Breast Cancers with a BRCA1-like DNA Copy Number Profile Recur Less Often Than Expected after High-Dose Alkylating Chemotherapy

This article is featured in Highlights of This Issue, p. 663

Inactivating germline mutations in the BRCA1 gene confer a high risk of developing breast cancer. Although these germline mutations are shown to be rare in the general population (1, 2), it is thought that a considerable subgroup of breast cancers has characteristics similar to BRCA1 germline mutated breast cancer but without harboring a BRCA1 mutation, so called BRCAness (3). BRCA1 is important in error-free DNA double-strand break (DSB) repair through the homologous recombination pathway (4, 5). In the absence of functional homologous recombination, DNA repair is mostly performed by error-prone nonhomologous end-joining (NHEJ). This in turn leads to genomic instability (6, 7). Genomic instability can be measured with DNA copy-number (CN) profiles. It has been established that DNA CN profiles of breast cancers that developed in BRCA1 mutation carriers have a specific signature (8, 9). To detect these tumors, classifiers have been trained that estimate the probability that a CN profile is similar to a BRCA1 mutated breast cancer (“BRCA1-like”; ref. 10). In addition, this classification can also identify tumors that have a defect in BRCA1 due to other causes than mutation, for example hypermethylation of the gene promoter (11, 12). This means that the biomarker (BRCA1-like signature) captures aspects of the broader concept BRCAness. Preclinical mechanistic studies indicate that deficiencies in this pathway strongly sensitize tumors to drugs inducing DNA damage through DNA DSBs, such as platinum compounds and bifunctional alkylators (3, 13–16). In general, these drugs are currently not standard therapy in breast cancer. High-dose regimens contain such agents. However, in the general population, these regimens did not confer a sufficient large survival advantage in the adjuvant or metastatic setting that could justify its toxicity profile. In addition, no known prognostic indicators could identify a subgroup of patients that benefitted substantially from this therapy. Young patients, however, seemed to have a better recurrence-free survival (RFS) than older patients (17, 18). We therefore investigated in a randomized clinical trial whether patients that classified as BRCA1-like based on their DNA CN profiles form a subgroup of patients that benefit substantially from DSB inducing high dose (HD) chemotherapy (19). The HD regimen in these previous studies contained four cycles of 5-fluorouracil, epirubicin, and cyclophosphamide (FEC, 500/90/500 mg/m²) followed by one cycle of carboplatin (1,600 mg/m²), thiotepa (480 mg/m²), and cyclophosphamide (6 g/m²). Patients in the control arm received five cycles of FEC (500/90/500 mg/m²) only. We found that patients with a BRCA1-like profile had an approximately 6-fold lower hazard for an event for disease-free survival (DFS), RFS, and overall survival (OS; refs. 20, 21).

In this study we investigated in an independent single institution cohort whether BRCA1-like patients also benefitted from a tandem HD regimen (22).

The patients in this study consist of a cohort of patients that has been treated with adjuvant HD therapy or conventional chemotherapy. High-risk breast cancer patients in the HD cohort were enrolled in a single-arm study that was conducted between 1992 and 1997 that evaluated safety and efficacy of the regimen. High risk was defined as stage II or III primary breast cancer with more than 10 tumor-positive lymph nodes, premenopausal patients with three to nine tumor-positive lymph nodes, or hormone receptor–negative disease. Conventionally treated patients from the same time period were identified by the cancer registry of the University of Heidelberg Gynecology and Obstetrics department. Conventionally treated patients were matched for nodal ratio, tumor size, and hormone receptor and HER2 status (22). We collected all available archival tissue from the archives of the Institute of Pathology Heidelberg. The HD regimen consisted of two cycles induction chemotherapy (total dose, 7,500 mg/m² ifosfamide, 120 mg/m² epirubicin every 3 weeks), followed by two cycles of 12 g/m² ifosfamide, 900 mg/m² carboplatin, and 180 mg/m² epirubicin every 4 to 6 weeks, with autologous stem-cell transplantation. Following chemotherapy, hormone receptor–positive premenopausal patients received gosereline as adjuvant endocrine therapy for 2 years and hormone receptor–positive postmenopausal patients received tamoxifen for 5 years. The conventional regimen consisted of 500 mg/m² cyclophosphamide, 40 mg/m² methotrexate, 600 mg/m² 5-fluorouracil (CMF) or 60/90 mg/m² epirubicin, 600 mg/m² cyclophosphamide (EC) or 600 mg/m² 5-fluorouracil, 60 mg/m² epirubicin, 600 mg/m² cyclophosphamide (FEC) chemotherapy. The Ethical Committee of the University of Heidelberg approved this study. Patients enrolled in the trials supplied written informed consent.

We collected archival FFPE tissue and selected tumor blocks that contained a region with 60% tumor cells. Tumor DNA isolation and Array Comparative Genomic Hybridization was performed as previously reported (23). One microgram of DNA was labeled using the Enzo Genomic Labeling Kit (Enzo Life Sciences) according to the manufacturer's instruction. The labeled DNA was then hybridized to a Nimblegen 135k whole-genome array containing 134,937 in situ synthesized oligonucleotides (Roche NimbleGen) according to the manufacturer's protocol. Slides of the arrays were scanned using a G2505C microarray scanner (Agilent Technologies). Images were analyzed using Nimblescan version 2.5.26 feature extraction software (Roche Nimblegen). Oligonucleotides were mapped according to the human genome build NCBI36.

Raw image files were processed using Nimblescan software and the DNAcopy algorithm. Unaveraged DNAcopy files were used as input for the BRCA1-like classifier as previously described (i.e., no training was performed, just application as described; ref. 24). The BRCA1-like classifier is a nearest shrunken centroids classifier (25) that was trained to distinguish CN profiles of BRCA1-mutated tumors from tumors without a BRCA1 mutation. Visualization of characteristic regions of CN alteration (10), and the actual locations and centroids of the classifier have been described before (20). This classifier assigns a class label to a CN profile, either BRCA1-like or non–BRCA1-like, based on the posterior probability of being non–BRCA1-like (probability, 0) or being BRCA1-like (probability, 1). The cutoff for optimal chemotherapy prediction was 0.63, as determined on a training cohort (metastatic breast cancer series), and validated on a randomized trial of high-dose chemotherapy versus conventional chemotherapy (20). We used these settings to investigate the current cohort. The data discussed in this publication have been deposited in NCBI's Gene Expression Omnibus (GEO; ref. 26) and are accessible through GEO Series accession number GSE50407 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE50407).

Calculations were performed in R v3.0.1 (27). Categorical clinical characteristics were compared with Fisher exact (2 × 2 Tables) or χ² tests (larger than 2 × 2 Tables). Survival was evaluated according to BRCA1-like status by Kaplan–Meier curves and log-rank statistics. Hazard ratios were estimated with a Cox Proportional Hazards model adjusting for relevant prognostic factors. To assess whether BRCA1-like status is a predictive marker, a test for interaction with chemotherapy regimen was performed (28). A description of this study according to the REporting recommendations for tumor MARKer prognostic studies (REMARK) guidelines is provided in Supplementary Table S1. Survival was calculated from start date of chemotherapy until date of event or date of last follow-up. Events were defined as any relapse or death (DFS), any relapse other than in the ipsi-lateral breast, chest wall, axillary or supraclavicular lymph nodes or death (DDFS) or death (OS). We assessed both DFS and DDFS to investigate whether a treatment benefit would be due to decreased systemic disease events, i.e., the adjuvant treatment is successful in its aim to kill occult micrometastases.

We collected the archival material for 82% of the patients that were treated at the University of Heidelberg. A flow diagram of patients included is shown in Fig. 1.

We generated CN profiles for 117 of these patients. The clinical characteristics for these patients are shown in Table 1. The treatment cohorts were not significantly different, except that patients treated with HD chemotherapy were more likely to have higher lymph node stage (non–BRCA1-like P < 0.001; BRCA1-like P = 0.05). BRCA1-like tumors have a higher N stage (P = 0.01), higher grade (P = 0.03), and are more often estrogen receptor (ER)–negative (P = 0.01) and triple negative (P < 0.001). A trend for fewer events in the conventionally treated non–BRCA1-like patients compared with high dose–treated non–BRCA1-like patients is present for DFS (P = 0.05) and DDFS (P = 0.07). Concluding, BRCA1-like patients treated with HD chemotherapy have more unfavorable prognostic factors than the other patients in this study.

The median follow-up for DFS and DDFS were 2.4 and 2.5 years, respectively. The median follow-up for OS was 8.5 years. Kaplan–Meier curves for non–BRCA1-like patients and BRCA1-like patients comparing the outcome of conventional and HD chemotherapy on DFS, distant DFS (DDFS), and OS are shown in Fig. 2.

When adjusted for the potential confounding factors, ER status, HER2 status, size, number of tumor-positive lymph nodes and grade, BRCA1-like patients had a better DFS (HR, 0.05; 95% CI, 0.01–0.38; P = 0.003); DDFS (HR, 0.06; 95% CI, 0.01–0.43; P = 0.01), and OS (HR, 0.15; 95% CI, 0.03–0.83; P = 0.03) with adjuvant HD chemotherapy than with conventional chemotherapy. No such benefit was observed in the non–BRCA1-like cases for DFS (HR, 0.74; 95% CI, 0.38–1.46; P = 0.39), DDFS (HR, 0.79; 95% CI, 0.41–1.52; P = 0.47), and OS (HR, 0.93; 95% CI, 0.52–1.64; P = 0.79), The P values for interaction were 0.01 (DFS), 0.01 (DDFS), and 0.045 (OS; Table 2). No evidence against proportionality of treatment-specific hazards was observed among BRCA1-like and non–BRCA1-like patients.

We investigated a single institution cohort treated with high-dose and conventional chemotherapy to assess whether BRCA1-like patients benefit from a tandem HD alkylating chemotherapy regimen (20). Patients with a BRCA1-like breast cancer significantly recurred less after the tandem HD alkylating chemotherapy regimen. BRCA1-like status may therefore serve as a predictive marker for tandem HD chemotherapy with iphosphamide, epirubicin, and carboplatin.

High-dose alkylating regimens in breast cancer are not used anymore after it has been shown that this regimen does not confer a survival benefit (17, 18). However, these regimens contain potent inducers of DNA DSBs, which are predicted beneficial in patients with tumors that do not have a functional homologous recombination, due to a defect in BRCA1 (4, 14–16, 29). To identify these patients, we used a BRCA1-like classifier, which assigns patients that share specific DNA CN aberrations with BRCA1-mutated tumors to the BRCA1-like class (8–10). The tumors assigned to be BRCA1-like comprise BRCA1-mutated and BRCA1–wild-type tumors that have another mechanism to inactivate BRCA1 function, for example BRCA1 promoter hypermethylation (11, 12, 20). Previously, we investigated this classifier in patients that were randomized to receive either five cycles of conventional adjuvant 5-FU, epirubicin, and cyclophosphamide (FEC) or four cycles of conventional FEC followed by one cycle of the HD regimen carboplatin (1,600 mg/m²), thiothepa (480 mg/m²), and cyclophosphamide (6 g/m²; ref. 19). In this study, patients with a BRCA1-like CN profile in their tumor derived a substantial benefit from another HD regimen compared with conventional chemotherapy. Patients with a non–BRCA1-like tumor that received HD chemotherapy did not benefit at all compared with conventional FEC. Although the HD regimen in the current study (2 cycles induction + 2 cycles of 12 g/m² ifosfamide, 900 mg/m² carboplatin, and 180 mg/m² epirubicin) is different from our previous study, we obtained similar results. Our data suggest that patients with a BRCA1-like tumor benefit from an HD alkylating regimen, whereas patients with a non–BRCA1-like tumor do not. Given the differences in drugs and doses, it may be that BRCA1-like tumors, generally are more sensitive to the dose or DNA crosslinking properties, rather than a specific agent.

Our present study is limited by several factors, which are shared by other investigations of BRCA1 deficiency as a potential biomarker. In clinical studies, there is to date neither agreement on different biomarkers that can be used to identify BRCA1 defects, nor any randomized evidence to demonstrate that selection of patients with a BRCA1 defect for a particular therapy results in improved treatment outcomes as has been reviewed extensively (3, 30–32). Clearly, our present study is limited by a small sample size and the nonrandomized design relying on a balanced patient cohort treated with several chemotherapy combinations during the same time period at the same institution. Strikingly, the patients expected to have a poor prognosis, (BRCA1-like patients associated with high N stage, high-grade, and ER-negative/triple-negative tumors) have a significantly better survival when treated with HD chemotherapy in comparison with conventional chemotherapy, whereas the non–BRCA1-like patients do not experience that HD treatment benefit.

Regarding the therapy regimens, patients in the control group received conventional chemotherapy consisting of either EC, CMF, or EC-CMF as they were treated in the pretaxane era. This allows for a unique view on biology but limits the conclusions on whether HD chemotherapy is better than a modern taxane-containing regimen. However, it has been suggested that BRCA1-mutated patients are relatively resistant to taxanes (33). Because we do not have an untreated control group, it is not possible to accurately model a potential prognostic effect that the BRCA1-like profile status may confer. Second, due to the combinations of different drugs, it is impossible to dissect whether the survival benefit could be due to a particular drug. Third, it could also be that the amount of DNA damage is greatly increased due to the higher dose, and that the cumulative dose is responsible for efficient killing of tumor cells.

Further research should therefore aim to overcome the limitations. First, by investigating other randomized trials between high-dose and conventional chemotherapy whether BRCA1-like patients benefit from the HD regimen. Single cohorts and especially nonrandomized cohorts do not reach the highest level of evidence and findings should therefore be confirmed in independent studies. (34, 35). These studies can strengthen the evidence while prospective, randomized evidence is being gathered in (neo-) adjuvant and metastatic breast cancer (http://clinicaltrials.gov: NCT01898117, NCT01057069, NCT01646034). Furthermore, identifying the role or contribution of specific agents, dose, and biomarker are important to determine the use of this information in the treatment of breast cancer. Specifically, the treatment burden associated with an HD chemotherapy regimen requires proper selection of patients. If possible, less toxic therapy should be investigated, for example, an intermediate dose or only the effective drug in this regimen. Furthermore, Poly(ADP)-Ribose Polymerase 1 (PARP1) inhibitors may prove effective with a more beneficial toxicity profile (36–39).

Concluding, in this single institution study, we observed that BRCA1-like CN profiles identify a group of patients that recur less often than expected after receiving HD than after conventional chemotherapy treatment. BRCA1-like CN profile status may therefore be a predictive biomarker for HD alkylating chemotherapy and could then be used to guide treatment choices to improve OS among patients with a poor prognosis.

S.C. Linn is an inventor on a BRCA1-like patent. No potential conflicts of interest were disclosed by the other authors.

Conception and design: P.C. Schouten, F. Marmé, A. Schneeweiss, S.C. Linn

Development of methodology: M. Hauptmann

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): P.C. Schouten, F. Marmé, S. Aulmann, H.-P. Sinn, A. Schneeweiss

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): P.C. Schouten, F. Marmé, S.C. Linn

Writing, review, and/or revision of the manuscript: P.C. Schouten, F. Marmé, M. Hauptmann, A. Schneeweiss, S.C. Linn

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): F. Marmé, S. Aulmann, H.F. van Essen, B. Ylstra, A. Schneeweiss

Study supervision: A. Schneeweiss, S.C. Linn

This work has been supported by a grant from A Sister's Hope (“BRCA-like CGH”).

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.