Expression Profiling of Purified Normal Human Luminal and Myoepithelial BreastCells: Identification of Novel Prognostic Markers for Breast Cancer
Chris Jones,1 Alan Mackay,2 Anita Grigoriadis,2 Antonio Cossu,3 Jorge S. Reis-Filho,1 Laura Fulford,1,4 Tim Dexter,1
Susan Davies,2 Karen Bulmer,1 Emily Ford,1 Suzanne Parry,1 Mario Budroni,5 Giuseppe Palmieri,6
A. Munro Neville,2 Michael J. O’Hare,2 and Sunil R. Lakhani1,7
1The Breakthrough Toby Robins Breast Cancer Research Centre, Institute of Cancer Research, London, United Kingdom; 2Ludwig Institute for Cancer Research/UniversityCollege London Breast Cancer Laboratory, University College London, London, United Kingdom; 3Istituto di Anatomia Patologica, Universita di Sassari, Italy; 4Ludwig Institutefor Cancer Research, University College London, United Kingdom; 5Servizio Epidemiologia, Sassari, Italy; 6Istituto di Chimica Biomolecolare, Alghero, Italy; and 7The RoyalMarsden Hospital, London, United Kingdom
ABSTRACT
The normal duct-lobular system of the breast is lined by two epithelialcell types, inner luminal secretory cells and outer contractile myoepithelialcells. We have generated comprehensive expression profiles of the twonormal cell types, using immunomagnetic cell separation and gene expres-sion microarray analysis. The cell-type specificity was confirmed at theprotein level by immunohistochemistry in normal breast tissue. Newprognostic markers for survival were identified when the luminal- andmyoepithelial-specific molecules were evaluated on breast tumor tissuemicroarrays. Nuclear expression of luminal epithelial marker galectin 3correlated with a shorter overall survival in these patients, and theexpression of SPARC (osteonectin), a myoepithelial marker, was an inde-pendent marker of poor prognosis in breast cancers as a whole. These dataprovide a framework for the interpretation of breast cancer molecularprofiling experiments, the identification of potential new diagnostic mark-ers, and development of novel indicators of prognosis.
INTRODUCTION
The terminal duct-lobular unit of the breast, the structure fromwhich the majority of breast cancers arise, is composed of two typesof epithelial cells. The inner or luminal cells, which are potential milksecreting cells, are surrounded by an outer basal layer of contractilemyoepithelial cells. Most breast carcinomas express phenotypic mark-ers that are consistent with an origin from luminal cells (1). Thebiology of normal luminal cells is the key to understanding breastcancer initiation, with genetic alterations occurring in both normalcells and epithelial hyperplastic lesions driving the earliest stages ofprogression (2–4).
Despite the luminal origin of breast tumors, a subset of invasive ductalcarcinomas in the breast also express markers specific for myoepithelialcells (5–10). Recent studies using cDNA microarray analysis of primaryhuman breast tumors have also identified a basal-like subset of invasiveductal carcinomas (11) based on their patterns of gene expression. Thesetumors exhibiting a basal phenotype have been reported to have anaggressive phenotype and poorer prognosis for the patient (12–14). Al-
though it is interesting to speculate on the cells from which these tumorsmay be derived, there is little evidence currently to suggest a myoepi-thelial origin for basal-like breast cancers.
Although tumor expression profiling data have begun to unravel thecomplexity of breast cancer, corresponding transcriptional studies ofthe two normal epithelial cell types have lagged behind. Immunomag-netic methods have been developed for large-scale purification ofnormal human luminal and myoepithelial breast cells from reductionmammoplasty samples, amenable to detailed molecular analysis (15).We report here the cDNA microarray analysis of separated normalluminal and myoepithelial cells. The objectives of this study were toprovide a baseline reference dataset to help understand preexistingand forthcoming tumor expression profiles, to determine whethernovel cell-type specific markers can be used for tumor subclassifica-tion and for differential diagnosis, and to identify new predictive andprognostic markers that could include potential targets for futuretherapy. Comparison of the transcriptional profiles of normal adulthuman luminal and myoepithelial cells does identify novel markers,including some which provide significant prognostic information forprimary breast cancers.
MATERIALS AND METHODS
Cell Preparations. Purified populations of �107 normal human breastluminal and myoepithelial cells were prepared from individual reductionmammoplasty samples (16) with modifications to enhance purity (15). Briefly,the different breast cells were immunomagnetically sorted from primary cul-tures using combined positive magnetic activated cell sorting selection usingantibodies against the luminal epithelial membrane marker EMA (rat mono-clonal ICR2) and the myoepithelial membrane antigen CD10 (mouse mono-clonal DAKO-CALLA clone SS2/36) followed by negative Dynabead selec-tion using mouse monoclonal antibodies against a different myoepithelialcell-surface antigen (Santa Cruz Biotechnology anti-�-4-integrin clone A9)and another luminal antigen (Dako BerEp-4 Epithelial Antigen).
cDNA Microarray Hybridizations. The cDNA microarrays used in thisstudy were constructed at the Sanger Centre as part of the Ludwig Institute forCancer Research/Cancer Research UK Microarray Consortium, containing9,930 sequence-validated cDNA clones representing �6000 unique humangene sequences (see web site for details and protocols).8 Clone annotation wasbased on the National Center for Biotechnology Information 34 assembly ofthe human genome with the Sanger Clone IDs mapping to Ensembl.9
RNA was prepared according to standard protocols (17), and preparationsfrom nine luminal and nine myoepithelial samples (four of which were pairedsamples from the same patient) were individually hybridized against a com-mon breast reference RNA in duplicated dye-swap experiments. The breastreference RNA was created by combining equal quantities of total RNA fromthe following breast cell lines, MDA-MB-361, MDA-MB-231, MDA-MB-435, BT20, HBL100, GI101, BT474, T47D, MCF7, SKBR3, ZR-75-1, andMDA-MB-468.
Image Processing and Data Analysis. Fluorescent images of hybridizedmicroarrays were captured using either the GenePix 4000 (Axon) dual color
Received 7/14/03; revised 1/26/04; accepted 2/20/04.Grant support: The microarray consortium is funded by the Wellcome Trust, Cancer
Research UK and the Ludwig Institute of Cancer Research. J. Reis-Filho is the recipientof the Gordon Signy International Fellowship Award and is partially supported by Ph.D.Grant SFRH/BD/5386/2001 from the Fundacao para a Ciencia e a Tecnologia, Portugal,and Programa Operacional Ciencia, Tecnologia e Inovacao POCTI/CBO/45157/2002.A. Cossu, M. Budroni, and G. Palmieri are partially funded by Regione Autonoma dellaSardegna.
The costs of publication of this article were defrayed in part by the payment of pagecharges. This article must therefore be hereby marked advertisement in accordance with18 U.S.C. Section 1734 solely to indicate this fact.
Note: C. Jones and A. Mackay contributed equally to this work. The humanI.M.A.G.E. cDNA clone collection was obtained from the Medical Research CouncilHuman Genome Mapping Project Resource Centre (Hinxton, United Kingdom). AllcDNA clone resequencing was performed by Team 56 at the Sanger Institute.
Requests for reprints: Sunil R. Lakhani, The Breakthrough Toby Robins BreastCancer Research Centre Institute of Cancer Research, Fulham Road, London SW3 6JB,United Kingdom. Phone: 020-7153-5525; Fax: 020-7153-5533; E-mail: [email protected].
8 Internet address: http://www.sanger.ac.uk/Projects/Microarrays/.9 Internet address: http://www.ensembl.org.
confocal laser scanner and GenePix software or a GSI Lumonics 4000 scannerand ScanArray software and quantitated and background subtracted using GSILumonics Quantarray 3.0 software. The log expression ratios were normalizedusing lowess local regression (18) using the statistical platform S-Plus version6.1 for Windows (Insightful). All raw fluorescence intensity data and microar-
ray image files have been deposited within the public repository for microarraybased gene expression data ArrayExpress,10 complying with minimum infor-
10 Internet address: http://www.ebi.ac.uk/arrayexpress.
Fig. 1. Unsupervised hierarchical clustering ofluminal and myoepithelial preparations. A, fullheatplot of 1896 gene list, ordered by clustering ofthe samples separating into two cell-type specificarms. The scale on the right of the dendrogramshows 1 minus correlation. B, magnified viewshowing a representative luminal gene cluster. C,magnified view showing a representative myoepi-thelial gene cluster.
Table 1 List of top 50 luminal-specific and top 50 myoepithelial specific genes as determined by Statistical Analysis of Microarrays analysis
Genes are ranked in order of fold change (myoepithelial over luminal) and are listed with their Sanger Institute Hver1.2.1 clone ID8 and Ensembl accession number.9 Geneshighlighted in italics formed part of the discriminator genes identified by Prediction Analysis of Microarrays.
Sanger ID Ensembl Gene ID Statistical Analysis of Microarray score Fold difference
mation about a microarray experiment (MIAME) standards (19), with theaccession number E-MEXP-36.
Genes were filtered from the total set of 9930 by exclusion because of lowmean intensity values (�20th percentile of highest intensity across all arrays),consistent local artifact, and low mean absolute deviation (�0.3). This resultedin a filtered gene list of 1896 targets for unsupervised and supervised analysis.Any remaining missing values were imputed using the k-mean nearest neigh-bor method in Statistical Analysis of Microarrays (SAM).
Unsupervised hierarchical clustering was carried out using “hclust” inS-Plus, as well as the Cluster package, and plotted with Treeview.11 Differ-entially expressed genes were identified by application of the SAM (version1.12) Excel add-in.12 Supervised analysis was carried by the nearest shrunkencentroid classification for class prediction using the Prediction Analysis ofMicroarrays package,13 implemented in R (1.6.2).14
Reverse Transcription-PCR. Ten �g total RNA was reverse transcribedfrom an oligo-dT primer under conventional conditions (Superscript II; LifeTechnologies, Inc.). The resulting reaction was diluted 10-fold in water, and 2�l were used as a template for PCR amplification. PCR was performed understandard conditions in 50 �l for 25–40 cycles (primer sequences and cyclenumbers are given in Supplementary Table S4). Products were resolved bystandard agarose gel electrophoresis. Differential expression was confirmed bydensitometry of ethidium bromide staining on conventional agarose gels usingNIH Image software.
Immunohistochemistry. Antibodies to differentially expressed genes wereobtained commercially where available. Sections were dewaxed in xyleneovernight, taken to ethanol (99.7–100% v/v), and blocked for endogenousperoxidase in methanol for 10 min. Sections were subjected to specific hightemperature antigen retrieval techniques, blocked in normal horse serum(2.5%; Vector Labs) for 20 min, and primary antibodies applied for 30 min.SPARC was subjected to 2 min of pressure cooking in citrate buffer (pH 6.0),1/5 dilution (Novocastra); S100A2 received 2 min of pressure cooking, dilu-tion 1/100 (Dako); maspin received 18 min of microwaving in Dako TargetRetrieval Solution (pH 6.0), dilution 1/100 (Novocastra); galectin 3 (LGALS3)received 2 min of pressure cooking, dilution 1/750 (Novocastra); CLDN4received 18 min of microwaving, dilution 1/100 (Zymed); CD24 received 3min of pressure cooking, dilution 1/100, (Serotech). All antibodies werediluted in Tris-buffered saline. The primary antibodies were rinsed off in 0.1%Tween 20 in Tris-buffered saline, developed using Vectastain Universal ABCkit (Vector Labs) and visualized with 3,3�-diaminobenzidine (Dako).
Tissue Microarrays. Breast tumors were selected from the archives of theIstituto di Anatomia Patologica (Sassari, Italy), with appropriate local ethicalcommittee approval. A total of 566 unselected primary breast cancers com-prising all grades and types was retrieved and reviewed by an experienced
pathologist. Up to 10 years clinical follow-up data were available for all cases(3–110 months, mean � 62 months, median � 73 months). The paraffin blockswere marked and punched with 0.6-mm2 tumor cores taken from the donorblocks for inclusion in duplicate recipient tissue array blocks using a precisiontissue array instrument (Beecher Instruments; Ref. 20).
Survival analysis was carried out using the statistical platform S-Plusversion 6.1 for Windows (Insightful) on our right-censored clinical follow-updata from the cases on the tissue microarray, using the log-rank test and theCox proportional hazards model.
RESULTS
cDNA Microarray Analysis. The microarray data from normalbreast luminal and myoepithelial cells were firstly analyzed in anunsupervised manner to determine the innate differences between thecell preparations and secondly using supervised algorithms to identifythe most discriminatory genes associated with each cell types. Usingthe normalized data from the 1896 gene list, unsupervised hierarchicalclustering on the normal breast luminal and myoepithelial prepara-tions was carried out (Fig. 1A). The sample dendrogram clearlyseparates two main branches each consisting of one of the two celltypes, exemplifying the inherent differences between the two epithe-lial cell types of the breast and the consistency of the cell separationand microarray analysis methods used. Clustering of the 1896 gene setalso identifies luminal and myoepithelial-specific gene clusters, rep-resentative regions of which are shown (Fig. 1, B and C).
A list of statistically significant genes which were differentiallyexpressed between the two cell types were identified using SAM.With a false discovery rate of 1%, 132 myoepithelial and 77 luminaldifferential genes were found (Supplementary Table S1). Expressionratios and SAM scores for the top 50 most differentially expressedgenes in each cell type are shown (Table 1). The data were alsoanalyzed by a supervised classification method to identify those geneswhich are the most predictive of each cell type, using a class predic-tion algorithm based on the nearest shrunken centroid method (21).Using Prediction Analysis of Microarrays, a classifier was first trainedusing the 1896 gene set, before cross-validation and plotting of thecross-validated error curves, to determine the threshold (amount ofshrinkage), which gives the minimum cross-validated error rate (Sup-plementary Figure S2). Applying a threshold of 2.7 gives a misclas-sification rate of 0 using 42 cDNA clones corresponding to 33 uniquegenes (Fig. 2A, Supplementary Figure S3). The cross-validated classprobabilities by sample type (Fig. 2B) demonstrate that these 33 genes
11 Internet address: http://rana.lbl.gov/EisenSoftware.htm.12 Internet address: http://www-stat.stanford.edu/�tibs/SAM/.13 Internet address: http://www-stat.stanford.edu/�tibs/PAM/.14 Internet address: http://www.r-project.org.
Table 1 Continued.
Sanger ID Ensembl Gene ID Statistical Analysis of Microarray score Fold difference
accurately classify all of the samples into their correct classes (celltype).
Confirmation of Differential Expression. To confirm the specificidentity and differential expression of our cell type specific markers,semiquantitative reverse transcription-PCR was carried out on thefour patient-matched luminal and myoepithelial samples used in ourmicroarray analysis. Differential expression was confirmed for 56 of62 (90%) genes by reverse transcription-PCR (primers, cycle num-bers, and confirmatory data for the 66 unique genes from Table 1 aregiven in Supplementary Table S4).
These genes were next examined at the protein level in paraffin-
embedded archival samples by immunohistochemistry, where theavailability of appropriate antibodies made this possible. Differentialluminal expression in normal breast lobules is shown for claudin 4(CLDN4), CD24, and LGALS3 proteins. CLDN4 staining shows astrong membrane component of luminal epithelial cells consistentwith its role in tight junction adhesion and does not stain the basementmembrane of these cells (Fig. 3A). CD24 stains the cytoplasmiccompartment of normal luminal epithelial cells as well as the apicalcell surface (Fig. 3B). LGALS3 stains the nucleus and cytoplasm ofluminal cells differentially compared with myoepithelial cells and alsostains intralobular fibroblasts in the breast (Fig. 3C).
Fig. 2. Supervised analysis using predictionanalysis of microarrays (PAM). A, centroid plotshowing a ranked list of the 42 most predictiveclones, corresponding to 33 individual genes. Thelength of the horizontal bar for a given gene isequivalent to the difference between the overallcentroid and the class-specific centroid. B, cross-validated probability plot with a threshold of 2.7,showing 0 misclassification error with the expres-sion profiles of the 42 predictor clones.
Differential myoepithelial expression in normal breast lobules isdemonstrated for S100A2, SERPINB5, and SPARC proteins. S100A2shows a strong nuclear and cytoplasmic staining specifically in themyoepithelial cells, with no expression in the stromal cells (Fig. 3E).Maspin (SERPINB5) expression is also restricted to myoepithelialcells, with strong nuclear and cytoplasmic staining (Fig. 3F). Os-teonectin (SPARC) stains the cytoplasmic compartment of myoepi-
thelial cells differentially to luminal cells and also stains inter- andintralobular fibroblasts (Fig. 3G).
Evaluation of Prognostic Significance Using Tissue Microar-rays. To evaluate whether the expression of the luminal and myoep-ithelial markers demonstrated any correlation with prognosis in breastcancer, immunohistochemistry was carried out on a tissue microarrayconsisting of 566 primary breast tumors of all types and grades for
Fig. 3. Immunohistochemistry of antibodiesraised against luminal and myoepithelial specificproteins. A, claudin 4, normal breast lobule (�40)showing luminal membrane staining. B, CD24, nor-mal breast lobule (�20) showing luminal cytoplas-mic and apical cell surface staining. C, LGALS3,normal breast lobule (�20) showing nuclear andcytoplasmic staining in luminal epithelial cells andalso intralobular fibroblasts. D, LGALS3, invasiveductal carcinoma on tissue microarray (�10) show-ing nuclear positivity. E, S100A2, normal breastlobule (�20) showing myoepithelial nuclear andcytoplasmic staining. F, maspin, normal breastlobule (�20) showing myoepithelial nuclear andcytoplasmic staining. G, SPARC, normal breastlobule (�20) showing myoepithelial cytoplasmicstaining as well as positivity in inter- and intralob-ular fibroblasts. H, SPARC, invasive ductal carci-noma on tissue microarray (�10) showing positivecytoplasmic staining in the tumor sample, as wellas a positive stromal reaction.
which outcome data in the form of overall survival was available. Asummary of the results of univariate analysis is given in Table 2.
The luminal epithelial maker LGALS3 showed a loss of expressionin approximately one-half of all assessable tumors on the tissuemicroarray, with 213 of 431 cases (49.4%) LGALS3 positive. Thisloss of expression did not correlate with prognostic outcome in alltumors (P � 0.597); however, when the subcellular localization ofLGALS3 was evaluated, tumors with nuclear positivity (9 of 431,2.1%, Fig. 3D) showed a statistically significant (P � 0.00895,log-rank test) poorer overall survival than negative cases or those forwhom expression was restricted to the cytoplasm (Fig. 4A). All ofthese nuclear positive cases were also positive for cyclin D1 (data notshown). There was no correlation between LGALS3 expression andage, grade, size, ER, progesterone receptor, or tumor type (data notshown). In multivariate analysis, loss of LGALS3 expression justfailed to reach formal statistical significance as an independent prog-nostic factor (P � 0.051).
Loss of expression from normal luminal epithelial cells to invasivecancer was also seen for other luminal markers tested. CLDN4 waspositive in 245 of 331 tumors (74.0%) and CD24 positive in 126 of426 tumors (29.6%). Although loss of expression shows a clearassociation with breast cancer development, neither marker conferredany independent prognostic information, nor were they correlatedwith age, grade, size, ER, progesterone receptor, or tumor type (datanot shown).
The myoepithelial marker osteonectin (SPARC) was found to be
positive in 17 of 350 (4.9%) assessable tumor cores (Fig. 3H). WhenKaplan-Meier survival curves for overall survival were plotted, a clearpoor prognosis was observed for SPARC-positive tumors. This wasfound to be statistically significant by the log-rank test in all tumors(P � 0.00844, Fig. 4B). By multivariate analysis (Table 3), SPARCwas found to be an independent prognostic factor (P � 0.0057, Coxproportional hazards), conferring the highest relative risk of all factorsfitted (6.88, 95% confidence interval 1.75–27.04), although the con-fidence interval is large, given the small number of positive tumors.
Other myoepithelial markers tested showed a proportion of breasttumors expressing these basal proteins. S100A2 was positive in 8 of443 cases (1.8%), whereas maspin (SERPINB5) was positive in 108of 333 cases (32.4%). S100A2 conferred no independent prognosticinformation, and whereas maspin expression appeared to indicate abetter overall survival, this did not reach statistical significance(P � 0.092). No association with clinicopathological variables orsurvival was observed with maspin expression or its subcellularlocalization.
DISCUSSION
Expression profiling of purified normal luminal epithelial and myo-epithelial cells in the breast provides a basis for interpretation of thelarge amount of microarray data currently being generated for breasttumors. Identification of subclassifications of breast tumors termedbasal-like and luminal-like (11), which differ in their clinical outcome(13) clearly demonstrates the need for accurate determination of thepatterns of gene expression in these normal cells of the breast. Ourobservations of myoepithelial-specific genes such as S100A2,LGALS7, CSTA, and BPAG1 in our cell preparations, which alsocluster together to define the basal-like group in these tumor studies(22), demonstrate the use of such an approach and will help toaccurately classify the proposed breast cancer stratifications. Ournormal luminal cell preparations are hormone receptor negative, incommon with the vast majority of normal luminal epithelial cells inthe breast. It is therefore not surprising that there is little overlapbetween our luminal epithelial profiles and those of the luminal-liketumors in these classifications (11, 13), which are almost exclusivelyER positive and the genes associated with them ER-responsive genes.The cell-type specific expression profiles in the normal breast alsoprovides a baseline for studies investigating breast cancer progression(23), outcome prediction (24, 25), and local (26) and distant metas-tasis (27).
Our data also help to clarify previous transcriptional studies usinghuman mammary epithelial cells (HMECs) as the normal component.Such cells were derived from cultures of unsorted normal breastepithelium. As it is the myoepithelially derived cells that have thegreatest proliferative potential in vitro (28), HMECs are essentiallymyoepithelial-like (i.e., basal) in phenotype (29). Consequently, whenHMECs are compared with (luminally derived) breast cancer cells orsolid tumors, the markers that emerge as differentially expressed areessentially those represented in our luminal versus myoepithelial lists,rather than specific tumor markers per se, as has been inferred (30).
Generation of a larger and more accurate panel of markers fordifferentiating the normal epithelial cell types will have a majorimpact in patient management. Routine histopathological discrimina-tion of in situ from invasive cancer in the breast uses the retention ofthe myoepithelial layer as a critical diagnostic criterion, with hugeimplications for planning appropriate surgery. Improving the differ-ential diagnosis from, for example, small needle core biopsies usingnovel myoepithelial-specific markers will make an important contri-bution to clinical practice.
Genes differentially expressed between luminal and myoepithelial
Table 2 Univariate analysis of clinicopathological and immunohistochemical data onthe 566 tumor tissue microarray
cells were also found to confer independent prognostic informationfor breast cancer patients. Galectin 3 is a gene expressed in normalluminal epithelial but not myoepithelial cells. Galectin 3 is thought toregulate many biological processes and has been associated withERBB2 expression (31). Down-regulation of galectin 3 has beenimplicated in breast cancer progression (32). Tissue microarray anal-ysis showed that loss of galectin 3 expression by malignant epithelialcells was seen in approximately one-half of all tumors assessed. Ofparticular interest is the correlation of nuclear LGALS3 expression
and poor outcome. Nuclear galectin 3 has been reported to have agrowth promoting activity through cyclin D1 induction (33), and all ofthese (9 of 431) tumors were also positive for cyclin D1. Thisobservation demonstrates the ability of our approach to identify lu-minal epithelial-specific markers in the normal breast and use them tomonitor breast cancer progression and establish their relationship withpatient prognosis.
SPARC (osteonectin) modulates cellular interaction with the extra-cellular matrix by its binding to structural matrix proteins such ascollagen and vitronectin. It was found by our cDNA microarrayanalysis to be differentially expressed between myoepithelial andluminal cells. Up-regulation of SPARC has been associated withincreased invasive potential in breast cancer cells in vitro (34) and hasbeen identified as a breast tumor marker by serial analysis of geneexpression analysis (35). SPARC was found to be positive in malig-nant epithelial cells in 4.9% of all breast tumors examined in ourtissue microarray study. SPARC-positive tumors showed a highlysignificant poorer overall survival in breast cancer patients by univa-riate analysis (P � 0.00844) and was an independent prognosticindicator by multivariate analysis (P � 0.0057). Aberrant expression
Fig. 4. Kaplan-Meier survival curves (months) from tissuemicroarray analysis. A, LGALS3 nuclear expression conferring apoor prognosis in all tumors compared with negative and cyto-plasmic positive patients (P � 0.00895). B, SPARC positivity isassociated with significantly shorter overall survival in all tumors(P � 0.00844).
Table 3 Multivariate analysis of the tissue microarray cohort using the Coxproportional hazards model
Only those statistically significant independent prognostic factors as determined by themodel are shown.
of myoepithelial proteins has been recognized in breast tumors forsome time, and data demonstrating poor prognosis of these tumorshave largely been associated with ER- and lymph node-negativetumors (13, 14). Here, we present the identification of a proteinexpressed in the myoepithelial cells but not the luminal cells of thenormal breast and whose expression in breast tumors confers a verypoor clinical outcome, regardless of ER or lymph node status.
The clinical use of expanding the list of normal cell-type specificgenes not only provides novel diagnostic and prognostic markers butwill assist in understanding the multistep progression from epithelialcells to invasive cancer in the breast. Recent studies of breast cancerstem cell phenotypes (36) have identified CD24-/CD44� cells ashaving tumorigenic potential, these markers being associated respec-tively with normal luminal and normal myoepithelial cells. At theother end of the progression spectrum, among 17 reporter genesassociated with metastasis (27) were up-regulation of COL1A1 anddown-regulation of MYLK, two of the most discriminant myoepithe-lial genes in the present study. Collectively, these observations high-light the importance of the normal breast cell expression profiles inunderstanding breast cancer. They also provide a unique dataset inwhich targets for future therapeutic intervention may be identified.
ACKNOWLEDGMENTS
We thank the staff of the Sanger Institute Microarray Facility for the supplyof arrays, lab protocols, and technical advice (David Vetrie, Cordelia Lang-ford, Adam Whittaker, Neil Sutton) and the staff of the Quantarray/Gene-Spring data files and all data analysis and databases relating to elements on thearrays (Kate Rice, Rob Andrews, Adam Butler, Harish Chudasama). We alsothank Professor Alan Ashworth for continued support and helpful discussionspertaining to this project.
REFERENCES
1. Taylor-Papadimitriou J, Lane EB. Keratin expression in the mammary gland. In:Neville MC, Daniel CW, editors. The mammary gland, development, regulation andfunction. New York: Plenum Press; 1987. p. 181–215.
2. Deng G, Lu Y, Zlotnikov G, Thor AD, Smith HS. Loss of heterozygosity in normaltissue adjacent to breast carcinomas Science (Wash. DC) 1996;274(5295):2057–9.
3. Lakhani SR, Chaggar R, Davies S., et al. Genetic alterations in ’normal’ luminal andmyoepithelial cells of the breast. J Pathol 1999;189(4):496–503.
4. Forsti A, Louhelainen J, Soderberg M, Wijkstrom H, Hemminki K. Loss of heterozy-gosity in tumour-adjacent normal tissue of breast and bladder cancer. Eur J Cancer2001;37(11):1372–80.
5. Gusterson BA, Warburton MJ, Mitchell D, Ellison M, Neville AM, Rudland PS.Distribution of myoepithelial cells and basement membrane proteins in the normalbreast and in benign and malignant breast diseases. Cancer Res 1982;42(11):4763–70.
6. Dairkee SH, Puett L, Hackett AJ. Expression of basal and luminal epithelium-specifickeratins in normal, benign, and malignant breast tissue. J Natl Cancer Inst (Bethesda)1988;80(9):691–5.
7. Gould VE, Koukoulis GK, Jansson DS, Nagle RB, Franke WW, Moll R. Coexpres-sion patterns of vimentin and glial filament protein with cytokeratins in the normal,hyperplastic, and neoplastic breast. Am J Pathol 1990;137(5):1143–55.
8. Wetzels RH, Kuijpers HJ, Lane EB, et al. Basal cell-specific and hyperproliferation-related keratins in human breast cancer. Am J Pathol 1991;138(3):751–63.
9. Tsuda H, Takarabe T, Hasegawa T, Murata T, Hirohashi S. Myoepithelial differen-tiation in high-grade invasive ductal carcinomas with large central acellular zones.Hum Pathol 1999;30(10):1134–9.
10. Jones C, Nonni AV, Fulford L, et al. CGH analysis of ductal carcinoma of the breastwith basaloid/myoepithelial cell differentiation. Br J Cancer 2001;85(3):422–7.
11. Perou CM, Sorlie T, Eisen MB, et al. Molecular portraits of human breast tumours.Nature (Lond.) 2000;406(6797):747–52.
12. Tsuda H, Takarabe T, Hasegawa F, Fukutomi T, Hirohashi S. Large, central acellularzones indicating myoepithelial tumor differentiation in high-grade invasive ductalcarcinomas as markers of predisposition to lung and brain metastases. Am J SurgPathol 2000;24(2):197–202.
13. Sorlie T, Perou CM, Tibshirani R, et al. Gene expression patterns of breast carcino-mas distinguish tumor subclasses with clinical implications. Proc Natl Acad Sci USA2001;98(19):10869–74.
14. van de Rijn M, Perou CM, Tibshirani R, et al. Expression of cytokeratins 17 and 5identifies a group of breast carcinomas with poor clinical outcome. Am J Pathol2002;161(6):1991–6.
15. Page MJ, Amess B, Townsend RR, et al. Proteomic definition of normal humanluminal and myoepithelial breast cells purified from reduction mammoplasties. ProcNatl Acad Sci USA 1999;96(22):12589–94.
16. Clarke C, Titley J, Davies S, O’Hare MJ. An immunomagnetic separation methodusing superparamagnetic (MACS) beads for large-scale purification of human mam-mary luminal and myoepithelial cells. Epithelial Cell Biol 1994;3(1):38–46.
17. Chomczynski P, Sacchi N. Single-step method of RNA isolation by acid guanidiniumthiocyanate-phenol-chloroform extraction. Anal Biochem 1987;162:156–9.
18. Yang YH, Dudoit S, Luu P, et al. Normalization for cDNA microarray data: a robustcomposite method addressing single and multiple slide systematic variation. NucleicAcids Res 2002;30(4):e15.
19. Brazma A, Hingamp P, Quackenbush J, et al. Minimum information about a microar-ray experiment (MIAME)-toward standards for microarray data. Nat Genet 2001;29(4):365–71.
20. Kononen J, Bubendorf L, Kallioniemi A, et al. Tissue microarrays for high-through-put molecular profiling of tumor specimens. Nat Med 1998;4(7):844–7.
21. Tibshirani R, Hastie T, Narasimhan B, Chu G. Diagnosis of multiple cancer typesby shrunken centroids of gene expression. Proc Natl Acad Sci USA 2002;99(10):6567–72.
22. Chung CH, Bernard PS, Perou CM. Molecular portraits and the family tree of cancer.Nat Genet 2002;32 Suppl:533–40.
23. Ma XJ, Salunga R, Tuggle JT, et al. Gene expression profiles of human breast cancerprogression. Proc Natl Acad Sci USA 2003;100(10):5974–9.
24. van ’t Veer LJ, Dai H, van de Vijver MJ, et al. Gene expression profiling predictsclinical outcome of breast cancer. Nature (Lond.) 2002;415(6871):530–6.
25. van de Vijver MJ, He YD, van’t Veer LJ, et al. A gene-expression signature as apredictor of survival in breast cancer. N Engl J Med 2002;347(25):1999–2009.
26. Huang E, Cheng SH, Dressman H, et al. Gene expression predictors of breast canceroutcomes. Lancet 2003;361(9369):1590–6.
27. Ramaswamy S, Ross KN, Lander ES, Golub TR. A molecular signature of metastasisin primary solid tumors. Nat Genet 2003;33(1):49–54.
28. O’Hare MJ, Ormerod MG, Monaghan P, Lane EB, Gusterson BA. Characterization invitro of luminal and myoepithelial cells isolated from the human mammary gland bycell sorting. Differentiation 1991;46(3):209–21.
29. DiRenzo J, Signoretti S, Nakamura N, et al. Growth factor requirements and basalphenotype of an immortalized mammary epithelial cell line. Cancer Res 2002;62(1):89–98.
30. Nacht M, Ferguson AT, Zhang W, et al. Combining serial analysis of gene expressionand array technologies to identify genes differentially expressed in breast cancer.Cancer Res 1999;59(21):5464–70.
31. Mackay A, Jones C, Dexter T, et al. cDNA microarray analysis of genes associatedwith ERBB2 (HER2/neu) overexpression in human mammary luminal epithelial cells.Oncogene 2003;22(17):2680–8.
32. Castronovo V, Van Den Brule FA, Jackers P, et al. Decreased expression of galectin-3is associated with progression of human breast cancer. J Pathol 1996;179(1):43–8.
33. Lin H-M, Pestell RG, Raz A, Kim H-RC. Galectin-3 enhances cyclin D1 promoteractivity through SP1 and a cAMP-responsive element in human breast epithelial cells.Oncogene 2002;21:8001–10.
34. Briggs J, Chamboredon S, Castellazzi M, Kerry JA, Bos TJ. Transcriptional up-regulation of SPARC, in response to c-Jun overexpression, contributes to increasedmotility and invasion of MCF7 breast cancer cells. Oncogene 2002;21(46):7077–91.
35. Porter DA, Krop IE, Nasser S, et al. A SAGE (serial analysis of gene expression) viewof breast tumor progression Cancer Res 2001;61(15):5697–702.
36. Al-Hajj M, Wicha MS, Benito-Hernandez A, Morrison SJ, Clarke MF. Prospectiveidentification of tumorigenic breast cancer cells. Proc Natl Acad Sci USA 2003;100(7):3983–8.
2004;64:3037-3045. Cancer Res Chris Jones, Alan Mackay, Anita Grigoriadis, et al. Markers for Breast CancerMyoepithelial Breast Cells: Identification of Novel Prognostic Expression Profiling of Purified Normal Human Luminal and
