Clinical Cancer Research Grants Advances in Breast Cancer
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
Cancer Research Clinical Cancer Research
Cancer Epidemiology Biomarkers & Prevention Molecular Cancer Therapeutics
Molecular Cancer Research Cancer Prevention Research
Cancer Prevention Journals Portal Cancer Reviews Online
Annual Meeting Education Book Meeting Abstracts Online
Clinical Cancer Research 14, 1744, March 15, 2008. doi: 10.1158/1078-0432.CCR-07-1833
© 2008 American Association for Cancer Research
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Supplementary Data
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Chanrion, M.
Right arrow Articles by Darbon, J.-M.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Chanrion, M.
Right arrow Articles by Darbon, J.-M.

Imaging, Diagnosis, Prognosis

A Gene Expression Signature that Can Predict the Recurrence of Tamoxifen-Treated Primary Breast Cancer

Maïa Chanrion1, Vincent Negre1, Hélène Fontaine1, Nicolas Salvetat2, Frédéric Bibeau1, Gaëtan Mac Grogan3, Louis Mauriac3, Dionyssios Katsaros4, Franck Molina2, Charles Theillet1 and Jean-Marie Darbon1

Authors' Affiliations: 1 U868 Institut National de la Sante et de la Recherche Medicale, Tumoral Identity and Plasticity, Cancer Research Center of Montpellier, Université Montpellier 1, CRLC Val d'Aurelle-Paul Lamarque; 2 Department of Informatics, UMR5160 Centre National de la Recherche Scientifique, Montpellier, France; 3 Department of Pathology, Institut Bergonié, Bordeaux, France; and 4 Department of Obstetrics and Gynecology, University of Turin, Turin, Italy

Requests for reprints: Jean-Marie Darbon, Institut National de la Sante et de la Recherche Medicale U868, Centre de Recherche en Cancérologie de Montpellier, CRLC Val d'Aurelle-Paul Lamarque, F-34298 Montpellier cedex 5, France. Phone: 33-467-612-556; Fax: 33-467-613-787; E-mail: jmdarbon{at}valdorel.fnclcc.fr.


   Abstract
Top
 Abstract
Patients and Methods
 Results
 Discussion
 References
 
Purpose: The identification of a molecular signature predicting the relapse of tamoxifen-treated primary breast cancers should help the therapeutic management of estrogen receptor–positive cancers.

Experimental Design: A series of 132 primary tumors from patients who received adjuvant tamoxifen were analyzed for expression profiles at the whole-genome level by 70-mer oligonucleotide microarrays. A supervised analysis was done to identify an expression signature.

Results: We defined a 36-gene signature that correctly classified 78% of patients with relapse and 80% of relapse-free patients (79% accuracy). Using 23 independent tumors, we confirmed the accuracy of the signature (78%) whose relevance was further shown by using published microarray data from 60 tamoxifen-treated patients (63% accuracy). Univariate analysis using the validation set of 83 tumors showed that the 36-gene classifier is more efficient in predicting disease-free survival than the traditional histopathologic prognostic factors and is as effective as the Nottingham Prognostic Index or the "Adjuvant!" software. Multivariate analysis showed that the molecular signature is the only independent prognostic factor. A comparison with several already published signatures demonstrated that the 36-gene signature is among the best to classify tumors from both training and validation sets. Kaplan-Meier analyses emphasized its prognostic power both on the whole cohort of patients and on a subgroup with an intermediate risk of recurrence as defined by the St. Gallen criteria.

Conclusion: This study identifies a molecular signature specifying a subgroup of patients who do not gain benefits from tamoxifen treatment. These patients may therefore be eligible for alternative endocrine therapies and/or chemotherapy.

Breast cancer is the most common female cancer in the Western world and the leading cause of death by cancer among women. Although the mortality rate is now stabilized or decreasing, breast cancer incidence is still on the rise through all European countries (1).

About two thirds of breast cancers are hormone (estrogen) dependent as they are positive for estrogen receptor (ER) and/or progesterone receptor (PR). Because estrogen is a major activator of proliferation in these tumors, its receptor and downstream signaling are excellent targets for the hormonal therapy in patients with ER+ (and/or PR+) breast cancers. Over the past three decades, the antiestrogen tamoxifen, which prevents the binding of estrogen to its receptor, has been the golden standard for the endocrine treatment of all stages of these cancers. In particular, large-scale randomized trials have shown that, in early-stage ER+ breast cancers, a 5-year course of tamoxifen, started immediately after surgery, reduces recurrence by 51% and mortality by 28% (2).

However, the success of tamoxifen therapy is limited by intrinsic or acquired tumor resistance. Approximately 40% of patients with ER+ breast cancers will not respond to tamoxifen. This is mostly because this selective ER modulator is not a pure antiestrogen and, indeed, shows some agonist activity. The balance between agonist and antagonist properties differs among cell types and seems to depend on several molecular events that can influence ER signaling. These include the level of coactivators and corepressors and the effect of cross-signaling in growth factor transduction pathways (3).

Recently, third-generation aromatase inhibitors have been shown to be more effective than tamoxifen to treat both advanced and early hormone-sensitive breast cancers in menopausal women (4). However, the absolute reduction in relapse-free survival for adjuvant aromatase inhibitors over tamoxifen is modest, and the issue of long-term adverse effects, in particular in bone, remains to be addressed. Moreover, aromatase inhibitors are also prone to develop resistance, although different mechanisms may be responsible for that.

As a consequence, the ability to accurately predict the response to tamoxifen should facilitate the choice of the best endocrine treatment and improve the management of primary ER+ breast cancers. Attempts to identify individual predictive markers failed to do that thus far. Recent publications on large-scale analysis of gene expression in breast cancers have underscored the considerable potential of DNA array technology. Hierarchical clustering of gene expression patterns has been successfully used to identify subtypes of breast tumors that exhibit distinct clinical behaviors (5–8). Supervised classifications of gene expression profiles have been done to identify poor prognostic signatures, predictive of recurrence in primary breast cancers (9–11). Such analysis has been used to identify genes that can help to discriminate between responders and nonresponders to chemotherapy agents in breast cancer (12, 13). Similar approaches have been tentatively used to define gene expression signatures that are predictive of recurrence in breast cancer patients treated with tamoxifen (14–16). However, the study by Jansen et al. concerned metastatic breast cancers (i.e., the response to a recurrent disease), whereas that of Ma et al. reported a two-gene signature whose performance has been debated (17–19). Both studies used a limited cohort of tumors as a training set (46 and 60 tumors, respectively). Finally, the study by Paik et al. (15) defined a recurrence score predictor for lymph node–negative breast cancer that also predicted the magnitude of chemotherapy benefit (20). However, this classifier was constructed from 250 candidate genes that were previously selected from the published literature. Finally, several studies have also reported molecular prognostic signatures able to classify ER+ breast cancer whatever the patient treatment (21–25).

To identify a gene expression signature that might predict recurrence of tamoxifen-treated primary breast cancers, we did a genome-wide microarray analysis of ER+ and/or PR+ primary breast tumors from 132 patients who had been treated with adjuvant tamoxifen. This analysis led to the identification of a 36-gene molecular signature that is highly predictive of clinical outcome.


   Patients and Methods
Top
 Abstract
 Patients and Methods
Results
 Discussion
 References
 
Patients and treatment. One hundred thirty-two primary breast carcinomas were analyzed in this study as a training set and 23 extra tumors as a validation set (Supplementary Table S1). These carcinomas were obtained from patients who had undergone initial surgery between 1989 and 2001 at the Cancer Research Center of Val d'Aurelle in Montpellier, the Bergonié Institute in Bordeaux, or the Department of Obstetrics and Gynecology of Turin. For the 155 patients, the median follow-up time was 65.9 months.

Fresh tissues were formalin-fixed and paraffin-embedded immediately after surgical removal. The remaining section of each tumor was snap frozen in liquid nitrogen and stored at –80°C. Frozen sections were stained with H&E to select samples with at least 50% of tumor cells. ER and PR status were determined by using a radioligand binding assay or by immunohistochemistry. All but eight tissue specimens were ER+, and six of the eight ER– tumors were PR+.

None of the patients received neoadjuvant systemic therapy. All patients were treated with tamoxifen (20 mg daily) for 5 y after surgery. One hundred twenty-one patients also received adjuvant radiotherapy. Recurrence was observed in 52 patients (48 distant metastases and 4 local recurrences) with median relapse time of 37.1 mo. Tumors from those patients were defined as R tumors (R for relapse), whereas tumors from patients who showed no recurrence were defined as RF tumors (RF for relapse-free).

RNA isolation, amplification, and labeling. Frozen breast samples (40 mg) were homogenized using the FastPrep System from Q-Biogene (MP Biomedicals). Total RNA was extracted and cleaned up from the lysate using the Qiagen RNeasy Mini Kit. RNA purity and integrity were controlled by the Bioanalyser 2100 (Agilent Technologies).

Gene expression profiling was done using 70-mer oligonucleotide microarrays. The 22,680 oligonucleotide probes (Oligo Set 2.1 from Qiagen-Operon), which represent 21,329 human specific genes, were spotted on Aminolink chips at the Genopole Montpellier-LR (26).

For each sample, 2 µg of total RNA were reverse transcribed and amplified by using the RNA amplification kit from Ambion. Fifteen micrograms of amplified RNA were labeled by direct chemical coupling to the Cy5 NHS ester (Amersham Biosciences). Labeled RNAs were purified, fragmented, and used as probes to hybridize microarrays. One-color design has been shown to perform as well as the two-color design, and it allowed easier normalization between samples (27).

Microarray gene expression analysis. Fluorescent images of hybridized microarrays were obtained with Axon 4000B scanner (Molecular Devices) and analyzed using Genepix 6.0. The mean of replicated spots was calculated for each gene by using the Acuity 4.0 software.

Gene expression signals were first normalized between arrays to the same median expression level (by dividing the intensity measurement for each gene on a given array by the median intensity of all genes across the array). Before analysis, a filter procedure eliminated noninformative genes on the basis of being significantly measured (i.e., expression level higher than 2-fold the mean expression of the negative control spots in at least 40% of the samples). Then, we selected the 5,415 genes whose expression varied by at least 3-fold from the median value in at least 1% of the samples. Expression data were then log 2 transformed and are accessible at the Gene Expression Omnibus5 (accession number GSE 9893).

The significance analysis of microarrays (SAM; ref. 28) was used to identify genes whose expression level best discriminated patients without (RF) and with relapse (R). This analysis was done on the 132 tumors of the training set (86 RF and 46 R). A thousand permutations were generated to estimate the false discovery rate value. Differences were considered as significant when the false discovery rate was <5% with a minimum 1.51-fold change and when the SAM score was greater than (or equal to) ±2.52.

A classifier that can predict recurrence under tamoxifen treatment was constructed by using the prediction analysis of microarrays (PAM; ref. 29) on the training set. A resampling approach was used as described in ref. 30. As shown in Supplementary Fig. S1, the training set was split into two parts, a learning set (85 patients) and a test set (47 patients). The learning set was used to construct the classifier, whereas the test set was used to estimate its performance. The split procedure was repeated 100 times. For each learning set, a minimal molecular signature was estimated by a leave-one-out cross-validation and its performance was estimated by the average proportion of misclassification for each associated test set. As the composition of the learning set changes in each iteration, the 100 minimal signatures obtained were obviously not identical. The final predictive signature corresponded to the most frequently selected genes (see Results).

To confirm the molecular signature obtained by PAM, we used another classifier, the K-nearest neighbor (31), with the same resampling of the training set.

A hierarchical pairwise average-linkage clustering (32) was done on the basis of the expression of the signature genes with median centered gene expression values and Pearson correlation as similarity metric.

We classified the training set tumors according to the five subgroups defined by Sorlie et al. (7) on the basis of the highest correlation to the respective centroid as indicated in Supplementary Methods.

To evaluate the performance of the consensus predictive signature, we used 23 independent tumors (17 RF and 6 R) as an external validation set as well as gene expression data from the study of Ma et al. (14). These data were obtained, like in our study, on 70-mer oligonucleotide microarrays and concerned 60 tamoxifen-treated patients (32 RF and 28 R). Forty-one patients (Supplementary Table S2) belonged to the intermediate-risk group according to the St. Gallen criteria (33). Raw data were downloaded from the Gene Expression Omnibus (accession number GSE1378). The Cy5 intensities were selected, normalized by median centering, and log 2 transformed.

To compare the performance of the 36-gene signature with those of intrinsic subtype (7), 70-gene profile (9), wound response (11), two-gene ratio (34), gene expression grade index (22), and molecular prognostic index (24), we evaluated the prognostic value of each expression-based model as indicated in Supplementary Methods, on both the training set (132 tumors) and a pooled validation set (83 tumors including the 23 tumors from our microarray study and the 60 tumors from Ma's study).

Statistical analyses. Specificity, sensitivity, and accuracy of the classifier as well as positive and negative predictive values were calculated as follows: Sensitivity = A / (A + C); specificity = D / (B + D); accuracy = (A + D) / (A + B + C + D). Positive predictive value = A / (A + B); negative predictive value = D / (C + D).


View this table:
[in this window]
[in a new window]

 
 
Univariate and multivariate analyses were done both on the training and the validation sets. All the input variables (tumor grade, ER and PR status, tumor size, patient age, lymph node status, Nottingham prognostic index, Adjuvant! software, and predictive classifier) were converted to a binary format (see below). Variables found to be significant in univariate analysis were selected to perform the multivariate analysis by logistical regression.

NPI score was calculated as follows: tumor size (cm) x 0.2 + grade + lymph node stage (negative nodes = 1; 1 to 3 positive nodes = 2; ≥4 positive nodes = 3), with 3.4 as threshold. Values for the Aduvant!6 were calculated for 10-year mortality with 0.2 as threshold.


   Results
Top
 Abstract
 Patients and Methods
 Results
Discussion
 References
 
Identification of the differentially expressed genes. Patients were classified in two groups according to the occurrence of relapse (R) or its absence (RF) within the 5 years of tamoxifen treatment. To identify which genes were differentially expressed in R tumors versus RF, we used a SAM analysis of the 5,415 filtered genes (see Patients and Methods). Three hundred one genes showed significant differences in their expression levels between R and RF tumors, with a false discovery rate below 5%. Then, the 48 most discriminating genes were selected on the basis of their fold change and SAM score values (Supplementary Table S3). Among those genes, 17 were overexpressed (positive SAM score) and 31 were underexpressed (negative SAM score) in tumors from patients with relapse.

Determination of a 36-gene predictive signature. To define a minimal expression signature, which could be used as a molecular classifier to predict recurrence of tamoxifen-treated patients, gene expression data were analyzed using a PAM algorithm. We used the 5,415-filtered genes instead of the 301 SAM-selected genes as it has been shown that building a classifier after selecting differentially expressed genes induced an overfitting bias (35). To determine the consensus molecular signature, we tested different signature lengths according to the level of gene occurrence in the 100 different minimal signatures. The error rate for R tumors decreased from 41% to 26% when the gene number was increased from 26 to 36 genes and remained steady from 36 to 71 genes (data not shown; for the 71-gene list, see Supplementary Table S4). Because contracted signatures are favored for predictive tests in clinical practice, we selected the 36 genes that were present in >60% of the 100 PAM iterative signatures as the optimal molecular signature (Table 1 ). This 36-gene signature classified the training set tumors with 80% sensitivity, 78% specificity, and 79% accuracy. The positive and negative predictive values of relapse were 87% and 68%, respectively.


View this table:
[in this window]
[in a new window]

  		Table 1. The list of 36 genes forming the minimal signature as defined by PAM analysis

 
Among the 36 signature genes, 26 belonged to the SAM selection of 48 genes (Table 1), with the remaining 10 genes coming out in the overall 301-gene list (not shown). A similar overlap between SAM and PAM analyses of expression data has been recently reported (36).

To confirm the reliability of this signature, we used a K-nearest neighbor classifier. As for PAM, a minimal predictive signature was defined for each iteration. The optimal consensus signature was obtained with 52 genes that were present in >47% of the 100 K-nearest neighbor iterative signatures (Supplementary Table S4). According to this 52-gene signature, the training set was classified with 83% sensitivity, 74% specificity, and 80% accuracy. Interestingly, 29 genes were common with those of the 36-gene PAM signature (Table 1), thus underlining the robustness of the 36-gene predictive signature.

We then did hierarchical clustering of the 132 tumors of the training set based on the 36-gene signature. As shown in Fig. 1 , the resulting dendrogram showed two main clusters: the R cluster of tumors from patients with relapse (34 of the 46 R tumors) and the RF cluster of tumors from relapse-free patients (71 of the 86 RF tumors). In conclusion, there was a strong agreement between the tumor classification obtained through the PAM classifier and the hierarchical clustering of the tumors through the predictive 36-gene set.


Figure 1
View larger version (43K):
[in this window]
[in a new window]
[Download PPT slide]
  		Fig. 1. Hierarchical clustering analysis of the 132 tumors from the training set using the 36-gene predictive signature. The 132 tumors from 86 relapse-free (RF) and 46 patients with relapse (R), as indicated on the top of the figure, were divided into two main clusters corresponding to the two branches of the dendrogram: the R and RF clusters. The 36 genes from the predictive signature are indicated on the right.

 
We have classified the 132 tumors according to the previously reported molecular subtypes. Forty-two tumors of the 132 tumors from the training set could not be classified in any subtype. Interestingly, 46 of the 58 tumors (79%) that were classified as luminal A were found in the relapse-free branch of the dendrogram, and 22 of the 31 tumors (71%) that were classified as luminal B were found in the relapse branch.

Validation of the 36-gene predictive signature. The 36-gene predictive signature, as defined by the PAM analysis, was validated by an external set of 23 patients (6 R and 17 RF) that were enrolled independently. Fourteen of the 17 RF (82% sensitivity) and 4 of the 6 R patients (67% specificity) were correctly classified with a 78% accuracy.

To confirm the performance of our classifier with a larger validation cohort, we used gene expression data obtained by Ma et al. (14) on tamoxifen-treated patients. We chose this data set as it was obtained on a 70-mer oligonucleotide microarray close to the one we used. We mapped our 36-gene signature on the Ma data set by using Hugo gene symbols. Thirty-five of the 36 signature genes were available in this study (BC015719 was missing). The omission of this gene did not affect prediction of clinical outcome that was evaluated on our training set. The 60 independent tumors from the Ma cohort were classified with a 63% accuracy (69% sensitivity, 57% specificity), confirming the relevance of our classifier.

Correlations to the clinical outcome. Univariate analysis of standard clinical prognostic factors (including the NPI and the Adjuvant! classifiers) and predictive 36-gene signature in terms of disease-free survival was done both on the training and validation sets. When the training set (132 patients) was used, the SBR grade, PR level, lymph node status, NPI, Adjuvant!, and the 36-gene signature were significantly associated to shortened disease-free survival (Table 2A ). Nevertheless, the predictive signature seemed to be the strongest predictive factor with two to four times higher odds ratios compared with other prognostic factors or indices. When done on the validation set (83 patients), only the 36-gene signature and the NPI and Adjuvant! classifiers were found to be significant with almost similar odds ratios (Table 2B).


View this table:
[in this window]
[in a new window]

  		Table 2. Univariate and multivariate analysis of 36-gene signature and usual prognostic factors in relation to disease-free survival

 
However, multivariate analysis that was done both on the training and validation sets by using the variables found to be significant in the univariate analysis showed that the 36-gene signature was the only independent prognostic factor (Table 2A and B).

These analyses confirmed that the molecular signature that we have identified was more efficient than the usual histopathologic variables to predict the clinical outcome of tamoxifen-treated breast cancers.

As shown on Fig. 2 , Kaplan-Meier analysis emphasized the prognostic power of the 36-gene signature both on the 83 patients from the validation set and on a subgroup of 60 of these 83 patients who exhibited an intermediate risk of recurrence according to the St. Gallen criteria (Supplementary Tables S1 and S2). This finding shows that the 36-gene signature may be useful to discriminate patients of good and poor prognosis in this intermediate-risk group of patients to define the best therapeutic approach.


Figure 2
View larger version (17K):
[in this window]
[in a new window]
[Download PPT slide]
  		Fig. 2. Kaplan-Meier survival curves for time to recurrence according to the 36-gene classifier. Kaplan-Meier disease-free survival curves based on the 36-gene signature: A, for the cohort of 83 patients from the validation set; B, for 60 of the 83 tumors that belong to the St. Gallen intermediate-risk group. The P values were calculated by using the log-rank test.

 
Comparison of the 36-gene signature with already published signatures. We have compared the performance of the present 36-gene signature with those of 6-gene expression–based models that have been described for breast cancer, namely, intrinsic subtype, 70-gene profile, wound response, two-gene ratio, gene expression grade index, and molecular prognostic index.

As shown on Table 3A , univariate analysis in terms of disease-free survival that was done on the training set (132 tumors) showed that five of the seven evaluated signatures correctly classified the tumors, the 36-gene, gene expression grade index, and molecular prognostic index signatures being the most efficient. Similar analysis done on the validation set (83 tumors) confirmed the performance of the 36-gene signature that was as efficient as the two-gene ratio to classify tumors (Table 3B).


View this table:
[in this window]
[in a new window]

  		Table 3. Comparison of the 36-gene signature with other molecular signatures

 
Functional analysis of the predictive signature. With the exception of one expressed sequence tag and 4 genes with unknown function, all of the 36 genes seemed relevant to cancer (Table 1). Indeed, the functional annotation showed that they were involved in the control of mitosis, cell cycle and cell proliferation, DNA replication and repair, cell signalization, adhesion/migration, cell death/survival, ER transcriptional activity, immune response, or metabolism.

Among these 36 genes, 23 were underexpressed and 13 were overexpressed in tumors from patients with recurrent disease in agreement with their putative function in oncogenesis. As shown in Table 4 , four of the underexpressed genes were involved in cellular adhesion or invasion, three could be implicated in immune response or inflammation, and two others were putative negative regulators of ER. Moreover, seven genes have been reported as candidate tumor suppressor genes, namely, AUTS2, GJA1/CX43, MTUS1/ATIP1, PCM1, ITM2B, LRRC17/P37NB, and LZTFL1.


View this table:
[in this window]
[in a new window]

  		Table 4. Functional classes of genes from the 36-gene predictive signature

 
Among the up-regulated genes, seven were involved in the control of mitosis and cell cycle, four have been implicated in DNA replication and/or DNA repair, and one gene, MMP1, has been shown to be a key promoter of cellular invasion.

Interestingly, CX43 and ITM2B, which have been proposed to have a proapoptotic function, were down-regulated in tumors from recurrent patients, whereas p15(PAF), which has been reported as a antiapoptotic factor, was up-regulated in these tumors. Also, CX43, FMO5, and PCM1 have been reported as members of an apoptotic gene module whose overexpression specified low-grade ER+ breast tumors from patients with a better survival outcome (37). All of them were found down-regulated in tumors from patients with relapse in this study.


   Discussion
Top
 Abstract
 Patients and Methods
 Results
 Discussion
References
 
The main problem encountered in gene expression profiling studies is the relatively small overlap between independently reported molecular signatures. It is noteworthy that the present 36-gene signature includes 11 genes (30%) that are members of a proliferation cluster present in several previously published classifiers. This major proliferation signature has been shown to specify poor prognosis subsets of ER+ breast cancer (15, 21–25). It includes AURKB, CCNB2, CDC2, PRC1, RRM2, TPX2, and UBE2C genes (22, 23), which are present in our signature, along with BIRC5/survivin, KI-67, MYBL2, STK6/15, and CCNB1, belonging to Paik's 21-gene classifier (15). Several members of this proliferation gene cluster are also found in van't Veer's 231-gene poor prognostic signature (9), namely, BIRC5, CCNB2, PRC1, RRM2, and STK6/15. Another gene, TK1, is found to be a member of this cluster both in this and other distinct studies (9, 21). ASF1B (38), ASPM (22, 38), and p15(PAF) (24, 39) have also been reported in proliferation gene clusters correlated to cancer aggressiveness (22, 38).

Interestingly, the similarity between these distinct gene signatures may further concern members of the same functional family or different partners of the same pathway(s). For example, CCNB2 and AURKB (STK12) may be substituted by CCNB1 and AURKA (STK6/15), respectively. Moreover, AURKB interacts with survivin (40) and TPX2 targets AURKA to the mitotic spindle (41).

Although the presence of such a proliferation cluster seems to be a hallmark of several prognostic signatures, the 11 members of this cluster from our 36-gene signature did not exhibit a prognostic power by itself (data not shown). This latter finding shows that the ER-related genes of the present signature could be essential to confer its prognostic power.

Indeed, our 36-gene signature included at least 18 genes (50%) that were related to estrogens either as ER targets or ER regulators. Six genes, namely, CRIM1, CX43, FMO5, P37NB/LRRC17, STC2, and WFDC2, have been shown to be induced by estrogens and the expression of these genes has been correlated with ER+ status (42–44). Interestingly, STC2 or FMO5 expression was associated with good prognosis in ER and/or PR+ breast cancer patients who were treated with adjuvant hormone therapy (43, 44). Conversely, TK1 overexpression, shown to occur in ER-negative breast tumors (45) and in our cohort of R tumors, has been reported to be a marker of poor clinical outcome of tamoxifen therapy (46). ABCC3 has been shown to be E2 repressed (47) and was overexpressed in tumors from patients with recurrence. Eight other genes of the 36-gene signature (i.e., AURKB/STK12, CCNB2, CDC2, MMP1, PRC1, RRM2, TPX2, and UBE2C) have been reported to discriminate ER+ from ER– breast cancers (8, 23, 42). Finally, two other genes, PNRC2 and PKIB, are possibly involved in the regulation of the ER activity. PNRC2 is mostly known to be a coactivator of nuclear receptors, including ER, and it has also been suggested to antagonize the growth factor–mediated mitogen-activated protein kinase activation of ER (48). On the other hand, down-regulation of PKIB, a protein kinase A inhibitor, might be associated with tamoxifen resistance. Indeed, such an association has been reported for PKAR1{alpha}, another protein kinase A–negative regulator whose down-regulation favored the phosphorylation of ER, converting tamoxifen from an ER antagonist into a growth stimulator (49). Also, OTUD7B encodes a deubiquinating enzyme that has been shown as a negative regulator of nuclear factor-{kappa}B, a prognostic marker associated to tamoxifen resistance (50).

In our study as in that by Paik et al. (15), the classifiers were able to predict the clinical outcome of tamoxifen-treated breast cancers and thus they could be considered as general prognostic classifiers. Whether our 36-gene signature may further specify responsiveness to tamoxifen remains to be investigated. However, the presence of several estrogen-related genes in this signature suggests that it could be the case. In any case, this molecular signature allows to discriminate a subset of patients who do not gain benefits from tamoxifen treatment. Those patients might be potential candidates for alternative endocrine therapies and/or chemotherapy. As a main finding, we show the prognostic power of our signature on a subgroup of patients who exhibited an intermediate risk of relapse according to the St. Gallen criteria. In other words, the 36-gene signature can be helpful in tailoring the therapeutic decision in this particular patient subset.

Despite their value, expression signatures, which have been obtained by studying retrospective cohorts, need to be confirmed by prospective studies.

In summary, our findings show the utility of large-scale gene expression profiling to define a molecular signature that can predict the recurrence of tamoxifen-treated primary breast cancer more efficiently than the usual clinical and histopathologic prognostic factors. Moreover, our data bring new insights on putative master genes involved in cancer progression and resistance to endocrine therapy.


   Acknowledgments
 
We thank Laurent Journot, Dany Séverac, and Cindy Aknin of the transcriptome platform facilities at the Genopole of Montpellier-LR; Thomas Filleron for his fruitful advices about statistical analysis; and Dr. Stéphane Culine for his help in collecting patient follow-up.


   Footnotes
 
Grant support: Institut National de la Sante et de la Recherche Medicale, GEFLUC Montpellier-Languedoc-Roussillon, and the Canceropole Grand-Sud-Ouest, France. M. Chanrion was a recipient of a fellowship from the Ligue contre le Cancer-Comité Hérault and V. Negre was a recipient of a fellowship from Institut National de la Sante et de la Recherche Medicale.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Note: Supplementary data for this article are available at Clinical Cancer Research Online (http://clincancerres.aacrjournals.org/).

5 http://www.ncbi.nlm.nih.gov/geo Back

6 http://www.adjuvantonline.com/index.jsp Back

Received 7/26/07; revised 11/30/07; accepted 12/11/07.


   References
Top
 Abstract
 Patients and Methods
 Results
 Discussion
 References
 

  1. Botha JL, Bray F, Sankila R, et al. Breast cancer incidence and mortality trends in 16 European countries. Eur J Cancer 2003;39:1718–29.[CrossRef][Medline]
  2. EBCTCG. Tamoxifen for early breast cancer: an overview of the randomised trials. Early Breast Cancer Trialists' Collaborative Group. Lancet 1998;351:1451–67.[CrossRef][Medline]
  3. Ali S, Coombes RC. Endocrine-responsive breast cancer and strategies for combating resistance. Nat Rev Cancer 2002;2:101–12.[CrossRef][Medline]
  4. Johnston SR, Dowsett M. Aromatase inhibitors for breast cancer: lessons from the laboratory. Nat Rev Cancer 2003;3:821–31.[CrossRef][Medline]
  5. Perou CM, Sorlie T, Eisen MB, et al. Molecular portraits of human breast tumours. Nature 2000;406:747–52.[CrossRef][Medline]
  6. Sorlie T, Perou CM, Tibshirani R, et al. Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implications. Proc Natl Acad Sci U S A 2001;98:10869–74.[Abstract/Free Full Text]
  7. Sorlie T, Tibshirani R, Parker J, et al. Repeated observation of breast tumor subtypes in independent gene expression data sets. Proc Natl Acad Sci U S A 2003;100:8418–23.[Abstract/Free Full Text]
  8. Sotiriou C, Neo SY, McShane LM, et al. Breast cancer classification and prognosis based on gene expression profiles from a population-based study. Proc Natl Acad Sci U S A 2003;100:10393–8.[Abstract/Free Full Text]
  9. van 't Veer LJ, Dai H, van de Vijver MJ, et al. Gene expression profiling predicts clinical outcome of breast cancer. Nature 2002;415:530–6.[CrossRef][Medline]
  10. Wang Y, Klijn JG, Zhang Y, et al. Gene-expression profiles to predict distant metastasis of lymph-node-negative primary breast cancer. Lancet 2005;365:671–9.[Medline]
  11. Chang HY, Nuyten DS, Sneddon JB, et al. Robustness, scalability, and integration of a wound-response gene expression signature in predicting breast cancer survival. Proc Natl Acad Sci U S A 2005;102:3738–43.[Abstract/Free Full Text]
  12. Chang JC, Wooten EC, Tsimelzon A, et al. Gene expression profiling for the prediction of therapeutic response to docetaxel in patients with breast cancer. Lancet 2003;362:362–9.[CrossRef][Medline]
  13. Ayers M, Symmans WF, Stec J, et al. Gene expression profiles predict complete pathologic response to neoadjuvant paclitaxel and fluorouracil, doxorubicin, and cyclophosphamide chemotherapy in breast cancer. J Clin Oncol 2004;22:2284–93.[Abstract/Free Full Text]
  14. 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;5:607–16.[CrossRef][Medline]
  15. 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;351:2817–26.[Abstract/Free Full Text]
  16. Jansen MP, Foekens JA, van Staveren IL, et al. Molecular classification of tamoxifen-resistant breast carcinomas by gene expression profiling. J Clin Oncol 2005;23:732–40.[Abstract/Free Full Text]
  17. Fan C, Oh DS, Wessels L, et al. Concordance among gene-expression-based predictors for breast cancer. N Engl J Med 2006;355:560–9.[Abstract/Free Full Text]
  18. Goetz MP, Suman VJ, Ingle JN, et al. A two-gene expression ratio of homeobox 13 and interleukin-17B receptor for prediction of recurrence and survival in women receiving adjuvant tamoxifen. Clin Cancer Res 2006;12:2080–7.[Abstract/Free Full Text]
  19. Reid JF, Lusa L, De Cecco L, et al. Limits of predictive models using microarray data for breast cancer clinical treatment outcome. J Natl Cancer Inst 2005;97:927–30.[Abstract/Free Full Text]
  20. Paik S, Tang G, Shak S, et al. Gene expression and benefit of chemotherapy in women with node-negative, estrogen receptor-positive breast cancer. J Clin Oncol 2006;24:3726–34.[Abstract/Free Full Text]
  21. Dai H, van't Veer L, Lamb J, et al. A cell proliferation signature is a marker of extremely poor outcome in a subpopulation of breast cancer patients. Cancer Res 2005;65:4059–66.[Abstract/Free Full Text]
  22. Sotiriou C, Wirapati P, Loi S, et al. Gene expression profiling in breast cancer: understanding the molecular basis of histologic grade to improve prognosis. J Natl Cancer Inst 2006;98:262–72.[Abstract/Free Full Text]
  23. Oh DS, Troester MA, Usary J, et al. Estrogen-regulated genes predict survival in hormone receptor-positive breast cancers. J Clin Oncol 2006;24:1656–64.[Abstract/Free Full Text]
  24. Teschendorff AE, Naderi A, Barbosa-Morais NL, et al. A consensus prognostic gene expression classifier for ER positive breast cancer. Genome Biol 2006;7:R101.[CrossRef][Medline]
  25. 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;25:1239–46.[Abstract/Free Full Text]
  26. Desagher S, Severac D, Lipkin A, et al. Genes regulated in neurons undergoing transcription-dependent apoptosis belong to signaling pathways rather than the apoptotic machinery. J Biol Chem 2005;280:5693–702.[Abstract/Free Full Text]
  27. Patterson TA, Lobenhofer EK, Fulmer-Smentek SB, et al. Performance comparison of one-color and two-color platforms within the MicroArray Quality Control (MAQC) project. Nat Biotechnol 2006;24:1140–50.[CrossRef][Medline]
  28. Tusher VG, Tibshirani R, Chu G. Significance analysis of microarrays applied to the ionizing radiation response. Proc Natl Acad Sci U S A 2001;98:5116–21.[Abstract/Free Full Text]
  29. Tibshirani R, Hastie T, Narasimhan B, et al. Diagnosis of multiple cancer types by shrunken centroids of gene expression. Proc Natl Acad Sci U S A 2002;99:6567–72.[Abstract/Free Full Text]
  30. Wessels LF, Reinders MJ, Hart AA, et al. A protocol for building and evaluating predictors of disease state based on microarray data. Bioinformatics 2005;21:3755–62.[Abstract/Free Full Text]
  31. Pomeroy SL, Tamayo P, Gaasenbeek M, et al. Prediction of central nervous system embryonal tumour outcome based on gene expression. Nature 2002;415:436–42.[CrossRef][Medline]
  32. Eisen MB, Spellman PT, Brown PO, et al. Cluster analysis and display of genome-wide expression patterns. Proc Natl Acad Sci U S A 1998;95:14863–8.[Abstract/Free Full Text]
  33. Goldhirsch A, Glick JH, Gelber RD, et al. Meeting highlights: international expert consensus on the primary therapy of early breast cancer 2005. Ann Oncol 2005;16:1569–83.[Abstract/Free Full Text]
  34. Ma XJ, Hilsenbeck SG, Wang W, et al. The HOXB13:IL17BR expression index is a prognostic factor in early-stage breast cancer. J Clin Oncol 2006;24:4611–9.[Abstract/Free Full Text]
  35. Simon R, Radmacher MD, Dobbin K, et al. Pitfalls in the use of DNA microarray data for diagnostic and prognostic classification. J Natl Cancer Inst 2003;95:14–8.[Free Full Text]
  36. Smid M, Wang Y, Klijn JG, et al. Genes associated with breast cancer metastatic to bone. J Clin Oncol 2006;24:2261–7.[Abstract/Free Full Text]
  37. Yu K, Ganesan K, Miller LD, et al. A modular analysis of breast cancer reveals a novel low-grade molecular signature in estrogen receptor-positive tumors. Clin Cancer Res 2006;12:3288–96.[Abstract/Free Full Text]
  38. Rosty C, Sheffer M, Tsafrir D, et al. Identification of a proliferation gene cluster associated with HPV E6/E7 expression level and viral DNA load in invasive cervical carcinoma. Oncogene 2005;24:7094–104.[CrossRef][Medline]
  39. Rhodes DR, Yu J, Shanker K, et al. Large-scale meta-analysis of cancer microarray data identifies common transcriptional profiles of neoplastic transformation and progression. Proc Natl Acad Sci U S A 2004;101:9309–14.[Abstract/Free Full Text]
  40. Bolton MA, Lan W, Powers SE, et al. Aurora B kinase exists in a complex with survivin and INCENP and its kinase activity is stimulated by survivin binding and phosphorylation. Mol Biol Cell 2002;13:3064–77.[Abstract/Free Full Text]
  41. Brunet S, Sardon T, Zimmerman T, et al. Characterization of the TPX2 domains involved in microtubule nucleation and spindle assembly in Xenopus egg extracts. Mol Biol Cell 2004;15:5318–28.[Abstract/Free Full Text]
  42. Abba MC, Hu Y, Sun H, et al. Gene expression signature of estrogen receptor {alpha} status in breast cancer. BMC Genomics 2005;6:37.[CrossRef][Medline]
  43. Bieche I, Girault I, Urbain E, et al. Relationship between intratumoral expression of genes coding for xenobiotic-metabolizing enzymes and benefit from adjuvant tamoxifen in estrogen receptor {alpha}-positive postmenopausal breast carcinoma. Breast Cancer Res 2004;6:R252–63.[CrossRef][Medline]
  44. Yamamura J, Miyoshi Y, Tamaki Y, et al. mRNA expression level of estrogen-inducible gene, {alpha} 1-antichymotrypsin, is a predictor of early tumor recurrence in patients with invasive breast cancers. Cancer Sci 2004;95:887–92.[CrossRef][Medline]
  45. O'Neill KL, Hoper M, Odling-Smee GW. Can thymidine kinase levels in breast tumors predict disease recurrence? J Natl Cancer Inst 1992;84:1825–8.[Abstract/Free Full Text]
  46. Foekens JA, Romain S, Look MP, et al. Thymidine kinase and thymidylate synthase in advanced breast cancer: response to tamoxifen and chemotherapy. Cancer Res 2001;61:1421–5.[Abstract/Free Full Text]
  47. Vendrell JA, Magnino F, Danis E, et al. Estrogen regulation in human breast cancer cells of new downstream gene targets involved in estrogen metabolism, cell proliferation and cell transformation. J Mol Endocrinol 2004;32:397–414.[Abstract]
  48. Zhou D, Chen B, Ye JJ, et al. A novel crosstalk mechanism between nuclear receptor-mediated and growth factor/Ras-mediated pathways through PNRC-Grb2 interaction. Oncogene 2004;23:5394–404.[CrossRef][Medline]
  49. Michalides R, Griekspoor A, Balkenende A, et al. Tamoxifen resistance by a conformational arrest of the estrogen receptor {alpha} after PKA activation in breast cancer. Cancer Cell 2004;5:597–605.[CrossRef][Medline]
  50. Zhou Y, Eppenberger-Castori S, Eppenberger U, et al. The NF{kappa}B pathway and endocrine-resistant breast cancer. Endocr Relat Cancer 2005;12 Suppl 1:S37–46.[Abstract/Free Full Text]



This article has been cited by other articles:


Home page
 Proc. Natl. Acad. Sci. USAHome page
S. P. Pitroda, N. N. Khodarev, M. A. Beckett, D. W. Kufe, and R. R. Weichselbaum
From the Cover: MUC1-induced alterations in a lipid metabolic gene network predict response of human breast cancers to tamoxifen treatment
PNAS, April 7, 2009; 106(14): 5837 - 5841.
[Abstract] [Full Text] [PDF]

Home page
 J Mol EndocrinolHome page
S. E Ghayad, J. A Vendrell, I. Bieche, F. Spyratos, C. Dumontet, I. Treilleux, R. Lidereau, and P. A Cohen
Identification of TACC1, NOV, and PTTG1 as new candidate genes associated with endocrine therapy resistance in breast cancer
J. Mol. Endocrinol., February 1, 2009; 42(2): 87 - 103.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Supplementary Data
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Chanrion, M.
Right arrow Articles by Darbon, J.-M.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Chanrion, M.
Right arrow Articles by Darbon, J.-M.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
Cancer Research Clinical Cancer Research
Cancer Epidemiology Biomarkers & Prevention Molecular Cancer Therapeutics
Molecular Cancer Research Cancer Prevention Research
Cancer Prevention Journals Portal Cancer Reviews Online
Annual Meeting Education Book Meeting Abstracts Online