WO2012097820A1

WO2012097820A1 - Method and assay for predicting long-term efficacy of tamoxifen treatment in estrogen receptor-positive breast cancer patients

Info

Publication number: WO2012097820A1
Application number: PCT/DK2012/050018
Authority: WO
Inventors: Henrik Ditzel; Maria LYNG; Anne-Vibeke LAENKHOLM
Original assignee: Syddansk Universitet; Region Syddanmark
Current assignee: Syddansk Universitet; Region Syddanmark
Priority date: 2011-01-20
Filing date: 2012-01-17
Publication date: 2012-07-26
Anticipated expiration: 2013-07-20

Abstract

Methods for prediction of resistance to anti-estrogen therapy in breast cancer patients with ER-positive breast tumor. The method concerns the use of quantitative measurements of the expression level of BCL2 and CDKN1A to determine i) the likelihood of a beneficial response to anti-estrogen therapeutic agents, such as tamoxifen; and ii) the potential magnitude of beneficial response to chemotherapy.

Description

Method and assay for predicting long-term efficacy of Tamoxifen treatment in estrogen receptor-positive breast cancer patients

FIELD OF THE INVENTION

The present invention relates to quantitative molecular indicators that can guide clinical decisions in breast cancer, such as estrogen receptor (ER)-positive breast cancer. In particular, the invention concerns certain genes, the varied expression of which indicates the likelihood of recurrence of surgically resected breast cancer in patients who are either treatment naive, or has been treated but to some medical concern cannot continue on the prescribed treatment, with a therapeutic agent in the neo-adjuvant, adjuvant or metastatic setting. I n addition, the invention concerns the use of quantitative measurement of the expression of certain genes to determine i) the likelihood of a beneficial response to anti- estrogen therapeutic agent, such as tamoxifen; and ii) the potential magnitude of beneficial response to chemotherapy.

BACKGROUND OF THE INVENTION

For patients with breast tumors expressing the estrogen receptor alpha protein (ER+) adjuvant anti-estrogen treatment with Tamoxifen significantly reduce the risk of recurrence and death in all age groups studied. In a meta-analysis of 37,000 women with breast cancer who took part in 55 trials of adjuvant Tamoxifen therapy, it was found that the proportional reduction in mortality was 26% after 10 years and the absolute improvements in 10-year overall survival were 10.9% in node-positive and 5.6% in node-negative patients Women with ER-negative disease exhibited no benefit from the treatment ². More recent drug development has lead to third generation aromatase inhibitors (Als) that have shown increased efficacy compared to Tamoxifen in post-menopausal women ^3"5. Tamoxifen remains the treatment modality for pre-menopausal breast cancer patients and patients resistant to Als. In addition, the various side-effects prevent some patients from receiving Als ^6"8. Furthermore, the majority of patients in many countries receive sequential treatment, e.g. a total of 5 years of endocrine treatment, half on Tamoxifen and half on an Al ⁹. Therefore, it is logical to continue the study of Tamoxifen as an adjuvant treatment. The expression of selected genes could provide important markers for predicting outcome in ER+ tumors. A few gene expression signatures have recently emerged that are associated with Tamoxifen response ^10"14. Furthermore, the rationale for using gene expression as a clinical tool is emphasized by the fact that two recently developed and validated assays are presently being employed as stratifiers in clinical trials, i.e. TAILORx and MINDACT ^15,16.

US 2006275844 A1 describes diagnostic markers of breast cancer treatment. A pair of molecular markers, TP53 (tumor protein 53) and BCL2 (B-cell lymphoma 2) are provided to predict the outcome in endocrine therapy (e.g. tamoxifen therapy) for breast cancer.

WO 2010002367 A1 describes diagnostic markers of breast cancer to predict response to adjuvant therapy. BCL2 and CDKN1 B (cyclin-dependent kinase inhibitor 1) are suggested from a list of proteomic markers for deducing a probability of response to e.g. tamoxifen treatment.

Meanwhile, the prior art does not disclose the specific pair of genetic markers, BCL2 and CDKN1A, used to predict the outcome of tamoxifen therapy in breast cancer patients

There is a great need for accurate, quantitative prognostic and predictive factors that can assist the practicing physician to make intelligent treatment choices, adapted to a particular patient's needs, based on well founded risk-benefit analysis.

SUMMARY OF THE INVENTION

Using quantitative PCR-based Low Density Array (LDA), gene signatures, marker genes, and methods were developed for predicting response or resistance to anti-estrogen, for example, Tamoxifen therapy and predicting outcome for breast cancer patients. Using a gene profile described herein, analysis of a patient's primary breast tumor against the gene profile is predictive of patient benefit of anti-estrogen, for example, Tamoxifen therapy for the treatment of recurring breast cancer.

Useful gene signatures for predicting outcome (response or resistance) and progression free survival in recurring breast cancer treated with anti-estrogen, for example, Tamoxifen therapy include the genes of the 2-gene, 8-gene, and 9-gene signatures shown in Figure 3. Gene expression levels can be determined using various known methods including nucleic acid hybridization in microarrays, nucleic acid amplification methods such as quantitative polymerase chain reaction (qPCR), and immunoassay of gene products, i.e. proteins of the genes of the predictive gene profile. Expression levels and expression level ratios of two or more genes of the predictive gene profile can be determined, for example, using real-time quantitative reverse-transcriptase PCR (qRT-PCR).

The gene signatures of the invention are useful in assays to predict response and/or outcome of anti-estrogen, for example, Tamoxifen therapy for breast cancer. In one embodiment, gene expression is analyzed in a primary breast tumor tissue sample and compared to the expressed gene signature determined from retrospective patient data as described in the Examples below. Specifically there is provided in a first aspect of the present invention a method of predicting resistance to anti-estrogen therapy in a breast cancer patient with an ER-positive breast tumor, comprising:

(i) obtaining a breast tumor tissue sample from said patient; and

(ii) determining the level of expression of the genes BCL2 and CDKN1A in said sample, wherein an altered, and preferably increased, level of expression relative to the level of expression of said genes in healthy tumor tissue is indicative of probable resistance to said endocrine therapy.

Preferably the anti-estrogen therapy is selected from the group consisting of tamoxifen, raloxifene, toremifene, fulvestrant, exemestane, letrozole or anastrozole. Most preferably the antiestrogen is tamoxifen.

Also there is provided a method of predicting resistance to anti-estrogen therapy in a breast cancer patient with an ER-positive breast tumor, comprising:

(i) obtaining a body fluid sample from said patient; and

(ii) immunologically detecting soluble gene products of BCL2 and CDKN1A in said sample, wherein soluble gene products of BCL2 and CDKN1A in said sample is indicative of probable resistance to said therapy. In an alternative embodiment there is provided a method of predicting resistance to an antiestrogen therapy in a breast cancer patient with an ER-positive breast tumor, comprising the use of a nucleic acid-based assay to detect BCL2 and CDKN1A preferably overexpressed in one or more tumors in said patient, wherein the detection of BCL2 and CDKN1A overexpression in said one or more tumors is indicative of probable resistance to said antiestrogen therapy.

The mentioned nucleic acid-based assay (or gene expression assay) comprises the use of PCR, RT-PCR, real-time PCR, or quantitative real-time RT-PCR. In still another embodiment there is provided a method of therapy selection for a breast cancer patient with an ER-positive breast tumor, comprising detecting the presence or absence of BCL2 and CDKN1A gene expression products in said tumor or in a metastasis of said tumor, wherein the presence of BCL2 and CDKN1A gene expression products in said tumor or in said metastasis is indicative of probable resistance to antiestrogen therapy, and wherein if BCL2 and CDKN1A gene expression product is present in said tumor or in said metastasis, said method further comprises administrating to said patient additional or alternative therapy to antiestrogen therapy. Preferably the antiestrogen therapy is tamoxifen.

ADDITIONAL EMBODIMENTS

In addition to the 2-gene signature the present invention also provides 8-gene (EMBODIMENT A) and 9-gene (EMBODIMENT B) signatures as shown in Figure 3.

EMBODIMENT A

1 a. A method of predicting resistance to anti-estrogen therapy in a breast cancer patient with an ER-positive breast tumor, comprising:

(i) obtaining a breast tumor tissue sample from said patient; and

(ii) determining the level of expression of the genes BCL2, PRKCE, PRKCE, EGFR, PRKCD, NRG1, NCOA1, ESR1, and IGF1R in said sample, wherein an altered, and preferably increased, level of expression relative to the level of expression of said genes in healthy tumor tissue is indicative of probable resistance to said endocrine therapy. 2a. The method of embodiment 1a, wherein said anti-estrogen therapy is selected from the group consisting of SERMs, SERDs and aromatase inhibitors, such as tamoxifen, raloxifene, toremifene, fulvestrant, exemestane, letrozole or anastrozole.

3a. The method of embodiment 2a, wherein said antiestrogen is tamoxifen.

4a. A method of predicting resistance to anti-estrogen therapy in a breast cancer patient with an ER-positive breast tumor, comprising:

(i) obtaining a body fluid sample from said patient; and

(ii) immunologically detecting soluble gene products of BCL2, PRKCE, EGFR, PRKCD, NRG1, NCOA 1, ESR1, and IGF1R in said sample, wherein soluble gene products of BCL2, PRKCE, EGFR, PRKCD, NRG1, NCOA1, ESR1, and IGF1R in said sample is indicative of probable resistance to said endocrine therapy.

5a. The method of embodiment 4a, wherein said anti-estrogen therapy is selected from the group consisting of SERMs, SERDs and aromatase inhibitors, such as tamoxifen, raloxifene, toremifene, fulvestrant, exemestane, letrozole or anastrozole. 6a. The method of embodiment 5a, wherein said antiestrogen is tamoxifen.

7a. A method of predicting resistance to an antiestrogen therapy in a breast cancer patient with an ER-positive breast tumor, comprising the use of a nucleic acid-based assay to detect BCL2, PRKCE, EGFR, PRKCD, NRG1, NCOA 1, ESR1, and IGF1R expression in one or more tumors in said patient, wherein the detection of BCL2, PRKCE, EGFR, PRKCD, NRG1, NCOA1, ESR1, and IGF1R preferably overexpressed relative to the level of expression of said genes in healthy tumor tissue is indicative of probable resistance to said antiestrogen therapy. 8a. The method of embodiment 7a, wherein said nucleic acid-based assay comprises the use of PCR, RT-PCR, real-time PCR, or quantitative real-time RT-PCR.

9a. A method of therapy selection for a breast cancer patient with an ER-positive breast tumor, comprising detecting the presence or absence of BCL2, PRKCE, EGFR, PRKCD, NRG1, NCOA1, ESR1, and IGF1R gene expression products in said tumor or in a metastasis of said tumor, wherein the presence of BCL2, PRKCE, EGFR, PRKCD, NRG1, NCOA 1, ESR1, and IGF1R gene expression products in said tumor or in said metastasis is indicative of probable resistance to antiestrogen therapy, and wherein if BCL2, PRKCE, EGFR, PRKCD, NRG1, NCOA1, ESR1, and IGF1R gene expression product is present in said tumor or in said metastasis, said method further comprises administrating to said patient additional or alternative therapy to antiestrogen therapy.

10a. The method of embodiment 9a, wherein said antiestrogen therapy is tamoxifen. 1 1a. A method of predicting resistance to anti-estrogen therapy in a breast cancer patient with an ER-positive breast tumor, comprising:

(i) obtaining a body fluid sample from said patient; and

(ii) immunologically detecting soluble gene products of BCL2, PRKCE, EGFR, PRKCD, NRG1, NCOA 1, ESR1, and IGF1R in said sample, wherein soluble gene products of BCL2, PRKCE, EGFR, PRKCD, NRG1, NCOA1, ESR1, and IGF1R in said sample is indicative of probable resistance to said therapy.

EMBODIMENT B

1 b. A method of predicting resistance to anti-estrogen therapy in a breast cancer patient with an ER-positive breast tumor, comprising:

(i) obtaining a breast tumor tissue sample from said patient; and

(ii) determining the level of expression of the genes BCL2, PRKCE, EGFR, AKT1, TFF1, NAT1, CKDN1A, BCAR3, and CGA in said sample, wherein an altered, and preferably increased, level of expression relative to the level of expression of said genes in healthy tumor tissue is indicative of probable resistance to said endocrine therapy.

2b. The method of embodiment 1 b, wherein said anti-estrogen therapy is selected from the group consisting of SERMs, SERDs and aromatase inhibitors, such as tamoxifen, raloxifene, toremifene, fulvestrant, exemestane, letrozole or anastrozole.

3b. The method of embodiment 2b, wherein said antiestrogen is tamoxifen. 4b. A method of predicting resistance to anti-estrogen therapy in a breast cancer patient with an ER-positive breast tumor, comprising:

(i) obtaining a body fluid sample from said patient; and

(ii) immunologically detecting soluble gene products of BCL2, PRKCE, EGFR, AKT1, TFF1, NAT1, CKDN1A, BCAR3, and CGA in said sample, wherein soluble gene products of BCL2,

PRKCE, EGFR, AKT1, TFF1, NAT1, CKDN1A, BCAR3, and CGA in said sample is indicative of probable resistance to said endocrine therapy.

5b. The method of embodiment 4b, wherein said anti-estrogen therapy is selected from the group consisting of SERMs, SERDs and aromatase inhibitors, such as tamoxifen, raloxifene, toremifene, fulvestrant, exemestane, letrozole or anastrozole.

6b. The method of embodiment 5b, wherein said antiestrogen is tamoxifen. 7b. A method of predicting resistance to an antiestrogen therapy in a breast cancer patient with an ER-positive breast tumor, comprising the use of a nucleic acid-based assay to detect BCL2, PRKCE, EGFR, AKT1, TFF1, NAT1, CKDN1A, BCAR3, and CGA expression in one or more tumors in said patient, wherein the detection of BCL2, PRKCE, EGFR, AKT1, TFF1, NAT1, CKDN1A, BCAR3, and CGA preferably overexpressed relative to the level of expression of said genes in healthy tumor tissue is indicative of probable resistance to said antiestrogen therapy.

8b. The method of embodiment 7b, wherein said nucleic acid-based assay comprises the use of PCR, RT-PCR, real-time PCR, or quantitative real-time RT-PCR.

9b. A method of therapy selection for a breast cancer patient with an ER-positive breast tumor, comprising detecting the presence or absence of BCL2, PRKCE, EGFR, AKT1, TFF1, NAT1, CKDN1A, BCAR3, and CGA gene expression products in said tumor or in a metastasis of said tumor, wherein the presence of BCL2, PRKCE, EGFR, AKT1, TFF1, NAT1, CKDN1A, BCAR3, and CGA gene expression products in said tumor or in said metastasis is indicative of probable resistance to antiestrogen therapy, and wherein if BCL2, PRKCE, EGFR, AKT1, TFF1, NAT1, CKDN1A, BCAR3, and CGA gene expression product is present in said tumor or in said metastasis, said method further comprises administrating to said patient additional or alternative therapy to antiestrogen therapy. 10b. The method of embodiment 9b, wherein said antiestrogen therapy is tamoxifen.

1 1 b. A method of predicting resistance to anti-estrogen therapy in a breast cancer patient with an ER-positive breast tumor, comprising:

(i) obtaining a body fluid sample from said patient; and

(ii) immunologically detecting soluble gene products of BCL2, PRKCE, EGFR, AKT1, TFF1, NAT1, CKDN1A, BCAR3, and CGA in said sample, wherein soluble gene products of BCL2, PRKCE, EGFR, AKT1, TFF1, NAT1, CKDN1A, BCAR3, and CGA in said sample is indicative of probable resistance to said therapy.

The following Example assesses the expression of a panel of genes that exhibited potential predictive utility in long-term Tamoxifen treatment using quantitative PCR. The patient samples analyzed comprised high-risk, post-menopausal, ER+ patients who had received adjuvant mono-therapy with Tamoxifen. Gene expression analysis revealed three profiles consisting of 2-, 8- and 9- genes, the predictive capability of which was evaluated in 4 previously published gene expression datasets. This independent validation provided data on a total of 503 breast cancer patients investigated on 7 microarray platforms with recurrence being the primary endpoint. Almost half exhibited accuracies of >70% regardless of the platform. Overall, the 2-gene profile (BCL2-CDKN1A) exhibited the strongest potential in both our quantitative PCR dataset and across the microarray platforms. The predictive value was further determined by comparing the ability of these genes to predict recurrence in an additional, previously-published, cohort consisting of Tamoxifen-treated (N=58, p = 0.015) and untreated patients (N=62, p = 0.25).

BRIEF DESCRIPTION OF THE DRAWINGS

Figure 1 shows A) The 2-, 8- and 9-gene profiles identified by various statistical analyses with potential predictive capability. B) AACt of the genes present in the 2-, 8- and 9-gene profiles. BCL2 overlap in all three, whereas CDKN1A is in the 2- and 9-gene profiles, and PRKCE and EGFR are in both the 8- and 9-gene profiles. A positive AACt_median value denotes that the expression of the gene is highest in the tumor sample from patients without recurrence, whereas negative values means the expression is higher in the tumor samples from patients with recurrence. Figure 2 shows joint distribution of the AACt values of BCL2 and CDNK1A. The diagonal line corresponds to the rule determined by conditional logistic regression. Pairs to the right of the line are correctly classified with respect to their outcome (recurrence/non-recurrence) (accuracy of 75%), whereas pairs left of the line are classified incorrectly.

Figure 3 shows the capabilities of the identified 2-, 8- and 9-gene profiles to predict recurrence in the independent gene expression data sets A) Summarized results of accuracy (%), along with sensitivity/specificity in parenthesis (both given as %), of the identified profiles to predict recurrence in the 6 independent gene expression studies. B-G) Dot-plots of the identified 2-gene profile (BCL2-CDKN1A) illustrating the probability of recurrence in the 6 independent gene expression datasets used for validation. The vertical line separates the cases, i.e. patients with recurrence (left of the line) from controls (right of the line). The horizontal line refers to the cut-point used, hence the upper left and lower right corners includes the correctly classified patients. B) GSE1378 C) GSE1379 D) GSE9893 E) GSE12093 F) GSE6532-GPL96 and G) GSE6532-GPL570.

Figure 4 shows Kaplan-Meier curves of recurrence-free survival according to model-based prediction of outcome using the 2-gene expression signature (BCL2-CDKN1A) for the independent gene expression dataset GSE2990. Grey line (top) indicates the good outcome signature, whereas the black line (bottom) indicates the poor outcome model. Only postmenopausal (>50 years) and ER+ were included in the analysis. Data was adjusted for clinical variables. A) Tamoxifen-treated patient samples B) Untreated patient samples.

DETAILED DESCRIPTION OF THE INVENTION

Definitions Unless defined otherwise, technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. For purposes of the present invention, the following terms are defined below.

Recurrence-free survival (RFS) refers to the time (in years) from primary surgery/diagnosis to the first local, regional, or distant recurrence. Distant recurrence-free survival (DFRS) refers to the time (in years) from primary surgery/diagnosis to the first anatomically distant recurrence. The calculation of these measures in practice may vary from study to study depending on the definition of events to be either censored or not considered. The term "microarray" refers to an ordered arrangement of hybridizable array elements, preferably polynucleotide probes, on a substrate.

The term "gene expression" describes the conversion of the DNA gene sequence information into transcribed RNA (the initial unspliced RNA transcript or the mature mRNA) or the encoded protein product. Gene expression can be monitored by measuring the levels of either the entire RNA or protein products of the gene or subsequences.

Prognostic factors are those variables related to the natural history of breast cancer, which influence the recurrence rates and outcome of patients once they have developed breast cancer. Clinical parameters that have been associated with a worse prognosis include, for example, lymph node involvement, increasing tumor size, and high grade tumors. Prognostic factors are frequently used to categorize patients into subgroups with different baseline relapse risks. In contrast, treatment predictive factors are variables related to the likelihood of an individual patient's beneficial response to a treatment, such as anti-estrogen or chemotherapy, independent of prognosis.

The term "prognosis" is used herein to refer to the likelihood of cancer-attributable death or cancer progression, including recurrence and metastatic spread of a neoplastic disease, such as breast cancer, during the natural history of the disease. Prognostic factors are those variables related to the natural history of a neoplastic diseases, such as breast cancer, which influence the recurrence rates and disease outcome once the patient developed the neoplastic disease, such as breast cancer. In this context, "natural outcome" means outcome in the absence of further treatment. For example, in the case of breast cancer, "natural outcome" means outcome following surgical resection of the tumor, in the absence of further treatment (such as, chemotherapy or radiation treatment). Prognostic factors are frequently used to categorize patients into subgroups with different baseline risks, such as baseline relapse risks.

The term "prediction" is used herein to refer to the likelihood that a patient will respond either favorably or unfavorably to a drug or set of drugs, and also the extent of those responses. Thus, treatment predictive factors are those variables related to the response of an individual patient to a specific treatment, independent of prognosis. The predictive methods of the present invention can be used clinically to make treatment decisions by choosing the most appropriate treatment modalities for any particular patient. The predictive methods of the present invention are valuable tools in predicting if a patient is likely to respond favorably to a treatment regimen, such as anti-estrogen therapy, such as Tamoxifen treatment alone or in combination with chemotherapy and/or radiation therapy.

The term "long-term" survival is used herein to refer to survival for at least 3 years, more preferably for at least 8 years, most preferably for at least 10 years following primary surgery/diagnosis or other treatment.

The term "node negative" cancer, such as "node negative" breast cancer, is used herein to refer to cancer that has not spread to lymph nodes.

EXAMPLE

Materials and methods

Selection of genes

The 59 candidate genes investigated were selected based on an extensive literature study using the PubMed database ¹⁷. with the search criteria: "breast cancer Tamoxifen", and either "resistance" (yielding 779 papers) or "prediction" (yielding 82 papers). Only papers reporting gene expression analysis of patient tumor samples were included (n=79). Papers identifying genes known to have prognostic implications, such as the proliferation marker Ki- 67, were excluded, and the 40 remaining papers served as the basis of the selected genes.

Patient material

All patients had received adjuvant Tamoxifen as mono-therapy and were extracted from Danish endocrine protocols of the Danish Breast Cancer Co-operative Group (DBCG) 89c and 99c ¹⁸ based on the existence of archival frozen tumor tissue stored at -80°C. Clinical information was obtained from DBCG. Inclusion: ER+ tumor, post-menopausal, treated with Tamoxifen for >3 months, tumor content >50% (haematoxylin- and eosin (HE)-stained cryosections).. Exclusion: bilateral breast cancer, recurrence <3 months of diagnosis, treatment with adjuvant chemotherapy/Als, secondary cancers (except for cancer cutis) or/and unavailable medical records. Recurrence was defined as a clinically-verified metastasis in the ipsilateral breast or distant organs. Follow-up was defined as time between diagnosis and date of last flow sheet for patients without recurrence, whereas patients with recurrence were censored at date of recurrence. The study was approved by the Ethical Committee of Funen and Vejle County (VF20040064), The Danish Data Protection Agency (2009-41-3928) and the DBCG.

Study design

Patients (IS 08) were collected nationwide). The criteria for matching patient pairs were based on the Nottingham prognostic index (NPI) ¹⁹. All 5 criteria were mandatory: 1) lymph node-negative or -positive sub-grouped as follows: 1 , 2, 3 or >4 metastatic axillary lymph nodes; 2) tumor size: ≤20 mm or >20 mm; 3) histological diagnosis: ductal or lobular invasive breast carcinomas; 4) malignancy grade: 1 , 2 or 3 (only graded for invasive ductal carcinomas); 5) duration of Tamoxifen: if ≤3 years; the treatment period could not differ by >6 months or, if both patients were treated >3 years, all treatment durations were acceptable within the matched pair. In addition, follow-up of the paired patient without recurrence had to be at least equal to the time-to-recurrence of the matched patient with recurrence. During review of the medical records, 6 patient-pairs were excluded. Clinical characteristics are listed in Table 1 and the data was collectively analyzed unless otherwise mentioned.

Table 1 : Characteristics of patients and their tumor included in the study.

a:Cut-off≥1 % staining of tumor cells.

Purification and evaluation of RNA

Total RNA was purified from a maximum of 35x10 μηι cryosections by Roche RNA isolation kits for tissue (MagNa Pure LC RNA isolation kit III tissue, Roche, Basel, Switzerland) using the MagNa Pure Robot (Roche). RNA concentration and purity was examined using the NanoDrop Spectrophotometer (Thermo Scientific, Wilmington, DE, USA). Samples were excluded from further analysis if the concentration was <10 ng/μ-. and/or if the purity ratio 260/280 was <1.8. The BioAnalyzer 2100 (Agilent Technologies, CA, USA) was used to evaluate samples from different centers. The average RNA integrity number (RIN) was 8.1 (range 6.4-9.5). cDNA synthesis

RNA (10 μΙ_) was reverse-transcribed to cDNA using random 9-mer oligonucleotide primers at 25 μΜ/reaction. RNA and primers were incubated for 5 min/70°C, placed on ice, and a reaction mixture of 1 mM dNTPs, 1 ΙΙηιί/μΙ_ RNase Inhibitor (Roche), 10 ΙΙηιί/μΙ_ Reverse Transcriptase (Invitrogen Life Technologies, Paisley, UK) and First Strand Buffer x5 (Invitrogen) was added. The material was incubated for 10 min/25°C, followed by 45 min/37°C, and finally 5 min/95°C. qPCR/LDA

TaqMan® Gene Expression Assays (Applied Biosystems (AB), Foster City, CA, USA) on Low Density Arrays (LDAs) were run for 2 min/50°C, 10 min/94.5°C, followed by 50 cycles of 30 sec/97°C and 1 min/59.7°C. All samples were run in triplicate on the ABI 7900HT system (AB). The genes listed in Table 1A were investigated in the first phase using a 63+1 LDA configuration (n=60). In the second phase (n=48), an LDA configuration of 31+1 was used to investigate the genes identified in the first phase (n=18) with p<0.15 (Wilcoxon signed rank sum test). The LDA configuration is pre-defined which left space for 9 additional genes. The resulting 27 genes investigated are marked with an asterisk in Table 1A.

Table 1A. Overview of the genes investigated in the first phase, n=59, and in the second phase, n=27 (marked with an asterisk (*)). The assay ID denotes manufacturers' (Applied Biosystems) internal reference ID. Furthermore, the cytoband location is provided. ^¥: optional reference genes (n=4).

Primary gene symbol as annotated Assay ID Location by manufacturer

CYP19A1 Hs00240671_m 15q21 .1

(N M_031226.1 , N M_000103.2) 1

* PRKCD Hs00178914_m 3p21 .31

(N M_212539.1 , N M_006254.3) 1

NPM3 (NM_006993.1) Hs00199625_m 10q24.31

1

NCOR2 (NM_006312.2) Hs00196955_m 12q24

1

EGR1 (NM_001964.2) Hs00152928_m 5q31 .1

1

TGFA (NM_003236.1) Hs00608187_m 2p13

1

YWHAZ (NM_003406.2) Hs00237047_m 8q23.1

1

* PAI1 (NM_000602.1) Hs00167155_m 7q21 .3.q2

1 2

MYC (NM_002467.3) Hs00153408_m 8q24.12- 1 .13

IGF1 (NM_000618.2) Hs00153126_m 12q22- 1 q23

CDKN1 B (NM_004064.2) Hs00153277_m 12p13.1 - 1 p12

PUM1 Hs00206469_m 1 p35.2

(NM_001020658.1 , NM_014676.2) 1

FOS (NM_005252.2) Hs00170630_m 14q24.3

1

NCOA3 Hs00180722_m 20q12

(NM_181659.1 ,NM_006534.2) 1

CGA (NM_000735.2) Hs00174938_m 6q12-q21

1

* NRG1 (NM_013956.1 , Hs00247624_m 8p21 -p12 NM_013957.1 , NM_013958.1 , 1

NM_013959.1 , NM_013961 .1 ,

NM_013962.1)

NCOR1 (NM_00631 1 .2) Hs00196920_m 17p1 1 .2

1

CCND1 (NM_053056.1) Hs00277039_m 1 1 q13

1

* EGFR (NM_005228.3) Hs00193306_m 7p12 1

TBP (NM_003194.3) Hs00427620_m 6q27

1

* IGF1 R (NM_000875.2) Hs00181385_m 15q26.3

1

ACTB (NM_001 101 .2) Hs99999903_m 7p15-p12

1

IL6 (NM_000600.1) Hs00174131_m 7p21

1

NRG2 (NM_004883.1 , Hs00171706_m 5q23-q33 NM_013981 .1 , NM_013982.1 , 1

NM_013983.1 , NM_013984.1 ,

NM_013985.1)

* NPM2 (NM_182795.1) Hs0040201 1_m 8p21 .3

1

* EGF (NM_001963.2) Hs00153181_m 4q25

1

* NCOA1 Hs00186661_m 2p23

(NM_147223.2, NM_147233.2, NM_0 1

03743.4)

* PGR (NM_000926.2) Hs00172183_m 1 1 q22- 1 q23

* PRKCE (NM_005400.2) Hs00178455_m 2p21

1

PRKCA (NM_002737.2) Hs00176973_m 17q22- 1 q23.2

* RARA (NM_000964.1) Hs00230907_m 17q21

1

* XBP1 (NM_005080.2) Hs00231936_m 22q12.1

1

* NAT1 (BC013732) Hs00377717_m 8p23.1 - 1 p21 .3

RPLPO Hs99999902_m 12q24.2

(NM_053275.3,NM_001002.3) 1

PRKAR1A Hs00267597_m 17q23-

(N M_212471 .1 , N M_212472.1 , N M_0 1 q24 02734.3)

IGF2 (NM_000612.2) Hs00171254_m 1 1 p15.5

1

CTSD (NM_001909.3) Hs00157205_m 1 1 p15.5

1

PLAUR Hs00182181_m 19q13

(NM_001005376.1 ,NM_001005377. 1

1 ,NM_002659.2) * ESR2 (ΝΜ_001437.1) Hs00230957_m 14q23.2

1

* ERBB4 (NM_005235.1) Hs00171783_m 2q33.3- 1 q34

* TNF (NM_000594.2) Hs00174128_m 6p21 .3

1

BCAR1 (NM_014567.2) Hs00183953_m 16q22- 1 q23

* ESR1 (NM_000125.2) Hs00174860_m 6q25.1

1

PLAU (NM_002658.2) Hs00170182_m 10q24

1

* HOXB13 (NM_006361 .4) Hs00197189_m 17q21 .2

1

* CDKN1A Hs00355782_m 6p21 .2

(NM_078467.1 ,NM_000389.2) 1

IRS1 (NM_005544.1) Hs00178563_m 2q36

1

* IL17RB Hs00218889_m 3p21 .1

(NM_018725.2, NM_172234.1) 1

BCAR3 (NM_003567.2) Hs00182488_m 1 p22.1

1

* AKT1 Hs00178289_m 14q32.32

(NM_001014431 .1 ,NM_001014432. 1

1 ,NM_005163.2)

* IRF1 (NM_002198.1) Hs00233698_m 5q31 .1

1

STS (NM_000351 .3) Hs00165853_m Xp22.32

1

AREG (NM_001657.2) Hs00155832_m 4q13-q21

1

CCNE1 Hs00233356_m 19q12

(N M_001238.1 ,NM_057182.1) 1

TGFB1 (NM_000660.3) Hs99999918_m 19q13.2- 1 .1

* AKT2 (NM_001626.2) Hs00609846_m 19q13.1 - 1 .2

* TFF1 (NM_003225.2) Hs00170216_m 21 q22.3

1

* ERBB3 Hs00176538_m 12q13

(NM_001005915.1 ,NM_001982.2) 1

TNFRSF1A (NM_001065.2) Hs01042313_m 12p13.2

1

* ERBB2 Hs00170433_m 17q21 .1 (N M_001005862.1 , N M_004448.2) 1

* BCL2 (NM_000633.2) Hs00608023_m 18q21 .33

1

BCAS3 (NM_017679.2) Hs00375126_m 17q23

1

DUSP6 (NM_001946.2) Hs00169257_m 12q22- 1 q23

Table 2: The most significant genes exhibiting altered expression in the recurrent vs. non-recurrent patient samples identified using single gene analysis, p<0.1 . The genes are ranked by the p-value of the Wilcoxon signed rank test (p_wil).

Data preparation

Q-PCR raw data were analyzed by SDS vers. 2.2 (AB). Criteria for objective removal of outliers: Ct<30: replicates must be within 0.5 Ct of each other, 30≤Ct≤33: replicates must be within 1.0 Ct of each other and 33≤Ct<37: all replicates were included. The Ct value for each target gene was determined by averaging the replicates. Measurements above Ct=37 were regarded as immeasurable. Target gene Ct-values were normalized to the average of 4 reference genes previously identified (TBP, RPLP0, PUM1 and ACTS) ²⁰, thereby obtaining the ACt Ct_ref,avg - Ct_target)- Alternate, at least two of the reference genes could be averaged using the before-mentioned reference genes, e.g. omitting ACTB, yielding TBP, RPLP0, and PUM1. The difference in gene expression for a given target gene between the matched patient-pairs: AACt = ACt_reCurrent - ACt_n0n-recurrent- In case one of the patients' ACt values were immeasurable, the AACt was computed using a value of 40 for the missing value. If the ACt value was immeasurable for both patients in a pair, no AACt value was computed. The primary endpoint for all statistical analyses was time from primary surgery to recurrence.

Statistics - real-time PCR data

The Wilcoxon singed rank sum test was used to assess the differential gene expression between patients and the overall significance considered Bonferroni corrected p-values.

To determine optimal pairs, triples and quadruples of genes a model building procedure based on conditional logistic regression was used. This yielded an equation to be used for estimating the StratifyER+ score for the 2-gene profile of BCL2-CDKN1A (StratifyER+ score = A* delta-Ct(CDKNIA) + B* delta-Ct(BCL2), where A and B are numerical values or equations yielding a numerical value). For the optimal model, we determined the rate of correct classification, i.e. accuracy, using cross validation, leave-one-pair-out. Additionally, a non-parametric bootstrap, based on 1000 x re-sampling of the pairs, was used to determine the stability. The 27 genes analyzed for all 54 patient-pairs were used in model building, cross validation and the bootstrap method. In addition, the genes were subjected to modified microarray-based statistics (Statistical Analysis of Microarray (SAM)) ²³.

The HOXB13:IL17BR ratio was investigated analogous to previously reported ¹¹. Twelve pairs with undetermined values for HOXB13 in both patients were excluded, as this would lead to an estimate of the effect of IL17BR alone. All statistical computations were conducted in Stata vs.10.1 (StataCorp, TX, USA) unless otherwise mentioned.

Statistics - microarray datasets

Four previously-published microarray datasets ^11"14, investigating patient samples treated with Tamoxifen and with characteristics similar to ours were used for independent examination (see Table 3). A complete list of our three signatures relative to the microarray datasets are provided in Table 2A. One of the studies included data from 3 microarray platforms ¹² and another included data from 2 ¹¹ , giving a total of 7 platforms for validation. One of these platforms (GSE6532-GPL97) was missing 15/27 genes, and was excluded from further analysis. The identified genes from the qPCR analysis were annotated to the probe IDs using Gene Symbol (Table 2A). The average detection of probes was used. Each signature-gene selected by the modified SAM procedure ²¹ was submitted to SVM. Performance of the selected gene signatures was assessed by leave-one-pair-out validation to obtain the accuracy, sensitivity and specificity. In all analyses of recurrence data, the mean accuracy was calculated as (number of patients with recurrence predicted as recurrent patients + number of patients without recurrence predicted as non-recurrent patients)/total number of patients). The cut-off threshold distinguishing recurrent from nonrecurrent patients was calculated as the proportion of tumors from patients with recurrence in the total sample.

Upon examination of the performance of the signatures in the four previously-published datasets mentioned above, BCL2-CDKN1A was determined to be the most promising. This signature was then evaluated for prognostic vs. predictive capabilities using survival statistics in a different dataset (GSE2990 ²²). Patient samples from the dataset were included in the analysis if they were >50 years and had an ER+ tumor, resulting in the following tumor characteristics: untreated patients were all N- and 41 % of the tumors were ≥20 mm, while the Tamoxifen-treated patients were 57% N+ and 61 % had tumors≥20 mm. Kaplan-Meier curves are used to show their differential survival with a p-value from Chi- square test, constructed after adjusting for clinical variables. Table 2A. A complete list of the genes identified by real-time PCR and the presence of corresponding probes in previously-published microarray datasets. Since the 9-gene signature was identified by using the 59 initially analyzed genes, two of these nine genes (CGA and BCAR3) were not part of the reduced gene set of 27 genes. The platform GSE6532-GPL97 was excluded due to the unavailability of several genes (grey text). NA: probe for the gene was not available. +: at least one probe for the gene was available.

Results

Patient material

Tumors from patients with high-risk, post-menopausal, ER+ breast cancer treated only with adjuvant Tamoxifen ¹⁸ were investigated, all diagnosed before 2001. A nested matched case-control study design was used to increase the likelihood of identifying genes associated with outcome beyond the parameters used for matching. The median clinical follow-up was 4.8 years (range 0.7 - 10.4 years; censored at time of recurrence or at the last date of clinical verification without recurrence). The patient and tumor characteristics are provided in Table 1.

Analysis of single genes

The gene expression of a panel of 59 genes was examined by qPCR. Gene expression values were obtained for 98%. The remaining 2% could reasonably be assumed to be genes not expressed since the reference genes were adequately expressed. Comparative analysis of the genes according to altered expression across the recurrent vs. non-recurrent patient samples (Table 2) identified BCL2 as the most significant (p=0.0002), and it remained significant upon Bonferroni correction. Construction of gene combinations

Cross validation was used to determine accuracy and the optimal combination for predicting outcome was identified to be the two-genes BCL2-CDNK1A (accuracy=75%), whereas the optimal 3-gene combination consisted of BCL2-CDNK1 A-NAT1 (accuracy=55%). BCL2 alone had an accuracy of 70%. The accuracy of the best-performing combination of BCL2- CDNK1A is depicted in Figure 2. The statistical stability of BCL2-CDNK1A was found to be sub-optimal as the 2 genes were found to rank highest in only 24.3% of the 1000 bootstrap sampling.

A modified microarray-based statistical analysis based on a machine-learning procedure combining a modified SAM analysis and SVM, identifying a 9-gene signature. This 9-gene signature enabled correct classification of 73% of the patient-pairs with regards to recurrence. A list of the identified genes is provided in Figure 1 along with the direction of expression, which was observed to primarily be higher in the tumors of patients that had not developed recurrence. Analysis of the H0XB13: IL17BR ratio as predictive score

The HOXB13:IL17BR ratio has been reported to predict outcome in early breast cancer patients treated with Tamoxifen ¹¹ , thus we evaluated the predictive value of this ratio in our data set using the same approach as previously reported. As found by Ma et al. (2004) ¹¹ , HOXB13 showed higher expression in tumor samples from patients with recurrence, and IL17BR had higher expression in tumor samples from patients without recurrence. The HOXB13:IL17BR ratio correctly classified 64%, and the Wilcoxon signed rank sum test, applied to the ratio values, yielded a p-value of 0.02. The study by Ma et al (2004) was mainly based on early stage cancers with few tumor-infiltrated lymph nodes, whereas our patient population consisted mainly of patients with several tumor-infiltrated lymph nodes at time of diagnosis (average was 4, and only 3/54 pairs had 0 tumor-infiltrated lymph nodes). Indeed, the predictive value of the ratio was higher in the 21 pairs with a maximum of 3 affected lymph nodes (71 % correctly classified, p=0.03) compared to the 21 pairs with >3 affected lymph nodes (57%, p=0.29).

Validation in independent microarrav datasets of Tamoxifen-treated patient samples

The identified 2-, 8-, and 9-gene expression signatures were examined for predictive capabilities in 6 microarray datasets (Table 2A) from four previously published studies (Table 3) ^11"14.

Table 3: Summary of the four previously-published gene expression datasets examining sam from patients treated with adjuvant Tamoxifen and used for examination of our 3 gene signatures.

No. of GSE accession Reference

patients number

a: A total of 255 patient samples were analyzed by Loi et al. [19], but only 152

samples were included in our analysis. The remaining 103 patients were either not

coupled to clinical data, were not treated with Tamoxifen or analyzed using

platform GPL97 (one of three platforms used in this study), which was excluded

since 79% of the probes for the genes of interest were missing. Overall, the accuracies for the 2-, 8- and 9-gene signatures were high (Fig. 3A), especially for the 2-gene signature, BCL2-CDKN1A, with half of the platforms having an accuracy of >70% (Fig. 3B-G). One of the studies also investigated tumor tissue from matched patient material (GSE1379), similar to our study design. The 2-gene signature performed even better in this independent population, exhibiting 85% accuracy vs. 75% in the qPCR dataset. Furthermore, in this microarray dataset (GSE1379), which led to the identification of the HOXB13:IL17BR ratio ¹¹ , BCL2-CDKN1A challenged the HOXB13:IL17BR ratio, as BCL2- CDKN1A exhibited an accuracy of 85% vs. 81 % reported accuracy for the HOXB13:IL17BR ratio.

For the 2- and 8-gene signatures, the sensitivity was higher than the specificity across all studies except GSE9893. This was most pronounced for BCL2-CDKN1A in GSE6532- GPL96, which showed the highest sensitivity across all studies (93%). GSE9893 appeared to be quite different from the others in that it had low sensitivities, but fair specificities, for all signatures.

Prognostic vs. predictive value of the BCL2-CDKN1A gene expression signature

The above-mentioned independent evaluation pointed to the 2-gene signature as the most promising (Fig. 3A). We therefore conducted survival analysis of the 2 genes in a new, independent, previously-published microarray cohort that contained both treatment-naive (N=62) and Tamoxifen-treated (N=58) patient samples (GSE2990 ²²). Kaplan-Meier curves were obtained after adjusting for clinical variables. Figure 4 shows that the 2-gene signature could separate the Tamoxifen-treated patients with respect to probability of recurrence-free survival (p=0.015), while it did not significantly separate the outcome of the untreated patient samples (p=0.25).

Discussion

Despite Tamoxifen being an effective drug for many ER+ breast cancer patients in the adjuvant setting, about a third will experience recurrence. To address this issue, we investigated the expression of 59 genes in early, high-risk, ER+, post-menopausal, breast cancer patients receiving adjuvant Tamoxifen mono-therapy.

We employed an approach that should increase the likelihood of identifying genes with predictive capabilities from the onset i.e. a nested matched case-control study ^{11 23"25}. Matching of patients must be conducted with care to avoid confounding or overmatching, but done correctly, this design exaggerates the association being examined (i.e. recurrence despite treatment) ²⁴'²⁶'²⁷.

Single-gene analysis revealed 8 genes (p<0.05) that could distinguish patients according to outcome. Overall, the genes where found to have higher expression in patients without recurrence than those with (Fig. 1 B). BCL2 seemed the most promising single-gene (Table 2). CDNK1A had a non-significant p-value as single-gene (p=0.07) but, in combination with BCL2, constituted the optimal gene-pair, exhibiting an accuracy of 75%. We also analyzed our qPCR data using a modified SAM procedure and identified a 9-gene signature exhibiting an accuracy of 73%. .

Other studies have also found a correlation between low expression of BCL2 and lack of response to Tamoxifen ¹⁰'^28"31 , as found in our study. This association of BCL2 and favorable outcome may relate to the fact that BCL2 is an estrogen-regulated gene ³², thus indicative of an intact pathway driving tumor growth and thereby sensitive to Tamoxifen. The full potential and clarification of BCL2's prognostic and/or predictive value remains to be determined, although it is clearly a promising marker.

The other gene in the 2-gene signature was CDKN1A, cyclin-dependent kinase inhibitor 1A, which encodes the protein p2i^{WAF1 clp1} ; the increased expression of which also previously has been found to be specifically associated with outcome after Tamoxifen ³³. p2i^{WAF1 clp1} has also been reported to be absent in a clinical case of Tamoxifen-stimulated growth ³⁴. p2^<|WAFi/cipi in eracts with several cell cycle regulators, but the precise mechanism(s) behind its role in Tamoxifen resistance remains to be elucidated. The gene combination of BCL2-CDKN1A outperformed the 8- and 9-gene signatures, showing an increased ability to correctly classify patients with recurrence (sensitivity up to 93%) despite Tamoxifen treatment. The microarray studies investigated patients with varying numbers of tumor-infiltrated lymph nodes at the time of diagnosis. The GSE12093 study, which examined only patients with lymph node-negative tumors, had the poorest accuracy, as expected, since the tumors we used for signature identification were from patients with many tumor-infiltrated lymph nodes (average of 4). This finding underscores the fact that large variations, even within the same cancer sub-types, are important, thus it seems plausible that different biomarkers are needed for high- vs. low-risk patients.

Since only patients treated with Tamoxifen were used to develop the signatures in this study, it was not possible to unequivocally distinguish whether the identified genes encompassed prognostic or/and predictive properties. However, matching with known prognostic factors implied that the genes identified as being associated with outcome provide information beyond the factors used for matching. Moreover, our demonstration that the most promising signature consisting of BCL2-CDKN1A could not significantly separate untreated patient samples according to outcome, but significantly separated Tamoxifen- treated patient samples in an independent dataset, supports it being a predictive signature. It should be noted that the patients in the untreated dataset were all N- since ethical considerations preclude denying treatment to N+ patients (the compared treated dataset was 57% N+). Otherwise, the dataset is directly comparable to our study.

In summary, we identified a 2-gene signature, BCL2-CKDN1A, which was, upon evaluation in independent datasets, found to be a potentially strong predictor of outcome for high-risk ER+ breast cancer patients treated with Tamoxifen.

References

1. Early Breast Cancer Trialists' Collaborative Group. Tamoxifen for early breast cancer: an overview of the randomised trials. Lancet. 1998; 351 :1451-1467

2. Knoop AS, Bentzen SM, Nielsen MM et al. Value of epidermal growth factor receptor, HER2, p53, and steroid receptors in predicting the efficacy of tamoxifen in high-risk postmenopausal breast cancer patients. J Clin Oncol. 2001 ; 19:3376-3384

3. Baum M, Budzar AU, Cuzick J et al. Anastrozole alone or in combination with tamoxifen versus tamoxifen alone for adjuvant treatment of postmenopausal women with early breast cancer: first results of the ATAC randomised trial. Lancet. 2002; 359:2131-2139

4. Coates AS, Keshaviah A, Thurlimann B et al. Five years of letrozole compared with tamoxifen as initial adjuvant therapy for postmenopausal women with endocrine-responsive early breast cancer: update of study BIG 1-98. J Clin Oncol. 2007; 25:486-492 5. Mouridsen HT, Giobbie-Hurder A, Mauriac L et al. BIG 1-98: A randomized double-blind phase III study evaluating letrozole and tamoxifen given in sequence as adjuvant endocrine therapy for postmenopausal women with receptor-positive breast cancer. SABCS, abstract. 2008; 13:webpage:abstracts2view.com/sabcs/view.php?nu=SABCS08L_553

6. Wong ZW, Ellis MJ. First-line endocrine treatment of breast cancer: aromatase inhibitor or antioestrogen? Br J Cancer. 2004; 90:20-25

7. Lonning PE, Geisler J. Aromatase inhibitors-socioeconomical issues. J Steroid Biochem Mol Biol. 2007; 106:55-61

8. Bertoldo F, Pancheri S, Zenari S et al. Emerging drugs for the management of cancer treatment induced bone loss. Expert Opin Emerg Drugs. 2010; 15:323-342 9. van d, V, Verma S, van Nes JG et al. Switching from tamoxifen to aromatase inhibitors for adjuvant endocrine therapy in postmenopausal patients with early breast cancer. Cancer Treat Rev. 2010; 36:54-62

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

1 1. Ma XJ, Wang Z, Ryan PD et al. A two-gene expression ratio predicts clinical outcome in breast cancer patients treated with tamoxifen. Cancer Cell. 2004; 5:607-616

12. Loi S, Haibe-Kains B, Desmedt C et al. Predicting prognosis using molecular profiling in estrogen receptor-positive breast cancer treated with tamoxifen. BMC Genomics. 2008; 9:239

13. Chanrion M, Negre V, Fontaine H et al. A gene expression signature that can predict the recurrence of tamoxifen-treated primary breast cancer. Clin Cancer Res. 2008; 14:1744- 1752

14. Zhang Y, Sieuwerts AM, McGreevy M et al. The 76-gene signature defines high-risk patients that benefit from adjuvant tamoxifen therapy. Breast Cancer Res Treat. 2009;

1 16:303-309

15. Sparano JA, Paik S. Development of the 21 -gene assay and its application in clinical practice and clinical trials. J Clin Oncol. 2008; 26:721-728

16. Cardoso F, van't VL, Rutgers E et al. Clinical application of the 70-gene profile: the MINDACT trial. J Clin Oncol. 2008; 26:729-735

17. NCBI. PubMed. webpage: pubmed gov. 2005; 18. Moller S, Jensen MB, Ejlertsen B et al. The clinical database and the treatment guidelines of the Danish Breast Cancer Cooperative Group (DBCG); its 30-years experience and future promise. Acta Oncol. 2008; 47:506-524

19. Elston CW, Ellis IO. Pathological prognostic factors in breast cancer. I. The value of histological grade in breast cancer: experience from a large study with long-term follow-up. Histopathology. 1991 ; 19:403-410

20. Lyng MB, Laenkholm AV, Pallisgaard N et al. Identification of genes for normalization of real-time RT-PCR data in breast carcinomas. BMC Cancer. 2008; 8:20

21. Tan Q, Thomassen M, Kruse TA. Feature Selection for Predicting Tumor Metastases in Microarray Experiments using Paired Design. Cancer Inform. 2007; 3:213-218

22. Loi S, Haibe-Kains B, Desmedt C et al. Definition of clinically distinct molecular subtypes in estrogen receptor-positive breast carcinomas through genomic grade. J Clin Oncol. 2007; 25:1239-1246

23. Sturmer T, Brenner H. Potential gain in precision and power by matching on strong risk factors in case-control studies: the example of laryngeal cancer. J Epidemiol Biostat. 2000;

5: 125-131

24. Wacholder S, Silverman DT, McLaughlin JK et al. Selection of controls in case-control studies. III. Design options. Am J Epidemiol. 1992; 135:1042-1050

25. Thomassen M, Tan Q, Eiriksdottir F et al. Comparison of gene sets for expression profiling: prediction of metastasis from low-malignant breast cancer. Clin Cancer Res. 2007; 13:5355-5360

26. Yanagawa T. Designing case-control studies. Environ Health Perspect. 1979; 32:143- 156 27. Bland JM, Altman DG. Matching. BMJ. 1994; 309: 1128

28. Linke SP, Bremer TM, Herold CD et al. A multimarker model to predict outcome in tamoxifen-treated breast cancer patients. Clin Cancer Res. 2006; 12: 1175-1 183

29. Vendrell JA, Robertson KE, Ravel P et al. A candidate molecular signature associated with tamoxifen failure in primary breast cancer. Breast Cancer Res. 2008; 10:R88

30. Daidone MG, Luisi A, Martelli G et al. Biomarkers and outcome after tamoxifen treatment in node-positive breast cancers from elderly women. Br J Cancer. 2000; 82:270-277

31. Ciocca DR, Elledge R. Molecular markers for predicting response to tamoxifen in breast cancer patients. Endocrine. 2000; 13: 1-10 32. Perillo B, Sasso A, Abbondanza C et al. 17beta-estradiol inhibits apoptosis in MCF-7 cells, inducing bcl-2 expression via two estrogen-responsive elements present in the coding sequence. Mol Cell Biol. 2000; 20:2890-2901

33. Pellikainen MJ, Pekola TT, Ropponen KM et al. p21WAF1 expression in invasive breast cancer and its association with p53, AP-2, cell proliferation, and prognosis. J Clin Pathol. 2003; 56:214-220

34. Abukhdeir AM, Vitolo Ml, Argani P et al. Tamoxifen-stimulated growth of breast cancer due to p21 loss. Proc Natl Acad Sci U S A. 2008; 105:288-293

Claims

1. A method of predicting resistance to anti-estrogen therapy in a breast cancer patient with an ER-positive breast tumor, comprising: determining the level of expression of the genes BCL2 and CDKN1A in a breast tumor tissue sample from said patient, wherein an altered, and preferably increased, level of expression relative to the level of expression of said genes in healthy tumor tissue is indicative of probable resistance to said endocrine therapy, said method not comprising the invasive step of collecting the tissue sample from said patient.

2. The method of claim 1 , wherein said anti-estrogen therapy is selected from the group consisting of SERMs, SERDs and aromatase inhibitors, such as tamoxifen, raloxifene, toremifene, fulvestrant, exemestane, letrozole or anastrozole.

3. The method of claim 2, wherein said antiestrogen is tamoxifen.

4. A method of predicting resistance to anti-estrogen therapy in a breast cancer patient with an ER-positive breast tumor, comprising: immunologically detecting soluble gene products of BCL2 and CDKN1A in a body fluid sample from said patient, wherein soluble gene products of BCL2 and CDKN1A in said sample is indicative of probable resistance to said endocrine therapy, said method not comprising the invasive step of collecting the body fluid sample from said patient.

5. The method of claim 4, wherein said anti-estrogen therapy is selected from the group consisting of SERMs, SERDs and aromatase inhibitors, such as tamoxifen, raloxifene, toremifene, fulvestrant, exemestane, letrozole or anastrozole.

6. The method of claim 5, wherein said antiestrogen is tamoxifen.

7. A method of predicting resistance to an antiestrogen therapy in a breast cancer patient with an ER-positive breast tumor, comprising the use of a nucleic acid-based assay to detect BCL2 and CDKN1A expression in a tumor sample from said patient, wherein the detection of altered BCL2 and CDKN1A expression, preferably over expression, relative to the level of expression of said genes in healthy tumor tissue is indicative of probable resistance to said antiestrogen therapy, said method not comprising the invasive step of collecting the tumor sample from said patient.

8. The method of claim 7, wherein said nucleic acid-based assay comprises the use of PCR, RT-PCR, real-time PCR, or quantitative real-time RT-PCR.

9. A method of therapy selection for a breast cancer patient with an ER-positive breast tumor, comprising detecting the presence or absence of BCL2 and CDKN1A gene expression products in tissue from said tumor or in a metastasis of said tumor, wherein the presence of BCL2 and CDKN1A gene expression products in said tumor or in said metastasis is indicative of probable resistance to antiestrogen therapy, and wherein if BCL2 and CDKN1A gene expression product is present in said tumor or in said metastasis, said method further comprises administrating to said patient additional or alternative therapy to antiestrogen therapy, said method not comprising the invasive step of collecting the tissue from said tumor or metastasis of said tumor.

10. The method of claim 9, wherein said antiestrogen therapy is tamoxifen.

1 1. An assay for predicting resistance to anti-estrogen therapy in a breast cancer patient with an ER-positive breast tumor, said assay configured and adapted to detect the gene signatures of claims 1 , 4, and 7.

12. An array comprising polynucleotide probes capable of hybridizing to transcription products of BCL2 and CDKN1A derived from a breast cancer cell, said array embodying polynucleotide probes being able to determine the expression level for said genes.