Estrogen-Regulated Genes Predict Survival in …...Estrogen-Regulated Genes Predict Survival in...

Estrogen-Regulated Genes Predict Survival in HormoneReceptor–Positive Breast CancersDaniel S. Oh, Melissa A. Troester, Jerry Usary, Zhiyuan Hu, Xiaping He, Cheng Fan, Junyuan Wu,Lisa A. Carey, and Charles M. Perou

A B S T R A C T

PurposeThe prognosis of a patient with estrogen receptor (ER) and/or progesterone receptor (PR) –positivebreast cancer can be highly variable. Therefore, we developed a gene expression–based outcomepredictor for ER� and/or PR� (ie, luminal) breast cancer patients using biologic differences amongthese tumors.

Materials and MethodsThe ER� MCF-7 breast cancer cell line was treated with 17�-estradiol to identify estrogen-regulated genes. These genes were used to develop an outcome predictor on a training set of 65luminal epithelial primary breast carcinomas. The outcome predictor was then validated on threeindependent published data sets.

ResultsThe estrogen-induced gene set identified in MCF-7 cells was used to hierarchically cluster a 65tumor training set into two groups, which showed significant differences in survival (P � .0004).Supervised analyses identified 822 genes that optimally defined these two groups, with thepoor-prognosis group IIE showing high expression of cell proliferation and antiapoptosis genes.The good prognosis group IE showed high expression of estrogen- and GATA3-regulated genes.Mean expression profiles (ie, centroids) created for each group were applied to ER� and/or PR�tumors from three published data sets. For all data sets, Kaplan-Meier survival analyses showedsignificant differences in relapse-free and overall survival between group IE and IIE tumors.Multivariate Cox analysis of the largest test data set showed that this predictor added significantprognostic information independent of standard clinical predictors and other gene expression–based predictors.

ConclusionThis study provides new biologic information concerning differences within hormone receptor–positivebreast cancers and a means of predicting long-term outcomes in tamoxifen-treated patients.

J Clin Oncol 24:1656-1664. © 2006 by American Society of Clinical Oncology

INTRODUCTION

Breast cancers are traditionally stratified into hor-mone receptor–positive and –negative groups to guidepatient management. This is because almost all hor-mone antagonist (ie, tamoxifen) -responsive breastcancers are estrogen receptor (ER) and/or progester-one receptor (PR) -positive.1 However, a subgroup ofthese patients experiences disease relapse irrespectiveof standard therapy with tamoxifen.2-4 In manycases, the resistance mechanisms are unknown.5-7 Amethod for identifying those ER� and/or PR� pa-tients who do not experience relapse in the presenceof tamoxifen versus those who do would be of sig-nificant value.

Gene expression profiling is a powerful methodfor breast cancer prognostication. With gene expres-

sion profiling, breast tumors can be classified intofive molecular subtypes (basal-like, HER2�/ER-,normal breast–like, and luminal A and B) with sig-nificant differences in patient outcome.8,9 The twoluminal subtypes are composed almost entirely ofER� and/or PR� tumors and are defined by thehigh expression of a gene set (luminal epithelialER� set) that includes ER and GATA3. Comparedwith luminal A tumors, luminal B tumors have pooroutcomes despite being clinically ER�.8,9

To date, several laboratories have developedgene expression–based predictors for ER� and/orPR� patients. All used supervised analyses basedon patient outcomes/tumor response.10-13 Here, wedeveloped a gene expression–based predictor usingan approach based solely on biologic characteris-tics of the tumors. Because of the importance of

From the Departments of Genetics,Medicine, and the Pathology & Labo-ratory Medicine, Lineberger Compre-hensive Cancer Center, University ofNorth Carolina, Chapel Hill, NC.

Submitted June 27, 2005; acceptedDecember 14, 2005.

Supported by funds from the NationalCancer Institute (Bethesda, MD) BreastSPORE program to the University ofNorth Carolina–Chapel Hill (Chapel Hill,NC; Grant No. P50-CA58223-09A1),National Institutes of Health Grant No.M01RR00046, and the National Insti-tute of Environmental Health SciencesGrant No. U19-ES11391-03; and byUniversity of North Carolina LinebergerCancer Control Education ProgramGrant No. R25 CA57726 (M.A.T.).

Terms in blue are defined in the glossary,found at the end of this article and onlineat www.jco.org.

Authors’ disclosures of potential con-flicts of interest and author contribu-tions are found at the end of thisarticle.

Address reprint requests to Charles M.Perou, PhD, Lineberger ComprehensiveCancer Center, University of NorthCarolina at Chapel Hill, Campus Box7295, Chapel Hill, NC 27599; e-mail:[email protected].

© 2006 by American Society of ClinicalOncology

0732-183X/06/2411-1656/$20.00

DOI: 10.1200/JCO.2005.03.2755

JOURNAL OF CLINICAL ONCOLOGY O R I G I N A L R E P O R T

VOLUME 24 � NUMBER 11 � APRIL 10 2006

1656

Copyright © 2006 by the American Society of Clinical Oncology. All rights reserved. Downloaded from www.jco.org on April 8, 2006 . For personal use only. No other uses without permission.

estrogen signaling in breast epithelial cell biology, we hypothesizedthat differential expression of estrogen-regulated genes would beuseful in predicting outcome.

MATERIALS AND METHODS

Cell Culture and Collection of mRNA

MCF-7 cells were maintained as described previously.14 Cells were platedin 150-mm dishes and grown until 50% confluence. Media was changed andcells maintained for 48 hours in estrogen-free medium (phenol red-freeRPMI-1640 with 10% charcoal-dextran-stripped fetal bovine serum) beforetreating for 2,4,8, or 24 hours with 10�6 M 17�-estradiol (Sigma-AldrichInc, St Louis, MO). Cells were harvested, and mRNA isolated using aMicro-FastTrack kit (Invitrogen Corp, Carlsbad, CA). A reference mRNAsample was collected from cells maintained for 48 hours in estrogen-freemedium (ie, estrogen-starved cells).

Microarray Experiments

Human whole-genome microarrays (Agilent Technologies Inc, PaloAlto, CA) were hybridized according to manufacturer’s protocol withCy3-CTP–labeled cRNA from estrogen-starved cells (2 �g/sample) andCy5-CTP–labeled cRNA from 17�-estradiol-treated cells (2 �g/sample),with dye-flip replicates for each time point. Microarrays were scannedand image files analyzed as described previously.15 All primary microarraydata are available via the University of North Carolina (Chapel Hill, NC;UNC-CH) Microarray Database (https://genome.unc.edu/) and the GeneExpression Omnibus (GEO; National Center for Biotechnology Informa-tion [NCBI], http://www.ncbi.nlm.nih.gov/geo/) with series numberGSE2740 (GSM52882-GSM52909, GSM34423-GSM34568).

Analysis of Microarray Data to Identify GATA3 and

Estrogen-Regulated Genes

Data from microarray experiments were calculated as described.15 Geneswere excluded from data analysis if they did not have signal intensity of at least30 in both channels for at least 70% of the experiments. To identify estrogen-regulated genes, we used one-class significance analysis of microarrays (SAM)to identify genes that changed in all estrogen-treated time points (as a singleclass) relative to the estrogen-starved cells.16 In our SAM analyses, we did notuse the fold-change cutoff option to avoid the fold-change associated compli-cations/pitfalls described by Larsson et al.17 Using a false discovery rate (FDR)of 0.04%, SAM identified 383 estrogen-induced and 574 estrogen-repressed genes; for subsequent estrogen-SAM analyses, only the 383 in-duced genes were used. Average linkage hierarchical cluster analysis wasconducted and the results visualized in TreeView (http://taxonomy.zoolo-gy.gla.ac.uk/rod/treeview.html).18

GATA3-induced genes were identified by microarray experiments on293T cells transfected with GATA3 gene constructs, as detailed in Usary et al.19

One-class SAM analysis (0.58% FDR) identified 407 genes that were inducedin the GATA3 samples (as a single class) relative to empty vector controls.

Analysis of Primary Breast Tumor Data Using the Estrogen-

Induced Gene Set

The primary breast tumor samples (collected with patient consentand UNC-CH Human Investigations Review Committee approval) used inthe training data set are described in Hu et al (submitted for publication),except for 14 new tumor samples. A total of 118 fresh frozen breast tumorand nine nontumor breast samples represented by 160 microarray experi-ments were analyzed using the 1,300-gene “breast intrinsic” gene set devel-oped by Hu et al (submitted), which identified 65 tumors as belonging tothe luminal subtype. These luminal tumors included 61 ER� and/or PR�tumors according to immunohistochemistry, three ER– and PR–, and onenot determined.

The 383-gene MCF-7 estrogen-induced gene list was used to hierarchi-cally cluster the 65 luminal tumors resulting in two groups, which we calledgroups I and II. We used a two-class, unpaired SAM analysis (with 1%FDR) to identify 822 genes (referred to as the estrogen-SAM list) that

optimally differentiated group I versus group II tumors.20,21 The 65 lumi-nal tumors were then clustered using the 822 genes, and two groups(groups IE and IIE) were evident.

By matching UniGene (NCBI, http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db�unigene) identifiers, data for as many as possible of the 822estrogen-SAM genes was obtained for ER� and/or PR� tumors (classified asprovided in the primary publications) from three independent test datasets.8,9,12,22 The Ma et al,12 Sorlie et al,8,9 and Chang et al22 data sets con-sisted of 60, 90, and 250 ER� and/or PR� tumors, respectively. Ma et altumors were uniformly treated with adjuvant tamoxifen alone. Sorlie et altumors received adjuvant tamoxifen, with some also receiving neoadjuvantchemotherapy. Chang et al tumors were heterogeneously treated; 24 of the250 tumors we used for this data set were discussed in earlier publications.23

To remove microarray platform/source systematic biases, we applieddistance-weighted discrimination (DWD)24 to the training and test data sets.From the DWD standardized luminal tumor training data set, centroids werecreated consisting of the average expression of the 822 estrogen-SAM genes forgroups IE and IIE. We then classified each ER� and/or PR� tumor in the testdata sets as group IE or IIE according to each sample’s nearest centroid asdetermined by Spearman correlation.

Survival Analyses

Kaplan-Meier survival plots were compared using the Cox-Mantellog-rank test in Winstat (R. Fitch Software, Staufen, Germany) for Excel(Microsoft Corp, Redmond, WA). Two-way contingency table analysis wasperformed using Winstat for Excel. For the Chang et al data set, multivariateCox proportional hazards analysis was performed using SAS (SAS Institute,Cary, NC).

RESULTS

Identification of Estrogen-Regulated Genes

To identify estrogen-regulated genes, we used the ER� hu-man breast tumor-derived MCF-7 cell line as a model system. Aone-class SAM supervised analysis16 with an FDR of 0.04% identi-fied 383 induced and 574 repressed genes in microarray experimentson MCF-7 cells treated with 17�-estradiol for 2, 4, 8, or 24 hours.Many genes identified were previously known to be ER-regulatedincluding CCND1, PR, RERG, CTSD, and PDZK1.25-28 Using theprogram EASE (Expression Analysis Systematic Explorer; http://david.niaid.nih.gov/david/ease.htm),29 the gene ontology (GO) cate-gories “sterol metabolism/biosynthesis,” “ribosome biogenesis/assembly,” and “cytoskeleton structural constituent” were over-represented relative to chance in the set of 383 estrogen-induced genes.

Estrogen and GATA3-Regulated Genes Are Present in

the Luminal/ER� Gene Cluster

The luminal/ER� expression cluster is a gene set identified inmany breast tumor profiling studies,23,30-32 includes GATA3 and ER,and is expressed in the luminal A and B tumor subtypes.8,9 To deter-mine whether estrogen-regulated genes are present in the luminal/ER� gene set, we first clustered 118 primary breast tumors using a1,300-gene “breast intrinsic” gene set developed by Hu et al (submit-ted; Fig 1). Figure 1B shows that many of the estrogen-regulated genesfrom our in vitro MCF-7 experiments were present in the tumordefined luminal/ER� gene cluster. To further define relationshipsamong genes in this cluster, we also ascertained the presence of genesregulated by GATA3, a transcription factor with an important role inER� breast cancer biology.19 Of the 407 genes induced by GATA3 invitro, many were present in the luminal/ER�gene cluster. Thus, genesidentified as being regulated by ER and/or GATA3 in vitro also cluster

Estrogen-Regulated Genes in ER�/PR� Breast Cancer

www.jco.org 1657


Fig 1. Genes regulated by estrogen and/or GATA3 in vitro are present in the primary tumor luminal epithelial/estrogen receptor–positive (ER�) gene cluster. (A)Scaled-down representation of 118 tumors hierarchically clustered using the 1,300-gene intrinsic list developed by Hu et al (submitted for publication). (B) Luminal/ER�gene cluster. Blue, luminal epithelial subtype; pink, HER2�/ER–; red, basal-like; green, nontumor breast–like.

Oh et al

1658 JOURNAL OF CLINICAL ONCOLOGY


Fig 2. Hierarchical cluster analysis of the 65 luminal tumors (identified in Fig 1) using the 822-gene estrogen–significance analysis of microarrays–derived list. (A)Scaled-down representation of the complete cluster diagram. Group IE and IIE tumors are indicated by blue and orange, respectively. Gene clusters containing (B) XBP1,(C) ribosomal genes, (D) progesterone receptor, (E) FOXA1, (F) MAGE genes, (G) proliferation signature, and (H) apoptosis and interferon-response genes.


www.jco.org 1659


near these transcription factors in vivo and help to define an expres-sion pattern seen in many studies.23,30,32,34

Analysis of Luminal Tumors Using

Estrogen-Induced Genes

We hypothesized that expression differences of estrogen-induced genes may define clinically relevant subgroups within clini-cally defined ER� and/or PR� tumors. To test this hypothesis, weclustered the 65 tumors identified as luminal in Figure 1 (blue dendro-gram branch) using the 383 MCF-7 estrogen-induced genes. Twomain groups resulted. Group I had higher expression of XBP1, PR, andTFF, which are all known ER targets. Group II had higher expressionof a cluster of estrogen-induced genes that included CTPS, E2F6, andFANCA. Kaplan-Meier survival analysis showed that group I patientshad significantly better relapse-free survival (RFS) outcomes thangroup II (P � .0004).

To further characterize the differences between group I and IItumors, we performed a supervised analysis (two-class SAM with 1%FDR) using the major dendrogram branch division to define the twosupervising groups. This analysis identified 822 genes for which groupI and II tumors showed significant differential expression. This geneset, the estrogen-SAM list, was then used to hierarchically cluster the65 luminal tumors (Fig 2), which as expected, resulted in a very similar

grouping of samples when compared with that using the 383 estrogen-induced genes. Kaplan-Meier analysis showed that using the estrogen-SAM list grouped the tumors into two groups (referred to as group IEand IIE) that predicted RFS (P � .019; Fig 3A).

Group IE tumors showed high expression of XBP1, FOXA1, PR,and many ribosomal genes (Fig 2B-E). According to EASE, the GOcategories “transcriptional regulation,” “DNA binding,” and “extra-cellular” were over-represented relative to chance in group IE tumors.Group IIE tumors showed the high expression of a prominent prolif-eration signature34,35 including Ki-67, MYBL2, Survivin, STK6, andCCNB2 (Fig 2G); these first four genes plus CCNB1 form the basisfor the proliferation portion of the Paik et al recurrence score predic-tor,13 which is a gene expression–based outcome predictor for ER�/node-negative, tamoxifen-treated patients. Recently, Dai et al10

performed a supervised analysis for genes that correlated withoutcomes in patients with high ER expression relative to age, andidentified this same proliferation signature as the main determi-nant for predicting patient outcomes; however, they identified fewgenes associated with good outcomes.

Group IIE tumors also showed high expression of a cluster ofMAGE-A genes (Fig 2F), which have been associated with an increasedrecurrence risk36 and poor tumor differentiation.37 Figure 2H shows

Fig 3. Kaplan-Meier survival curves of estrogen receptor– and/or progesterone receptor–positive tumors classified as groups IE or IIE using the 822-geneestrogen–significance analysis of microarrays–derived list. Survival curves are shown for (A) the 65 Luminal epithelial tumor training data set, (B) the Ma et al,12 (C)Sorlie et al,8,9 and (D) Chang et al22 data sets. P values were calculated using the log-rank test.

Oh et al



that group IIE tumors have high expression of genes with functions inthe interferon pathway and apoptosis, such as FLIP/CFLAR, which isan inhibitor of tumor necrosis factor receptor–mediated apoptosis.38

Several antiapoptosis genes including FLIP, AVEN, Survivin, andBCL2A1 showed high expression in group IIE, suggesting an impairedability to undergo cell death. Recent reports have shown that highexpression of FLIP39 or BCL2A140,41 can directly contribute tochemoresistance, suggesting that functional inhibition of these pro-teins may provide a therapeutic target for group IIE patients. Accord-ing to EASE, the GO categories “cell cycle/mitosis,” “antiapoptosis,”and “MHC-I” were over-represented relative to chance in group IIE.

Group IE-IIE Classification Predicts Outcome in ER�and/or PR� Tumors

To test the group IE-IIE classification as a clinically relevantoutcome predictor, we analyzed ER� and/or PR� tumors from threepublished breast tumor microarray data sets.9,12,22 We used a single-sample prediction algorithm to classify tumors in each test data set,which involved creating group IE and IIE centroids/average profilesfrom the training data set (described in Analysis of Primary Breast

Tumor Data Using the Estrogen-Induced Gene Set, under Materialsand Methods). Kaplan-Meier analysis (Fig 3B-D) showed that groupIE tumors had significantly better RFS in all test data sets. The groupIE-IIE classification was also a significant predictor of overall survival(OS) for the test data sets in which OS data was available.9,22 Further-more, by decreasing the FDR in SAM, we were able to define groups IEand IIE using a reduced estrogen-SAM list of 113 genes without anyloss of predictive ability (Appendix A1, online only).

Multivariate Analysis

Multivariate Cox proportional hazards analysis was performedon the Chang et al data set (Table 1). Using RFS and OS as the endpoints, multivariate analysis showed that classifying tumors asgroup IE or IIE provided significant prognostic power indepen-dent of standard clinical factors (RFS P � .0001; OS P � .001). Thegroup IE-IIE designation had the strongest association of all vari-ables with RFS and OS.

In multivariate analyses that included Chang et al’s22 wound-response signature and van’t Veer et al’s23 70-gene signature alongwith the clinical variables, the group IE-IIE classification continued to

Table 1. Multivariate Cox Proportional Hazards Analysis of Various Prognostic Factors in Relation to Relapse-Free Survival and Overall Survival for ER�and/or PR� Tumors in the Chang et al22 Data Set

Variable

Relapse-Free Survival Overall Survival

Hazard Ratio 95% CI P Hazard Ratio 95% CI P

Group IIE v IE� 2.90 1.71 to 4.92 < .0001 3.64 1.67 to 7.95 .001

Age† 0.48 0.31 to 0.74 .001 0.53 0.30 to 0.93 .028

Size‡ 1.59 1.01 to 2.48 .044 1.45 0.80 to 2.64 .22Tumor grade 2 or 3 v 1 1.80 0.99 to 3.3 .056 3.57 1.24 to 10.23 .02

Node status§ 2.11 1.08 to 4.11 .028 1.85 0.74 to 4.61 .19Hormonal or chemotherapy v no adjuvant therapy 0.36 0.18 to 0.71 .003 0.47 0.19 to 1.19 .11

NOTE. Boldfacing indicates variables found to be significant (P � .05) in the Cox proportional hazards model.Abbreviations: ER, estrogen receptor; PR, progesterone receptor.�Tumors were classified as group IE or IIE using the estrogen–significance analysis of microarrays–derived list.†Age was a continuous variable formatted as decade-years.‡Size was a binary variable: 0 � diameter � 2 cm; 1 � diameter � 2 cm.§Node status was a binary variable: 0 � no positive nodes; 1 � one or more positive nodes.

Table 2. Multivariate Cox Proportional Hazards Analysis for ER� and/or PR� Tumors in the Chang et al22 Data Set Using Various Prognostic FactorsIncluding the Group IE-IIE Classification, the van’t Veer et al23 70-Gene Signature, and the Chang et al Wound-Response Signature

Variable

Relapse-Free Survival Overall Survival

Hazard Ratio 95% CI P Hazard Ratio 95% CI P

Group IIE v IE� 2.01 1.15 to 3.49 .014 2.31 1.03 to 5.19 .042

70-gene signature (poor v good) 2.76 1.50 to 5.06 .001 4.17 1.62 to 10.73 .003

Wound-response signature (activated v quiescent) 2.30 1.09 to 4.85 .028 2.80 0.82 to 9.55 .10Age† 0.56 0.36 to 0.87 .010 0.64 0.36 to 1.14 .13Size‡ 1.45 0.93 to 2.28 .10 1.34 0.74 to 2.45 .34Tumor grade 2 or 3 v 1 0.93 0.47 to 1.82 .82 1.62 0.53 to 4.93 .40Node status§ 1.72 0.89 to 3.33 .11 1.51 0.62 to 3.70 .36Hormonal or chemotherapy v no adjuvant therapy 0.37 0.19 to 0.74 .005 0.46 0.18 to 1.14 .095

NOTE. Boldfacing indicates variables found to be significant (P � .05) in the Cox proportional hazards model. The 70-gene signature and the wound-responsesignature classifications were taken exactly as calculated in Chang et al,22 and their performances in multivariate analysis may be optimistic.

�Tumors were classified as group IE or IIE using the estrogen–significance analysis of microarrays–derived list.†Age was a continuous variable formatted as decade-years.‡Size was a binary variable: 0 � diameter � 2 cm; 1 � diameter � 2 cm.§Node status was a binary variable: 0 � no positive nodes; 1 � one or more positive nodes.


www.jco.org 1661


provide significant prognostic power independent of other variablesin the model (RFS P � .014; OS P � .042; Table 2). The performanceof the 70-gene and wound-response signatures in this multivariateanalysis may be optimistically high because a subset of the patients inthe Chang et al data set was used to train/optimize these two signa-tures; therefore, the ability of the group IE-IIE classification to showindependent prognostic power in a model containing these two pre-dictors indicates its usefulness in predicting outcomes.

Group IE-IIE Associations With Clinical and

Biologic Parameters

To examine the hypothesis that group IE may be more differen-tiated than group IIE tumors, we determined whether an associationexisted between this classification and histologic grade. Two-way con-tingency table analysis showed significant association between gradeand group IE-IIE class (Appendix Table A2, online only), with grade 1and 3 tumors more likely to be classified as group IE and IIE, respec-tively. Cramer’s V statistic, which measures the strength of associationbetween two variables in a contingency table, indicated a substantialassociation (Cramer’s V � 0.36) between grade and group IE-IIE classfor all data sets. For the Sorlie et al data set, p53 mutation data wasavailable, and a two-way contingency table analysis showed a signifi-cant association between p53 status and group IE-IIE class, with groupIIE more likely to be p53 mutant (P � .0019; Cramer’s V � 0.44).

Comparison of Group IE-IIE Classification to Luminal

A/B classification

We compared group IE-IIE classification to the luminal A/Bclassification.8,9 To identify Luminal A and B tumors in the three testdata sets, we used the single-sample predictor developed in Hu et al(submitted), which employs centroids for each of the five breast tu-mor “intrinsic subtypes.” We then reclassified luminal A and B tumorsfrom each data set as group IE or IIE. Kaplan-Meier analyses showedthat the group IE-IIE classification did equally well or slightly bettercompared with the luminal A/B classification in separating luminaltumors into two groups with different survival outcomes (AppendixTable A3, online only).

DISCUSSION

The search for markers that predict long-term outcomes in hormonereceptor-positive tamoxifen-treated patients has been an intense areaof study. Genomic analyses have contributed to this area, with thedevelopment of several predictive gene sets and assays based on theselection of genes that directly correlate with patient/tumoroutcomes.10-13,23 We took a different approach and selected genesusing no knowledge of outcomes, and instead selected genes on thebasis of regulation by estrogen and their natural patterns of expressionin primary tumors. The 822-gene estrogen-SAM list identified manygenes that may help explain the outcome differences seen in ER�

and/or PR� patients. Good-outcome group IE tumors tended to bemore differentiated, and highly expressed a subset of estrogen- andGATA3-regulated genes. Conversely, poor-outcome group IIE tu-mors were more likely to be poorly differentiated. Association of thegroup IE-IIE profile with grade is expected because grade includes ameasure of proliferation, which is an important determinant of out-comes in ER� and/or PR� patients.8,10,13,19 However, because thegroup IE-IIE distinction was significant in a multivariate analysis withgrade included, this distinction adds prognostic information beyondwhat grade provides.

We used three published data sets as test sets and confirmed thatthe Group IE-IIE classification was a significant predictor in ER�and/or PR� patients. We note, however, that the relapse rates differedbetween data sets, and that the Group IE tumors showed 7% to 40%relapse rates depending on the data set (Fig 3). This underscores thefact that relapse rates are dependent on the characteristics of thepatient set used. For example, comparing relapse rates in the Changet al22 data set to those observed in Paik et al13 may not be validbecause the Paik et al data set comprised tamoxifen-treated, node-negative patients, whereas the majority of the Chang et al patientsreceived no adjuvant therapy and many were node positive. However,the multivariate analysis we performed on the Chang et al data setindicated that our predictor had significant prognostic value inde-pendent of standard clinical factors and other gene expression–based predictors, and a hazard ratio of 2.90 for group IIE versus IEindicates that our predictor has potential clinical utility. By limit-ing the Chang et al data set to those patients who received adjuvanttherapy and were stage I-II, we observed a relapse rate for groupIE patients of 12% (P � .007) and significance for overall survivaloutcomes (data not shown). This indicates that given a patientpopulation similar to Paik et al, our predictor’s “good group” canachieve outcomes similar to the Paik et al low-risk group.

An important unanswered question is whether the group IE-IIEdistinction predicts pure prognosis, responsiveness to endocrine ther-apy, or both. From analyses of patient subsets in the Chang et al dataset, it is clear that the group IE-IIE distinction predicts outcome inER� and/or PR� patient subsets either receiving or not receivingadjuvant hormone therapy (data not shown). Paik et al observedsimilar results for their predictor.13 This is not surprising because half(eight of 16) of the Paik et al genes were present in the estrogen-SAMgene set. However, an advantage of our analysis is that it providesadditional biologic information (eg, antiapoptosis genes) that the Paiket al and other predictors did not. Paik et al’s finding that their predic-tor also predicts benefit of chemotherapy42 may also apply to ours.Thus, the most pressing questions remaining regarding the groupIE-IIE classification are (1) whether group IE and IIE gain similarbenefits from chemotherapy, and (2) because group IIE tumors dopoorly in the presence of tamoxifen, might they do better if adminis-tered alternative endocrine therapies.

REFERENCES

1. Osborne CK, Yochmowitz MG, Knight WA III,et al: The value of estrogen and progesterone recep-tors in the treatment of breast cancer. Cancer 46:2884-2888, 1980

2. Early Breast Cancer Trialists’ CollaborativeGroup (EBCTCG): Effects of chemotherapy and hor-

monal therapy for early breast cancer on recurrenceand 15-year survival: An overview of the randomisedtrials. Lancet 365:1687-1717, 2005

3. Bardou VJ, Arpino G, Elledge RM, et al: Pro-gesterone receptor status significantly improvesoutcome prediction over estrogen receptor statusalone for adjuvant endocrine therapy in two largebreast cancer databases. J Clin Oncol 21:1973-1979, 2003

4. Osborne CK: Tamoxifen in the treatmentof breast cancer. N Engl J Med 339:1609-1618,1998

5. Ellis MJ, Coop A, Singh B, et al: Letrozole ismore effective neoadjuvant endocrine therapy thantamoxifen for ErbB-1- and/or ErbB-2-positive, estro-gen receptor-positive primary breast cancer: Evi-dence from a phase III randomized trial. J Clin Oncol19:3808-3816, 2001

Oh et al



6. Osborne CK, Bardou V, Hopp TA, et al: Role ofthe estrogen receptor coactivator AIB1 (SRC-3) andHER-2/neu in tamoxifen resistance in breast cancer.J Natl Cancer Inst 95:353-361, 2003

7. Shou J, Massarweh S, Osborne CK, et al:Mechanisms of tamoxifen resistance: Increased es-trogen receptor-HER2/neu cross-talk in ER/HER2-positive breast cancer. J Natl Cancer Inst 96:926-935, 2004

8. Sorlie T, Perou CM, Tibshirani R, et al: Geneexpression patterns of breast carcinomas distin-guish tumor subclasses with clinical implications.Proc Natl Acad Sci U S A 98:10869-10874, 2001

9. Sorlie T, Tibshirani R, Parker J, et al: Repeatedobservation of breast tumor subtypes in indepen-dent gene expression data sets. Proc Natl Acad SciU S A 100:8418-8423, 2003

10. Dai H, van’t Veer L, Lamb J, et al: A cellproliferation signature is a marker of extremely pooroutcome in a subpopulation of breast cancer pa-tients. Proc Natl Acad Sci U S A 65:4059-4066, 2005

11. Jansen MP, Foekens JA, van Staveren IL, etal: Molecular classification of tamoxifen-resistantbreast carcinomas by gene expression profiling.J Clin Oncol 23:732-740, 2005

12. Ma XJ, Wang Z, Ryan PD, et al: A two-geneexpression ratio predicts clinical outcome in breastcancer patients treated with tamoxifen. Cancer Cell5:607-616, 2004

13. Paik S, Shak S, Tang G, et al: A multigeneassay to predict recurrence of tamoxifen-treated,node-negative breast cancer. N Engl J Med 351:2817-2826, 2004

14. Troester MA, Hoadley KA, Sorlie T, et al:Cell-type-specific responses to chemotherapeuticsin breast cancer. Cancer Res 64:4218-4226, 2004

15. Hu Z, Troester M, Perou CM: High reproduc-ibility using sodium hydroxide-stripped long oligonu-cleotide DNA microarrays. Biotechniques 38:121-124, 2005

16. Tusher VG, Tibshirani R, Chu G: Significanceanalysis of microarrays applied to the ionizing radia-tion response. Proc Natl Acad Sci U S A 98:5116-5121, 2001

17. Larsson O, Wahlestedt C, Timmons JA: Con-siderations when using the significance analysis ofmicroarrays (SAM) algorithm. BMC Bioinformatics6:129, 2005

18. Eisen MB, Spellman PT, Brown PO, et al:Cluster analysis and display of genome-wide expres-

sion patterns. Proc Natl Acad Sci U S A 95:14863-14868, 1998

19. Usary J, Llaca V, Karaca G, et al: Mutation ofGATA3 in human breast tumors. Oncogene 23:7669-7678, 2004

20. Bair E, Tibshirani R: Semi-supervised meth-ods to predict patient survival from gene expressiondata. PLoS Biol 2:E108, 2004

21. Bullinger L, Dohner K, Bair E, et al: Use ofgene-expression profiling to identify prognostic sub-classes in adult acute myeloid leukemia. N EnglJ Med 350:1605-1616, 2004

22. Chang HY, Nuyten DS, Sneddon JB, et al:Robustness, scalability, and integration of a wound-response gene expression signature in predictingbreast cancer survival. Proc Natl Acad Sci U S A102:3738-3743, 2005

23. van’t Veer LJ, Dai H, van de Vijver MJ, et al:Gene expression profiling predicts clinical outcomeof breast cancer. Nature 415:530-536, 2002

24. Benito M, Parker J, Du Q, et al: Adjustment ofsystematic microarray data biases. Bioinformatics20:105-114, 2004

25. Altucci L, Addeo R, Cicatiello L, et al: 17beta-Estradiol induces cyclin D1 gene transcription,p36D1-p34cdk4 complex activation and p105Rbphosphorylation during mitogenic stimulation ofG(1)-arrested human breast cancer cells. Oncogene12:2315-2324, 1996

26. Finlin BS, Gau CL, Murphy GA, et al: RERG isa novel ras-related, estrogen-regulated and growth-inhibitory gene in breast cancer. J Biol Chem 276:42259-42267, 2001

27. Ghosh MG, Thompson DA, Weigel RJ: PDZK1and GREB1 are estrogen-regulated genes ex-pressed in hormone-responsive breast cancer. Can-cer Res 60:6367-6375, 2000

28. Jakowlew SB, Breathnach R, Jeltsch JM, etal: Sequence of the pS2 mRNA induced by estrogenin the human breast cancer cell line MCF-7. NucleicAcids Res 12:2861-2878, 1984

29. Hosack DA, Dennis G Jr, Sherman BT, et al:Identifying biological themes within lists of geneswith EASE. Genome Biol 4:R70, 2003

30. Gruvberger S, Ringner M, Chen Y, et al:Estrogen receptor status in breast cancer is associ-ated with remarkably distinct gene expression pat-terns. Cancer Res 61:5979-5984, 2001

31. Hoch RV, Thompson DA, Baker RJ, et al:GATA-3 is expressed in association with estrogen

receptor in breast cancer. Int J Cancer 84:122-128,1999

32. Sotiriou C, Neo SY, McShane LM, et al: Breastcancer classification and prognosis based on geneexpression profiles from a population-based study.Proc Natl Acad Sci U S A 100:10393-10398, 2003

33. West M, Blanchette C, Dressman H, et al:Predicting the clinical status of human breast cancerby using gene expression profiles. Proc Natl AcadSci U S A 98:11462-11467, 2001

34. Perou CM, Jeffrey SS, van de Rijn M, et al:Distinctive gene expression patterns in humanmammary epithelial cells and breast cancers. ProcNatl Acad Sci U S A 96:9212-9217, 1999

35. Whitfield ML, Sherlock G, Saldanha AJ, et al:Identification of genes periodically expressed in thehuman cell cycle and their expression in tumors.Mol Biol Cell 13:1977-2000, 2002

36. Otte M, Zafrakas M, Riethdorf L, et al:MAGE-A gene expression pattern in primary breastcancer. Cancer Res 61:6682-6687, 2001

37. Kavalar R, Sarcevic B, Spagnoli GC, et al:Expression of MAGE tumour-associated antigens isinversely correlated with tumour differentiation ininvasive ductal breast cancers: An immunohisto-chemical study. Virchows Arch 439:127-131, 2001

38. Srinivasula SM, Ahmad M, Ottilie S, et al:FLAME-1, a novel FADD-like anti-apoptotic moleculethat regulates Fas/TNFR1-induced apoptosis. J BiolChem 272:18542-18545, 1997

39. Abedini MR, Qiu Q, Yan X, et al: Possible roleof FLICE-like inhibitory protein (FLIP) in chemoresis-tant ovarian cancer cells in vitro. Oncogene 23:6997-7004, 2004

40. Wang CY, Guttridge DC, Mayo MW, et al:NF-kappaB induces expression of the Bcl-2 homo-logue A1/Bfl-1 to preferentially suppress chemotherapy-induced apoptosis. Mol Cell Biol 19:5923-5929, 1999

41. Cheng Q, Lee HH, Li Y, et al: Upregulation ofBcl-x and Bfl-1 as a potential mechanism of che-moresistance, which can be overcome by NF-kappaB inhibition. Oncogene 19:4936-4940, 2000

42. Paik S, Shak S, Tang G, et al: Expression ofthe 21 genes in the Recurrence Score assay andprediction of clinical benefit from tamoxifen inNSABP study B-14 and chemotherapy in NSABPstudy B-20. Presented at the 27th Annual San An-tonio Breast Cancer Symposium, San Antonio, TX,December 8-11, 2004 (abstr 24)

■ ■ ■

Acknowledgment

We thank Phil Bernard, MD, Juan Palazzo, MD, and Funmi Olopade, MD, for providing tumor materials and advice, and Lynda Sawyer andAmy Drobish for collection of patient-associated data.

Appendix

The Appendix is included in the full-text version of this article, available online at www.jco.org. It is not included in the PDF version(via Adobe® Reader®).

Authors’ Disclosures of Potential Conflicts of InterestThe authors indicated no potential conflicts of interest.


www.jco.org 1663


Author Contributions

Conception and design: Daniel S. Oh, Charles M. PerouFinancial support: Charles M. PerouProvision of study materials or patients: Lisa A. Carey, Charles M. PerouCollection and assembly of data: Daniel S. Oh, Jerry Usary, Zhiyuan Hu, Xiaping He, Cheng Fan, Junyuan Wu, Charles M. PerouData analysis and interpretation: Daniel S. Oh, Melissa A. Troester, Jerry Usary, Cheng Fan, Charles M. PerouManuscript writing: Daniel S. Oh, Melissa A. Troester, Jerry Usary, Lisa A. Carey, Charles M. PerouFinal approval of manuscript: Daniel S. Oh, Zhiyuan Hu, Lisa A. Carey, Charles M. Perou

GLOSSARY

Centroid: Mean expression profile of a group of samples for agiven set of genes.

Cramer’s V statistic: This statistic provides a quantitativemeasure of the strength of association between the two variablesin a contingency table (which cannot be obtained from the Pvalue). Cramer’s V values range from 0 to 1, with 0 indicating norelationship and 1 indicating perfect association. Traditionally,values between 0.36 and 0.49 indicate a substantial relationship,and values greater than 0.50 indicate a very strong relationship.The V statistic is a generalization of the more familiar � statisticto non–2 � 2 contingency tables, and for 2 � 2 tables the V sta-tistic is equal to the � statistic.

DWD (distance-weighted discrimination): A methodto identify and adjust systematic biases that are present withinmicroarray datasets. These systematic biases are a result of manydifferent experimental features including different microarrayplatforms and different production lots of microarrays. DWD isuseful in merging/comparing two tumor microarray datasetscompleted on different microarray platforms.

EASE (Expression Analysis Systematic Explorer): Acomputer program used for rapid biologic interpretation of genelists that result from the analysis of microarray or other high-throughput genomic data. The main function of EASE is to per-form theme discovery for a given list of genes.

Gene ontology: Allows for annotating genes and their productswith a limited set of attributes, with the three organizing principles beingmolecular function, biological process, and cellular component. Thedevelopment of structured, controlled vocabularies (ontologies) thatdescribe gene products in terms of these organizing principles in aspecies-independent manner is a constantly evolving process.

Luminal epithelial tumors: Human breast tumors that are char-acterized by the expression of genes typically expressed in the cells thatline the ducts of normal mammary glands; these genes include the estro-gen receptor (ER), GATA3, X-box binding protein 1, FOXA1, and cyto-keratins 8 and 18.

SAM (significance analysis of microarrays): A statisticaltechnique using established software that determines the significance inchanges of gene expression seen in microarray analysis (eg, cDNA andoligonucleotide microarrays), which measures the expression of thou-sands of genes and identifies changes in expression between differentbiologic states. On the basis of changes in gene expression relative to thestandard deviation of repeated measurements, SAM assigns a score toeach gene. When scores are greater than an adjustable threshold, permu-tations of repeated measurements are used by SAM to estimate the per-centage of such genes identified by chance, the false discovery rate(FDR). In addition, SAM correlates gene expression data to a wide rangeof clinical parameters, including treatment, diagnosis categories, andsurvival time.

Supervised analysis: Analysis of gene expression profiling data inwhich external information such as survival status is used as a guide toselect genes from microarray data.

Two-way contingency table: A tabular representation of cate-gorical data, it shows the frequencies for particular combinations of val-ues of two discrete variables X and Y. Each cell in the table represents amutually exclusive combination of X-Y values. A �2-based P value canbe calculated for a contingency table and is used to determine whetherthere is a statistically significant association between the two variables inthe contingency table. A 2 � 2 contingency table refers to a table inwhich X and Y are binary variables.

Oh et al



Functional Analysis of the Breast Cancer GenomeMatthew J. Ellis, Division of Oncology, Department of Internal Medicine, Washington University in St Louis, St Louis, MO

Clinicians have long recognized that a diagnosis of breastcancer encompasses different tumor types with very different clin-ical outcomes. The first critical step toward a biomarker-basedsubclassification schema to aid disease management occurred withthe widespread introduction of tumor estrogen-receptor (ER)measurement three decades ago. It is worth revisiting the litera-ture of the time, because the introduction of ER testing was verycontroversial. In particular, investigators worried that a benefitfrom endocrine therapy in ER– disease could not be excluded.1

It took the weight of a meta-analysis to resolve this issue,2 buteven today we worry about excluding a patient from appropri-ate endocrine therapy because of false-negative ER results. Thesingle-sample prediction problem continues to be central to thecurrent debate regarding new and more complex biomarkerapproaches. Apparently robust group predictions typical ofthese analyses does not necessarily require great measurementaccuracy. Prospective clinical testing, on the other hand, needsto be highly precise to avoid the potentially tragic consequencesof tumor misclassification.

Further subclassification for ER� disease to prospectivelyidentify endocrine therapy–resistant ER� tumors is of longstand-ing interest. The introduction of HER2 testing has already taken ussome way down this road. It is clear that the activation of HER2 bygene amplification drives endocrine therapy resistance,3 and theintroduction of the HER2-targeting monoclonal antibody trastu-zumab into the adjuvant setting improves outcomes for patientswith ER� HER2� tumors.4 HER2 gene amplification does not,however, explain all or even most cases of endocrine therapy fail-ure, and new insights and predictive models are needed. Onelongstanding idea is to consider the ER in the context of an entirepathway; as our insight into the role of estrogen in tumorigenesisdeepens, we will be able to move beyond reliance on progesteronereceptor (PgR) measurement for ER activity assessment. The lateWilliam McGuire, who championed progesterone receptor mea-surement in breast cancer,5 would therefore have been very inter-ested in the paper by Oh et al6 in this issue of the Journal of ClinicalOncology. The University of North Carolina–Chapel Hill (ChapelHill, NC) team assessed the biomarker potential of hundreds ofestrogen-regulated genes (including PgR) encoding proteins with awide variety of cellular functions.

To briefly summarize their findings, Oh et al identified anextensive list of genes that were found to be regulated by estradiol atmultiple time points in the MCF7 breast cancer cell line. Althoughthey found both estrogen-repressed genes and estrogen-inducedgenes, their subsequent work in the paper focused on only upregu-

lated genes. As an interesting side issue in the article, Oh et alcompared this list of genes with a list induced by transfection of theGATA3 transcription factor into a GATA3 null cell line and foundconsiderable overlap. Thus GATA3 and ER must work in concertto regulate this large gene repertoire. There was also a good deal ofoverlap with genes found to be expressed in ER� breast cancers inearlier studies (referred to as “luminal” in keeping with Oh et al’searlier work on breast cancer classification) as well as genes repre-sented in the only clinically available multigene test that predictstamoxifen resistance.7

To ascertain the role these estrogen-induced genes might playin clinical outcomes, Oh et al first performed hierarchical cluster-ing of the estrogen-induced genes in a 64 –luminal tumor trainingset and identified two major groups, groups I and II. All of the geneinformation available from these cases was subsequently comparedbetween groups I and II (a supervised analysis), with a 1% false-discovery rate, to produce an 822-gene list. The tumors were thenreclustered with the 822 differentially expressed genes to reproducetwo groups (called IE and IIE). Information on as many of the 822genes as possible was then sought from three publicly available,clinically annotated breast cancer microarray databases. After datamanipulation to remove “platform and source systemic bias” anaverage expression index or “centroid” was used to classify clini-cally ER� and/or PgR� tumors into two groups with significantdifferences in survival. From a functional genomics standpoint, itis interesting to note that the favorable-prognosis group IE gene listwas largely populated by transcription factors XBP1, FOXA1, andPgR and ribosomal genes. Unfavorable Group IIE tumors wererich in proliferation genes (Ki-67, MYBL2, Survivin, STK6, andCCNB2), as well as apoptosis regulators FLIP/CFLAR, AVEN, andBCL2A1. These gene signatures imply, not surprisingly, that endo-crine therapy resistance is associated with a deregulated cell cycleand resistance to programmed cell death.

So what are we to make of these data? The first issue, of course,is that Oh et al’s extensive gene list is far from a clinical test.Although the signature is robust, and survives the data averagingalgorithms necessary for cross-platform assessment, clinical testingneeds to be formulated into a reproducible technology that can berepeated thousands of times in a cost-effective manner. Here, weare faced with a dilemma that is becoming more acute every day.Although the systematic evaluation of expression from 822 genesin a clinical environment sounds expensive, actually, by usingsmaller customized microarrays, it could be achieved for a couplehundred dollars with a turnaround of 1 or 2 days. However, thestarting material for RNA extraction (at least currently) must be a

JOURNAL OF CLINICAL ONCOLOGY E D I T O R I A L

VOLUME 24 � NUMBER 11 � APRIL 10 2006

1649Journal of Clinical Oncology, Vol 24, No 11 (April 10), 2006: pp 1649-1650DOI: 10.1200/JCO.2005.05.3520

Copyright © 2006 by the American Society of Clinical Oncology. All rights reserved. Information downloaded from www.jco.org and provided by UNIV OF NC/ACQ SRVCS on April 8, 2006 from 152.2.181.221.

snap-frozen tumor biopsy whose quality and tumor content havebeen assessed by histopathology. On the other hand, the measure-ment of 822 genes by quantitative polymerase chain reaction fromRNA extracted from a paraffin block is impractical at this stage, anda step to reduce the gene number closer to 100 or so without loss ofpredictive abilities would be required for the clinical implementa-tion of Oh et al’s signature. We may have to face the fact that, likeour colleagues who treat lymphoma, breast cancer physicians willhave to insist on frozen tumor acquisition in the near future. Weachieved fresh tissue acquisition for biochemical ER testing beforeand I predict that, despite the barriers, we will be doing it again. Anobvious move in this direction is the prospective evaluation of theAmsterdam 70-gene prognostic signature that requires frozen tis-sue RNA analysis.8

One of the most interesting aspects of the Oh et al article con-cerns the glimpses of the molecular complexities of the biologic systemthat we are trying to understand. Certain genes, such as STK6, keepsurfacing in breast cancer profiling studies. This gene is present in theNational Surgical Adjuvant Breast and Bowel Project (NSABP) recur-rence score published by Paik et al.7 As a side note, gene nomenclatureis extremely frustrating: STK6 is actually assigned to the mouse gene,whereas the human homolog is Aurora-A/AURKA/STK15/BTAK. STK15 is deregulated in a wide spectrum of poor-prognosiscancers9-11 andof interest,Aurorakinase inhibitorsare indevelopment.12

The power of the HER2-trastuzumab paradigm is that wehave successfully paired a somatic mutation with a targeted ther-apy. It seems likely that the final formulation of a biologically basedclassification of breast cancer will have a strong bias toward geneticabnormalities that create opportunities for targeted treatments.The functional annotation of the breast cancer genome required

for this vision has only just begun, and the excitement in the breastcancer research community is palpable.

© 2006 by American Society of Clinical Oncology

REFERENCES1. Chamness GC, Mercer WD, McGuire WL: Are histochemical methods for

estrogen receptor valid? J Histochem Cytochem 28:792-798, 19802. Tamoxifen for early breast cancer: An overview of the randomised trials—

Early Breast Cancer Trialists’ Collaborative Group. Lancet 351:1451-1467, 19983. De Laurentiis M, Arpino G, Massarelli E, et al: A meta-analysis on the

interaction between HER-2 expression and response to endocrine treatment inadvanced breast cancer. Clin Cancer Res 11:4741-4748, 2005

4. Piccart-Gebhart MJ, Procter M, Leyland-Jones B, et al: Trastuzumab afteradjuvant chemotherapy in HER2-positive breast cancer. N Engl J Med 353:1659-1672, 2005

5. McGuire WL, Horwitz KB, Pearson OH, et al: Current status of estrogen andprogesterone receptors in breast cancer. Cancer 39:2934-2947, 1977

6. Oh DS, Troester MA, Usary J, et al: Estrogen-regulated genes predictsurvival in hormone receptor-positive breast cancers. J Clin Oncol 24:10.1200/JCO.2005.03.2755

7. Paik S, Shak S, Tang G, et al: A multigene assay to predict recurrence oftamoxifen-treated, node-negative breast cancer. N Engl J Med 351:2817-2826, 2004

8. van de Vijver MJ, He YD, van’t Veer LJ, et al: A gene-expression signatureas a predictor of survival in breast cancer. N Engl J Med 347:1999-2009, 2002

9. Zhou H, Kuang J, Zhong L, et al: Tumour amplified kinase STK15/BTAKinduces centrosome amplification, aneuploidy and transformation. Nat Genet20:189-193, 1998

10. Forozan F, Mahlamaki EH, Monni O, et al: Comparative genomic hybrid-ization analysis of 38 breast cancer cell lines: A basis for interpreting comple-mentary DNA microarray data. Cancer Res 60:4519-4525, 2000

11. Tanner MM, Grenman S, Koul A, et al: Frequent amplification of chromo-somal region 20q12-q13 in ovarian cancer. Clin Cancer Res 6:1833-1839, 2000

12. Gadea BB, Ruderman JV: Aurora kinase inhibitor ZM447439 blockschromosome-induced spindle assembly, the completion of chromosome con-densation, and the establishment of the spindle integrity checkpoint in Xenopusegg extracts. Mol Biol Cell 16:1305-1318, 2005

■ ■ ■

Acknowledgment

Supported by National Cancer Institute (Bethesda, MD) Grants No. U01 CA114722 and R01 CA095614.

Author’s Disclosures of Potential Conflicts of InterestThe author indicated no potential conflicts of interest.

Author Contributions

Manuscript writing: Matthew J. EllisFinal approval of manuscript: Matthew J. Ellis

Matthew J. Ellis


Copyright © 2006 by the American Society of Clinical Oncology. All rights reserved. Information downloaded from www.jco.org and provided by UNIV OF NC/ACQ SRVCS on April 8, 2006 from 152.2.181.221.

Estrogen-Regulated Genes Predict Survival in …...Estrogen-Regulated Genes Predict Survival in...

Documents

Transcript of Estrogen-Regulated Genes Predict Survival in …...Estrogen-Regulated Genes Predict Survival in...