UBC Faculty Research and Publications

PAM50 Breast Cancer Subtyping by RT-qPCR and Concordance with Standard Clinical Molecular Markers Bastien, Roy R; Rodríguez-Lescure, Álvaro; Ebbert, Mark T; Prat, Aleix; Munárriz, Blanca; Rowe, Leslie; Miller, Patricia; Ruiz-Borrego, Manuel; Anderson, Daniel; Lyons, Bradley; Álvarez, Isabel; Dowell, Tracy; Wall, David; Seguí, Miguel Á; Barley, Lee; Boucher, Kenneth M; Alba, Emilio; Pappas, Lisa; Davis, Carole A; Aranda, Ignacio; Fauron, Christiane; Stijleman, Inge J; Palacios, José; Antón, Antonio; Carrasco, Eva; Caballero, Rosalía; Ellis, Matthew J; Nielsen, Torsten O; Perou, Charles M; Astill, Mark; Bernard, Philip S; Martín, Miguel Oct 4, 2012

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

Item Metadata


52383-12920_2012_Article_315.pdf [ 587.1kB ]
JSON: 52383-1.0132661.json
JSON-LD: 52383-1.0132661-ld.json
RDF/XML (Pretty): 52383-1.0132661-rdf.xml
RDF/JSON: 52383-1.0132661-rdf.json
Turtle: 52383-1.0132661-turtle.txt
N-Triples: 52383-1.0132661-rdf-ntriples.txt
Original Record: 52383-1.0132661-source.json
Full Text

Full Text

RESEARCH ARTICLE Open AccessPAM50 Breast Cancer Subtyping by RT-qPCRand Concordance with Standard ClinicalMolecular MarkersRoy RL Bastien1, Álvaro Rodríguez-Lescure2, Mark TW Ebbert1, Aleix Prat3,4, Blanca Munárriz5, Leslie Rowe1,Patricia Miller1, Manuel Ruiz-Borrego6, Daniel Anderson1, Bradley Lyons1, Isabel Álvarez7, Tracy Dowell1,David Wall1, Miguel Ángel Seguí8, Lee Barley1, Kenneth M Boucher9, Emilio Alba10, Lisa Pappas9, Carole A Davis11,Ignacio Aranda12, Christiane Fauron1, Inge J Stijleman11, José Palacios13, Antonio Antón14, Eva Carrasco15,Rosalía Caballero15, Matthew J Ellis16, Torsten O Nielsen17, Charles M Perou3, Mark Astill1,Philip S Bernard1,11,19*† and Miguel Martín18†AbstractBackground: Many methodologies have been used in research to identify the “intrinsic” subtypes of breast cancercommonly known as Luminal A, Luminal B, HER2-Enriched (HER2-E) and Basal-like. The PAM50 gene set is oftenused for gene expression-based subtyping; however, surrogate subtyping using panels of immunohistochemical(IHC) markers are still widely used clinically. Discrepancies between these methods may lead to different treatmentdecisions.Methods: We used the PAM50 RT-qPCR assay to expression profile 814 tumors from the GEICAM/9906 phase IIIclinical trial that enrolled women with locally advanced primary invasive breast cancer. All samples were scored at asingle site by IHC for estrogen receptor (ER), progesterone receptor (PR), and Her2/neu (HER2) protein expression.Equivocal HER2 cases were confirmed by chromogenic in situ hybridization (CISH). Single gene scores by IHC/CISHwere compared with RT-qPCR continuous gene expression values and “intrinsic” subtype assignment by the PAM50.High, medium, and low expression for ESR1, PGR, ERBB2, and proliferation were selected using quartile cut-pointsfrom the continuous RT-qPCR data across the PAM50 subtype assignments.Results: ESR1, PGR, and ERBB2 gene expression had high agreement with established binary IHC cut-points (areaunder the curve (AUC)≥ 0.9). Estrogen receptor positivity by IHC was strongly associated with Luminal (A and B)subtypes (92%), but only 75% of ER negative tumors were classified into the HER2-E and Basal-like subtypes.Luminal A tumors more frequently expressed PR than Luminal B (94% vs 74%) and Luminal A tumors were lesslikely to have high proliferation (11% vs 77%). Seventy-seven percent (30/39) of ER-/HER2+ tumors by IHC wereclassified as the HER2-E subtype. Triple negative tumors were mainly comprised of Basal-like (57%) and HER2-E(30%) subtypes. Single gene scoring for ESR1, PGR, and ERBB2 was more prognostic than the corresponding IHCmarkers as shown in a multivariate analysis.(Continued on next page)* Correspondence: philip.bernard@hci.utah.edu†Equal contributors1The ARUP Institute for Clinical and Experimental Pathology, Salt Lake City,UT, USA11Department of Pathology, University of Utah Health Sciences Center/Huntsman Cancer Institute, Salt Lake City, UT, USAFull list of author information is available at the end of the article© 2012 Bastien et al.; licensee BioMed Central Ltd. 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 cited.Bastien et al. BMC Medical Genomics 2012, 5:44http://www.biomedcentral.com/1755-8794/5/44(Continued from previous page)Conclusions: The standard immunohistochemical panel for breast cancer (ER, PR, and HER2) does not adequatelyidentify the PAM50 gene expression subtypes. Although there is high agreement between biomarker scoring byprotein immunohistochemistry and gene expression, the gene expression determinations for ESR1 and ERBB2 statuswas more prognostic.BackgroundFor over a decade, research studies have used gene expres-sion to classify invasive breast cancers into biologically andclinically distinct subtypes that have become known asLuminal A, Luminal B, HER2-Enriched (HER2-E) andBasal-like [1-3]. Subtype information has repeatedly shownto be an independent predictor of survival in breast cancerwhen used in multivariate analyses with standard clinical-pathological variables [3-6]. In 2009, Parker et al. derived aminimal gene set (PAM50) for classifying “intrinsic”subtypes of breast cancer [3,7]. The PAM50 gene set hashigh agreement in classification with larger “intrinsic”gene sets previously used for subtyping [1,3,4,8], and isnow commonly employed [9-12].There are several multi-gene expression tests clinicallyavailable for determining risk of relapse in early stagebreast cancer, including the 21-gene recurrence score[13] (Oncotype DxW, Genomic Health Inc, Redwood City,CA, www.oncotypedx.com), the 14-gene distant metastasissignature [14] (BreastOncPx™, US Labs, Irvine, CA,www.uslabs.net), the 97-gene histologic grade predictor[15] (MapQuant Dx™ Genomic Grade, Ipsogen, Marseilles,France and New Haven, CT, USA, www.ipsogen.com), andthe 70-gene prognosis signature [16] (MammaPrintW,Agendia, Irvine, CA, www.agendia.com). The molecu-lar signature of proliferation is perhaps the strongestvariable in all these tests for determining outcome inER + breast cancer.In addition to gene expression profiling by micro-array or RT-qPCR [2-4,8,17,18], many studies haveused immunohistochemical panels to identify subtypes[19-21]. For example, high grade ER+/HER2- tumorsand ER+/HER2+ tumors are often considered LuminalB, while ER-/HER2+ are considered HER2-E subtypeand triple negative tumors are considered Basal-like. Inthis study, we assess agreement between histopathology/IHC status and PAM50 classification for subtype, ESR1,PGR, ERBB2, and proliferation.MethodsSamples and clinical dataThere was ethical review and approval for all protocolsused in this study from the respective centers involvedand all subjects gave written informed consent to par-ticipate. A training set was developed using 171 breastsamples, comprised of 16 “normal” breast tissue samplesfrom reduction mammoplasties or grossly uninvolvedbreast tissue and 155 primary invasive breast cancers.These samples were collected from 2005–2009 underIRB approved protocols at the University of Utah andthe University of North Carolina at Chapel Hill. Clinical-pathological information associated with the samples isbased on the College of American Pathology (CAP) andAmerican Joint Committee on Cancer (AJCC) standardsat the time of collection (Additional file 1). Subtype classifi-cation and single and meta-gene (proliferation) scoreswere predicted on an independent test set of 814 sam-ples from the GEICAM/9906 clinical trial, a rando-mized Phase 3 trial of fluorouracil, epirubicin, andcyclophosphamide alone or followed by paclitaxel[22]. Patients that were hormone receptor positive(ER and/or PR positive by IHC) were given adjuvanttamoxifen. The hormone receptor status for thesesamples was evaluated at a single site (Department ofPathology, Hospital General Universitario de Alicante)using immunohistochemistry (IHC) for progesteronereceptor (PR) (clone PgR636, DAKO, Glostrup, Den-mark) and estrogen receptor (ER) (clone 1D5, DAKO,Glostrup, Denmark) (Additional file 2). The scores forthe proportion of dyed cells and intensity were summed toobtain a total Allred Score [23]. Measurement of HER2 ex-pression was performed by Herceptest™ (DAKO, Glostrup,Denmark) and samples with scores of 2+ by IHC wereconfirmed by CISH, following the ASCO/CAP guide-lines [24]. The clinical data for the training set andGEICAM/9906 test set are summarized in Table 1.Measurement of PCR efficiency, limits of detection, andlimits of quantificationBreast cancer cell lines (BT474, MCF7, MDA-MB-231,MDA-MB-436, MDA-MB-453, MDA-MB-468, SKBR3and T47D) were cultured, pelleted and processed intoFFPE tissue blocks. The RNA was extracted, pooled, re-verse transcribed, and serially diluted at 2-fold incrementsfrom 2.56μg to 0.039ng per assay, which corresponds to arange of 7.11ng to 108.51fg of cDNA per reaction well.Each gene was measured in triplicate per RT-qPCRrun on the Roche LC480 and 2 runs were performedfor each of the 17 dilutions. A detailed description ofmethods used to calculate PCR efficiency, limits of de-tection and limits of quantification can be found inAdditional file 3.Bastien et al. BMC Medical Genomics 2012, 5:44 Page 2 of 12http://www.biomedcentral.com/1755-8794/5/44Table 1 Patient characteristicsVariable Training set data n = 154 Total (%) Variable Test set data n = 814 Total (%)Age (years) Median 55.5 Age (years) Median 50.4(range) 26 – >92 (range) 23.1 – 76.2Menopausal status Pre 49 (31.8) Menopausal status Pre 439 (53.9)Post 101 (65.6) Post 375 (46.1)Unknown 4 (2.6)Primary tumor size T1 63 (40.9) Primary tumor size T1 338 (41.5)T2 69 (44.8) T2 430 (52.8)T3 17 (11.0) T3 46 (5.7)Unknown 1 (0.6)Reduction Mamoplasty 4 (2.6)Nodal status 0 95 (61.7) Nodal status 0 01 – 3 54 (35.1) 1 – 3 503 (61.8)> 3 0 (0) > 3 311 (38.2)Unknown 1 (0.6)Reduction Mamoplasty 4 (2.6)Histopathologic grade* G1 23 (14.9) Histopathologic grade* G1 107 (13.1)G2 45 (29.2) G2 335 (41.2)G3 80 (51.9) G3 313 (38.5)GX 2 (1.3) GX 59 (7.2)Reduction Mamoplasty 4 (2.6)Estrogen receptor^ Positive 100 (64.9) Estrogen receptor^ Positive 644 (79.1)Negative 49 (31.8) Negative 170 (20.9)Unknown 1 (0.6)Reduction Mamoplasty 4 (2.6)Progesterone receptor^^ Positive 82 (53.2) Progesterone receptor^^ Positive 567 (69.7)Negative 67 (43.5) Negative 247 (30.3)Unknown 1 (0.6)Reduction Mamoplasty 4 (2.6)HER2 status Positive 37 (24.0) HER2 status Positive 116 (14.3)Negative 111 (72.1) Negative 698 (85.7)Unknown 2 (1.3)Reduction Mamoplasty 4 (2.6)Ki67 IHC Unknown 154 (100) Ki67 IHC Positive 236 (29.6)Negative 561 (70.4)PAM50 Intrinsic Subtype Luminal A 53 (34.4) PAM50 Intrinsic Subtype Luminal A 277 (34.0)Luminal B 27 (17.5) Luminal B 261 (32.1)HER2-enriched 32 (20.8) HER2-enriched 174 (21.4)Basal-like 38 (24.7) Basal-like 70 (8.6)Normal-like 4 (2.6) Normal-like 32 (3.9)*Grade based on Nottingham-Bloom-Richardson scoring.^ER positive required at least 10% staining nuclei.^^PR positive required at least 10% staining nuclei.#HER2 positive were 3+ IHC or 2+ and CISH confirmed.Bastien et al. BMC Medical Genomics 2012, 5:44 Page 3 of 12http://www.biomedcentral.com/1755-8794/5/44Selection of prototype samples for the RT-qPCR trainingsetTraining set samples were run across 3 batches of PCRplates manufactured at ARUP Laboratories (ARUP Labora-tories, Salt Lake City, UT, www.aruplab.com). The methodto identify prototype samples representing the subtypes hasbeen previously described [3]. Briefly, hierarchical clustering(median centered by gene, Pearson correlation, centroid-linkage) [25] was performed on the RT-qPCR data and Sig-Clust was run at each node of the dendrogram beginning atthe root and stopping when the test was no longer signifi-cant (p> 0.001). A “centroid” was generated for each sub-type in the training set using the average expression foreach gene across all prototype samples of a given subtype.Single sample subtype prediction was performed bycalculating a Spearman rank correlation coefficientbetween the gene expression values of an individualsample compared to each of the centroid gene valuesfor Luminal A, Luminal B, HER2-Enriched, Basal-like,and Normal. The subtype classification for the newsample is assigned to the centroid with the highestcorrelation.10-Fold cross validation to determine stability of selectedprototypesThe 154 prototype samples identified by SigClust wererandomly split into 10 groups. Nine of the 10 groupswere used to calculate new centroids for each of the 5possible subtype assignments. Each sample from theremaining group was then assigned a subtype basedon closest proximity to the newly calculated cen-troids using Spearman's Rho. The process of calculat-ing centroids using 9 of the 10 groups and predictingon the remaining group was repeated leaving out adifferent group each time.Measurements of assay reproducibilityReproducibility of the PAM50 assay was determinedusing 3 cell lines (MCF7, ME16C and SKBR3) and apool of Luminal A prototype samples that were each run12 times (3 runs across 4 batches of PAM50 plates) over30 days. Variation in each gene measurement wasassessed using the difference between the mean calibratorcrossing point (CP) and each sample replicate CP (ΔCP).The square root of the mean CV2 for ΔCP was used to es-timate the variation for each gene within plate, withinbatch, and across batches. Higher gene CVs may be due tolower concentration of a single gene within a sample. Weused the technical variability in measuring each gene tofurther assess the stability of the categorical subtype call inthe GEICAM/9906 test set samples. Since the biology be-tween subtypes is a continuum and some samples mayhave close proximity to more than 1 prototypic sub-type, we used a Monte-Carlo simulation to introducerandom error into the call to determine the frequencyof switching subtype [26].Scaling single and Meta-Gene scoresThe PAM50 subtype assay can also provide quantitativeand qualitative gene expression scores for the standardbiomarkers usually measured semi-quantitatively byIHC: ESR1/ER, PGR/PR and ERBB2/HER2. In addition,the PAM50 contains many cell cycle regulated genes thatcan be combined into a meta-gene for proliferation(CENPF, ANLN, CDC20, CCNB1, CEP55, MYBL2, MKI67,UBE2C, RRM2, and KIF2C). The meta-gene for prolif-eration were selected because they had strong correl-ation within the associated dendrogram of the trainingset cluster. The quantitative scale of 1–10 for the singlegenes and proliferation was derived by rescaling theoriginal log-expression ratios from the training set andincluded a 10% buffer on either side of the originalvalues to allow for values that were higher or lowerthan what was encountered in the training set. Anynew values that were less than 0 or greater than 10were truncated at 0 and 10, respectively.Fixed cut-points (low vs. intermediate/high) for thesingle genes (ESR1, PGR, and ERBB2) and proliferationwere directly applied from the training set to the GEI-CAM/9906 test set. Receiver Operator Characteristic(ROC) curves were generated by dichotomizing IHC dataand treating RT-qPCR data as a continuous variable.ResultsTraining set, subtype stability, and classificationaccuracyWe identified 154 prototypic samples from the RT-qPCR data by hierarchical clustering of the PAM50 clas-sifier genes, and statistical selection from the dendro-gram by SigClust [27]. The training set was comprisedof 53 Luminal A, 27 Luminal B, 32 HER2-enriched, 38Basal-like and 4 Normal-like (Figure 1). The 10-foldcross validation had 91.6% concordance (multi-raterkappa score of 0.885) with the initial SigClust subtypeassignments (Additional file 4).Interference from normal breast tissue contaminationA major source of subtype misclassification comes fromhaving normal tissue within the tumor sample [28].We assessed the effect of having contaminating normaltissue within the tumor sample by diluting out RNAfrom tumor subtypes with pooled RNA from “normal”reduction mammoplasties (0%, 25%, 50% and 75%).Primary tumors were pooled to represent Luminal Aand HER2-E samples while cell lines were used torepresent Luminal B (MCF7) and Basal-like (ME16C).The changes in subtype classification occurred in asystematic fashion with all subtypes switching directlyBastien et al. BMC Medical Genomics 2012, 5:44 Page 4 of 12http://www.biomedcentral.com/1755-8794/5/44to a classification of Normal, with the exception ofLuminal B, which switched to Luminal A. The switchfrom Luminal B to Luminal A required 50% contribu-tion from the normal breast tissue signature. Interfer-ence data from the introduction of normal breasttissue RNA into each of the subtypes is provided inAdditional file 5. During the dilution series for HER2-Ewith “normal” there was switching in the ESR1 scorebetween intermediate and low suggesting that bothsamples had similar ESR1 expression near the cut-offfor those scores.Subtype, immunohistochemistry, and RT-qPCR genescoresThe RT-qPCR values for ESR1, PGR, ERBB2, and prolif-eration were evaluated across prototypic samples in thetraining set. High, intermediate, and low cut points weremade based on the continuous distribution of expressionacross the tumor subtypes. The cut-points for each ofthe scores and how they were determined is presentedin Table 2. Figure 2 shows the expression and cut-pointsfor ESR1 in the training set and how these comparewithin the GEICAM/9906 test set. Additional single andmeta-gene cut-points for the training and test sets canbe found in Additional file 6. Comparisons between thegene expression and IHC data for GEICAM/9906 gavegood overall agreement with a high area under the curve(AUC) for ESR1/ER (AUC= 0.90), PGR/PR (AUC=0.90),and ERBB2/HER2 (AUC= 0.95) (Figure 3). Rather thanre-optimize the cut-points on the test set, the fixed cut- FGFR4  ERBB2  GRB7  BLVRA  BAG1  BCL2  CXXC5  ESR1  GPR160  FOXA1  MLPH  NAT1  SLC39A6  MAPT  PGR  MDM2  TMEM45B  MMP11  ACTR3B  CDC6  CCNE1  EXO1  CDCA1  KNTC2  BIRC5  CENPF  ANLN  CDC20  CCNB1  CEP55  MYBL2  MKI67  UBE2C  RRM2  KIF2C  MELK  TYMS  PTTG1  ORC6L  UBE2T  CDH3  EGFR  KRT17  KRT14  KRT5  FOXC1  MIA  SFRP1  PHGDH  MYC Figure 1 Clinical PAM50 RT-qPCR breast cancer training set. Hierarchical clustering of RT-qPCR data for the PAM50 classifier genesnormalized to the 5 control genes using 171 FFPE procured breast samples. Statistical selection using SigClust identified the 5 significant groupspreviously identified and designated as Luminal A (dark blue), Luminal B (light blue), HER2-E (pink), Basal-like (red), and Normal (green). The 16non-neoplastic samples (grey), from reduction mammoplasty and grossly uninvolved breast tissues, all Clustered together and away from theinvasive cancers. SigClust identified 4 reduction mammoplasty samples (green) that were used to train the Normal subtype.Table 2 Cut-points for quantitative gene scoresScore rangesGenes/Meta-genes Low Intermediate HighESR1 (ER)† 0 - 5.2 >5.2 - 7.6 >7.6 - 10PGR (PR){ 0 - 5.1 >5.1 - 7.4 >7.4 - 10ERBB2 (HER2)} 0 - 5.6 >5.6 - 7.5 >7.5 - 10Proliferation* 0 - 3.9 >3.9 - 5.3 >5.3 - 10†high ESR1 = above median expression for Luminal A; low ESR1=belowmedian expression for HER2-enriched.{high PGR= above median expression for Luminal A; low PGR=below medianexpression for Luminal B.}high ERBB2= above median expression for HER2-enriched; low ERBB2=belowlowest quartile expression for HER2-enriched.*high Proliferation = above highest quartile expression for Luminal A; lowProliferation = below lowest quartile expression for Luminal A.Bastien et al. BMC Medical Genomics 2012, 5:44 Page 5 of 12http://www.biomedcentral.com/1755-8794/5/44points based on the training set were used and it showedhigh sensitivity/specificity, although a slightly higherfalse positive rate for ERBB2 than would have beenselected by eye.Ninety-two percent (497/538) of Luminal (A/B)tumors were ER+ by IHC and 99% (530/538) had anintermediate-high ESR1 score (Tables 3 and 4). LuminalA tumors more frequently expressed PR/PGR thanLuminal B tumors using either IHC (94% vs 74%) orqPCR (95% vs 61%).Although the HER2-E subtype is often thought of asbeing ER-, only 36% (63/174) were ER- by IHC and 44%(76/174) were low ESR1 score. Seventeen percent ofHER2-E samples were called triple-negative. Of theclinically HER2+ group by IHC/CISH, approximatelytwo-thirds (69/113 = 61%) were HER2-E and one-thirdwere Luminal B (37/113= 33%) subtype (Figure 4). Usingthe qPCR cut-off for ERBB2 expression, we found that98% (609/624) of samples that were low ERBB2 were alsoHER2- by IHC/CISH, while 53% (109/190) of tumors withFigure 2 ESR1 score cut-offs using training set and the GEICAM/9906 testing set. The ESR1 score is provided as a qualitative call of high,intermediate, or low. The cut-offs were based on the continuous expression of ESR1 across prototype samples in the training set. Each circle on thebox plot represents an individual sample that is color coded according to IHC status. The cut-points between high, intermediate, and low classes wereindividually derived from the training set samples (A). Data from ESR1 gene expression over the GEICAM 9906 samples (B) are plotted on the samescale as the training set. Samples are colored according to ER IHC positivity (red) or negativity (blue) determined at a central facility.False positive rateTrue positive rate0.0 0.2 0.4 0.6 0.8 0.90Sensitivity: 0.84Specificity: 0.85B)False positive rateTrue positive rate0.0 0.2 0.4 0.6 0.8 0.90Sensitivity: 0.96Specificity: 0.74A)False positive rateTrue positi ve r ate0.0 0.2 0.4 0.6 0.8 0.95Sensitivity: 0.94Specificity: 0.85C)ER/ESR1 PR/PGR HER2/ERBB2Figure 3 Receiver Operator Characteristic (ROC) curves for ESR1, PGR, and ERBB2 for the GEICAM/9906 test set. ROC curves for theGEICAM/9906 test set were generated using the clinical IHC status (positive vs. negative) for ER, PR, and HER2/neu as compared to the continuousRT-qPCR data for ESR1, PGR, and ERBB2. The cut-points for sensitivity/specificity are based on the training set.Bastien et al. BMC Medical Genomics 2012, 5:44 Page 6 of 12http://www.biomedcentral.com/1755-8794/5/44intermediate-high ERBB2 expression were HER2+. How-ever, analyses just within the HER2-E subtype showed 71%(66/93) of tumors with high ERBB2 gene expression wereHER2+ by IHC/CISH.Ninety percent (63/70) of Basal-like tumors were ER-by IHC and 96% (67/70) were low ESR1 score. Further-more, 81% (57/70) of Basal-like tumors were triple nega-tive (ER-/PR-/HER2-) by IHC/CISH and 86% (60/70)were low in all 3 genes by qPCR (Table 5). Conversely,only 56% (57/101) and 67% (60/90) of triple negativesdefined by IHC/CISH or qPCR were Basal-like, respect-ively. There was no difference (p > 0.05) in ESR1, PGR orERBB2 expression by qPCR in Basal-like tumors, regard-less of being triple-negative or non-triple negative byIHC/CISH (Figure 5).Additional file 7 shows a comparison of unsupervisedhierarchical clustering with supervised subtype assign-ment and single marker scores for GEICAM/9906. Ingeneral, the supervised classification agreed with thesample associated dendrogram clusters. The side branchesof the dendrogram clusters are less correlated to othersamples and reflect the continuum in the biology, espe-cially between Luminal A, Luminal B and HER2-E sub-types. The HER2-E and Basal-like subtypes clusteraway from the Luminal tumors and have similar geneexpression profiles overall; however, standard IHC/CISH biomarkers poorly define these subtypes.Prognostic significance of gene expression versusstandard methods for ER and HER2 statusAlthough there was high agreement between IHC/CISHand RT-qPCR measurements for ER/ESR1 and HER2/ERBB2, we wanted to assess whether the two differentmethods provided equivalent prognostic information.When tested in a multivariate Cox model for overall sur-vival, only the RT-qPCR assignments were selected inthe final Cox model in the GEICAM/9906 test set(Table 6). When all patients with locally advanced breastcancer were stratified, regardless of chemotherapy regimen(FEC vs FEC-T), both classifications for assessing ER/ESR1and HER2/ERBB2 status were significantly associated withoutcome (Figure 6). Since endocrine therapy was based onER status determined by IHC, those ER + samples thatwere ESR1- (29/154= 19%) would have received adjuvanttamoxifen and conversely those patients with ER- tumorsthat were ESR1+ (45/660 =7%) would not have receivedtherapy. When separating outcome based on agreementand disagreement between the methods, we find thatwomen with ER+/ESR1+ tumors have similar outcomesto women with ER-/ESR1+ tumors, and women withER-/ESR1- tumors have similar outcomes to womenwith ER+/ESR1- tumors. This shows that the RT-qPCRassignment is more prognostic and accurate than IHCfor ER.DiscussionMany studies have tried to identify the gene expression-based “intrinsic” subtypes using a variety of methods forthe sake of simplicity, cost, and available technologies.Methods that can be used from formalin-fixed, paraffin-embedded tissues are optimal since this is how samplesare procured and archived in most pathology departments.Table 3 Histological scoring across PAM50 subtypesGrade ER PR HER2LumA G1-68 (25%) Neg-19 (7%) Neg-16 (6%) Neg-273 (99%)n=277 G2-142 (51%) Pos-258 (93%) Pos-261 (94%) Pos-4 (1%)G3-39 (14%)GX-28 (10%)LumB G1-25 (10%) Neg-22 (8%) Neg-68 (26%) Neg-224 (86%)n=261 G2-111 (43%) Pos-239 (92%) Pos-193 (74%) Pos-37 (14%)G3-111 (43%)GX-14 (5%)HER2-E G1-6 (3%) Neg-63 (36%) Neg-93 (53%) Neg-105 (60%)n=174 G2-65 (37%) Pos-111 (64%) Pos-81 (47%) Pos-69 (40%)G3-96 (55%)GX-7 (4%)Basal G1-0 (0%) Neg-63 (90%) Neg-62 (89%) Neg-67 (96%)n= 70 G2-4 (6%) Pos-7 (10%) Pos-8 (11%) Pos-3 (4%)G3-61 (87%)GX-5 (7%)Table 4 Single gene scores and proliferation acrossPAM50 subtypesProliferation ESR1^ PGR^ ERBB2^LumA Low-104 (38%) Neg-0 (0%) Neg-15 (5%) Neg-254(92%)n=277 Intermediate-116(42%)Pos-277(100%)Pos-262(95%)Pos-23 (8%)High-57 (11%)LumB Low-4 (2%) Neg-8 (3%) Neg-101(39%)Neg-194(74%)n=261 Intermediate-55(21%)Pos-253(97%)Pos-160(61%)Pos-67(26%)High-202 (77%)HER2-ELow-15 (9%) Neg-76(44%)Neg-112(64%)Neg-81(47%)n=174 Intermediate-60(34%)Pos-98 (56%) Pos-62(36%)Pos-93(53%)High-99 (57%)Basal Low-0 (0%) Neg-67(96%)Neg-67(96%)Neg-66(94%)n= 70 Intermediate-3 (4%) Pos-3 (4%) Pos-3 (4%) Pos-4 (6%)High-67 (96%)^ RT-qPCR status based on intermediate-high (positive) versus low (negative).Bastien et al. BMC Medical Genomics 2012, 5:44 Page 7 of 12http://www.biomedcentral.com/1755-8794/5/44The two preferred technologies for gene expression profil-ing from FFPE tissues are RT-qPCR [17,18] and Nano-string nCounter [29]. The nCounter system uses color-coded probes that bind directly to the RNA transcriptwithout reverse transcription and PCR amplification.While these methods have high agreement for genequantification, other methodologies may lead to differ-ent conclusions and treatment decisions. For instance,in the NCIC.CTG MA.12 clinical trial that randomizedpre-menopausal women with primary breast cancer totamoxifen versus placebo it was found that a panel 6IHC antibodies for subtyping was not prognostic butthe PAM50 RT-qPCR subtypes were prognostic [30]. Inanother randomized study (NCIC.CTG MA.5) thatassessed PAM50 subtype sensitivity to anthracycline-based chemotherapy, it was shown that the HER2-Esubtype received the most benefit, while women withBasal-like tumors had no benefit from this aggressivetreatment [31]. This study and the MA.5 trial found thatonly about two-thirds of clinically Her2+ tumors areclassified as HER2-E and about the same percent oftriple negatives are classified as Basal-like. Thus, only asubset of the IHC defined groups overlap with PAM50subtype classification, which may have ramifications forclinical trial findings and predicting therapy benefit.Receiver Operator Characteristic (ROC) curves arecommonly used in medicine to optimize the sensitivity/specificity of an assay depending on the purpose of thetest (i.e. screening, monitoring, prognosis, etc.) [32]. Inclinical pathology, ROC curves are often used to validatea new methodology against an existing “gold” standard.A major limitation to this approach is that cut-offs arethen determined by comparison to an often less thanperfect reference. We used an approach for selectingsingle (ESR1, PGR, ERBB2) and meta-gene (prolifera-tion) cut-offs that was based on the distribution ofexpression of these markers across the different sub-types. This method showed to be reproducible in anindependent test set.The ROC curves showed high agreement between RT-qPCR and the standard IHC biomarkers. ESR1 had highsensitivity although the cut-off for ER+ status was 10%positive staining nuclei, whereas the new recommendationfor determining ER status is 1% [33]. These borderlinecases for ER positivity may be better characterized by theoverall subtype biology. For ERBB2, there was highspecificity, which is optimal since confirmatory CISHor FISH would only be performed when it was un-certain if the gene was truly amplified [34]. It hasbeen suggested that the use of single gene RT-qPCRmeasurement for ERBB2 is insufficient for determin-ing HER2 positive samples that may benefit fromtrastuzumab/HerceptinW therapy [35]. Dabbs et al.Table 5 Surrogate subtyping by 3-marker scoringLuminal A Luminal B HER2-E Basal-likeER+/PR+/HER2- (n = 471) 244 (52%) 170 (36%) 53 (11%) 4 (1%)ESR1+/PGR+/ERBB2-(n = 397)239 (60%) 127 (32%) 31 (8%) 0 (0%)ER+/PR-/HER2- (n = 81) 12 (15%) 47 (58%) 19 (23%) 3 (4%)ESR1+/PGR-/ERBB2-(n = 101)15 (15%) 65 (64%) 18 (18%) 3 (3%)ER-/PR-/HER2+ (n = 39) 0 (0%) 7 (18%) 30 (77%) 2 (5%)ESR1-/PGR-/ERBB2+(n = 51)0 (0%) 6 (12%) 41 (80%) 4 (8%)ER-/PR-/HER2- (n = 101) 4 (4%) 10 (10%) 30 (30%) 57(56%)ESR1-/PGR-/ERBB2- (n = 90) 0 (0%) 2 (2%) 28 (31%) 60(67%)ER+/PR+/HER2+ (n = 45) 2 (4%) 18 (40%) 25 (56%) 0 (0%)ESR1+/PGR+/ERBB2+(n = 80)23 (29%) 33 (41%) 24 (30% 0 (0%)ER+/PR-/HER2+ (n = 18) 0 (0%) 4 (22%) 14 (78%) 0 (0%)ESR1+/PGR-/ERBB2+(n = 53)0 (0%) 28 (53%) 25 (47%) 0 (0%)ER-/PR+/HER2- (n = 25) 14 (56%) 4 (16%) 3 (12%) 4 (16%)ESR1-/PGR+/ERBB2- (n = 7) 0 (0%) 0 (0%) 4 (57%) 3 (43%)ER-/PR+/HER2+ (n = 2) 1 (50%) 1 (50%) 0 (0%) 0 (0%)ESR1-/PGR+/ERBB2+(n = 3)0 (0%) 0 (0%) 3 (100%) 0 (0%)Figure 4 Association between HER2 status and “intrinsic” subtype. Figure (A) shows the subtype distribution within HER2+ samples by IHC/CISH. Figure (B) shows the ER/HER2 status for samples only within the HER2-E subtype.Bastien et al. BMC Medical Genomics 2012, 5:44 Page 8 of 12http://www.biomedcentral.com/1755-8794/5/44found that the negative predictive value for determin-ing HER2/ERBB2 status was high between the Her-cepTest and the GHI Oncotype Dx qPCR assay(99%); but the concordance for positive HER2/ERBB2samples was only 28%. In contrast, we showed thatthe concordance between HER2 (IHC/CISH) andERBB2 (RT-qPCR) is greater than 90% whenrestricted to the HER2-E subtype.In order to determine if there was a prognostic dif-ference between the RT-qPCR and IHC we includedboth methods in a Cox proportional hazards modeland showed that gene expression remained significantin the multivariate analysis and replaced IHC. Fur-thermore, the outcome plots for women with tumorsscored positive for ER by IHC but negative for ESR1had outcomes similar to women that were ER-/ESR1-.Conversely, women with ER- tumors by IHC butpositive for ESR1 had similar outcomes to womenwith ER+/ESR1+ disease. Thus, despite the fact thatpatients were treated in favor of the IHC diagnosis(i.e. ER + disease was treated with adjuvant tamoxifen)the course of disease was in agreement with the geneexpression determination. The better prognosis seenin the ESR1+ but ER- subtype is curious since thesepatients would not have been given adjuvantendocrine blockade therapy. However, gene expressionfor ESR1 may be identifying the “true” luminal originof these tumors which have a better prognosis, re-gardless of therapy [30]. In addition, the patientsincluded in the test set were locally advanced andreceived chemotherapy that can cause chemotherapyinduced amenorrhea and a reduction in ovarian func-tion [36], which again may benefit the luminal sub-type most.The Normal subtype was developed from reductionmammoplasty “normal” breast tissue and serves as aquality control measure since these cases would be con-sidered to have an insufficient amount of tumor tissueto make a tumor subtype call. Interference studiesshowed that the introduction of “normal” breast tissueRNA caused a systematic shift in subtype assignmentwith subtypes switching to Normal, except Luminal Bwhich changed to Luminal A.None of the assignment switches occurred until theintroduction of 50% “normal” breast tissue RNA. Thegreatest risk of misclassification would come from Lu-minal B subtypes masquerading as Luminal A tumorsbecause of “normal” tissue contamination [28]; however,these tumors maintain a high proliferation score sug-gesting they are still a high risk Luminal tumor.A) B) C)0246810 p=0.24nonTNBC TNBC0246810 p=0.572nonTNBC TNBC0246810 p=0.168nonTNBC TNBCFigure 5 Relative transcript abundance for ESR1, PGR, and ERBB2 in the Basal-like subtype. There was no difference (p > 0.05) in (A) ESR1,(B) PGR, or (C) ERBB2 expression by qPCR in Basal-like tumors, regardless of being called triple-negative or non-triple negative by IHC/CISH.Table 6 Univariate and multivariate analyses of prognostic factors in GEICAM/9906MVA analysis for OS Univariate analysis Multivariate analysis (backward/forward stepwise selection)Signatures HR Lower 95% Upper 95% p-value HR Lower 95% Upper 95% p-valueARM FEC-P vs. FEC 0.708 0.528 0.948 0.021 0.734 0.543 0.993 0.045Grade 3 vs. 1-2 1.745 1.297 2.346 <0.001 1.335 0.962 1.853 0.084Nodes >3 vs. 1-3 2.103 1.574 2.808 <0.001 1.882 1.391 2.546 0.000Tumor size >2 cm vs. ≤2cm 2.089 1.510 2.890 <0.001 1.724 1.224 2.427 0.002Age >50 vs ≤50 1.189 0.890 1.589 0.242 1.012 0.999 1.026 0.078ER status by IHC+ vs. - 0.619 0.449 0.855 0.004 - - - -ER status by GE + vs. - 0.536 0.387 0.741 <0.001 0.630 0.438 0.906 0.013Clinical HER2 status + vs. - 1.389 0.9532 2.024 0.0872 - - - -Bastien et al. BMC Medical Genomics 2012, 5:44 Page 9 of 12http://www.biomedcentral.com/1755-8794/5/44A fifth tumor type that has often been referred toas “Normal-like” has been suggested to be an artifactof having too few tumor cells and a large backgroundof normal breast cells in the sample. Our mixingexperiments here support this hypothesis and showthat when increasing amounts of “normal” tissueRNA is added to a tumor it switches into theNormal-like group. It is, however, suspected thatsome tumors now called Normal-like may be put intothe recently described Claudin-low classification [37].The Claudin-low subtype is mostly triple-negative,shares biomarkers in common with normal breastepithelial cells and Basal-like tumors, and may becaused by deficiency in either BRCA1or p53, or both;however there is no clinical indication for Claudin-low, and most are typically classified as Basal-like.There are now many more groups of tumors beingidentified with transcriptome and copy number vari-ance analyses [38,39]. The overlap between these newgroups, existing subtypes, and standard biomarkersalready in practice should allow for more personalizedtreatments and better outcomes in the future.ConclusionsCompiling small biomarker panels for the purpose of“intrinsic” subtyping is of limited value in identifyingPAM50 based subtypes. Gene expression scoring forESR1 and ERBB2 has good agreement with the corre-sponding protein biomarkers (ER and HER2) and mayhave more prognostic power.Additional filesAdditional file 1: Clinical-pathological information associated withtraining set subtypes. Clinical-pathological information associated withthe 171 samples included in the training set.Additional file 2: Clinical-pathological information and PAM50 dataassociated with GEICAM/9906 test set. Clinical-pathologicalinformation and PAM50 RT-qPCR results associated with the 814 samplesincluded in the GEICAM9906 test set.Additional file 3: Additional materials and methods. Methods forplate manufacturing, PCR, calculation of log-expression ratios, PCR-efficiency, limits of detection, and limits of quantification are described(Additional files 8 and 9).Additional file 4: 10-fold cross validation of training set. Each genewas measured in triplicate per RT-qPCR run on the Roche LC480 and 2runs were performed for each of the 17 dilutions. The prototype samplesidentified by SigClust were split into 10 groups and nine of the 10groups were used to calculate new centroids for each of the 5 possiblesubtype assignments.Additional file 5: Interference in subtype call and single/meta-genescores from normal contamination. Interference by normal cellcontamination of subtype call and single and meta-gene classes is shown.The changes in subtype classification occurred in a systematic fashion withall subtypes switching to a classification of Normal/Insufficient, with theexception of Luminal B, which switched to Luminal A.Additional file 6: Single and meta-gene cutoffs. Data from single andmeta-gene expression score over the GEICAM 9906 samples are plottedon the 1–10 scale. The cut-points between high, intermediate, and lowclasses were individually derived from the training set. Samples are color-coded according to immunohistochemistry positivity (red) or negativity(blue), except in the case of the training set proliferation score wheresamples are colored by high, intermediate or low proliferation class.Luminal score samples are colored as being ER+/PR+, ER + orPR + (positive, red), and ER-/PR- (negative, blue).Additional file 7: Hierarchical clustering for GEICAM 9906. Acomparison of unsupervised hierarchical clustering with supervisedsubtype assignment and single marker scores for GEICAM 9906.0 2 4 6 8 2 4 6 8 yearsGEICAM 9906 Overall SurvivalESR1-/ERBB2-ESR1-/ERBB2+ESR1+/ERBB2+ESR1+/ERBB2-RT-qPCR IHCOverall Survival ProportionOverall Survival ProportionLog Rank p=0.00247Log Rank p=0.001490 2 4 6 8 versus IHCOverall Survival ProportionLog Rank p=0.00173ESR1-/ER-ESR1-/ER+ESR1+/ER-ESR1+/ER+A) B) C)ER-/Her2-ER-/Her2+ER+/Her2+ER+/Her2-Figure 6 Kaplan-Meier plots of overall survival in GEICAM 9906 data set. When stratifying all patients with locally advanced breast cancer,regardless of chemotherapy regimen (FEC vs FEC-T), both RT-qPCR (A) and IHC/CISH (B) molecular classifications for assessing ESR1/ER and ERBB2/Her2 status were significant. However, the separation of the survival curves suggests that ER-status as assessed by qPCR has prognostic superiorityto IHC (C).Bastien et al. BMC Medical Genomics 2012, 5:44 Page 10 of 12http://www.biomedcentral.com/1755-8794/5/44Additional file 8: PCR Efficiency, limits of detection, and limits ofquantification. Supplemental table listing the efficiency of PCR, limits ofdetection, and limits of quantification for the 50 classifier and 5housekeeper genes of the PAM50. Data are from 34 runs across 17dilutions from a mixture of 8 breast cancer cell lines.Additional file 9: Reproducibility of PAM50 gene measurements.Within plate, within plate batch and across plate batch coefficient ofvariation for the 50 classifier and 5 housekeeper genes of the PAM50were calculated using cell lines and a tumor samples.Competing interestsARUP Laboratories Inc. has a financial interest in the commercial offering ofthe subject matter. PSB receives research funding from the ARUP Institute forClinical and Experimental Pathology, although he is not an employee ofARUP. PSB, CMP, and MJE have equity interest in Bioclassifier LLC, which hassublicensed the PAM50 signature from the University of Utah.Authors’ contributionsRRLB participated in design of the study, generating data, and drafting themanuscript. ARL participated in recruiting patients, collecting samples andclinical data, and reviewing the manuscript. MTWE participated in design ofthe study, bioinformatics and statistical analysis, and drafting the manuscript.AP participated in bioinformatics and statistical analysis. BM participated inrecruiting patients, collecting samples and clinical data, and reviewing themanuscript. LR participated in design of the study. PM participated in designof the study. MRB participated in recruiting patients, collecting samples andclinical data, and reviewing the manuscript. DA participated in design of thestudy. BL participated in manufacturing of the PCR plates. IA performed IHQ/CISH on GEICAM/9906 samples and participated in reviewing the manuscript.TD participated in manufacturing of the PCR plates. DW participated inmanufacturing of the PCR plates. MAS participated in recruiting patients,collecting samples and clinical data, and reviewing the manuscript. LBparticipated in manufacturing of the PCR plates. KMB participated instatistical analysis. EA participated in recruiting patients, collecting samplesand clinical data, and reviewing the manuscript. LP participated in statisticalanalysis. CAD participated in sample preparation and organization. IAparticipated in recruiting patients, collecting samples and clinical data, andreviewing the manuscript. CF participated in robotics design. IJS participatedin data generation and reviewing the manuscript. JP participated inrecruiting patients, collecting samples and clinical data, and reviewing themanuscript. AA participated in recruiting patients, collecting samples andclinical data, and reviewing the manuscript. EC participated in design of thestudy and reviewing the manuscript. RC participated in design of the study,managing the collection of samples and central laboratory activity, andreviewing the manuscript. MJE participated in reviewing the manuscript.TON participated in reviewing the manuscript. CMP participated in collectingsamples and clinical data, and reviewing the manuscript. MA participated inthe conceiving and design of the study. PSB participated in conceiving anddesign of the study, and drafting the manuscript. MM participated in designof the study, recruiting patients, collecting samples and clinical data, andreviewing the manuscript. All authors read and approved the finalmanuscript.AcknowledgementsThis work was supported by the Huntsman Cancer Institute (HCI)/Foundation, the ARUP Institute for Clinical and Experimental Pathology, andNCI grants U01 CA114722-01 and P30 CA42014-19. A. Prat is supported bythe Translational Oncology fellowship of the Sociedad Española deOncologia Médica (SEOM). We thank the TRAC facility and ResearchInformatics at HCI, and the UNC Tissue Procurement Facility for contributingsamples. We also appreciate the guidance of Dr. Joel S. Parker.Author details1The ARUP Institute for Clinical and Experimental Pathology, Salt Lake City,UT, USA. 2Department of Medical Oncology, Hospital Universitario de Elche,Elche, Spain. 3Lineberger Comprehensive Cancer Center and Department ofGenetics and Department of Pathology & Laboratory Medicine, University ofNorth Carolina at Chapel Hill, Chapel Hill, NC, USA. 4Department of Medicine,Universitat Autónoma de Barcelona, Barcelona, Spain. 5Department ofMedical Oncology, Hospital Universitario La Fe, Valencia, Spain. 6Departmentof Medical Oncology, Hospital Universitario Virgen del Rocío, Sevilla, Spain.7Department of Medical Oncology, Hospital de Donostia, San Sebastián,Spain. 8Department of Medical Oncology, Corporatiò Sanitaria Parc Taulí,Sabadell, Spain. 9Department of Oncological Sciences, Huntsman CancerInstitute, Salt Lake City, UT, USA. 10Department of Medical Oncology, HospitalUniversitario Virgen de la Victoria, Málaga, Spain. 11Department of Pathology,University of Utah Health Sciences Center/Huntsman Cancer Institute, SaltLake City, UT, USA. 12Department of Pathology, Hospital General Universitariode Alicante, Alicante, Spain. 13Department of Pathology, Hospital Virgen delRocio, Sevilla, Spain. 14Department of Medical Oncology, HospitalUniversitario Miguel Servet, Zaragoza, Spain. 15Spanish Breast CancerResearch Group, GEICAM, Madrid, Spain. 16Department of Oncology,Washington University, St. Louis, MO, USA. 17Department of AnatomicalPathology, University of British Columbia, Vancouver, Canada. 18Departmentof Medical Oncology, Hospital General Universitario Gregorio Marañón,Universidad Complutense, Madrid, Spain. 19Huntsman Cancer Institute, 2000Circle of Hope, Salt Lake City, UT 84112, USA.Received: 6 April 2012 Accepted: 31 August 2012Published: 4 October 2012References1. Perou CM, Sorlie T, Eisen MB, van de Rijn M, Jeffrey SS, Rees CA, Pollack JR,Ross DT, Johnsen H, Akslen LA, et al: Molecular portraits of human breasttumours. Nature 2000, 406(6797):747–752.2. Sorlie T, Perou CM, Tibshirani R, Aas T, Geisler S, Johnsen H, Hastie T, EisenMB, van de Rijn M, Jeffrey SS, et al: Gene expression patterns of breastcarcinomas distinguish tumor subclasses with clinical implications. ProcNatl Acad Sci U S A 2001, 98(19):10869–10874.3. Parker JS, Mullins M, Cheang MC, Leung S, Voduc D, Vickery T, Davies S,Fauron C, He X, Hu Z, et al: Supervised risk predictor of breast cancerbased on intrinsic subtypes. J Clin Oncol 2009, 27(8):1160–1167.4. Hu Z, Fan C, Oh DS, Marron JS, He X, Qaqish BF, Livasy C, Carey LA,Reynolds E, Dressler L, et al: The molecular portraits of breast tumors areconserved across microarray platforms. BMC Genomics 2006, 7:96.5. Nielsen TO, Parker JS, Leung S, Voduc D, Ebbert M, Vickery T, Davies SR,Snider J, Stijleman IJ, Reed J, et al: A comparison of PAM50 intrinsicsubtyping with immunohistochemistry and clinical prognostic factors intamoxifen-treated estrogen receptor-positive breast cancer. Clin CancerRes 2010, 16(21):5222–5232.6. Cheang MCU, Voduc KD, Dongsheng T, Jiang S, Leung S, Chia SK, ShepherdL, Levine MN, Pritchard KI, Davies SR, et al: Responsiveness of intrinsicsubtypes to adjuvant anthracycline substitution in the NCIC.CTG MA.5randomized trial. Clin Cancer Res in press.7. Tibshirani R, Hastie T, Narasimhan B, Chu G: Diagnosis of multiple cancertypes by shrunken centroids of gene expression. Proc Natl Acad Sci U S A2002, 99(10):6567–6572.8. Sorlie T, Tibshirani R, Parker J, Hastie T, Marron JS, Nobel A, Deng S, JohnsenH, Pesich R, Geisler S, et al: Repeated observation of breast tumorsubtypes in independent gene expression data sets. Proc Natl Acad Sci US A 2003, 100(14):8418–8423.9. Ellis MJ, Suman VJ, Hoog J, Lin L, Snider J, Prat A, Parker JS, Luo J,DeSchryver K, Allred DC, et al: Randomized phase II neoadjuvantcomparison between letrozole, anastrozole, and exemestane forpostmenopausal women with estrogen receptor-rich stage 2 to 3 breastcancer: clinical and biomarker outcomes and predictive value of thebaseline PAM50-based intrinsic subtype--ACOSOG Z103. J Clin Oncol2011, 29(17):2342–2349.10. Esserman LJ, Berry DA, Cheang MC, Yau C, Perou CM, Carey L, DeMichele A,Gray JW, Conway-Dorsey K, Lenburg ME, et al: Chemotherapy responseand recurrence-free survival in neoadjuvant breast cancer depends onbiomarker profiles: results from the I-SPY 1 TRIAL (CALGB 150007/150012; ACRIN 6657). Breast Cancer Res Treat 2012, 132(3):1049–1062.11. Gonzalez-Angulo AM, Iwamoto T, Liu S, Chen H, Do KA, Hortobagyi GN,Mills GB, Meric-Bernstam F, Symmans WF, Pusztai L: Gene expression,molecular class changes, and pathway analysis after neoadjuvantsystemic therapy for breast cancer. Clin Cancer Res 2012, 18(4):1109–1119.12. Iwamoto T, Booser D, Valero V, Murray JL, Koenig K, Esteva FJ, Ueno NT,Zhang J, Shi W, Qi Y, et al: Estrogen receptor (ER) mRNA and ER-relatedgene expression in breast cancers that are 1% to 10% ER-positive byimmunohistochemistry. J Clin Oncol 2012, 30(7):729–734.Bastien et al. BMC Medical Genomics 2012, 5:44 Page 11 of 12http://www.biomedcentral.com/1755-8794/5/4413. Paik S, Shak S, Tang G, Kim C, Baker J, Cronin M, Baehner FL, Walker MG,Watson D, Park T, et al: A multigene assay to predict recurrence oftamoxifen-treated, node-negative breast cancer. N Engl J Med 2004, 351(27):2817–2826.14. Tutt A, Wang A, Rowland C, Gillett C, Lau K, Chew K, Dai H, Kwok S, Ryder K,Shu H, et al: Risk estimation of distant metastasis in node-negative,estrogen receptor-positive breast cancer patients using an RT-PCR basedprognostic expression signature. BMC Cancer 2008, 8:339.15. Sotiriou C, Wirapati P, Loi S, Harris A, Fox S, Smeds J, Nordgren H, Farmer P,Praz V, Haibe-Kains B, et al: Gene expression profiling in breast cancer:understanding the molecular basis of histologic grade to improveprognosis. J Natl Cancer Inst 2006, 98(4):262–272.16. van de Vijver MJ, He YD, van't Veer LJ, Dai H, Hart AA, Voskuil DW, Schreiber GJ,Peterse JL, Roberts C, Marton MJ, et al: A gene-expression signature as apredictor of survival in breast cancer. N Engl J Med 2002, 347(25):1999–2009.17. Perreard L, Fan C, Quackenbush JF, Mullins M, Gauthier NP, Nelson E, MoneM, Hansen H, Buys SS, Rasmussen K, et al: Classification and riskstratification of invasive breast carcinomas using a real-time quantitativeRT-PCR assay. Breast Cancer Res 2006, 8(2):R23.18. Mullins M, Perreard L, Quackenbush JF, Gauthier N, Bayer S, Ellis M, Parker J,Perou CM, Szabo A, Bernard PS: Agreement in breast cancer classificationbetween microarray and quantitative reverse transcription PCR fromfresh-frozen and formalin-fixed, paraffin-embedded tissues. Clin Chem2007, 53(7):1273–1279.19. Nielsen TO, Hsu FD, Jensen K, Cheang M, Karaca G, Hu Z, Hernandez-Boussard T, Livasy C, Cowan D, Dressler L, et al: Immunohistochemical andclinical characterization of the basal-like subtype of invasive breastcarcinoma. Clin Cancer Res 2004, 10(16):5367–5374.20. Morrison DH, Rahardja D, King E, Peng Y, Sarode VR: Tumour biomarkerexpression relative to age and molecular subtypes of invasive breastcancer. Br J Cancer 2012.21. Caudle AS, Yu TK, Tucker SL, Bedrosian I, Litton JK, Gonzalez-Angulo AM,Hoffman K, Meric-Bernstam F, Hunt KK, Buchholz TA, et al: Local-regionalcontrol according to surrogate markers of breast cancer subtypes andresponse to neoadjuvant chemotherapy in breast cancer patientsundergoing breast conserving therapy. Breast Cancer Res 2012, 14(3):R83.22. Martin M, Rodriguez-Lescure A, Ruiz A, Alba E, Calvo L, Ruiz-Borrego M,Munarriz B, Rodriguez CA, Crespo C, de Alava E, et al: Randomizedphase 3 trial of fluorouracil, epirubicin, and cyclophosphamide aloneor followed by Paclitaxel for early breast cancer. J Natl Cancer Inst2008, 100(11):805–814.23. Allred DC, Harvey JM, Berardo M, Clark GM: Prognostic and predictivefactors in breast cancer by immunohistochemical analysis. Modern Pathol1998, 11(2):155–168.24. Wolff AC, Hammond ME, Schwartz JN, Hagerty KL, Allred DC, Cote RJ,Dowsett M, Fitzgibbons PL, Hanna WM, Langer A, et al: American Societyof Clinical Oncology/College of American Pathologists guidelinerecommendations for human epidermal growth factor receptor 2 testingin breast cancer. J Clin Oncol 2007, 25(1):118–145.25. Eisen MB, Spellman PT, Brown PO, Botstein D: Cluster analysis and displayof genome-wide expression patterns. Proc Natl Acad Sci U S A 1998, 95(25):14863–14868.26. Ebbert MT, Bastien RR, Boucher KM, Martin M, Carrasco E, Caballero R,Stijleman IJ, Bernard PS, Facelli JC: Characterization of uncertainty in theclassification of multivariate assays: application to PAM50 centroid-basedgenomic predictors for breast cancer treatment plans. J Clin Bioinforma2011, 1:37.27. Liu Y, Hayes DN, Nobel A, Marron JS: Statistical Significance of Clusteringfor High-Dimension, Low-Sample Size Data. J Am Stat Assoc 2008, 103(483):1281–1293.28. Elloumi F, Hu Z, Li Y, Parker JS, Gulley ML, Amos KD, Troester MA:Systematic Bias in Genomic Classification Due to Contaminating Non-neoplastic Tissue in Breast Tumor Samples. BMC Med Genomics 2011, 4:54.29. Geiss GK, Bumgarner RE, Birditt B, Dahl T, Dowidar N, Dunaway DL, Fell HP,Ferree S, George RD, Grogan T, et al: Direct multiplexed measurement ofgene expression with color-coded probe pairs. Nat Biotechnol 2008, 26(3):317–325.30. Chia SK, Bramwell VH, Tu D, Shepherd LE, Jiang S, Vickery T, Mardis E, LeungS, Ung K, Pritchard KI, et al: A 50-Gene Intrinsic Subtype Classifier forPrognosis and Prediction of Benefit from Adjuvant Tamoxifen. ClinCancer Res 2012, 18(16):4465–4472.31. Cheang MC, Voduc KD, Tu D, Jiang S, Leung S, Chia SK, Shepherd LE, LevineMN, Pritchard KI, Davies S, et al: Responsiveness of intrinsic subtypes toadjuvant anthracycline substitution in the NCIC.CTG MA.5 randomizedtrial. Clin Cancer Res 2012, 18(8):2402–2412.32. Soreide K: Receiver-operating characteristic curve analysis indiagnostic, prognostic and predictive biomarker research. J ClinPathol 2009, 62(1):1–5.33. Hammond ME, Hayes DF, Dowsett M, Allred DC, Hagerty KL, Badve S,Fitzgibbons PL, Francis G, Goldstein NS, Hayes M, et al: AmericanSociety of Clinical Oncology/College Of American Pathologistsguideline recommendations for immunohistochemical testing ofestrogen and progesterone receptors in breast cancer. J Clin Oncol2010, 28(16):2784–2795.34. Garcia-Caballero T, Grabau D, Green AR, Gregory J, Schad A, Kohlwes E, EllisIO, Watts S, Mollerup J: Determination of HER2 amplification in primarybreast cancer using dual-colour chromogenic in situ hybridization iscomparable to fluorescence in situ hybridization: a European multicentrestudy involving 168 specimens. Histopathology 2010, 56(4):472–480.35. Dabbs DJ, Klein ME, Mohsin SK, Tubbs RR, Shuai Y, Bhargava R: High False-Negative Rate of HER2 Quantitative Reverse Transcription PolymeraseChain Reaction of the Oncotype DX Test: An Independent QualityAssurance Study. J Clin Oncol 2011, 29(32):4279–4285.36. Swain SM, Jeong JH, Geyer CE Jr, Costantino JP, Pajon ER, Fehrenbacher L,Atkins JN, Polikoff J, Vogel VG, Erban JK, et al: Longer therapy, iatrogenicamenorrhea, and survival in early breast cancer. N Engl J Med 2010, 362(22):2053–2065.37. Prat A, Parker JS, Karginova O, Fan C, Livasy C, Herschkowitz JI, He X, PerouCM: Phenotypic and molecular characterization of the claudin-lowintrinsic subtype of breast cancer. Breast Cancer Res 2010, 12(5):R68.38. Ha G, Roth A, Lai D, Bashashati A, Ding J, Goya R, Giuliany R, Rosner J,Oloumi A, Shumansky K, et al: Integrative analysis of genome-wideloss of heterozygosity and mono-allelic expression at nucleotideresolution reveals disrupted pathways in triple negative breastcancer. Genome Res 2012,: . Epub.39. Curtis C, Shah SP, Chin SF, Turashvili G, Rueda OM, Dunning MJ, Speed D,Lynch AG, Samarajiwa S, Yuan Y, et al: The genomic and transcriptomicarchitecture of 2,000 breast tumours reveals novel subgroups. Nature2012, 486(7403):346–352.doi:10.1186/1755-8794-5-44Cite this article as: Bastien et al.: PAM50 Breast Cancer Subtyping by RT-qPCRand Concordance with Standard Clinical Molecular Markers. BMC Medical Genomics 2012 5:44.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 redistributionSubmit your manuscript at www.biomedcentral.com/submitBastien et al. BMC Medical Genomics 2012, 5:44 Page 12 of 12http://www.biomedcentral.com/1755-8794/5/44


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



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"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items