EndoPredict predicts for the response to neoadjuvant chemotherapy in ER-positive, HER2-negative breast cancer

Cancer Letters 355 (2014) 70–75

Contents lists available at ScienceDirect

Cancer Letters j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / c a n l e t

Original Articles

EndoPredict predicts for the response to neoadjuvant chemotherapy in ER-positive, HER2-negative breast cancer François Bertucci a,b,*, Pascal Finetti a, Patrice Viens b,c, Daniel Birnbaum a a Département d’Oncologie Moléculaire, Centre de Recherche en Cancérologie de Marseille (CRCM), Institut Paoli-Calmettes (IPC), UMR1068 Inserm, 13009 Marseille, France b Faculté de Médecine, Aix-Marseille Université, 13001 Marseille, France c Département d’Oncologie Médicale, CRCM, IPC; UMR1068 Inserm, France

A R T I C L E

I N F O

Article history: Received 25 July 2014 Received in revised form 4 September 2014 Accepted 4 September 2014 Keywords: Breast cancer EndoPredict Neoadjuvant Response to chemotherapy

A B S T R A C T

The EndoPredict (EP) signature is a prognostic 11-gene expression signature specifically developed in ER+/HER2– node-negative/positive breast cancer. It is associated with relapse-free survival in patients treated with adjuvant hormone therapy, suggesting that EP low-risk patients could be treated with adjuvant hormone therapy alone whereas high-risk patients would deserve addition of adjuvant chemotherapy. Thus, it is important to determine whether EP high-risk patients are or are not more sensitive to chemotherapy than low-risk patients. Here, we have assessed the EP predictive value for pathological complete response to neoadjuvant chemotherapy in ER+/HER2– breast cancer. We gathered gene expression and histoclinical data of 553 pre-treatment ER+/HER2– breast carcinomas treated with anthracycline-based neoadjuvant chemotherapy. We searched for correlation between the pathological complete response (pCR) and the EP score-based classification. The overall pCR rate was 12%. Fifty-one percent of samples were classified as low-risk according to the EP score and 49% as high-risk. EP classification was associated with a pCR rate of 7% in the low-risk group and 17% in the high-risk group (p < 0.001). In multivariate analysis, the EP score remained significantly associated with pCR. Many genes upregulated in the highrisk tumours were involved in cell proliferation, whereas many genes upregulated in the low-risk tumours were involved in ER-signalling and stroma. Despite higher chemosensitivity, the high-risk group was associated with worse disease-free survival. In conclusion, EP high-risk ER+/HER2– breast cancers are more likely to respond to anthracycline-based chemotherapy. © 2014 Elsevier Ireland Ltd. All rights reserved.

Introduction During the last decade, the survival of patients with breast cancer improved thanks to mass screening and adjuvant systemic therapy. A crucial issue regarding adjuvant chemotherapy is to improve the prognostic features to avoid overtreatment of patients – notably those without or less than four pathologically involved axillary lymph nodes – who receive post-operative chemotherapy but would be cured by surgery and radiotherapy alone. That is particularly critical for patients with oestrogen receptor-positive (ER+)/HER2– tumour, who

Abbreviations: cT, clinical tumour size; cN, clinical axillary lymph node status; DAVID, Database for Annotation, Visualisation and Integrated Discovery; DFS, diseasefree survival; EP, EndoPredict; ER, oestrogen receptor; FFPE, formalin-fixed paraffinembedded; GO, GeneOntology; pCR, pathological complete response; pN, pathological axillary lymph node status; PR, progesterone receptor; pT, pathological tumour size; RMA, Robust Multichip Average; ROC, receiver operating characteristic; RT-PCR, reverse transcription–polymerase chain reaction; SAM, Significance Analysis of Microarrays; SBR, Scarff–Bloom–Richardson. * Corresponding author. Tel.: +33 4 91 22 35 37; fax: +33 4 91 22 36 70. E-mail address: [email protected] (F. Bertucci). http://dx.doi.org/10.1016/j.canlet.2014.09.014 0304-3835/© 2014 Elsevier Ireland Ltd. All rights reserved.

represent the largest subset of early breast cancer patients. Indeed, if adjuvant chemotherapy is currently recommended for most women with HER2+ tumours or with triple-negative tumours [1], its indications are much more challenging in patients with ER+/HER2– tumours that are either candidate for adjuvant hormone therapy alone or for both hormone therapy and chemotherapy. During the last decade, gene expression profiling identified promising prognostic multigene signatures in breast cancer [2]. In the setting of ER+/HER2– tumours, several signatures including Mammaprint [3], Recurrence Score [4], or EndoPredict [5] have been proposed to improve the selection of patients candidate or not to adjuvant chemotherapy [1]. They were developed in heterogeneous cohorts of patients with respect to ER and HER2 status, except EndoPredict (EP), which was specifically developed in ER+/HER2– node-negative/positive tumours [5]. EP is based on mRNA expression of 8 genes of interest and 3 reference genes measured using quantitative RT-PCR on formalin-fixed paraffin-embedded (FFPE) samples. In contrast with Mammaprint and Recurrence Score, EP is suitable for decentralised testing in molecular pathology laboratories instead of a single reference laboratory [6]. EP independent prognostic value was first validated in a series of 1702 indepen-

F. Bertucci et al./Cancer Letters 355 (2014) 70–75

dent tumours from two prospective trials of adjuvant hormone therapy. EP segregated patients into two groups according to their distant relapse risk: for example, in the ABCSG-8 prospective trial, the distant recurrence rate was 6% in the low-risk group and 15% in the high-risk group, suggesting that low-risk patients could be treated with adjuvant hormone therapy without any chemotherapy whereas high-risk patients would deserve addition of adjuvant chemotherapy. Thus, it is important to determine whether EP highrisk patients are or are not more sensitive to chemotherapy than low-risk patients. This has already been shown with other signatures such as Recurrence Score [7], Mammaprint [8] and Genomic Grade Index [9]. We thus hypothesised that patients with breast cancer classified as high-risk of relapse according to EP would respond better to chemotherapy. To prove this issue, we retrospectively analysed the predictive value of EP signature in terms of pathological response to chemotherapy in a large pooled series of breast cancer patients treated with anthracycline-based neo-adjuvant chemotherapy and whose tumour had been previously profiled using DNA microarrays. We show that the EP signature is associated with response to chemotherapy. Materials and methods

“metagene EP score” as the difference of mean expression profile between upregulated and downregulated genes; a threshold equal to 0 classified the samples in two groups as “high-risk predicted” and “low-risk predicted”. This “metagene EP score” was then applied to two independent validation sets separately (the two largest ones after Hatzis’ set): Hess’ set including 77 ER+/HER2– tumours [13] and Tabchy’s set including 93 tumours [16]. Correlation of the predicted EP classification with the observed EP classification was assessed using receiver operating characteristic (ROC) curve of the “metagene EP score” and using Fisher’s exact test. Statistical analysis Correlations between tumour groups and histoclinical features were calculated with Fisher’s exact test or Student’s t-test when appropriate. Primary endpoint was the degree of pathological response to neoadjuvant chemotherapy defined as categorical variable (pCR versus non-pCR). Disease-free survival (DFS) was calculated from the date of diagnosis to the date of first loco-regional or metastatic relapse, or death. The follow-up was measured to the date of last news for event-free patients. Survival curves were obtained using the Kaplan–Meier method and compared with the log-rank test. Univariate and multivariate analyses for pCR prediction were done using generalised linear models. Variables tested in univariate analysis included patients’ age (≤50 versus >50 years), prechemotherapy clinical tumour size (cT) and axillary lymph node status (cN), pathological SBR grade (3 versus 2 versus 1), and PR status (positive versus negative). Multivariate analysis was applied to variables significant in the univariate analysis. All statistical tests were two-sided at the 5% level of significance. Analyses were done using the survival package (version 2.30), in the R software (version 2.15.2). Our analysis adhered to the REporting recommendations for tumour MARKer prognostic studies (REMARK) [21].

Patients’ selection

Results

We collected seven retrospective gene expression data sets of breast cancer samples profiled using DNA microarrays (Supplementary Table S1), including our own set [10] and six public sets [11–16]. Gene expression and histoclinical data were retrieved from the NCBI GEO database and authors’ website. Selection criteria of data sets were the presence of at least 10 profiled samples representing pre-treatment invasive breast carcinoma, treated with anthracycline-based neoadjuvant chemotherapy followed by surgery including lumpectomy or mastectomy and axillary lymph node dissection, and documentation of pathological response. The seven selected data sets included a total of 1739 breast cancer samples, including 553 cases that were ER+/HER2– as defined using mRNA expression level (see below). These 553 samples were retained for the present study. Pathological response to neoadjuvant chemotherapy was defined on surgical specimens and scored on both the primary tumour and the lymph nodes using Chevallier grading [17]. Grades 1 and 2 were considered as pCR, and grades 3 and 4 as non-pCR.

Patients’ characteristics

Gene expression data analysis Gene expression data were processed before analysis. First, each of the seven data sets, all profiled using Affymetrix microarrays, was normalised separately by using Robust Multichip Average (RMA) [18] with the non-parametric quantile algorithm as normalisation parameter. Normalisation was done in R using Bioconductor and associated packages. Second, we mapped hybridisation probes across the different Affymetrix DNA microarrays represented using NetAffx Annotation files (www.affymetrix.com; release from 01/12/2008) to update the Affymetrix gene chips annotations and both SOURCE (http://smd.stanford.edu/cgi-bin/source/sourceSearch), and EntrezGene (Homo sapiens gene information db, release from 09/12/2008, ftp:// ftp.ncbi.nlm.nih.gov/gene/) to map the probes based on their EntrezGeneID. When multiple probes mapped to the same GeneID, we selected the one with the highest variance in a particular data set to represent the GeneID. In each data set separately, oestrogen receptor (ER), progesterone receptor (PR) and HER2 status were defined at the mRNA level using gene expression data of their respective gene, ESR1, PGR and HER2/ERBB2. The bimodal distribution of expression levels of these genes allowed us to identify a threshold of positivity, common to all sets, for each gene. Tumour samples with expression levels lower than this threshold were classified as negative; the others were classified as positive [19]. We then applied the EP classifier to each data set separately. EP risk score was measured using the previously published algorithm [5], and analysed as dichotomised variable (low risk and high risk) using the predefined cut-off [5] and as continuous variable. The predicted risk of relapse was also defined in each data set separately according to the Mammaprint [3] and Recurrence Score [4] signatures as previously reported. To explore the biological pathways linked to the EP classifier, we applied a supervised analysis to the largest data set, the Hatzis data set [12] that contained 279 ER+/HER2– tumours, including 146 classified as low-risk and 133 as high-risk. We compared the expression profiles of 7260 genes between low-risk and high-risk tumours using Significance Analysis of Microarrays (SAM) [20] with a Δ-value equal to 2.23 and a false discovery rate inferior to 0.001. Ontology analysis of the resulting gene list was focused on GO biological processes of the Database for Annotation, Visualisation and Integrated Discovery (DAVID; david.abcc.ncifcrf.gov/). To test the robustness of this list of differentially expressed genes, we computed for each tumour sample a

71

We collected personal and public gene expression and histoclinical data of 553 pre-treatment ER+/HER2– invasive breast carcinomas with available pathological response assessed on the surgical specimen after anthracycline-based neoadjuvant chemotherapy. Their characteristics are summarised in Table 1. Median age of patients was 49 years. Clinically, the tumours were classified T2–T4

Table 1 Histoclinical characteristics of 553 patients with ER+/HER2– breast cancer. Characteristics

N (%)

Age, n = 553 < = 50 years >50 years Clinical tumour size (cT), n = 552 cT1 cT2-4 Clinical axillary lymph node status (cN), n = 519 cN0 cN1-3 Histological type, n = 232 IDC ILC Other Grade, n = 511 1 2 3 PR status (mRNA), n = 553 0 1 Pathological complete response (pCR), n = 553 No Yes EndoPredict-based classification, n = 553 Low-risk High-risk Median follow-up, months (range), n = 299 DFS, n = 302 No event Event 5-year DFS, n = 299

298 (54%) 255 (46%)

40 (7%) 512 (93%) 183 (35%) 336 (65%) 179 (77%) 10 (4%) 43 (19%) 47 (9%) 272 (53%) 192 (38%) 317 (57%) 236 (43%) 489 (88%) 64 (12%) 283 (51%) 270 (49%) 40 (6–165) 259 (86%) 43 (14%) 81% (CI95 75–87)

IDC, invasive ductal cancer; ILC, invasive lobular cancer; DFS, disease-free survival.

72

F. Bertucci et al./Cancer Letters 355 (2014) 70–75

in 93%, and the axillary lymph node status was positive in 65% of patients. Most tumours were ductal type and SBR grade 2 (53%) or 3 (38%). Anthracycline-based neoadjuvant chemotherapy included taxane in more than 75% of patients. The overall pCR rate was 12% (64 out of 553 patients). With a median follow-up of 40 months, the 5-year DFS was 81% (CI95 75–87). EndoPredict-based classification and histoclinical correlations All cases were assigned to relapse risk categories according to the EP score: 283 (51%) were classified as low-risk and 270 (49%) as high-risk. Associations between EP categories and histoclinical features are shown in Table 2. High-risk tumours were associated (Fisher’s exact test) with poor-prognosis features: large tumour size (p = 0.004), positive lymph node status (p = 0.001), PR-negative status (p < 0.001), high tumour grade (p < 0.001), and as expected poorer survival with 5-year DFS equal to 73% (CI95 63–85) versus 88% (CI95 81–95) in the low-risk group (p = 0.015, log-rank test). EndoPredictbased classification was also associated with classifications based on the two other major prognostic signatures: when compared with EP low-risk tumours, EP high-risk tumours were more often defined (p < 0.001) as high-risk tumours according to Mammaprint and Recurrence Score. EndoPredict-based classification and response to chemotherapy EP signature, as categorical variable and as continuous variable, correlated with the pathological response to chemotherapy. EP classification was associated with a pCR rate lower in the

low-risk group (7%; 19 out of 283) than in the high-risk group (17%; 45 out of 270; p < 0.001, Fisher’s exact test). Seventy percent (45 out of 64) of patients with pCR were assigned to the EP high-risk group, versus 48% of patients without pCR (225 out of 468). The correlation persisted when EP score was analysed as continuous variable (p < 0.001, Student’s t-test; Fig. 1). In univariate analysis (Table 3), high tumour grade was also associated with pCR. No association was found with patients’ age, prechemotherapy tumour size, lymph node status, histological type, and PR status. In multivariate analysis (Table 3), including the EP score as continuous value and tumour grade, the EP score remained associated with pCR, whereas grade did not.

EndoPredict-based classification and associated biological processes Using SAM algorithm, we identified 930 genes differentially expressed in the Hatzis data set between the high-risk tumours and the low-risk tumours as defined by EP: 478 genes were upregulated and 452 were downregulated in the high-risk samples (Supplementary Table S2). The robustness of this gene list was confirmed in two independent validation sets including respectively 77 and 93 tumours (Fig. 2). Ontology analysis of these 930 genes (Supplementary Table S3) revealed that high-risk tumours overexpressed genes involved in cell proliferation and cycle, whereas low-risk tumours overexpressed genes involved in cell differentiation, adhesion, stroma and ER-signalling. Table 4 shows the top 10 GO biological processes associated with the upregulated and downregulated genes, including at least 10 genes and significant with a p-value lower than 0.001.

Table 2 EP-based classification and histoclinical correlations. Characteristics

Age, n = 553 < = 50 years >50 years Clinical tumour size (cT), n = 552 cT1 cT2-4 Clinical axillary lymph node status (cN), n = 519 cN0 cN1-3 Histological type, n = 232 IDC ILC Other Grade, n = 511 1 2 3 PR status (mRNA), n = 553 0 1 Mammaprint classification, n = 553 Low-risk High-risk Recurrent score classification, n = 553 Low/intermediate-risk High-risk Pathological complete response (pCR), n = 553 No Yes Median follow-up, months (range), n = 299 DFS, n = 302 No event Event 5-year DFS, n = 299

N

EndoPredict classification Low-risk (n = 283)

High-risk (n = 270)

298 255

163 (58%) 120 (42%)

135 (50%) 135 (50%)

40 512

25 (9%) 257 (91%)

15 (6%) 255 (94%)

183 336

114 (42%) 159 (58%)

69 (28%) 177 (72%)

179 10 43

86 (75%) 8 (7%) 20 (18%)

93 (79%) 2 (2%) 23 (19%)

47 272 192

38 (15%) 163 (63%) 59 (23%)

9 (4%) 109 (43%) 133 (53%)

317 236

119 (42%) 164 (58%)

198 (73%) 72 (27%)

210 343

167 (59%) 116 (41%)

43 (16%) 227 (84%)

496 57

281 (99%) 2 (1%)

215 (80%) 55 (20%)

489 64 299

264 (93%) 19 (7%) 42 (9–88)

225 (83%) 45 (17%) 38 (6–165)

259 43

139 (91%) 14 (9%) 88% (81–95)

120 (81%) 29 (19%) 73% (63–85)

p-value

Statistic

0.088

1.4 [0.96–1.93]

0.143

1.6 [0.82–3.45]

1.27E-03

1.8 [1.25–2.71]

0.14

1.64E-13

IDC, invasive ductal cancer; ILC, invasive lobular cancer; DFS, disease-free survival.

8.82E-14

0.26 [0.18–0.38]

1.81E-26

7.6 [4.99–11.65]

2.56E-16

36 [9.27–306.15]

2.88E-04

2.8 [1.54–5.18]

0.3936 1.31E-02

1.56E-02

2.4 [1.16–5.14]

F. Bertucci et al./Cancer Letters 355 (2014) 70–75

B

A

15

15

EP, high-risk (N = 270, 17% pCR)

10

EndoPredict score

EndoPredict score

73

5

p = 4.17E-06

10

5

EP low-risk (N = 283, 7% pCR) 0

0

0

100

200

300

400

no

500

Sample

yes pCR

Fig. 1. Correlation between the EP-based classification and the pathological response to chemotherapy. (A) All samples are ordered according to their EP score. The orange horizontal line delimitates the EP low-risk group (below) and the EP high-risk group (above). Patients with pCR are indicated by red dots and patients without pCR by black dots. (B) Box-plot of the EP score according to the response to chemotherapy (Student’s t-test). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Discussion We searched for correlations between the EP classification and the response to chemotherapy in ER+/HER2– breast cancer. Through a retrospective analysis of gene expression data of 553 pre-treatment samples, we show that tumours with an EP high-risk score are associated with higher pCR rate to chemotherapy than tumours with a low-risk score. To our knowledge, this is the first demonstration of correlation between the EP signature and the response to chemotherapy. The EP signature was developed as a multigene signature to predict the risk of metastatic relapse in ER+/HER2– breast cancers treated with adjuvant hormone therapy [5]. Two scores were defined: the EP score, based on mRNA expression levels of 11 genes, and the EPclin score, combining the EP score and two pathological tumour features defined on the operative specimen (tumour size pT, and axillary lymph node status pN). Retrospective analysis of two prospective clinical trials of adjuvant hormone therapy in postmenopausal patients showed that the EP score provided additional prognostic information to classical prognostic features [22], both for early and late relapses [23]. But its predictive value for the benefit of chemotherapy is unknown. This was the aim of our study. In this context, our primary endpoint was the pathological complete response to neoadjuvant chemotherapy, a rapid endpoint that is associated with increased long-term survival in breast cancer [24].

The neoadjuvant setting, and consequently the absence of information for pT and pN, led us to assess the EP score rather than the EPclin score. Importantly, it has already been shown that the measurements of EP score are comparable between core biopsies and surgical specimens [25]. Although different chemotherapy regimens were delivered in our patients series, all regimens were anthracycline-based and more than three-quarters were anthracycline and taxane-based. As expected in the neoadjuvant setting, our population showed poor-prognosis criteria, and the proportion of tumours with EP low-risk of relapse was 51%, a little less than the proportion (63%) reported in the adjuvant setting [22]. As expected, the EP-based classification was associated with classical prognostic features and survival, but also with the prognostic classifications based on the Mammaprint and Recurrence Score signatures. More importantly, the EP-based classification was also associated with pathological response to chemotherapy with a pCR rate 2.4 times larger in the high-risk group (17%) than in the lowrisk group (7%). Of all tested histoclinical variables, only tumour grade was also associated with pCR, but lost its predictive value in multivariate analysis, whereas EP remained significant. A possible mechanistic biological link between EP and chemosensitivity was proposed by our supervised analysis, which compared the gene expression profiles of EP low-risk versus EP high-risk ER+/HER2– tumours. Many genes upregulated in the high-risk tumours were involved in cell cycle and proliferation, classically associated with

Table 3 Univariate and multivariate analysis for pCR. Univariate

Age, > 50 years versus < = 50 years Histological type, ILC versus IDC Histological type, other versus IDC Clinical tumour size, cT2-4 versus cT1 Clinical axillary lymph node status, cN1-3 versus cN0 PR status, 1 versus 0 Grade, 2 versus 1 Grade, 3 versus 1 EP-based classification score IDC, invasive ductal cancer; ILC, invasive lobular cancer.

Multivariate

N

Odds ratio [CI95]

p-value

N

Odds ratio [CI95]

p-value

553 232

0.62 [0.39–0.97] 2.77 [0.75–8.66] 0.85 [0.33–1.91] 0.72 [0.35–1.66] 1.33 [0.8–2.27] 0.73 [0.46–1.14] 1.22 [0.4–5.53] 5.92 [2.05–26.01] 1.27 [1.17–1.38]

0.08 0.16 0.76 0.49 0.36 0.25 0.796 1.68E-02 6.28E-07

511 511 511

1.03 [0.33–4.71] 3.93 [1.29–17.7] 1.13 [1.04–1.24]

0.97 0.07 1.99E-02

552 519 553 511 553

74

F. Bertucci et al./Cancer Letters 355 (2014) 70–75

Hess’ set (N = 77)

A

Metagene EP score ~ EP groups

Metagene EP score ~ EP groups 1.0

True positive rate (sensitivity)

1.0 True positive rate (sensitivity)

Tabchy’s set (N = 93)

B

0.8 0.6 0.4 0.2 AUC = 0.93

0.0

0.8 0.6 0.4 0.2 AUC = 0.89

0.0 0.0

0.2

0.4

0.6

0.8

1.0

0.0

False positive rate (1-specificity)

low-risk high-risk

0.4

0.6

0.8

1.0

False positive rate (1-specificity)

EP classifications

Predicted

0.2

EP classifications

Observed low-risk high-risk 36 5 6 30

Predicted

low-risk high-risk

Accuracy = 86% p = 1.6E-10

Observed low-risk high-risk 38 9 10 36 Accuracy = 80% p = 9.4E-09

Fig. 2. Independent validation of the EP high-risk/low-risk gene expression signature. (A) Hess validation set. Top, ROC curve of the “metagene EP score”. The high area under curve (AUC = 0.93) reflects the strong positive correlation between the predicted and observed EP classifications. Bottom, cross-tables between the two classifications (Fisher’s exact test). (B) Similar to A, but in the Tabchy validation set.

chemosensitivity [26], whereas many genes upregulated in the low-risk tumours were involved in cell differentiation, adhesion, ER-signalling and stroma known to be associated with poor chemosensitivity [27,28]. In agreement with this observation, EP remained significant (continuous value, p < 0.001) in multivariate analysis for pCR (data not shown) when confronted with the mRNA expression of Ki67 proliferation marker (continuous value, p = 0.31), which was however very significant in univariate analysis (p < 0.001). Our results thus suggest that patients with poor-prognosis EP profile are more likely to achieve pCR than patients with goodprognosis EP profile. A similar correlation has been reported with

other prognostic signatures such as Recurrence Score [7], Mammaprint [8] and Genomic Grade Index [9], but in tumour series heterogeneous in terms of ER and HER2 status, whereas our present series included only ER+/HER2– samples. Interestingly, this relation was confirmed in the adjuvant setting for Recurrence Score [29,30] and Mammaprint [31] where high-risk tumours showed greater benefit from adjuvant chemotherapy than low-risk tumours. Recently, the Spanish GEICAM group [32] showed that EP was an independent prognostic parameter in node-positive, ER+/HER2– breast cancer patients treated with adjuvant chemotherapy followed by hormonal therapy in the GEICAM/9906 trial. Similarly, in

Table 4 Top 10 ontologies associated with the genes differentially expressed between EP high-risk versus EP low-risk breast cancers. Term ID

GO:0000278 GO:0022403 GO:0000279 GO:0000280 GO:0007067 GO:0000087 GO:0048285 GO:0051301 GO:0006260 GO:0006259 GO:0032989 GO:0007243 GO:0007167 GO:0042060 GO:0022610 GO:0007155 GO:0042127 GO:0002526 GO:0009611 GO:0006414

Term

Mitotic cell cycle Cell cycle phase M phase Nuclear division Mitosis M phase of mitotic cell cycle Organelle fission Cell division DNA replication DNA metabolic process Cellular component morphogenesis Protein kinase cascade Enzyme linked receptor protein signalling pathway Wound healing Biological adhesion Cell adhesion Regulation of cell proliferation Acute inflammatory response Response to wounding Translational elongation

Up in EP high-risk versus low-risk

Down in EP high-risk versus low-risk

N° genes

p-value

N° genes

p-value

65 63 55 42 42 42 42 45 33 49

2.31E-32 9.98E-28 3.59E-26 3.47E-22 3.47E-22 7.05E-22 1.67E-21 9.06E-20 3.23E-16 1.99E-13 23 22 21 15 35 35 38 11 31 13

4.56E-04 4.49E-04 4.19E-04 3.38E-04 1.82E-04 1.81E-04 1.78E-04 1.74E-04 2.87E-05 8.39E-06

F. Bertucci et al./Cancer Letters 355 (2014) 70–75

our series of patients treated with neoadjuvant chemotherapy and adjuvant hormone therapy, EP was associated with prognosis with 5-year DFS equal to 73% in the high-risk group versus 88% in the low-risk group, despite higher sensitivity to chemotherapy in the high-risk group. It is thus probable that the increased chemosensitivity of high-risk tumours – despite a 70% positive predictive value observed here – does not sufficiently compensate their high risk of relapse and putative lower sensitivity to hormone therapy (which may be related to their lower expression of ER signalling). In conclusion, we show that ER+/HER2– breast cancers with a poor-prognosis EP profile are more sensitive to anthracyclinebased chemotherapy than cancers with a good-prognosis EP profile. This result reinforces the justification to avoid adjuvant chemotherapy in EP good-prognosis patients who not only display a lower risk of relapse but also display a low probability to respond to chemotherapy. In addition to its originality, our study presents several strengths: the size of our series (to date, 553 samples represent the largest series of ER+/HER2– breast cancers profiled using DNA microarrays and analysed with respect to pCR to chemotherapy), the use of anthracycline-based chemotherapy for all patients, and the use of a consistent technological platform (Affymetrix). Limitations of our study come from its retrospective nature and the neoadjuvant setting. The next step to demonstrate the value of the EP signature for predicting the benefit of adjuvant chemotherapy in ER+/HER2– early breast cancer will be to retrospectively test the signature in randomised prospective clinical trials of adjuvant chemotherapy by testing the hypothesis that the difference of DFS between the patients treated and those untreated with chemotherapy is larger in the EP poor-prognosis group than in the EP goodprognosis group. Such study will be greatly facilitated by the possibility to work from RNA extracted from FFPE samples. Conflict of interest The authors declare that they have no conflict of interest.

[8]

[9]

[10]

[11]

[12]

[13]

[14]

[15]

[16]

[17]

[18]

[19]

[20]

[21]

Acknowledgments [22]

Our work is supported by Institut Paoli-Calmettes, Inserm, Institut National du Cancer (AOPL2010 IVOIRES), Ligue Nationale contre le Cancer (label DB) and SIRIC INCa-DGOS-Inserm 6038.

[23]

Appendix: Supplementary material

[24]

Supplementary data to this article can be found online at doi:10.1016/j.canlet.2014.09.014.

[25]

References

[26]

[1] A. Goldhirsch, E.P. Winer, A.S. Coates, R.D. Gelber, M. Piccart-Gebhart, B. Thurlimann, et al., Personalizing the treatment of women with early breast cancer: highlights of the St Gallen International Expert Consensus on the Primary Therapy of Early Breast Cancer 2013, Ann. Oncol. 24 (2013) 2206–2223. [2] R. Sabatier, A. Goncalves, F. Bertucci, Personalized medicine: present and future of breast cancer management, Crit. Rev. Oncol. Hematol. (2014) In press. [3] M.J. van de Vijver, Y.D. He, L.J. van’t Veer, H. Dai, A.A. Hart, D.W. Voskuil, et al., A gene-expression signature as a predictor of survival in breast cancer, N. Engl. J. Med. 347 (2002) 1999–2009. [4] S. Paik, S. Shak, G. Tang, C. Kim, J. Baker, M. Cronin, et al., A multigene assay to predict recurrence of tamoxifen-treated, node-negative breast cancer, N. Engl. J. Med. 351 (2004) 2817–2826. [5] M. Filipits, M. Rudas, R. Jakesz, P. Dubsky, F. Fitzal, C.F. Singer, et al., A new molecular predictor of distant recurrence in ER-positive, HER2-negative breast cancer adds independent information to conventional clinical risk factors, Clin. Cancer Res. 17 (2011) 6012–6020. [6] R. Kronenwett, K. Bohmann, J. Prinzler, B.V. Sinn, F. Haufe, C. Roth, et al., Decentral gene expression analysis: analytical validation of the Endopredict genomic multianalyte breast cancer prognosis test, BMC Cancer 12 (2012) 456. [7] L. Gianni, M. Zambetti, K. Clark, J. Baker, M. Cronin, J. Wu, et al., Gene expression profiles in paraffin-embedded core biopsy tissue predict response to

[27]

[28]

[29]

[30]

[31]

[32]

75

chemotherapy in women with locally advanced breast cancer, J. Clin. Oncol. 23 (2005) 7265–7277. M.E. Straver, A.M. Glas, J. Hannemann, J. Wesseling, M.J. van de Vijver, E.J. Rutgers, et al., The 70-gene signature as a response predictor for neoadjuvant chemotherapy in breast cancer, Breast Cancer Res. Treat. 119 (2010) 551–558. C. Liedtke, C. Hatzis, W.F. Symmans, C. Desmedt, B. Haibe-Kains, V. Valero, et al., Genomic grade index is associated with response to chemotherapy in patients with breast cancer, J. Clin. Oncol. 27 (2009) 3185–3191. R. Sabatier, P. Finetti, J. Adelaide, A. Guille, J.P. Borg, M. Chaffanet, et al., Down-regulation of ECRG4, a candidate tumor suppressor gene, in human breast cancer, PLoS ONE 6 (2011) e27656. H. Bonnefoi, A. Potti, M. Delorenzi, L. Mauriac, M. Campone, M. Tubiana-Hulin, et al., Validation of gene signatures that predict the response of breast cancer to neoadjuvant chemotherapy: a substudy of the EORTC 10994/BIG 00–01 clinical trial, Lancet Oncol. 8 (2007) 1071–1078. C. Hatzis, L. Pusztai, V. Valero, D.J. Booser, L. Esserman, A. Lluch, et al., A genomic predictor of response and survival following taxane-anthracycline chemotherapy for invasive breast cancer, JAMA 305 (2011) 1873–1881. K.R. Hess, K. Anderson, W.F. Symmans, V. Valero, N. Ibrahim, J.A. Mejia, et al., Pharmacogenomic predictor of sensitivity to preoperative chemotherapy with paclitaxel and fluorouracil, doxorubicin, and cyclophosphamide in breast cancer, J. Clin. Oncol. 24 (2006) 4236–4244. T. Iwamoto, G. Bianchini, D. Booser, Y. Qi, C. Coutant, C.Y. Shiang, et al., Gene pathways associated with prognosis and chemotherapy sensitivity in molecular subtypes of breast cancer, J. Natl Cancer Inst. 103 (2011) 264–272. V. Popovici, W. Chen, B.G. Gallas, C. Hatzis, W. Shi, F.W. Samuelson, et al., Effect of training-sample size and classification difficulty on the accuracy of genomic predictors, Breast Cancer Res. 12 (2010) R5. A. Tabchy, V. Valero, T. Vidaurre, A. Lluch, H. Gomez, M. Martin, et al., Evaluation of a 30-gene paclitaxel, fluorouracil, doxorubicin, and cyclophosphamide chemotherapy response predictor in a multicenter randomized trial in breast cancer, Clin. Cancer Res. 16 (2010) 5351–5361. B. Chevallier, P. Chollet, Y. Merrouche, H. Roche, P. Fumoleau, P. Kerbrat, et al., Lenograstim prevents morbidity from intensive induction chemotherapy in the treatment of inflammatory breast cancer, J. Clin. Oncol. 13 (1995) 1564–1571. R.A. Irizarry, B. Hobbs, F. Collin, Y.D. Beazer-Barclay, K.J. Antonellis, U. Scherf, et al., Exploration, normalization, and summaries of high density oligonucleotide array probe level data, Biostatistics 4 (2003) 249–264. B.D. Lehmann, J.A. Bauer, X. Chen, M.E. Sanders, A.B. Chakravarthy, Y. Shyr, et al., Identification of human triple-negative breast cancer subtypes and preclinical models for selection of targeted therapies, J. Clin. Invest. 121 (2011) 2750–2767. V.G. Tusher, R. Tibshirani, G. Chu, Significance analysis of microarrays applied to the ionizing radiation response, Proc. Natl Acad. Sci. U.S.A. 98 (2001) 5116–5121. L.M. McShane, D.G. Altman, W. Sauerbrei, S.E. Taube, M. Gion, G.M. Clark, et al., Reporting recommendations for tumour MARKer prognostic studies (REMARK), Br. J. Cancer 93 (2005) 387–391. P. Dubsky, M. Filipits, R. Jakesz, M. Rudas, C.F. Singer, R. Greil, et al., EndoPredict improves the prognostic classification derived from common clinical guidelines in ER-positive, HER2-negative early breast cancer, Ann. Oncol. 24 (2013b) 640–647. P. Dubsky, J.C. Brase, R. Jakesz, M. Rudas, C.F. Singer, R. Greil, et al., The EndoPredict score provides prognostic information on late distant metastases in ER+/HER2- breast cancer patients, Br. J. Cancer 109 (2013a) 2959–2964. P. Cortazar, L. Zhang, M. Untch, K. Mehta, J.P. Costantino, N. Wolmark, et al., Pathological complete response and long-term clinical benefit in breast cancer: the CTNeoBC pooled analysis, Lancet 384 (2014) 164–172. B.M. Muller, J.C. Brase, F. Haufe, K.E. Weber, J. Budzies, C. Petry, et al., Comparison of the RNA-based EndoPredict multigene test between core biopsies and corresponding surgical breast cancer sections, J. Clin. Pathol. 65 (2012) 660–662. M. Colleoni, G. Viale, A. Goldhirsch, Lessons on responsiveness to adjuvant systemic therapies learned from the neoadjuvant setting, Breast 18 (Suppl. 3) (2009) S137–S140. P. Farmer, H. Bonnefoi, P. Anderle, D. Cameron, P. Wirapati, V. Becette, et al., A stroma-related gene signature predicts resistance to neoadjuvant chemotherapy in breast cancer, Nat. Med. 15 (2009) 68–74. M. Ignatiadis, S.K. Singhal, C. Desmedt, B. Haibe-Kains, C. Criscitiello, F. Andre, et al., Gene modules and response to neoadjuvant chemotherapy in breast cancer subtypes: a pooled analysis, J. Clin. Oncol. 30 (2012) 1996–2004. K.S. Albain, W.E. Barlow, S. Shak, G.N. Hortobagyi, R.B. Livingston, I.T. Yeh, et al., Prognostic and predictive value of the 21-gene recurrence score assay in postmenopausal women with node-positive, oestrogen-receptor-positive breast cancer on chemotherapy: a retrospective analysis of a randomised trial, Lancet Oncol. 11 (2010) 55–65. S. Paik, G. Tang, S. Shak, C. Kim, J. Baker, W. Kim, et al., Gene expression and benefit of chemotherapy in women with node-negative, estrogen receptorpositive breast cancer, J. Clin. Oncol. 24 (2006) 3726–3734. M. Knauer, S. Mook, E.J. Rutgers, R.A. Bender, M. Hauptmann, M.J. van de Vijver, et al., The predictive value of the 70-gene signature for adjuvant chemotherapy in early breast cancer, Breast Cancer Res. Treat. 120 (2010) 655–661. M. Martin, J.C. Brase, L. Calvo, K. Krappmann, M. Ruiz-Borrego, K. Fisch, et al., Clinical validation of the EndoPredict test in node-positive chemotherapytreated ER+/HER2- breast cancer patients: results from the GEICAM/9906 trial, Breast Cancer Res. 16 (2014) R38.