Access the full text.
Sign up today, get DeepDyve free for 14 days.
www.nature.com/npjbcancer All rights reserved 2374-4677/15 ARTICLE OPEN A multifactorial ‘Consensus Signature’ by in silico analysis to predict response to neoadjuvant anthracycline-based chemotherapy in triple-negative breast cancer 1,4 2,4 2 1,3 2,5 1,5 Natalie Turner , Mattia Forcato , Simona Nuzzo , Luca Malorni , Silvio Bicciato and Angelo Di Leo BACKGROUND: Owing to the complex processes required for anthracycline-induced cytotoxicity, a prospectively deﬁned multifactorial Consensus Signature (ConSig) might improve prediction of anthracycline response in triple-negative breast cancer (TNBC) patients, whose only standard systemic treatment option is chemotherapy. AIMS: We aimed to construct and evaluate a multifactorial signature, comprising measures of each function required for anthracycline sensitivity in TNBC. METHODS: ConSigs were constructed based on ﬁve steps required for anthracycline function: drug penetration, nuclear topoisomerase IIα (topoIIα) protein location, increased topoIIα messenger RNA (mRNA) expression, apoptosis induction, and immune activation measured by, respectively, HIF1α or SHARP1 signature, LAPTM4B mRNA, topoIIα mRNA, Minimal Gene signature or YWHAZ mRNA, and STAT1 signature. TNBC patients treated with neoadjuvant anthracycline-based chemotherapy without taxane were identiﬁed from publicly available gene expression data derived with Affymetrix HG-U133 arrays (training set). In silico analyses of correlation between gene expression data and pathological complete response (pCR) were performed using receiver-operating characteristic curves. To determine anthracycline speciﬁcity, ConSigs were assessed in patients treated with anthracycline plus taxane. Speciﬁcity, sensitivity, positive and negative predictive value, and odds ratio (OR) were calculated for ConSigs. Analyses were repeated in two validation gene expression data sets derived using different microarray platforms. RESULTS: In the training set, 29 of 147 patients had pCR after anthracycline-based chemotherapy. Various combinations of components were evaluated, with the most powerful anthracycline response predictors being ConSig1: (STAT1+topoIIα mRNA +LAPTM4B) and ConSig2: (STAT1+topoIIα mRNA+HIF1α). ConSig1 demonstrated high negative predictive value (85%) and high OR for no pCR (3.18) and outperformed ConSig2 in validation sets for anthracycline speciﬁcity. CONCLUSIONS: With further validation, ConSig1 may help reﬁne selection of TNBC patients for anthracycline chemotherapy. npj Breast Cancer (2015) 1, 15003; doi:10.1038/npjbcancer.2015.3; published online 2 June 2015 INTRODUCTION single biomarker may not be sufﬁcient to predict anthracycline response, and a multifactorial approach using gene signatures Preferred breast cancer (neo)adjuvant chemotherapy regimens are 7,10 7 might be required. In the TOP trial of 149 patients with generally anthracycline based, given the improved outcomes estrogen receptor (ER)-negative early or locally advanced breast compared with cyclophosphamide/methotrexate/ﬂuorouracil. cancer treated with a single agent epirubicin, TOP2A ampliﬁcation However, across all anthracycline-treated patients, only a small was signiﬁcantly associated with pathological complete response percentage actually receives beneﬁt while these agents are (pCR). In addition, a multifactorial anthracycline sensitivity score associated with signiﬁcant toxicities. Breast cancer is well recognized as a heterogeneous disease and therefore treating (the A-score), comprised of three gene signatures, was evaluated, demonstrating a very high negative predictive value (NPV), all breast cancers with the same chemotherapeutic agents could be considered illogical. Of considerable use would be a predictive although a much lower positive predictive value (PPV). Gene signatures are often deﬁned through retrospective marker of response to distinguish patients likely to receive beneﬁt analyses of tumor tissue gene expression patterns correlated with from those who are not, sparing predicted ‘poor responders’ from patient outcomes. Although concordance in outcome prediction associated toxicities. Unfortunately, suitable predictive biomarkers between signatures has been demonstrated, signatures derived for chemotherapeutic agents have remained elusive to date. in this manner often contain few genes in common as well as Topoisomerase IIα gene (TOP2A) is a putative marker of anthracycline sensitivity, with its gene product being the direct numerous genes of unknown function, making their clinical target of anthracyclines. TOP2A ampliﬁcation has been shown to relevance less certain. It might be possible to improve clinical 2–7 predict increased sensitivity to anthracyclines in several studies, relevance of a signature by identifying the molecular processes 8,9 although this ﬁnding has not been entirely consistent. Indeed, a required for a speciﬁc cellular function, such as anthracycline- 1 2 ‘Sandro Pitigliani Medical Oncology Department, Prato Hospital, Istituto Toscano Tumori, Prato, Italy; Department of Biomedical Sciences, Centre for Genome Research, University of Modena and Reggio Emilia, Modena, Italy and Translational Research Unit, Prato Hospital, Istituto Toscano Tumori, Prato, Italy. Correspondence: A Di Leo (email@example.com) These authors contributed equally to this work. These authors contributed equally to this work. Received 17 March 2015; accepted 20 March 2015 © 2015 Breast Cancer Research Foundation/Macmillan Publishers Limited Anthracyclines in triple-negative breast cancer N Turner et al induced cytotoxicity, and construct a signature containing mea- For each step, a representative gene or gene signature was selected, with gene signatures chosen over single genes where possible. In some sures of each of these functions. On the basis of this hypothesis, cases, more than one potential genes/gene signatures was evaluated for we aimed to construct and evaluate a multifactorial Consensus each step, to compare their relative utility. Details of the marker(s) Signature for predicting anthracycline sensitivity in triple-negative evaluated, their association with anthracycline response, and rationale for breast cancer (TNBC). The term Consensus Signature was chosen their selection are listed in Table 1. to reﬂect the concept that, by each selected component acting as a surrogate marker of the different steps required for anthracy- Quantiﬁcation of genes and gene signatures cline cytotoxicity, included components (the genes and gene 14 14 SHARP1 signature, HIF1α hypoxia signature (HIF), and the Minimal signatures) would work synergistically to provide an overall Gene signature were quantiﬁed as previously described. Brieﬂy, each measure of effective anthracycline function. signature was calculated by summarizing the standardized expression We focused speciﬁcally on TNBC for the following reasons. First, levels of the genes in the signature into a combined score with zero mean. there has been substantial work performed already evaluating the AURKA, STAT1, and PLAU signatures were computed as previously predictive role of TOP2A in HER2 breast cancer, due to the known described using genefu R package. Brieﬂy, for each sample, the signature relationship between TOP2A ampliﬁcation and HER2 ampliﬁcation. was quantiﬁed as: s = Σ ω ξ /Σ |ω | where ξ is the expression of a gene i i i i i i I Anthracyclines are commonly used in TNBC and appear to have included in the set of genes of interest and ω is either +1 or − 1 depending activity. We wanted to assess this without the confounding factor on the sign of the association under study. Gene expression lists for each signature are included in Supplementary Table 2. of HER2 overexpression. Second, with treatment options in TNBC LAPTM4B, AURKA, YWHAZ, and topoIIα mRNA expression levels were limited to chemotherapy, more effective use of chemotherapy calculated using the corresponding probe sets or the median expression if would be of considerable beneﬁt. Finally, with growing under- multiple probe sets were available for each gene. For the NKI data set, standing of breast cancer biological diversity, evaluation of a probes were ﬁltered on the basis of their quality, keeping only probes predictive biomarker within a speciﬁc subtype might be prefer- classiﬁed as ‘perfect’ and ‘good’ in the Bioconductor illumina Human v3.db able, as positive results could otherwise be masked if evaluated annotation package. across a heterogeneous combined cohort. Quantiﬁcation of the Consensus Signature The Consensus Signature score was calculated in a continuous form, MATERIALS AND METHODS consisting of a linear combination of the various components/signatures. Data set Prior to combination, each component/signature was scaled to have the The construction and evaluation of consensus signatures (ConSigs) were interquartile range equal to 1 and the median equal to 0. On the basis of carried out using a retrospective cohort study design, with in silico analyses the association with pCR (Table 1), we hypothesized that a high Consensus of previously collected genetic data, clinical characteristics, and responses. Signature score would predict for increased pCR rate, whereas a low score The data set comprised gene expression proﬁles of patients who had should predict anthracycline resistance. received neoadjuvant anthracycline-based chemotherapy without a taxane. As the ConSigs were designed to be speciﬁc for anthracyclines, Statistical analysis taxane use was considered a confounding factor. All statistical analyses were performed in R version 2.15.1. Odds ratios Data derived from Affymetrix (Santa Clara, CA, USA) gene expression (ORs) were used to compare pCR rates between groups deﬁned by arrays based on build 133 of UniGene database (HG-U133) were combined different clinical and molecular characteristics (stats R package). The area and evaluated as a single group, designated the ‘breast compendium’ under the curve (AUC) was used to assess the prediction performance of (details on the construction and composition of the breast compendium any signature score (ROCR R package). AUC was estimated through the are reported in Supplementary Table 1). The subset of samples treated concordance index (survcomp R package) under the alternative hypothesis with anthracycline-based neoadjuvant chemotherapy was used as training that AUC was greater than 0.5, as each signature score was designed to set to derive the ConSigs, whereas the cohort of patients treated with have positive AUC. Its signiﬁcance and conﬁdence interval were estimated anthracycline plus taxane-based neoadjuvant chemotherapy served as a assuming asymptotic normality. Because of the differences in array design control group to assess ConSig speciﬁcity for anthracyclines. In addition, and technology between the training and the two validation sets, the two cohorts of patients treated with anthracycline-based neoadjuvant threshold ConSig score for each cohort was calculated using the score chemotherapy with gene expression data derived from different micro- value that corresponded to the 75th percentile of the score distribution for array platforms served as validation sets. The European Organisation for that cohort. P values of o0.05 were considered signiﬁcant. Research and Treatment of Cancer (EORTC)/BIG00-01 data set included gene expression data obtained with the Affymetrix X3P array from patients with locally advanced, inﬂammatory, or large operable breast cancers RESULTS treated with either ﬂuorouracil/epirubin/cyclophosphamide or docetaxel Data set followed by docetaxel/epirubicin under the auspices of the EORTC 10994 trial. The data set was available in the Gene Expression Omnibus The training set was derived from a breast cancer data set repository under accession number GSE6861. The Netherlands Cancer originally consisting of 4,600 samples collected in 27 different Institute (NKI) data set (accessible under accession number GSE34138) studies. After exclusion of duplicate samples (n = 939) and used the Illumina (San Diego, CA, USA) HumanWG 6 v3.0 expression 17 adjusting for batch effect using ComBat, 1,069 samples (29%) beadchip for gene expression proﬁling and included patients with were classiﬁed as TNBC by the SCMOD2 subtype clustering intermediate or high-risk ER-breast cancer treated with neoadjuvant 13 classiﬁer (subtype clustering model) contained in genefu R dose-dense doxorubicin/cyclophosphamide. package. Among these samples, 491 had information about For this study, only TNBC patients in the training and control subsets of neoadjuvant chemotherapy. Eleven samples were from patients the compendium and in the validation sets were considered. Further treated with taxane but not anthracycline and were excluded, details on the construction of the breast compendium and on the deﬁnition of TNBCs are reported in the Supplementary Methods. whereas 147 and 333 samples were from patients treated with anthracycline-based therapy without taxane and anthracycline- based therapy with taxane, respectively (Figure 1). Design of the consensus signature Clinical and tumor characteristics for the patients in the training In order for anthracyclines to be effective, we postulated that the following set (n = 147) treated with anthracycline-based chemotherapy are steps must occur: (1) penetration of the drug into the tumor bed, (2) listed in Table 2. The samples were originally contained in four location of the target (topoIIα protein) within the nucleus, (3) increased 7,20–22 different data sets, and included 29 (19.7%) pCRs and 118 topoIIα messenger RNA (mRNA) expression above that related to proliferation alone, (4) induction of apoptosis, and (5) active immune/ (80.3%) samples with residual disease. pCR was deﬁned as ypT0/is, stromal function. ypN0 in all included studies. npj Breast Cancer (2015) 15003 © 2015 Breast Cancer Research Foundation/Macmillan Publishers Limited Anthracyclines in triple-negative breast cancer N Turner et al Table 1. Consensus Signature components based on putative steps required for effective anthracycline-induced cytotoxicity Step Surrogate marker Association with Rationale pCR Penetration of drug into SHARP1 signature Negative Hypoxia, promoted by HIFs, is a well-known contributor to decreased drug the tumor bed Hypoxia signature Negative penetration, and chemoresistance. Montagner et al. recently described a (HIF) hypoxia signature of 22 genes, with increased expression correlated with increased HIF activity. A direct interaction between SHARP1 (a downstream target of the tumor suppression gene p63) and HIF1α and HIF2α was demonstrated, with a signature of low SHARP1 activity in TNBC conferring increased HIF function and increased hypoxia. With the SHARP1 signature measuring low SHARP activity and thus increased HIF function, it has a negative association with pCR. Location of topoIIα protein LAPTM4B Negative In order to work effectively, the target of anthracyclines, topoIIα protein, must within the nucleus have access to nuclear DNA; thus, it must be located in the nucleus. Nuclear 3,4 export of topoIIα protein may contribute to anthracycline resistance. topoIIα protein nuclear location might be inferred using the expression level of LAPTM4B. LAPTM4B gene resides on chromosome 8q22, with overexpression shown to increase sequestration of anthracyclines in the cytoplasm. Increased levels of LAPTM4B mRNA have been correlated with increased anthracycline resistance, whereas selective depletion of LAPTM4B signiﬁcantly increased sensitivity to anthracycline, but not cisplatin or taxane, chemotherapy. Increased expression topoIIα mRNA Positive TOP2A transcription can be enhanced by proliferation signals independently of of topoIIα mRNA, topoIIα mRNA: AURKA Positive gene aberrations and topoIIα protein is strongly inﬂuenced by proliferation. independent of topoIIα mRNA: AURKA Positive Increased expression of topoIIα protein therefore may be seen in the setting of proliferation signature highly proliferating tumors, without correlating with an increased likelihood of response speciﬁcally to anthracyclines. By determining the ratio of expression of topoIIα mRNA relative to that of a known proliferation marker (Aurora kinase A 7,8 gene, AURKA, or AURKA gene signature), tumors with increased topoIIα mRNA independent of proliferation might be determined. Induction of apoptosis YWHAZ Negative The anti-apoptotic gene YWHAZ (coding for 14-3-3ζ) resides on chromosome Minimal gene Positive 8q22 close to LAPTM4B gene and may promote de novo anthracycline signature (MS) resistance. Increased expression has been associated with increased doxorubicin resistance in breast cancer cell lines, and early relapses after anthracycline chemotherapy. siRNA knockdown of YWHAZ in breast cancer cell lines signiﬁcantly increased doxorubicin-induced apoptosis. An alternate marker of apoptosis is the MS, comprising two genes, SHARP1 and CCNG2.As with SHARP1, CCNG2 is a downstream target of p63. As p63 is inhibited by mutant p53, lack of MS expression implies dysfunction in the p53 pathway, the major apoptotic pathway in the presence of oncogenic stress, and may be a suitable surrogate for lack of apoptosis. Active immune function Immune function Positive Both innate and adaptive immune responses are important in anthracycline 10–13 signature (STAT1) toxicity. Anthracyclines trigger immunogenic cell death by eliciting tumor- Stromal signature Negative speciﬁcIFNγ CD8 cytotoxic T lymphocytes, thus an anthracycline-induced (PLAU) anticancer immune response can help eradicate residual cancer cells, or maintain 14,15 residual cells in state of dormancy. Moreover, immune module scores have been associated with higher probability of achieving pCR after anthracycline± taxane chemotherapy among all breast cancer subtypes when deﬁned by immunohistochemistry. Closely related to immune function, stromal signatures 15,17 may also be useful in predicting anthracycline sensitivity or resistance. Abbreviations: HIF, hypoxia-inducible factor; IFNγ, interferon gamma; LAPTM4B, lysosomal-associated protein transmembrane 4B gene; mRNA, messenger RNA; MS, minimal gene signature; pCR, pathological complete response; siRNA, small interfering RNA; TNBC, triple-negative breast cancer. References for this table are listed in Supplementary Materials. Clinical characteristics pCR status (Supplementary Table 3; Supplementary Figure 1a). All other components, when considered individually, were not All clinical variables were tested for their ability to predict pCR, signiﬁcantly correlated with pCR. TopoIIα mRNA corrected for with no signiﬁcant association between any clinical characteristics proliferation with either AURKA mRNA or AURKA signature and pCR found (data not shown). showed no increased correlation with pCR compared with topoIIα mRNA alone (data not shown). Predictive power of single gene or gene signatures Using receiver-operating characteristic curves, the ability of any Predictive power of ConSigs single component (gene or gene signature) to discriminate ConSigs was constructed using various combinations of compo- patients with pCR from patients with residual disease in the nents, with the starting point being components shown to training set was assessed. STAT1 was signiﬁcantly associated with have signiﬁcant or near-signiﬁcant predictive capability when © 2015 Breast Cancer Research Foundation/Macmillan Publishers Limited npj Breast Cancer (2015) 15003 Anthracyclines in triple-negative breast cancer N Turner et al 27 BC studies with gene expression profiles N=4600 Duplicates N=939 N=3661 Non-TNBC N=2592 TNBC/Basal-like N=1069 No NAC data N=578 NAC data available N=491 T without A N=11 A-based NAC A + T-based NAC N=147 N=333 Response to NAC unknown N=34 A + T-based NAC with known response N=299 pCR No pCR pCR No pCR N=29 N=118 N=101 N=198 Figure 1. Consort diagram for selection of samples in the training set. A, anthracycline-based chemotherapy; BC, breast cancer; NAC, neoadjuvant chemotherapy; pCR, pathological complete response; T, taxane; TNBC, triple-negative breast cancer. used alone, that is, STAT1, topoIIα, HIF, and LAPTM4B. Using Classiﬁcation performance of ConSigs a continuous score to quantify ConSig expression level, all A threshold score that could be used to classify a patient in the combinations of core components demonstrated a signiﬁcant training set as a putative responder or as resistant was determined correlation with pCR in patients in the training set treated with for each ConSig by selecting the score value corresponding to the anthracycline-based chemotherapy without taxane. The two 75th percentile of the score distribution. PPV and NPVs, sensitivity, most predictive combinations were designated ConSig1: (STAT1 speciﬁcity, and OR were then calculated. For ConSig1 NPV was +topoIIα+LAPTM4B) with AUC 0.70 (Supplementary Figure 1b), high (85%) as was OR for lack of pCR (OR = 3.18, P = 0.008) − 5 P = 3.9 × 10 , and ConSig2: (STAT1+topoIIα+HIF) with AUC 0.71, (Table 4). PPV, however, was modest (PPV = 35%). ConSig2 − 6 P = 4.2 × 10 . High correlation with pCR was maintained with the performed similarly, with high NPV and OR, but modest PPV. In addition of further component genes/gene signatures to either the control group of 299 patients treated with anthracycline plus ConSig1 or ConSig2, but overall predictive power was not better taxane, NPVs for ConSig1 and ConSig2 were lower, at 66 and 67%, than with three components (Table 3). The combination of STAT1 respectively, and ORs for lack of pCR were no longer statistically +PLAU, the components for TNBC of another multifactorial scoring signiﬁcant for either ConSig (Table 4), further supporting the signature, the A-score, was correlated with pCR, although less speciﬁcity of these ConSigs for anthracycline response. strongly than other combinations. Substituting topoIIα mRNA with topoIIα mRNA corrected for proliferation did not improve the Evaluation of ConSig1 and ConSig2 in two independent ‘validation’ performance of any of the ConSigs. data sets To assess speciﬁcity of ConSigs for anthracycline response Two data sets, NKI and EORTC/BIG00-01, were selected as compared with other chemotherapy regimens, we analyzed their validation sets. The EORTC/BIG00-01 data set comprised 161 respective performances in a control group of patients who received samples, 85 of which were TNBC, with 46 patients treated with taxane in addition to anthracycline (n = 333), 299 of whom had anthracycline without taxane (18 pCRs) and 39 treated with information about response and 101 with pCR. For the ConSigs with anthracycline plus taxane (18 pCRs). The NKI data set included the highest predictive power in the training set, i.e., ConSig1 and 178 ER-negative breast cancer samples, 52 of which were TNBC. Of ConSig2, neither was correlated with pCR in this control group these 52 patients, 24 had pCR. All patients received anthracycline- (Supplementary Figure 1c and d). Although (STAT1+PLAU) had based neoadjuvant chemotherapy without taxane. predicted response to anthracycline-based chemotherapy, it did not In the NKI data set, both ConSig1 and ConSig2 were signi- appear to be anthracycline speciﬁc, performing similarly in anthracycline+taxane-treated patients (Table 3). ﬁcantly correlated with pCR for patients receiving anthra- npj Breast Cancer (2015) 15003 © 2015 Breast Cancer Research Foundation/Macmillan Publishers Limited Anthracyclines in triple-negative breast cancer N Turner et al cycline without taxane (Table 5). In the EORTC/BIG00-01 DISCUSSION data set, ConSig1 remained speciﬁc for anthracyclines in the Overall, the best performing combination of components was data set, whereas ConSig2 was correlated with pCR for both ConSig1 (STAT1+topoIIα+LAPTM4B). Our results suggest that Con- anthracycline-based and anthracycline+taxane-based neoadju- Sig1 has excellent ability to predict anthracycline resistance within a vant chemotherapy. cohort of anthracycline-treated TNBC patients. This is clinically relevant, as, if further validated, ConSig1 couldbeusedtoidentify TNBC patients for whom the addition of anthracycline is likely to add toxicity without beneﬁt, and thus for whom an alternate chemotherapy regimen might be selected. When evaluated in TNBC Table 2. Clinical and tumor characteristics of TNBC patients in the patients who received anthracycline and taxane, the predictive training set treated with anthracycline-based neoadjuvant ability of ConSig1 was lost. Although not conclusive, a lack of chemotherapy without taxane (n= 147) predictive utility in patients who also received taxanes supports the anthracycline speciﬁcity of ConSig1. ConSig2 was most strongly Characteristic No. of patients (%) correlated with pCR in the training set; however, in the validation set ConSig2 did not show discrimination in performance between Age (years) o40 12 (8) patients treated with anthracycline-based and anthracycline plus 40–60 40 (27) taxane-based neoadjuvant chemotherapy. 460 7 (5) Although we considered that ﬁve main processes should take NA 88 (60) place for anthracyclines to cause cell death, markers for two of these processes, that is, induction of apoptosis and hypoxia/drug Tumor size penetration, did not appear to contribute signiﬁcantly to the T1 15 (10) T2 90 (61) predictive power of ConSig1. Given that immune function (STAT1) T3 22 (15) was a powerful contributor to ConSig1 and ConSig2, we postulated T4 19 (13) that in the setting of a highly active immune response, intact NA 1 (1) apoptotic pathways as measured by Minimal Gene signature or Nodal status N0 14 (9) N1 13 (9) Table 4. Performance of ConSig1 and ConSig2 in the training set for N2 6 (4) predicting pathological complete response NA 114 (78) ConSig NAC NPV (%) PPV (%) OR 95% CI P value Histologic grade − 3 G1 2 (1) ConSig1 A 85 35 3.18 1.34–7.54 8.2 × 10 G2 25 (17) A+T 66 35 1.05 0.60–1.82 0.85 G3 112 (76) − 2 NA 8 (6) ConSig2 A 85 32 2.63 1.10–6.20 2.8 × 10 A+T 67 37 1.23 0.71–2.12 0.45 pCR Abbreviations: A, anthracycline-based chemotherapy; A+T, anthracycline Yes 29 (20) +taxane chemotherapy; CI, conﬁdence interval; ConSig, Consensus No 118 (80) Signature; NAC, neoadjuvant chemotherapy; NPV, negative predictive Abbreviations: NA, not available; pCR, pathological complete response; value; OR, odds ratio for lack of pathological complete response; PPV, TNBC, triple-negative breast cancer. positive predictive value. Table 3. Correlation with pCR for various combinations of ConSig components in the training set ConSig combination Type of NAC AUC 95% CI P value − 3 STAT1+PLAU A 0.63 0.52–0.73 9.0 × 10 − 2 A+T 0.56 0.50–0.63 3.1 × 10 − 4 STAT1+topoII A 0.68 0.57–0.78 7.6 × 10 − 3 A+T 0.60 0.54–0.67 1.2 × 10 − 5 STAT1+topoIIα+LAPTM4B (ConSig1) A 0.70 0.60–0.80 3.9 × 10 A+T 0.55 0.48–0.61 0.08 − 6 STAT1+topoIIα+HIF1 (ConSig2) A 0.71 0.62–0.80 4.2 × 10 A+T 0.52 0.45–0.58 0.33 − 4 STAT1+topoIIα+YWHAZ A 0.66 0.56–0.76 6.8 × 10 A+T 0.54 0.48–0.61 0.10 − 4 STAT1+topoIIα+LAPTM4B+YWHAZ A 0.66 0.56–0.76 8.0 × 10 A+T 0.51 0.45–0.58 0.34 − 2 STAT1+topoIIα+LAPTM4B+MS+PLAU A 0.62 0.51–0.72 1.3 × 10 − 2 A+T 0.56 0.49–0.62 4.2 × 10 Abbreviations: AUC, area under the ROC curve; CI, conﬁdence interval; ConSig, Consensus Signature; HIF, hypoxia-inducible factor; LAPTM4B, lysosomal- associated protein transmembrane 4B gene; MS, minimal gene signature; NAC, neoadjuvant chemotherapy; pCR, pathological complete response; ROC, receiver-operating characteristic. For A-based chemotherapy, n= 147 with 29 pCRs. For A+T chemotherapy, n= 299 with 101 pCRs. A: anthracycline-based neoadjuvant chemotherapy without taxane; A+T: anthracycline plus taxane neoadjuvant chemotherapy. © 2015 Breast Cancer Research Foundation/Macmillan Publishers Limited npj Breast Cancer (2015) 15003 Anthracyclines in triple-negative breast cancer N Turner et al Table 5. Performance of ConSig1 and ConSig2 in the validation sets for predicting pathological complete response Validation set ConSig NAC AUC 95% CI P value NPV (%) PPV (%) OR 95% CI P value − 3 − 3 EORTC/BIG00-01 ConSig1 A 0.70 0.55–0.85 5.2 × 10 74 75 8.33 2.00–44.44 6.0 × 10 A+T 0.60 0.43–0.77 0.12 62 70 3.81 0.87–20.75 0.09 − 2 ConSig2 A 0.65 0.50–0.80 2.4 × 10 62 42 1.15 0.29–4.40 0.83 − 2 A+T 0.65 0.49–0.81 3.6 × 10 59 60 2.13 0.50–9.92 0.31 − 4 − 2 NKI ConSig1 A 0.74 0.61–0.87 1.3 × 10 64 77 5.95 1.53–29.99 1.6 × 10 − 2 ConSig2 A 0.64 0.49–0.78 3.1 × 10 59 62 2.30 0.65–8.86 0.20 Abbreviations: A, anthracycline-based chemotherapy; A+T, anthracycline+taxane chemotherapy; AUC, area under the ROC curve; CI, conﬁdence interval; ConSig, Consensus Signature; EORTC, European Organisation for Research and Treatment of Cancer; NAC, neoadjuvant chemotherapy; NKI, Netherlands Cancer Institute; NPV, negative predictive value; OR, odds ratio for lack of pathological complete response; PPV, positive predictive value; ROC, receiver-operating characteristic. YWHAZ might be less important. However, predictive power was PPV. Indeed, in the validation sets, where pCR rates are 39% (EORTC/BIG00-01) and 46% (NKI), PPV is higher than 75% (Table 5). not improved using either Minimal Gene signature or YWHAZ in Finally, our sample size was relatively limited. The training set absence of the immune/stromal components (STAT1 and PLAU) (data not shown). Although the inclusion of a component marker used contained all publicly available data on the same platform for of hypoxia had the highest correlation with pCR in the training set TNBC treated with anthracycline-based chemotherapy. However, (ConSig2), this combination did not show consistent anthracycline this incorporated only 147 TNBC patients. Thus, further validation of ConSig1 in independent TNBC cohorts is necessary. speciﬁcity, with similar performance in anthracycline- and In conclusion, this project demonstrated the feasibility of deﬁning anthracycline plus taxane-treated patients in the EORTC validation a multifactorial Consensus Signature for predicting anthracycline set. This might be because hypoxia is a critical factor in the response and, when applied in a TNBC patient cohort, ConSig1 was function of other chemotherapy agents, not only anthracyclines. highly predictive for anthracycline resistance. With further validation, Interestingly, although topoIIα protein expression is known to this signature may provide clinicians with ausefultoolfor improved relate to proliferation, correcting topoIIα mRNA expression level selection of TNBC patients for anthracyclines, potentially leading to for proliferation made no difference to the predictive utility of the better treatment tolerance and more effective therapy. ConSigs. The lack of discriminating ability of the proliferative markers, AURKA or AURKA signature, might relate to the fact that the majority of tumors in our data set (76%) were grade 3 ACKNOWLEDGMENTS with most of the rest being grade 2, and thus were all moderately We acknowledge the generous support of the Italian Association for Cancer Research to highly proliferating. With no comparator group of low- (AIRC) that provided funding for this project under the auspices of the AIRC Special proliferating tumors, the inﬂuence of a proliferation marker Program Molecular Clinical Oncology ‘5 per mille’ research initiative. cannot easily be evaluated. Furthermore, when considering a cohort of TNBC, proliferation is characteristically high, making the CONTRIBUTIONS ability to measure variability of topoIIα protein expression due to All the authors contributed to the conception and design of the manuscript and to low versus high proliferation arguably less critical. Although the the data analysis and interpretation. NT, MF, and SN contributed also to the collection inﬂuence of proliferation itself in the results is uncertain, the and assembly of the data. All authors have written and approved the manuscript. speciﬁcity of ConSig1 for anthracycline response over and above ADL and SB are the manuscript’s guarantors. that of proliferation is supported by the fact that it is not predictive in patients treated with anthracyclines plus taxanes, a situation where proliferation is still important. COMPETING INTERESTS Differently from the A-score, which evaluated ER-/HER2 and The authors declare no conﬂict of interest. ER-/HER2 tumors, we focused speciﬁcally on TNBC, a group where treatment is limited to chemotherapy, and thus where FUNDING optimization of chemotherapy regimen would be of considerable utility. Interestingly, the predictive ability of ConSig1 for TNBC The authors declare that no funding was received. patients treated with anthracycline-based chemotherapy without taxane appeared to be better than the combination of STAT1 REFERENCES +PLAU, the component gene signatures in the A-score for TNBC. 1 Peto R, Davies C, Godwin J, Gray R, Pan HC, Clarke M et al. Comparisons between By constructing a biologically relevant multifactorial ConSig,we different polychemotherapy regimens for early breast cancer: meta-analyses of hypothesized that this should predict anthracycline response, as long-term outcome among 100,000 women in 123 randomised trials. Lancet 2012; well as resistance; however, PPV of all ConSigs in the training set 379:432–444. were lower than anticipated, and only around 35% for ConSig1. 2 Di Leo A, Desmedt C, Bartlett JM, Piette F, Ejlertsen B, Pritchard KI et al. HER2 and Conversely, ConSig1 predicts likelihood of lack of response (NPV), TOP2A as predictive markers for anthracycline-containing chemotherapy regi- which is still a ﬁnding of considerable clinical utility if further mens as adjuvant treatment of breast cancer: a meta-analysis of individual validated. In general, high PPV for biomarkers has previously been patient data. Lancet Oncol 2011; 12: 1134–1142. 3 Di Leo A, Gancberg D, Larsimont D, Tanner M, Jarvinen T, Rouas G et al. HER-2 shown to be hard to achieve. Considering two well-established ampliﬁcation and topoisomerase IIalpha gene aberrations as predictive markers in biomarkers and the only two routinely used in breast cancer node-positive breast cancer patients randomly treated either with an management, ER and HER2, positivity predicts treatment response anthracycline-based therapy or with cyclophosphamide, methotrexate, and 5- in only around 50% of patients for endocrine therapy and less ﬂuorouracil. Clin Cancer Res 2002; 8: 1107–1116. than 40% for trastuzumab as a single agent. However, it is worth 4 Press MF, Sauter G, Buyse M, Bernstein L, Guzman R, Santiago A et al. Alteration of noting that the small number of pCRs in the training set (29/147; topoisomerase II-alpha gene in human breast cancer: association with respon- 19.7%) might have contributed to less than stable results for siveness to anthracycline-based chemotherapy. J Clin Oncol 2011; 29:859–867. npj Breast Cancer (2015) 15003 © 2015 Breast Cancer Research Foundation/Macmillan Publishers Limited Anthracyclines in triple-negative breast cancer N Turner et al 5 Slamon D, Eiermann W, Robert N, Pienkowski T, Martin M, Press M et al. Adjuvant 17 Johnson WE, Li C, Rabinovic A. Adjusting batch effects in microarray expression trastuzumab in HER2-positive breast cancer. N Engl J Med 2011; 365: 1273–1283. data using empirical Bayes methods. Biostatistics 2007; 8:118–127. 6 Arriola E, Rodriguez-Pinilla SM, Lambros MB, Jones RL, James M, Savage K et al. 18 Wirapati P, Sotiriou C, Kunkel S, Farmer P, Pradervand S, Haibe-Kains B et al. Meta- Topoisomerase II alpha ampliﬁcation may predict beneﬁt from adjuvant anthra- analysis of gene expression proﬁles in breast cancer: toward a uniﬁed under- standing of breast cancer subtyping and prognosis signatures. Breast Cancer Res cyclines in HER2 positive early breast cancer. Breast Cancer Res Treat 2007; 106: 2008; 10: R65. 181–189. 19 Haibe-Kains B, Schroeder M, Bontempi G, Sotiriou C, J Quackenbush genefu: 7 Desmedt C, Di Leo A, de Azambuja E, Larsimont D, Haibe-Kains B, Selleslags J et al. Relevant function for gene expression analysis, especially in breast cancer. Multifactorial approach to predicting resistance to anthracyclines. J Clin Oncol R package version 1.6.1; http://www.bioconductor.org/packages/release/bioc/ 2011; 29: 1578–1586. vignettes/genefu/inst/doc/genefu.pdf (Accessed on 31st October 2013). 8 Bartlett JM, Munro A, Cameron DA, Thomas J, Prescott R, Twelves CJ. Type 1 20 Hatzis C, Pusztai L, Valero V, Booser DJ, Esserman L, Lluch A et al. A genomic receptor tyrosine kinase proﬁles identify patients with enhanced beneﬁt from predictor of response and survival following taxane-anthracycline chemotherapy anthracyclines in the BR9601 adjuvant breast cancer chemotherapy trial. J Clin for invasive breast cancer. JAMA 2011; 305: 1873–1881. Oncol 2008; 26: 5027–5035. 21 Tabchy A, Valero V, Vidaurre T, Lluch A, Gomez H, Martin M et al. Evaluation 9 Martin M, Romero A, Cheang MC, López García-Asenjo JA, García-Saenz JA, Oliva B of a 30-gene paclitaxel, ﬂuorouracil, doxorubicin, and cyclophosphamide chemo- et al. Genomic predictors of response to doxorubicin versus docetaxel in primary therapy response predictor in a multicenter randomized trial in breast cancer. breast cancer. Breast Cancer Res Treat 2011; 128:127–136. Clin Cancer Res 2010; 16: 5351–5361. 10 Di Leo A, Moretti E, Oakman C, Biganzoli L, Santarpia L. Predictive moleular 22 Iwamoto T, Bianchini G, Booser D, Qi Y, Coutant C, Shiang CY et al. Gene pathways markers of anthracycline effectiveness in early breast cancer. Eur J Cancer 2011; associated with prognosis and chemotherapy sensitivity in molecular subtypes of Supp 9:16–21. breast cancer. J Natl Cancer Inst 2011; 103: 264–272. 11 Fan C, Oh DS, Wessels L, Weigelt B, Nuyten DS, Nobel AB et al. Concordance 23 Li Y, Zou L, Li Q, Haibe-Kains B, Tian R, Li Y et al. Ampliﬁcation of LAPTM4B and among gene-expression-based predictors for breast cancer. N Engl J Med 2006; YWHAZ contributes to chemotherapy resistance and recurrence of breast cancer. 355: 560–569. Nat Med 2010; 16:214–218. 12 Farmer P, Bonnefoi H, Anderle P, Cameron D, Wirapati P, Becette V et al. A stroma- 24 Oakman C, Moretti E, Di Leo A. Re-searching anthracycline therapy. Breast Cancer related gene signature predicts resistance to neoadjuvant chemotherapy in Res Treat 2010; 123:171–175. breast cancer. Nat Med 2009; 15:68–74. 25 Schiff R, Massarweh S, Shou J, Osborne CK. Breast cancer endocrine resistance: 13 deRonde JJ, Lips EH, Mulder L, Vincent AD, Wesseling J, Nieuwland M et al. how growth factor signaling and estrogen receptor coregulators modulate SERPINA6, BEX1, AGTR1, SLC26A3, and LAPTM4B are markers of resistance to response. Clin Cancer Res 2003; 9: 447S–454S. neoadjuvant chemotherapy in HER2-negative breast cancer. Breast Cancer Res 26 Nahta R, Esteva FJ. HER2 therapy: molecular mechanisms of trastuzumab resis- Treat 2013; 137: 213–223. tance. Breast Cancer Res 2006; 8: 215. 14 Montagner M, Enzo E, Forcato M, Zanconato F, Parenti A, Rampazzo E et al. SHARP1 suppresses breast cancer metastasis by promoting degradation of hypoxia-inducible factors. Nature 2012; 487: 380–384. This work is licensed under a Creative Commons Attribution- 15 Adorno M, Cordenonsi M, Montagner M, Dupont S, Wong C, Hann B et al. NonCommercial-NoDerivatives 4.0 International License. The images A Mutant-p53/Smad complex opposes p63 to empower TGFbeta-induced or other third party material in this article are included in the article’s Creative Commons metastasis. Cell 2009; 137:87–98. license, unless indicated otherwise in the credit line; if the material is not included under 16 Desmedt C, Haibe-Kains B, Wirapati P, Buyse M, Larsimont D, Bontempi G et al. the Creative Commons license, users will need to obtain permission from the license Biological processes associated with breast cancer clinical outcome depend on holder to reproduce the material. To view a copy of this license, visit http:// the molecular subtypes. Clin Cancer Res 2008; 14: 5158–5165. creativecommons.org/licenses/by-nc-nd/4.0/ Supplementary Information accompanies the paper on the npj Breast Cancer website (http://www.nature.com/npjbcancer) © 2015 Breast Cancer Research Foundation/Macmillan Publishers Limited npj Breast Cancer (2015) 15003
npj Breast Cancer – Springer Journals
Published: Jun 2, 2015
Access the full text.
Sign up today, get DeepDyve free for 14 days.