:Figure S2B and Additional file 7: Table S2C).The increase and decrease in cell numbers observed inthe primary and secondary screens likely resulted from aRECEPTOR-LIGANDCOMPLEXESBIND G PROTEINS(R)CLASS A/1 (RHODOPSIN-LIKERECEPTORS)(R)INTERACTION(K)SIGNALINGAURORAKINASES(Figure 4 Network analysis of genes important in epithelial cell growtprimary screen data that reduced growth to 25% or less of the control conplugin in Cytoscape v2.8.1. APC/C, Anaphase-promoting complex; GPCR, GPI3K, Phosphatidylinositol 3-kinase; TNF, Tumour necrosis factor.SOME(K)FORMATIONANDMATURATIONOFMRNATRANSCRIPT(R)TABOLISM OF ROTEINS(R)BASALTRANSCRIPTIONFACTORS(K)TRANSLATION(R)INFLUENZA LIFE CYCLE(R)HIV LIFE CYCLE(R)TRANSCRIPTION(R)PARKINSONDISEASE(P)CALCIUMSIGNALINGPATHWAY(K)CHROMATINREMODELING BY HSWI/SNFATP-DEPENDENTB CELL LING SED(N) mix of proliferation (cell division) and apoptosis. Todetermine the relationship and relative contribution ofthese two factors, we quantified caspase 3/7 activity andproliferation simultaneously by using high-content livecell imaging (see IncuCyte ZOOM description in theMethods section). We observed that there was aninverse correlation (Additional file 8: Figure S4B (rankcorrelation −0.954) and Additional file 9: Figure S5(rank correlation, −0.959)) between caspase activityand the rate of cell increase (determined as AUC overtime). We verified that cell proliferation measured byhigh-content live cell imaging as AUC over timewas positively correlated with the fraction of cells inS-phase, which we determined by EdU incorporation(Additional file 8: Figure S4A (rank correlation,0.735)). The relationship between caspase activationand proliferation was similar, regardless of whether theepithelial 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 rankcorrelation, 0.892; P < 1 × 10−5 in randomization test).Taken together, these data show that the genes of greatestCOMPLEXES(B)RECEPTORSIGNALINGPATHWAY(K) NICOTINICACETYLCHOLINERECEPTORSIGNALINGPATHWAY(P)TNFALPHA/NF-KB(C) BY N)h. Enrichment map of the top-ranking clusters of genes within thedition, as determined with the Reactome Functional Interactomeprotein–coupled receptor; IL2, Interleukin 2; NF-KB, Nuclear factor κB;Burleigh et al. Breast Cancer Research  (2015) 17:4 Page 10 of 21effect on cell growth tend to affect both proliferation(cell cycle) and caspase activation (apoptosis). Someexceptions were noted, such as the EDG (LPAR3) receptor,where the effect on proliferation was higher in rank thanthe degree of apoptosis.Functional annotation clustering was performed usingthe DAVID Bioinformatics Database, with enrichmentbased upon the Gene Ontology biological process forthe 47 genes required for 184-hTERT-L9 growth [39].Enrichment scores were generated based upon thefunctional classification of these genes and denote therelatedness of a seemingly heterogeneous group ofgenes (Additional file 7: Table S2F). Notably, a number ofguidance factors required for neuronal development arenecessary for fibroblast-driven mammary epithelial cellgrowth [40,41]. Also enriched are clusters of genesinvolved in axon guidance, signal transduction throughprotein kinase cascades, intracellular ion homeostasis,GPCR signalling and cell migration (Additional file 10:Figure S6B). Additional file 10: Figure S6A displays theaxon guidance pathway in which SEMA3C, and part of itsreceptor complex, PLXNA2, ROBO3, EFNA4, NTN1 andNTN2L from our target gene list, are all highlighted. Axonguidance molecules are increasingly recognized to playa role in mammary gland development and breast tumouri-genesis [42], and our data significantly strengthen this asso-ciation. We considered the possibility that some of the 47genes identified may have differential expression in theluminal and basal developmental lineages, and therefore weinspected flow-sorted RNA-seq libraries of mammarylumen and myoepithelium (three independent pairs fromthe RNA-seq library; see Additional file 11: Table S3 anddiscussion of NIH Canadian Roadmap Epigenomicsmammary epithelial cell RNA-seq libraries in the Methodssection). However, no statistically significant differenceswere observed (Holm-Bonferroni-adjusted P-values >0.05).Distinct requirements for transmembrane and/orextracellular genes in three-dimensional epithelial cellgrowth and differentiationThe response of mammary epithelial cells to mitogensand cellular signalling can differ dramatically when cellsare grown on solid supports or feeder layers, as opposedto within a three-dimensional context embedded inextracellular matrix (Matrigel) [43,44]. Using short- andlong-term cultures (with defined media and no BPEsupplements) in Matrigel assays, we sought to determinewhich of the 47 gene transcripts identified as modulatingepithelial cell growth in the co-culture assay were alsorequired for three-dimensional growth and differentiation.First, using acinar formation in short-term culture as agrowth measure (that is, counting three-dimensionalstructures formed after plating cells at low density, asin a two-dimensional colony-forming assay), we determinedthe requirement for each gene using siRNA knockdown. AssiRNA transfection complexes cannot effectively penetrateMatrigel, we transfected cells in culture for 48 hours priorto harvesting and replating them in Matrigel [45]. Wedetermined that siRNA silencing was effective for up to10 days in Matrigel cultures of 184-hTERT-L9. Thenumber of acini compared to a nontargeting siRNAcontrol was enumerated after 8 days in culture (Figure 5).All but four genes (ADCY4, PCDHB13, KCNJ5 andFLOT2) showed significant (P < 0.001 with Dunnett’sadjustment) reductions of growth in three dimensions,and three genes (HSD17B2, PROCR and SNN) showedmore than tenfold reductions in the number of acini. Wenoted that, despite having the greatest effects on three-dimensional growth, PROCR (31.6-fold decrease in threedimensions; 95% confidence interval (CI), 14.0 to 163.8)and HSD17B2, a transmembrane-linked enzyme requiredfor oestrogen and testosterone steroid biosynthesis(27.3-fold decrease in three dimensions; 95% CI, 10.0to 163.9) were consistently in the third or fourthquartile of effects on two-dimensional growth in boththe primary and secondary two-dimensional screensin comparison to the final 47 target genes. In all genes, themagnitude of growth modulation in two dimensions didnot correlate with the magnitude of three-dimensionalgrowth change (Additional file 7: Table S2C).The requirements for three-dimensional growth aredistinct from monolayer cultures, and the process ofacinar formation can be disrupted for many reasons.In the second approach, we sought to determine ifthree-dimensional effects were limited solely to growth bylooking at acinar formation after 21 days when polarizationand lumen formation have occurred. For a 21-day culture,long-term RNAi is required, and we therefore exam-ined shRNA stable lines in which growth could berescued (for initial clone generation) by BPE mediasupplements. Among the three genes showing thegreatest effects on acinar formation, SNN, HSD17B2 andPROCR (Additional file 7: Table S2C), the shRNAsinhibited growth to a degree precluding derivation ofstable clones for all but PROCR. The PROCR stableclone exhibited a 14.1-fold (99% CI, 7.16 to 28.57;RQ Manager software; Applied Biosystems, FosterCity, CA, USA) compared to nontargeting controls(Additional file 7: Table S2D). When placed into aquantitative two-dimensional co-culture assay, growthwas decreased 3.73-fold (95% CI, 3.38 to 4.16; Student’st-test) in comparison to a stable cell line generatedwith a nontargeting shRNA construct.After plating control and PROCR-knockdown clones inMatrigel, the number of acinar structures was decreasedcompared to the control nontargeting shRNA. Those acinithat formed did so abnormally. These acini lacked thenormal multilayered epithelial organization and did notPhalloidinNon-targetingshRNAPROCRshRNANuclei GFP MergeFigure 6 PROCR expression is necessary for cell organization and luminal clearance in three-dimensional culture. 184-hTERT-L9 cell lineswith stable integration of pGIPZ short-hairpin RNA (shRNA) lentiviral constructs against PROCR or a nontargeting control were seeded intothree-dimensional Matrigel culture and fixed after 21 days of growth. Staining was performed with Alexa Fluor 546–conjugated phalloidin andDRAQ5 nuclear stain prior to imaging on a Nikon confocal laser scanning microscope. GFP, Green fluorescent protein.Fold changeADCY4PCDHB13KCNJ5FLOT2SEMA3CROBO3PLA2G2FMMP28SLC6A4RHCESCARB2CD79AGPR182BDKRB2TMEM9BSLC7A7PDCD1GPR80GPR39SERPINH1OPRS1PLUNCCOL9A3NKAIN4LPAR3FZD2CTNNA1FLJ30634LGALS1PARD3BST1NTN2LSAA1ACE2NTN1LTBP3MMP24TMEM14CNPTX1EFNA4TUFT1SNNHSD17B2PROCR35 33 31 29 27 25 23 21 19 17 15 13 11 9 8 7 6 5 4 3 2 1 2Down | UpFigure 5 Genes that modulate two-dimensional co-culture growth are also important in anchorage-independent three-dimensionalgrowth. Three-dimensional Matrigel culture assays were performed with 184-hTERT-L9 cells after they were transfected with small interfering RNA(siRNA) pools targeting the 47 genes required for mammary epithelial cell growth. The acini (with an acinus defined as a ball of 50 cells or more)formed per well were counted 8 days postplating, and surviving fraction estimates were generated in comparison to both transfection reagent-onlyand nonsilencing siRNA controls. Acini count fold changes are plotted as a rank relative to the control mean. Fold changes were calculated as meannumber of acini when the gene was silenced in three-dimensional cultures divided by the mean number of acini when siRNA and transfection reagentcontrols were used in three-dimensional cultures. Error bars represent 95% confidence intervals (n = 4).Burleigh et al. Breast Cancer Research  (2015) 17:4 Page 11 of 21in vitro growth of human primary mammary progenitorcells. The strategy employed was to transduce primaryepithelium–enriched, but otherwise unsorted, cellsobtained from dissociated reduction mammoplasties(see the Methods section for details) and to assay theluminal and basal cell epithelial lineages by scoringcolony types in colony-forming assays [10,47-50]. Allof the shRNA constructs tested produced statisticallysignificant silencing of the mRNA transcript of interest inepithelial cells (Additional file 7: Table S2D). Transductionefficiency in unsorted primary tissue, as gauged by greenfluorescent protein (GFP) expression from the pGIPZlentiviral vector backbone, was 28.3% (95% CI, 20.2% to36.1%). Three patient samples were evaluated in technicalduplicates for each target gene to compare the effect ofFigure 7 LPAR3, but not LPAR1 or LPAR2, is necessary fortwo-dimensional and three-dimensional mammary growth.Equal numbers of 184-hTERT-L9 cells transfected with small interferingRNA (siRNA) pools against three lysophosphatidic acid (LPA) receptors(LPAR1, LPAR2 and LPAR3), transfection reagent alone and PLK1 (as apositive control) were plated in two-dimensional colony-formingassays (n = 3) (A) and three-dimensional Matrigel cultures (n = 4)(B). Cells were counted after 8 days in culture. Error bars represent95% confidence intervals.Burleigh et al. Breast Cancer Research  (2015) 17:4 Page 12 of 21form a hollow lumen, as seen in structures formedwith the nontargeting shRNA (Figure 6), indicating arole for PROCR in mammary epithelial differentiation,organization and growth. To further explore the effects onthree-dimensional differentiation and epithelial organization,we quantified the degree of polarization and lumenformation for PROCR, EFNA4, LGALS1 and NTN1.Knockdown of all four genes resulted in increased disrup-tion of GM130 localization (70% to −88% nonpolarizedstructures compared to 14% for nontargeting siRNA)(Additional file 1: Figure S7 and Additional file 2:Figures S8A and S8B); however, only LGALS1 knockdownshowed significant disruption of lumen formation (50%nonhollow compared to 8% to 17% for other genes andnontargeting siRNA) (Additional file 1: Figure S7 andAdditional file 2: Figures S8A and S8B). PROCR wasthe only gene for which knockdown resulted in significantdisruption of CD49f basal localization (100% structures dis-rupted) (Additional file 1: Figure S7 and Additional file 2:Figures S8A and S8B). Taken together, these resultssuggest that several of the genes identified may haveroles in differentiation and development of multilayeredmammary epithelium.Strikingly, both the two-dimensional and three-dimensional assays showed that several GPCRs appearedto be required for epithelial cell growth, including LPAR3,one of three major receptors for lysophosphatidic acid(LPA). In transgenic mouse models, overexpression ofeach of these individual LPA receptors (LPAR1, LPAR2and LPAR3) under the control of the mouse mammarytumour virus long terminal repeat (MMTV-LTR)promoter led to the formation of late-onset mammarycarcinomas [46]. However, our screen of nontransformedhuman epithelium did not initially identify a role for eitherLPAR1 or LPAR2 in mammary epithelial cell growth. Toverify this, we used siRNA pools to silence LPAR1, LPAR2and LPAR3 expression in 184-hTERT-L9 cells prior toplating the cells in colony-forming assays. In keeping withthe primary siRNA screen results, LPAR3 was required formammary epithelial cell growth, with moderate growtheffects in the colony-forming assay identified when LPAR2was silenced (Figure 7A). Additionally, in three-dimensionalculture conditions, silencing of LPAR1 or LPAR2 did notaffect acinar formation, whereas we observed a completeabrogation of three-dimensional growth when LPAR3 wassilenced in 184-hTERT cells (Figure 7B).Requirement of screen-identified transmembrane and/orsecreted genes for primary mammary progenitor cellgrowthTo address the question whether the growth dependencyidentified in the cell lines is mirrored in primarymammary epithelial cells, we tested 29 of the 47 targetgenes described above for their ability to diminish thethe target shRNA to that of a nonsilencing controlshRNA. The total number of colonies produced from bothBurleigh et al. Breast Cancer Research  (2015) 17:4 Page 13 of 21luminal and myoepithelial progenitor cells was used tocalculate the fraction of surviving colonies (see theMethods section for details) after silencing of the targetgenes (Figure 8). The high patient-to-patient variability inprimary epithelial cell growth is a feature that constrainedthe quantitative assessment of primary cell growth effects;for example, LPAR3 did not show a statistically significantgrowth difference. Nevertheless, the data showed a rangeof quantitative effects on colony formation of cells derivedFigure 8 Genes modulating 184-hTERT-L9 co-culture growthalso affect primary human mammary epithelial cell growth.Single-cell suspensions of dissociated human primary mammary cellswere infected with individual short-hairpin RNA (shRNA) constructstargeting our screen-identified gene set prior to being plated incolony-forming assays. The fraction of surviving colonies after silencingof the target genes relative to a nontargeting shRNA control is plotted.Error bars represent 95% confidence intervals (n= 3 patient samples).from primary epithelium, with more than two-thirdsof the genes exhibiting a median decrease in colonyformation and two genes showing a statistically significantdecrease in surviving fraction (GPR39 and NTN1; 2.19-foldand 3.33-fold decreases, respectively).Expression of screen-identified genes among breastcancer subtypesGenes regulating growth and differentiation are importantcomponents of the cell type landscape on which malignanttransformation occurs in human cancers. One indication ofthis is the cosegregation of gene expression with biologicalsubtypes of cancer. To address this question for the genesidentified in relation to epithelial cell growth, we examinedthe distribution of transcript expression by subtype across1,998 breast cancer patients [25] for whom mRNA expres-sion has been measured, alongside other features of thegenome and with clinical outcomes. Within the last year,the METABRIC study and related genomic landscapeshave redefined the number of biologically distinct primarybreast cancer subtypes [22,25,51-55]. Using these data and10 recently defined biological primary breast cancersubgroups, we examined the expression levels of the 47screen-identified genes (Figure 9A and B, Additional file 7:Table S2C and Additional file 12: Figure S9). The relation-ship to the five PAM50 breast cancer subgroups was alsocalculated (Additional file 13: Figure S10).We examined each gene individually for subtype-specificexpression. Significant differences in expression distributionwere seen for 40 of the 47 genes across the 10 METABRICdatasets (Figure 9B and Additional file 12: Figure S9)(Kruskal-Wallis-adjusted P-value <0.05, Benjamini-Hochbergmultiple-comparisons method) and for 39 of the 47genes in the PAM50 groups (Additional file 13: Figure S10)[25]. Several genes (for example, RIPK2, EFNA4 andTMEM9B) (Figure 9B, Additional file 12: Figure S9 andAdditional file 13: Figure S10) showed differential expres-sion in subtypes associated with high proliferation, such asIntClust 5 (predominantly HER2/ERBB2-amplified cancers)and IntClust 10 (predominantly basal expression typecancers). 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 significantenrichment for over- and underexpression of several genes.We noted that 21 of 47 genes required for 184-hTERT cellslie within chromosomal hotspots for copy number amplifi-cation (Additional file 7: Table S2C). All 47 genes have pre-viously identified mutations in human tumours, and 18 ofthese 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 arandomization simulation study to assess the strength ofassociation demonstrated by this set of 47 genes withinthe 10 IntClust subtypes. In 10,000 simulation runs,                                                           Down        Fold Change              Up SLC6A4LPAR3HSD17B2ACE2GPR80PDCD1CD79ASLC7A7RIPK2PCDHB13CTNNA1TUFT1EFNA4OPRS1COL9A3SNNFLOT2SAA1ROBO3ADCY4TMEM14CFZD2SERPINH1LGALS1BST1PROCRMMP28NPTX1MMP24LTBP3NTN2LPARD3GPR182RHCEPLA2G2FPLUNCNKAIN4GPR39KCNJ5SEMA3CSCARB2BDKRB2NTN1FIT1TMEM9BKTELC11      2           3                   4                5        6       7                8            9         100.4 0.6 0.8 1.0 1.2 1.4 1.6FLOT2 *CD79A **++TMEM14C **NPTX1 *SEMA3C *KTELC1FIT1MMP24PLA2G2FADCY4BST1SNNPDCD1SCARB2PLUNCPARD3NTN1RHCESAA1NKAIN4TMEM9BMMP28COL9A3GPR80OPRS1LTBP3GPR39GPR182ROBO3LGALS1PCDHB13NTN2LLPAR3SLC6A4FZD2HSD17B2EFNA4BDKRB2ACE2 *CTNNA1PROCRRIPK2  +SLC7A7SERPINH1 **++TUFT1 *KCNJ5 **−4−20246RIPK2intClust GroupGene Expression Log2 Z1 2 3 4 5 6 7 8 9 10K−W   p = 9.9e−128−4−20246TUFT1intClust GroupGene Expression Log2 Z1 2 3 4 5 6 7 8 9 10K−W   p = 2.7e−111−4−20246EFNA4intClust GroupGene Expression Log2 Z1 2 3 4 5 6 7 8 9 10K−W   p = 3.5e−109−4−20246TMEM9BintClust GroupGene Expression Log2 Z1 2 3 4 5 6 7 8 9 10K−W   p = 7.9e−109Figure 9 (See legend on next page.)Burleigh et al. Breast Cancer Research  (2015) 17:4 Page 14 of 21sslaymapsnceatploerewiingesanBurleigh et al. Breast Cancer Research  (2015) 17:4 Page 15 of 21selecting 47 genes at random from among the 4,103transmembrane/extracellular gene set, we found that211 (0.21%) of the 10,000 of the random sets showed40 or more genes to be significantly associated withthe 10 METABRIC groups (null hypothesis simulationKruskal-Wallis P = 0.021). For the PAM50 subtypes,637 (0.637%) of 10,000 of the random sets of 47 genesshowed 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 simplya randomly assembled set, suggesting that this gene set isenriched with regard to association with breast cancersubtype.Finally, we asked whether expression differences inany of the 47 genes exhibit independent disease outcomeassociations (overall survival) in the 1,998-patient dataset.This was tested using two methods of multivariableanalysis to determine if the prognostic significance of thegene expression was independent of that already carriedby the breast cancer subtype (see the Methods section fordetails). The ranked proportional hazards with respect tooverall survival (Figure 9C) for each gene showed arange of effect sizes. Two gene targets showed significantdifferences in survival within IntClust breast cancer subtypegroups as assessed by Kaplan-Meier G-rho-stratifiedanalysis and by Cox proportional hazards analysis,(See figure on previous page.)Figure 9 Genes identified though screening are differentially expregenes found to be required for mammary epithelial cell growth are dispMETABRIC dataset. Gene expression was determined on the Illumina Huare clustered according to ten biological primary breast cancer subgroudepicting target gene expression in the ten biological primary breast caobservations are shown as small horizontal lines in a one-dimensional scand the average indicated by the long horizontal line. (C) Rank-orderedhigh gene expression relative to low gene expression demonstrates difftargets are marked with identifiers for binarized expression variables shoby ** or ++) and adjusted P-values <0.1 (FDR, 0.1) (denoted by * or +). Scomparisons using the G-rho model test (ρ = 1). Single plus sign indicatan omnibus test from a Cox model with biomarker and IntClust subtypeafter adjustment for multiple comparisons using themethod of Benjamini and Hochberg [26]. Elevated CD79Aexpression showed improved survival for IntClusts 8 and10 (Additional file 14: Figure S11). Elevated SERPINH1expression showed poorer survival for IntClusts 6, 9 and10. An additional two gene targets showed significant dif-ferences as assessed by the Kaplan-Meier G-rho-stratifiedanalysis, after adjustment for multiple comparisons.Elevated KCNJ5 expression showed poorer survival forIntClusts 4 and 8. Elevated TMEM14C expression showedimproved survival for IntClust 4.DiscussionThe extracellular factors and transmembrane signalsthat regulate mammary epithelial growth and differentiationremain poorly understood. In part, this is due to a lack ofmethods for systematic, genome-wide, genetic andfunctional interrogation of genes in relation to mammaryepithelial growth and differentiation. Although genome-wide functional screens using RNAi methods have provensuccessful in many similar instances (for example, see[58,59]), these have mostly been undertaken with trans-formed, somatically mutated epithelial cell types, wherekey extracellular interactions that modulate growth cannotbe recapitulated. Important features of the endogenousmilieu, such as growth stimulation by fibroblast stromahave been undersampled as a consequence. To overcomethis, we isolated and characterized a cloned, diploid, non-transformed mammary epithelial cell line, 184-hTERT-L9,which retains both fibroblast growth dependence and thecapacity to differentiate in three-dimensional growth con-ditions. We used this cell line in a genome-wide RNAiscreen to identify, in a systematic manner, genes requiredfor mammary progenitor cell growth and differentiation.The 184-hTERT-L9 clone described here is derived fromprimary mammary epithelium immortalized by hTERTtransfection after limited initial passages. The clonal lineretains important properties, such as fibroblast-dependentgrowth and the ability to differentiate in three-dimensionalcultures, and is diploid and nontransformed.Fibroblasts are known to possess an instructive roleed across breast cancer subtypes. (A) Expression levels of the 47ed on a heat map representing 1,998 breast cancer patients within thenHT-12 Expression BeadChip array (Illumina, San Diego, CA, USA). Casesdescribed in Additional file 2: Figure S8 (also see [25]). (B) Beanplotsr subgroups for RIPK2, TUFT1, EFNA4 and TMEM9B. The individualterplot with the estimated density of the distributions shown in colourt of hazard estimates and unadjusted 95% confidence intervals fornces in survival based upon the expression of the target genes. Geneng adjusted P-values <0.05 (false discovery rate (FDR), 0.05) (denotedle asterisk indicates a significant finding after adjustment for multiplea significant finding after adjustment for multiple comparisons usingd their interaction terms. K-W, Kruskal-Wallis test.in regulating mammary epithelial cells in normal develop-ment and oncogenesis [6,60,61]. More specifically, thein vitro growth of bipotent progenitor cells is reliant uponthe presence of fibroblasts as feeder cells. The 184-hTERT-L9cells mimic the growth of bipotent progenitor cells whenplated with increasing densities of irradiated NIH 3T3 feedercells in a well-defined colony-forming assay. This is notobserved in epithelial cell lines with genomic aberrationsand/or additional adaptation events, such as MCF10A cells(Figure 2) and transformed epithelial cells derived frommalignancies. Thus, 184-hTERT-L9 cells provided agenomically characterized model system, also amenableto RNAi transfection and image-based, high-contentscreening, whereby we could replicate in vitro the fibroblast-dependent growth environment of mammary progenitorTable 1 Relative effects of the 47 target genes in two-dimensional and three-dimensional culturesaGenenameRefSeq ID Relative rank, primaryscreen (N = 47)Median relative rank,secondary screen (N = 47)Relative rank,three-dimensionalgrowth assay (N = 44)Three-dimensionalacinar formation,fold decrease (siRNA)shRNA knockdowneffect on cloneACE2 NM_021804 22 30 11 5.169 LethalADCY4 NM_139427 44 17 44 0.878 LethalBDKRB2 NM_000623 28 44 31 2.484 Growth postselectionBST1 NM_004334 7 6 14 4.406 LethalCD79A NM_001783 13 16 33 2.366 Growth postselectionCOL9A3 NM_001853 16 4 22 3.489 LethalCTNNA1 NM_001903 37 29 18 4.047 LethalEFNAA NM_005227 38 37 5 7.338 Growth postselectionFIT1 NM_203402 4 22 No data No data Growth postselectionFL_J30634 NM_153014 15 30 17 4.055 Growth postselectionFLOT2 NM_004475 23 31 41 1.618 LethalFZD2 NM_001466 6 25 19 3.7126 LethalGPR182 NM_007264 9 20 32 2.483 Growth postselectionGPR39 NM_001508 10 22 26 3.159 Growth postselectionGPR80 NM_080818 32 14 27 3.137 LethalHSD17B2 NM_002153 27 38 2 27.328 LethalKCNJ5 NM_000890 3 35 42 1.494 Growth postselectionKTELC1 NM_020231 8 17 No data No data Growth postselectionLGALS1 NM_002305 36 7 16 4.120 Growth postselectionLPAR3 NM_012152 1 5 20 3.530 Growth postselectionLTBP3 NM_021070 39 28 9 5.291 Growth postselectionMMP24 NM_00690 40 5 8 5.447 Growth postselectionMMP28 NM_024302 18 25 37 2.075 LethalNKAIN4 NM_152864 20 9 21 3.505 LethalNPTX1 NM_002522 45 18 6 7.082 LethalNTN1 NM_004822 47 22 10 5.175 Growth postselectionNTN2L NM_006181 5 30 13 4.691 Growth postselectionOPRS1 NM_005866 29 26 24 3.243 LethalPARD3 NM_019619 30 31 15 4.221 LethalPCDHB13 NM_018933 46 27 43 1.214 Growth postselectionPDCD1 NM_00518 21 36 28 2.997 Growth postselectionPLA2G2F NM_022819 19 22 38 2.028 LethalPLUNC NM_016583 31 30 23 3.280 Growth postselectionPROCR NM_006404 35 36 1 31.637 Growth postselectionRHCE NM_020485 42 31 35 2.209 LethalRIPK2 NM_003821 41 23 No data No data LethalROBO3 NM_022370 11 35 39 2.012 Growth postselectionSAA1 NM_000331 14 21 12 5.167 LethalSCARB2 NM_005506 34 9 34 2.317 LethalSEMA3C NM_006379 12 25 40 1.936 Growth postselectionSERPINH1 NM_001235 26 32 25 3.187 LethalSLC6A4 NM_001045 17 29 36 2.142 Growth postselectionSNN NM_003498 24 30 3 11.523 LethalBurleigh et al. Breast Cancer Research  (2015) 17:4 Page 16 of 21nsreeseorl orBurleigh et al. Breast Cancer Research  (2015) 17:4 Page 17 of 21cells in order to investigate the signalling pathways involvedin regulating mammary epithelial and progenitor cell growth.Among the 21,121 siRNA pools tested in the primaryscreen, 2,337 demonstrated statistically significant abro-gation of growth to less than 25% of the control condi-tion. This was a stringent selection criterion, given thatknockdown of receptors for two of the four definedmedium components (insulin, EGF, transferrin andisoproterenol) did not achieve this level of growth inhib-ition. Surprisingly, GPCR and associated signallingproteins were also found amongst this list, suggestingan underappreciated role of this class of receptorswithin mammary gland biology.To identify novel external regulators and signal transduc-ers, we focused our in-depth analysis on cell surface andsecreted genes. After rescreening including an independ-ently cloned sister cell to assess reproducibility, and afterqPCR assessment for on-target siRNA activity, we identified47 transmembrane genes for follow-up examination bydeconvolution and by assaying growth using independentlytargeted shRNA constructs (summarized in Table 1).Moreover, we quantified the relative influence of prolifera-tion and apoptosis for each gene, which indicated ageneral inverse correlation between these two functions.Although these 47 genes have diverse functions, theyare strikingly enriched for both GPCRs (LPAR3, FZD2,ADMR, BDKRB2, GPR39, GPR80 and GPR182) (notedin the primary screen) and axonal guidance molecules(SEMA3C, PLXNA2, ROBO3, EFNA4, NTN1 and NTN2L).For many of these genes, we provide the first description ofa role in growth regulation or mammary biology.To better understand the roles of these genes in growthTable 1 Relative effects of the 47 target genes in two-dimeTMEM14C NM_016462 33 32TMEM9B NM_020644 43 23TUFT1 NM_020127 25 20/47 /47aThe relative rank of target genes in the primary small interfering RNA (siRNA) scrank in the three-dimensional acinar growth assay are displayed. The fold decreato the control condition. The effect of infection with an independent lentiviral shpuromycin (resistance marker within lentiviral construct) as a binary effect (lethaand differentiation (reviewed in [62]), we assessed therequirement of the 47 validated target genes for growth inthree-dimensional culture. The silencing of all but four ofthe genes (ADCY4, PCDHB13, KCNJ5 and FLOT2)decreased three-dimensional acinar formation to a levelcomparable to that seen with PLK1 silencing (which isessential for mitosis). Intriguingly, we have shown thatLPAR3 is required for two-dimensional and three-dimensional mammary growth. Given the establishedimportance of LPAR1, LPAR2 and LPAR3 in mammarytumourigenesis, we wanted to confirm that LPAR1and LPAR2 are indeed irrelevant for growth of normalepithelium (as determined in the primary geneticscreen). Colony formation in two-dimensional assaysand three-dimensional acinar formation within Matrigelstill occur upon silencing of LPAR1 and LPAR2, suggestingthat they are not required for the growth of normal,nontransformed epithelial cells. LPAR3 has a greaterbinding affinity for 2-acyl-LPA with unsaturated fattyacids, whereas LPAR1 and LPAR2 are more responsive tosaturated acyl chains [63]. With responsiveness to asimilar ligand, it is possible that compensatory redundancyexists between LPAR1 and LPAR2.Some of the genes identified as regulating growthin two dimensions also affected differentiation whenepithelial cells were grown in three-dimensions, withSNN, HSD17B2 and PROCR showing greater than tenfoldreduction in acinar formation. The requirement forlong-term cultures (21 days) and the lethality of theshRNAs for SNN and HSD17B2 precluded analysis ofthese two genes; however, reduction of PROCR inlong-term cultures was associated with absence oflumen formation and disorganized epithelial growth.We quantified the relative effects on epithelialorganization and lumen formation and observed thatdisordered differentiation was also present for two axonalpathfinding associated genes, EFNA4 and NTN1, andfor LGALS1 (Additional file 1: Figure S7, Figure 8),with differential effects on polarization and epithelialorganization. PROCR has been implicated as a receptorfor protease-cleaved substrates in breast cancer migration[64] and as a marker of colony-forming cells in malignantcell lines [65]. Here we show for the first time, to ourknowledge, a role in growth and differentiation of primaryional and three-dimensional culturesa (Continued)7 5.884 Lethal30 2.556 Lethal4 7.399 Lethal/44n, the median relative rank in the secondary siRNA screen and the relativein structure formation for each target gene is displayed for each siRNA relativet-hairpin (shRNA) clone against each target gene is shown after selection withgrowth postselection).breast epithelium. Loss of NTN1 causes disorganization inthe terminal end buds, an effect that is proposed to occurthrough the loss of cellular adhesions [66]. It has also beenshown that implantation of NTN1-secreting pellets intomammary glands during pregnancy increases the numberof alveolar structures that develop [67]. In the presentstudy, we show a role for NTN1 in both luminal andbipotent progenitor cell growth in that it was required forcolony formation in our in vitro assays.Finally, genes required for growth and differentiationare often implicated in tumourigenesis. In this study, weidentified a subset of genes that have not previously beenscreening parameters and external validation of results. (A) Total nuclearBurleigh et al. Breast Cancer Research  (2015) 17:4 Page 18 of 21implicated in mammary gland growth or development.We sought to determine if expression of these genescorrelated with any of the breast cancer subtypes.Significant, nonrandom differences in expression distribu-tion across the 10 METABRIC datasets was seen for 40 ofthe 47 genes, with several genes (for example, RIPK2,EFNA4 and TMEM9B) differentially expressed in breastcancer subtypes with high proliferation. Furthermore, wewere able to demonstrate independent prognostic signifi-cance for CD79A (with elevated expression improvingsurvival in two of the ten METABRIC subtypes) andSERPINH1 (with elevated expression decreasing survival inthree of the ten METABRIC subtypes). The possible rolesof these genes in the tumour subtypes studied requiresfuture functional studies in representative models; however,it is notable that all of the genes studied here are accessibleby virtue of solubility or membrane location, making thema practical choice for intervention.ConclusionsThis work shows for the first time, to our knowledge,the diversity of transmembrane signals and/or proteinsrequired for the growth of nontransformed mammaryepithelial cells in the physiological state where extracellularsignals from fibroblasts are required. In doing so, wedemonstrate the functional requirement of severaltransmembrane and extracellular proteins in normalmammary growth in multiple well-accepted models ofin vitro mammary physiology. Taken further, theseproteins were differentially associated with breast cancersubtypes, which were examined in 1,998 patients,indicating that these proteins may be associated with thebiology of breast cancer subtypes. This is of particularnote, as the location of these proteins makes themamenable to therapeutic intervention.Additional filesAdditional file 1: Figure S7. Target gene expression is necessary fornormal acinar formation in three-dimensional culture. 184-hTERT-L9 celllines with stable integration of pGIPZ shRNA lentiviral constructs againstEFNA4, LGALS1, NTN1, PROCR or a nontargeting control were seeded intothree-dimensional Matrigel culture and fixed after 21 days of growth.Staining was performed with Alexa Fluor 546–conjugated phalloidin andDRAQ5 nuclear stain prior to imaging on a Nikon confocal laser scanningmicroscope. For E-cadherin, the percentage of structures that were hol-low versus filled was quantified for each condition, with LGALS1 showinga significant increase in the percentage of filled structures relative to thecontrol. For CD49f (marker of basal polarity), the percentage of structuresthat were polarized versus nonpolarized was quantified, with PROCRshowing a complete reversal of polarization relative to the control andother siRNAs. For GM130 (a marker of apical polarity), the percentage ofstructures that were polarized versus nonpolarized was quantified, withall of the siRNAs tested showing a decrease in polarized structures rela-tive to the nontargeting siRNA control. Examples of the patterns scoredin each case are shown in the upper panel series below the table (whiteindicates higher fluorescence intensity). The middle panel represents thesame images with the greyscale inverted to better reveal the antibodypattern (dark indicates higher fluorescence intensity). The presence ofapoptotic cells in the siNTN1 structures is shown in the bottom panel(dark reddish-brown indicates caspase by immunohistochemistry).Additional file 2: Figure S8. Epithelial organization is disrupted inthree-dimensional culture with silencing of target genes. (A) 184-hTERT-L9cell lines with stable integration of pGIPZ shRNA lentiviral constructs againstEFNA4, LGALS1, NTN1, PROCR or a nontargeting control were seeded intothree-dimensional Matrigel culture and fixed after 21 days of growth. Stain-ing was performed with Alexa Fluor 546–conjugated phalloidin and DRAQ5nuclear stain prior to imaging on a Nikon confocal laser scanning micro-scope. Magnified views of representative structures for each condition arepresented for CD49f, E-cadherin and GM130 staining. (B) The sameimages shown in (A) are depicted with greyscale inverted for visual clarity.Additional file 3: Figure S1. 184-hTERT-L9 cells are a cytogeneticallynormal cloned human mammary epithelial cell line. (A) Southern blotsof HindIII- and BglII-digested DNA with a cDNA probe targeting theneomycin resistance gene present on a lentiviral construct used toimmortalize the cells showing one band per digestion, suggesting asingle integration site and a single population of cells after cloning in threerepresentative cell lines (184-hTERT-L9, 184-hTERT-L5 and 184-hTERT-L2). (B)Immunohistochemistry of formalin-fixed, paraffin-embedded 184-hTERT-L9cell blocks shows ubiquitous expression of keratin 5/6, epidermal growthfactor receptor, the epithelial cell marker E-cadherin and wild-type p53.There is no detectable expression of oestrogen receptor α (ERα) or humanepidermal growth factor receptor 2 (HER2). (C) Western blots of clonal184-hTERT-L9, 184-hTERT-L5, 184-hTERT-L9 and primary unsorted mammaryepithelial cells (HMECs) with antibodies raised against β-actin, keratin 18,keratin 5, p63 and p16. (D) Array-based comparative genomic hybridizationwas performed using the Affymetrix GeneChip SNP 6.0 array (Affymetrix,Santa Clara, CA, USA). Log2 ratios of signal intensity for 184-hTERT cell linescompared to normal human female reference DNA are plotted in relationto chromosomal position.Additional file 4: Table S1. Contrasting properties of the 184-hTERT-L9[20] and MCF10A [17,68,69] cell lines.Additional file 5: Figure S3. Irradiated fibroblasts are not susceptible tosiRNA-mediated RNA interference. (A) Irradiated murine NIH 3T3 cells donot efficiently mediate RNAi with a universal cell-lethal siRNA. IrradiatedNIH 3T3 cells at a density of 30,000 cells/cm2 were transfected withincreasing concentrations of transfection reagent. After 48 hours, live-deaddiscrimination was performed (calcein AM/ethidium homodimer 1) andenumerated using the IN Cell Analyzer and IN Cell Developer software(GE Healthcare Bio-Sciences). The percentage of viable cells was determinedin comparison to nontransfected control wells. Error bars represent standarddeviation (n = 2). (B) Irradiated fibroblasts do not show protein targetknockdown. Irradiated NIH 3T3 cells at a density of 10,000 cells/cm2 weretransfected with 50 nM pooled siRNAs targeting β-actin complexed withincreasing transfection reagent concentrations. After 96 hours, β-actin(green) and lamin C (red) levels were detected using the LI-COR Odysseyimaging system (LI-COR Biotechnology, Lincoln, NE, USA). (C) Irradiatedfibroblasts do not show protein target knockdown. Irradiated IMR-90 humanfibroblasts at a density of 10,000 cells/cm2 were transfected with 50 nMpooled siRNAs targeting human GAPDH, with transfection reagent alone(Lipofectamine 2000) and with a nontargeting siRNA (siCon3) as controls.After 72 hours, Western blotting was performed with both GAPDH and thenβ-actin (control). (D) The cell killing abilities of siPLK1 and siTOX werecompared in irradiated and nonirradiated NIH 3T3 cells. Cell number wasassessed (confluence over time) on the IncuCyte ZOOM live cell microscopefor irradiated and nonirradiated NIH 3T3 cells with increasing concentrationsof siTOX and mouse siPLK1. Nontransfected cells (Lipofectamine 2000) wereincluded as the control. Cell number was not above starting and/or controlconditions for the irradiated fibroblasts, as expected, given their arrestedstate. Application of siTOX to the irradiated cells had no significant effect onirradiated cell viability (no difference from controls; Dunnett-adjustedP > 0.05). In contrast, in the nonirradiated cells, both siPLK1 and siTOX resultin cell killing compared to the control (Dunnett-adjusted P < 0.02).Additional file 6: Figure S2. Optimization of genome-wide siRNAcount accurately reflects the 184-hTERT count in the co-culture screeningassay. Comparison of total nuclear count and GFP-positive count inBurleigh et al. Breast Cancer Research  (2015) 17:4 Page 19 of 21co-cultures treated with a siRNA library targeting cell cycle genes showedthat the total count can be used as a surrogate for the GFP-positivecount. Three thousand irradiated NIH 3T3 cells were plated with fourhundred 184-hTERT-GFP cells in 96-well plates prior to transfection with0.3 μl of Lipofectamine 2000 reagent and 30 nM of siRNA per well.Twenty-one fields of view were acquired after 5 days of growth using a10× lens objective on the IN Cell Analyzer. The GFP-negative count wasobtained by subtracting the GFP-positive count (green diamond) fromthe total nuclear count (red square) and represents the number of irradiatedfibroblasts present in each well (blue triangle). (B) Deconvolution of siRNApools and RNA interference with lentiviral shRNA constructs externallyvalidates different subsets of genes. Genes listed in blue and purple withinthe Venn diagram represent genes with which 184-hTERT cells infected withan independently designed lentiviral shRNA construct were unable to growin culture after puromycin was applied to the cells as a means of selectionfor stable integration of the lentiviral construct. Genes listed in red andpurple within the Venn diagram represent genes with which knockdownwith three or four of the individual siRNAs comprising the original siRNApool led to a statistically significant decrease in cell growth (adjustedP < 0.05 by Benjamini-Hochberg multiple-comparisons method).Additional file 7: Table S2. (A) Ranked growth effect of the 21,121genes evaluated in the primary siRNA screen. (B) Ranked growth effect ofthe 388 extracellular and plasma membrane genes evaluated in thesecondary siRNA screen. (C) Relative effects of the 47 target genes intwo-dimensional and three-dimensional cultures, as well as their status inthe METABRIC and COSMIC databases. (D) Relative transcript levels afterlentiviral shRNA silencing in 184-hTERT-L9 cells. (E) RT-qPCR primers andcorresponding Roche Universal ProbeLibrary (Roche Applied Science,Penzberg, Germany) probe number utilized for RT-qPCR. (F) Functionalenrichment scores for 47 target genes. Functional annotation clusteringwas performed on the 47 genes required for 184-hTERT-L9 cell growthusing the Gene Functional Classification tool in the DAVID BioinformaticsDatabase [39]. The data used to derive the annotation clustering heatmaps in Additional file 10: Figure S6 are listed.Additional file 8: Figure S4. Relative contribution of proliferation andapoptosis of target genes in two-dimensional culture. (A) Cell proliferationas measured by AUC over time by high-content imaging is positivelycorrelated with the fraction of cells in S-phase, determined by EdUincorporation. EdU incorporation was performed in 24-well plates containing184-hTERT-L9 cells stably transfected with NucLight Red. Cells weretransfected with 30 nM siRNA 24 hours after plating and labelled with EdU68 hours posttransfection. Labelling was detected with the Click-iT EdU AlexaFluor 488 Flow Cytometry Assay Kit and analysed by flow cytometry. (B) Cellproliferation (AUC over time, median value in red) and relative apoptosis(caspase-3/7 activity, median value in green) in the target genes byrank-ordered effects on proliferation display an inverse correlation. 184-hTERTcells stably infected with NucLight Red were plated without feeder cells andtransfected 24 hours later with 30 nM of siRNA to the respective target genes.CellPlayer Caspase-3/7 reagent was used to mark apoptotic cells after anadditional 24 hours. Proliferation was measured every 4 hours for 84 hours,with respective AUCs for serial measurements calculated for each conditionfor red (proliferating) or dual-labelled (apoptotic) cells.Additional file 9: Figure S5 Relative contributions of proliferationand apoptosis of target genes in two-dimensional culture is un-changed by the presence of feeder cells. Cell proliferation (AUC overtime, median value in red) and relative apoptosis (caspase-3/7 activity,median value in green) in the target genes by rank-ordered effects onproliferation display an inverse correlation. 184-hTERT cells stably infectedwith NucLight Red were plated with irradiated NIH 3T3 feeder cells andtransfected 24 hours later with 30 nM of siRNA to the respective targetgenes. CellPlayer Caspase-3/7 reagent was used to mark apoptotic cellsafter an additional 24 hours. Proliferation was measured every 4 hours for84 hours, with respective AUCs for serial measurements calculated foreach condition for red (proliferating) or dual labelled (apoptotic) cells.Genes with the greatest effect on growth affect both proliferation andapoptosis. (PNG 148 kb)Additional file 10: Figure S6. Genes identified through screening andfunctionally related. (A) Functional annotation clustering was performedon the 47 genes required for 184-hTERT-L9 cell growth using the GeneFunctional Classification tool in the DAVID Bioinformatics Database [39].The cluster of genes involved in axon guidance are highlighted by redstars on the Kyoto Encyclopedia of Genes and Genomes (KEGG) humanaxon guidance pathway [40,41]. (B) Functional annotation clustering wasperformed on the 47 genes required for 184-hTERT-L9 cell growth usingthe Gene Functional Classification tool in the DAVID BioinformaticsDatabase [39]. Enrichment clusters are depicted for genes with theircorresponding associated Gene Ontology biological process terms. Posi-tive gene term associations are represented in green, and currently unre-ported gene term associations are represented in black.Additional file 11: Table S3. Target genes are not differentiallyexpressed in freshly sorted RNA-seq libraries of enriched luminal versusenriched myoepithelial cells taken from human reduction mammoplastysamples. Data from three independent paired libraries of mammaryluminal and myoepithelium for each target gene are shown along withluminal:basal expression ratios and the P-values calculated with theWilcoxon and Student’s t tests and the Holm-Bonferroni adjustment.No statistically significant differences (Holm-Bonferroni-adjustedP-values >0.05) were observed.Additional file 12: Figure S9. Beanplots depicting target geneexpression in the ten biological primary breast cancer subgroups. Theindividual observations are shown as small horizontal lines in aone-dimensional scatterplot with the estimated density of the distributionsshown in colour and the average indicated by the long horizontal line.Additional file 13: Figure S10. Beanplots depicting target geneexpression in the PAM50 breast cancer subgroups. The individualobservations are shown as small horizontal lines in a one-dimensionalscatterplot with the estimated density of the distributions shown incolour and the average indicated by the long horizontal line.Additional file 14: Figure S11. Independent disease outcomeassociations are present with select target genes. Individual genes shownin Figure 9C are depicted with significant survival differences with low(green) versus high (red) gene expression for both integrative cluster(IntClust) breast cancer subtype groups and PAM50 subtypes. Associationswere assessed by Kaplan-Meier G-rho-stratified analysis and by Coxproportional hazards analysis, after adjustment for multiple comparisonsusing the method of Benjamini and Hochberg [26]. CD79A expressionshows improved survival for IntClusts 8 and 10 and HER2. KCNJ5 expressionshows poorer survival for IntClusts 4 and 8. SERPINH1 expression showspoorer survival for IntClusts 6, 9 and 10. TMEM14C expression showsimproved survival for IntClust 4 and luminal B.AbbreviationsAIC: Akaike information criterion; AUC: Area under the curve; BPE: Bovinepituitary extract; CI: Confidence interval; DAPI: 4′,6-diamidino-2-phenylindole;E-cad: E-cadherin; EdU: 5-ethynyl-2′-deoxyuridine; ERα: Oestrogen receptor α;EpCAM: Epithelial cell adhesion molecule; FACS: Fluorescence-activated cellsorting; FDR: False discovery rate; FISH: Fluorescence in situ hybridization;GFP: Green fluorescent protein; GPCR: G protein–coupled receptor;HER2: Human epidermal growth factor receptor 2; IntClust: Integrativecluster; MEBM: Mammary epithelial cell basal medium; MEGM: Mammaryepithelial cell growth medium; MOI: Multiplicity of infection; RNAi: RNAinterference; shRNA: Short-hairpin RNA; siRNA: Small interfering RNA.Competing interestsThe authors declare that they have no competing interests.Authors’ contributionsAB, CE and SA conceived of and designed the experiments. AB, VN, AW, JB,DY, PE and LP performed the experiments. AB, SM, SP, PE, DY, CC and SAanalysed the data. LA and CB provided intellectual input on the cloning of184 polyclonal cells and reviewed the manuscript. SM, SP, LA, CB and CEcontributed reagents, materials, and analytical tools. AB, SM and SA wrotethe manuscript. All authors reviewed the final draft of the manuscript.All authors read and approved the final manuscript.AcknowledgementsThe authors acknowledge excellent technical contributions from Ken Fong,John Fee, Darcy Wilkinson, Gavin Ha and the staff of the Flow Cytometry7. Haslam SZ. Mammary fibroblast influence on normal mouse mammaryBurleigh et al. Breast Cancer Research  (2015) 17:4 Page 20 of 21epithelial cell responses to estrogen in vitro. Cancer Res. 1986;46:310–6.8. Lühr I, Friedl A, Overath T, Tholey A, Kunze T, Hilpert F, et al. Mammaryfibroblasts regulate morphogenesis of normal and tumorigenic breastepithelial cells by mechanical and paracrine signals. Cancer Lett.2012;325:175–88.9. Darcy KM, Zangani D, Shea-Eaton W, Shoemaker SF, Lee PP, Mead LH, et al.Mammary fibroblasts stimulate growth, alveolar morphogenesis, andfunctional differentiation of normal rat mammary epithelial cells. In Vitro CellDev Biol Anim. 2000;36:578–92.10. Stingl J, Eaves CJ, Zandieh I, Emerman JT. Characterization of bipotentmammary epithelial progenitor cells in normal adult human breast tissue.Breast Cancer Res Treat. 2001;67:93–109.11. van Amerongen R, Bowman AN, Nusse R. Developmental stage and timedictate the fate of Wnt/β-catenin-responsive stem cells in the mammarygland. Cell Stem Cell. 2012;11:387–400.12. Makarem M, Spike BT, Dravis C, Kannan N, Wahl GM, Eaves CJ. Stem cellsand the developing mammary gland. J Mammary Gland Biol Neoplasia.2013;18:209–19.13. Silva JM, Marran K, Parker JS, Silva J, Golding M, Schlabach MR, et al.Profiling essential genes in human mammary cells by multiplex RNAiscreening. Science. 2008;319:617–20.14. Bishop CL, Bergin AM, Fessart D, Borgdorff V, Hatzimasoura E, Garbe JC,et al. Primary cilium-dependent and -independent Hedgehog signalinginhibits p16INK4A. Mol Cell. 2010;40:533–47.Facility of the Terry Fox Laboratory and the staff at the Centre forTranslational and Applied Genomics at the BC Cancer Agency. Mammoplastytissue was obtained with the assistance of Drs. Jane Sproul, Peter Lennox,Nancy Van Laeken and Richard Warren. We are grateful to Martin Hirst andmembers of the NIH and Canadian Roadmap Epigenomics programs foraccess to RNA-seq libraries. We thank Dr. Sarah Mullaly for editing themanuscript.Author details1Department of Pathology and Laboratory Medicine, University of BritishColumbia, and BC Cancer Agency, 675 West 10th Avenue, Vancouver, BC V5Z1L3, Canada. 2Centre for Translational and Applied Genomics, BC CancerAgency, 600 West 10th Avenue, Vancouver, BC V5Z 4E6, Canada. 3Chromatinand Gene Expression Section, Research Triangle Park, Durham, NC 27709,USA. 4Laboratory of Molecular Carcinogenesis, National Institute ofEnvironmental Health Sciences, National Institutes of Health, ResearchTriangle Park, Durham, NC 27709, USA. 5Cancer Research UK CambridgeResearch Institute and Department of Oncology, University of Cambridge, LiKa Shin Centre, Cambridge CB2 0RE, UK. 6Terry Fox Laboratory, BC CancerAgency, Vancouver, BC V5Z 1L3, Canada.Received: 31 October 2013 Accepted: 18 December 2014References1. Visvader JE. Keeping abreast of the mammary epithelial hierarchy andbreast tumorigenesis. Genes Dev. 2009;23:2563–77.2. Eirew P, Stingl J, Raouf A, Turashvili G, Aparicio S, Emerman JT, et al. Amethod for quantifying normal human mammary epithelial stem cells within vivo regenerative ability. Nat Med. 2008;14:1384–9.3. Lim E, Vaillant F, Wu D, Forrest NC, Pal B, Hart AH, et al. Aberrant luminalprogenitors as the candidate target population for basal tumordevelopment in BRCA1 mutation carriers. Nat Med. 2009;15:907–13.4. Smith HS, Lan S, Ceriani R, Hackett AJ, Stampfer MR. Clonal proliferation ofcultured nonmalignant and malignant human breast epithelia. Cancer Res.1981;41:4637–43.5. Kuperwasser C, Chavarria T, Wu M, Magrane G, Gray JW, Carey L, et al.Reconstruction of functionally normal and malignant human breast tissuesin mice. Proc Natl Acad Sci U S A. 2004;101:4966–71.6. Parmar H, Young P, Emerman JT, Neve RM, Dairkee S, Cunha GR. A novelmethod for growing human breast epithelium in vivo using mouse andhuman mammary fibroblasts. Endocrinology. 2002;143:4886–96.15. Borgdorff V, Lleonart ME, Bishop CL, Fessart D, Bergin AH, Overhoff MG, et al.Multiple microRNAs rescue from Ras-induced senescence by inhibitingp21Waf1/Cip1. Oncogene. 2010;29:2262–71.16. Zhang Y, Smolen P, Baxter DA, Byrne JH. The sensitivity of memoryconsolidation and reconsolidation to inhibitors of protein synthesis andkinases: computational analysis. Learn Mem. 2010;17:428–39.17. Annab LA, Terry L, Cable PL, Brady J, Stampfer MR, Barrett JC, et al.Establishment and characterization of a breast cell strain containing aBRCA1 185delAG mutation. Gynecol Oncol. 2000;77:121–8.18. Horikawa I, Cable PL, Mazur SJ, Appella E, Afshari CA, Barrett JC.Downstream E-box-mediated regulation of the human telomerase reversetranscriptase (hTERT) gene transcription: evidence for an endogenousmechanism of transcriptional repression. Mol Biol Cell. 2002;13:2585–97.19. Bayani J, Squire J. Multi-color FISH techniques. Curr Protoc Cell Biol.2004;24:Chapter 22:Unit 22.5.20. Debnath J, Muthuswamy SK, Brugge JS. Morphogenesis and oncogenesis ofMCF-10A mammary epithelial acini grown in three-dimensional basementmembrane cultures. Methods. 2003;30:256–68.21. Poon SS, Wong JT, Saunders DN, Ma QC, McKinney S, Fee J, et al. Intensitycalibration and automated cell cycle gating for high-throughput image-basedsiRNA screens of mammalian cells. Cytometry A. 2008;73:904–17.22. Shah SP, Roth A, Goya R, Oloumi A, Ha G, Zhao Y, et al. The clonal andmutational evolution spectrum of primary triple-negative breast cancers.Nature. 2012;486:395–9.23. Smoot ME, Ono K, Ruscheinski J, Wang PL, Ideker T. Cytoscape 2.8: newfeatures for data integration and network visualization. Bioinformatics.2011;27:431–2.24. Wu G, Feng X, Stein L. A human functional protein interaction network andits application to cancer data analysis. Genome Biol. 2010;11:R53.25. Curtis C, Shah SP, Chin SF, Turashvili G, Rueda OM, Dunning MJ, et al. Thegenomic and transcriptomic architecture of 2,000 breast tumours revealsnovel subgroups. Nature. 2012;486:346–52.26. Benjamini HY. Controlling the false discovery rate: a practical and powerfulapproach to multiple hypothesis testing. J Roy Stat Soc Stat Soc.1995;57:289–300.27. Harrington DP, Fleming TR. A class of rank test procedures for censoredsurvival-data. Biometrika. 1982;69:553–66.28. Canadian Epigenetics, Environment and Health Research Consortium(CEEHRC) Platform. Data downloads. http://www.epigenomes.ca/downloads.html. Accessed 13 Jan 2015.29. Canadian Epigenetics, Environment and Health Research Consortium(CEEHRC). Protocols and standards. http://www.epigenomes.ca/protocols.html. Accessed 13 January 2015.30. Roy S, Gascard P, Dumont N, Zhao J, Pan D, Petrie S, et al. Rare somatic cellsfrom human breast tissue exhibit extensive lineage plasticity. Proc Natl AcadSci U S A. 2013;110:4598–603.31. Kampstra P. Beanplot: a boxplot alternative for visual comparison ofdistributions. J Stat Softw. 2008;28:Code Snippet 1.32. Raouf A, Brown L, Vrcelj N, To K, Kwok W, Huntsman D, et al. Genomicinstability of human mammary epithelial cells overexpressing a truncatedform of EMSY. J Natl Cancer Inst. 2005;97:1302–6.33. Hammond SL, Ham RG, Stampfer MR. Serum-free growth of humanmammary epithelial cells: rapid clonal growth in defined medium andextended serial passage with pituitary extract. Proc Natl Acad Sci U S A.1984;81:5435–9.34. Lasfargues EY, Moore DH. A method for the continuous cultivation ofmammary epithelium. In Vitro. 1971;7:21–5.35. White MT, Hu AS, Hamamoto ST, Nandi S. In vitro analysis of proliferatingepithelial cell populations from the mouse mammary gland: fibroblast-freegrowth and serial passage. In Vitro. 1978;14:271–81.36. Petersen OW, Rønnov-Jessen L, Howlett AR, Bissell MJ. Interaction withbasement membrane serves to rapidly distinguish growth anddifferentiation pattern of normal and malignant human breast epithelialcells. Proc Natl Acad Sci U S A. 1992;89:9064–8.37. Ashburner M, Ball CA, Blake JA, Botstein D, Butler H, Cherry JM, et al. GeneOntology: tool for the unification of biology. Nat Genet. 2000;25:25–9.38. Paddison PJ, Silva JM, Conklin DS, Schlabach M, Li M, Aruleba S, et al. Aresource for large-scale RNA-interference-based screens in mammals.Nature. 2004;428:427–31.39. Huang DW, Sherman BT, Lempicki RA. Systematic and integrative analysis oflarge gene lists using DAVID bioinformatics resources. Nat Protoc.2008;4:44–57.40. Kanehisa M, Goto S. KEGG: Kyoto Encyclopedia of Genes and Genomes.Nucleic Acids Res. 2000;28:27–30.Nickerson JA. The ultrastructure of MCF-10A acini. J Cell Physiol.2006;208:141–8.69. Chua HL, Bhat-Nakshatri P, Clare SE, Morimiya A, Badve S, Nakshatri H. NF-κBrepresses E-cadherin expression and enhances epithelial to mesenchymaltransition of mammary epithelial cells: potential involvement of ZEB-1 andZEB-2. Oncogene. 2007;26:711–24.Burleigh et al. Breast Cancer Research  (2015) 17:4 Page 21 of 2141. Kanehisa M, Goto S, Sato Y, Furumichi M, Tanabe M. KEGG for integrationand interpretation of large-scale molecular data sets. Nucleic Acids Res.2012;40:D109–14.42. Harburg GC, Hinck L. Navigating breast cancer: axon guidance molecules asbreast cancer tumor suppressors and oncogenes. J Mammary Gland BiolNeoplasia. 2011;16:257–70.43. Yang J, Balakrishnan A, Hamamoto S, Beattie CW, Das Gupta TK,Wellings SR, et al. Different mitogenic and phenotypic responses of humanbreast epithelial cells grown in two versus three dimensions. Exp Cell Res.1986;167:563–9.44. Wang F, Weaver VM, Petersen OW, Larabell CA, Dedhar S, Briand P, et al.Reciprocal interactions between β1-integrin and epidermal growth factorreceptor in three-dimensional basement membrane breast cultures:a different perspective in epithelial biology. Proc Natl Acad Sci U S A.1998;95:14821–6.45. Koch A, Poirier F, Jacob R, Delacour D. Galectin-3, a novelcentrosome-associated protein, required for epithelial morphogenesis.Mol Biol Cell. 2010;21:219–31.46. Liu S, Umezu-Goto M, Murph M, Lu Y, Liu W, Zhang F, et al. Expression ofautotaxin and lysophosphatidic acid receptors increases mammarytumorigenesis, invasion, and metastases. Cancer Cell. 2009;15:539–50.A published erratum appears in. Cancer Cell. 2009;16:172.47. Stoker M, Perryman M, Eeles R. Clonal analysis of morphological phenotypein cultured mammary epithelial cells from human milk. Proc R Soc Lond BBiol Sci. 1982;215:231–40.48. Edwards PA, Brooks IM, Monaghan P. Expression of epithelial antigens inprimary cultures of normal human breast analysed with monoclonalantibodies. Differentiation. 1984;25:247–58.49. O’Hare MJ, Ormerod MG, Monaghan P, Lane EB, Gusterson BA.Characterization in vitro of luminal and myoepithelial cells isolated from thehuman mammary gland by cell sorting. Differentiation. 1991;46:209–21.50. Stingl J, Eaves CJ, Kuusk U, Emerman JT. Phenotypic and functionalcharacterization in vitro of a multipotent epithelial cell present in thenormal adult human breast. Differentiation. 1998;63:201–13.51. Banerji S, Cibulskis K, Rangel-Escareno C, Brown KK, Carter SL, Frederick AM,et al. Sequence analysis of mutations and translocations across breastcancer subtypes. Nature. 2012;486:405–9.52. Ellis MJ, Ding L, Shen D, Luo JQ, Suman VJ, Wallis JW, et al. Whole-genomeanalysis informs breast cancer response to aromatase inhibition. Nature.2012;486:353–60.53. 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;486:400–4.54. Koboldt DC, Fulton RS, McLellan MD, Schmidt H, Kalicki-Veizer J, McMichaelJF, et al. Comprehensive molecular portraits of human breast tumours.Nature. 2012;490:61–70.55. Dawson SJ, Rueda OM, Aparicio S, Caldas C. A new genome-drivenintegrated classification of breast cancer and its implications. EMBO J.2013;32:617–28.56. Forbes SA, Bhamra G, Bamford S, Dawson E, Kok C, Clements J, et al. TheCatalogue of Somatic Mutations in Cancer (COSMIC). Curr Protoc HumGenet. 2008;57:Chapter 10:Unit 10.11.57. Bamford S, Dawson E, Forbes S, Clements J, Pettett R, Dogan A, et al. TheCOSMIC (Catalogue of Somatic Mutations in Cancer) database and website.Br J Cancer. 2004;91:355–8.58. Whitehurst AW, Bodemann BO, Cardenas J, Ferguson D, Girard L, Peyton M,et al. Synthetic lethal screen identification of chemosensitizer loci in cancercells. Nature. 2007;446:815–9.59. Simpson KJ, Selfors LM, Bui J, Reynolds A, Leake D, Khvorova A, et al.Identification of genes that regulate epithelial cell migration using an siRNAscreening approach. Nat Cell Biol. 2008;10:1027–38.60. Bhowmick NA, Chytil A, Plieth D, Gorska AE, Dumont N, Shappell S, et al.TGF-β signaling in fibroblasts modulates the oncogenic potential ofadjacent epithelia. Science. 2004;303:848–51.61. Bhowmick NA, Neilson EG, Moses HL. Stromal fibroblasts in cancer initiationand progression. Nature. 2004;432:332–7.62. Schmeichel KL, Bissell MJ. Modeling tissue-specific signaling and organ func-tion in three dimensions. J Cell Sci. 2003;116:2377–88.63. Choi JW, Herr DR, Noguchi K, Yung YC, Lee CW, Mutoh T, et al. LPAreceptors: subtypes and biological actions. Annu Rev Pharmacol Toxicol.2010;50:157–86.Submit your next manuscript to BioMed Centraland take full advantage of: • Convenient online submission• Thorough peer review• No space constraints or color figure charges• Immediate publication on acceptance• Inclusion in PubMed, CAS, Scopus and Google Scholar• Research which is freely available for redistribution64. Beaulieu LM, Church FC. Activated protein C promotes breast cancer cellmigration through interactions with EPCR and PAR-1. Exp Cell Res.2007;313:677–87.65. Hwang-Verslues WW, Kuo WH, Chang PH, Pan CC, Wang HH, Tsai ST, et al.Multiple lineages of human breast cancer stem/progenitor cells identifiedby profiling with stem cell markers. PLoS One. 2009;4:e8377.66. Srinivasan K, Strickland P, Valdes A, Shin GC, Hinck L. Netrin-1/neogenininteraction stabilizes multipotent progenitor cap cells during mammarygland morphogenesis. Dev Cell. 2003;4:371–82.67. Strizzi L, Mancino M, Bianco C, Raafat A, Gonzales M, Booth BW, et al.Netrin-1 can affect morphogenesis and differentiation of the mousemammary gland. J Cell Physiol. 2008;216:824–34.68. Underwood JM, Imbalzano KM, Weaver VM, Fischer AH, Imbalzano AN,Submit your manuscript at www.biomedcentral.com/submit

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.52383.1-0223272/manifest

Comment

Related Items