ERCC1 Is a Predictor of Anthracycline Resistance and Taxane Sensitivity in Early Stage or Locally Advanced Breast Cancers

Genomic instability could be a beneficial predictor for anthracycline or taxane chemotherapy. We interrogated 188 DNA repair genes in the METABRIC cohort (n = 1980) to identify genes that influence overall survival (OS). We then evaluated the clinicopathological significance of ERCC1 in early stage breast cancer (BC) (mRNA expression (n = 4640) and protein level, n = 1650 (test set), and n = 252 (validation)) and in locally advanced BC (LABC) (mRNA expression, test set (n = 2340) and validation (TOP clinical trial cohort, n = 120); and protein level (n = 120)). In the multivariate model, ERCC1 was independently associated with OS in the METABRIC cohort. In ER+ tumours, low ERCC1 transcript or protein level was associated with increased distant relapse risk (DRR). In ER−tumours, low ERCC1 transcript or protein level was linked to decreased DRR, especially in patients who received anthracycline chemotherapy. In LABC patients who received neoadjuvant anthracycline, low ERCC1 transcript was associated with higher pCR (pathological complete response) and decreased DRR. However, in patients with ER−tumours who received additional neoadjuvant taxane, high ERCC1 transcript was associated with a higher pCR and decreased DRR. High ERCC1 transcript was also linked to decreased DRR in ER+ LABC that received additional neoadjuvant taxane. ERCC1 based stratification is an attractive strategy for breast cancers.


Introduction
Anthracycline and taxane based adjuvant and neoadjuvant chemotherapies are standard approaches in the management of early stage or locally advanced breast cancers to reduce distant recurrence and improve survival [1][2][3]. Moreover, the recent development of multi-parameter gene-expression assays, largely based on proliferation biomarkers, has facilitated the selection of patients who are most likely to benefit from systemic chemotherapy [4]. However, despite the genomic based selection, not all patients benefit from chemotherapy. In addition, chemotherapy related toxicity (such as anthracycline induced cardiotoxicity/leukaemia and taxol induced irreversible peripheral

ERCC1 Transcript Is a Predictor of Tumour Grade and Chromosomal Instability in Early Stage breast cancers (BCs)
A large body of clinical evidence confirms that high grade BCs is associated with chromosomal instability. Given the critical role of ERCC1 in NER, DSB repair, ICL repair and chromosomal stability, we evaluated ERCC1 transcripts in the METABRIC Cohort. A low level of ERCC1 transcript was significantly associated with higher grade cancer, whereas low grade tumours were common in high ERCC1 tumours (Table 1) (p values < 0.0001). Low ERCC1 transcript was also associated with Ki67 positivity, ER−, PAM50 Luminal B, Pam50 Her2, Pam50 basal and Genufu ER+/HER− (high proliferation) tumours. On the other hand, low grade, Ki67 negative, PAM50 Luminal A and Genufu ER+/HER− (low proliferation) were more common in tumours with high ERCC1 transcript (all p values < 0.0001). To evaluate associations with chromosomal stability we investigated ERCC1 in various integrative molecular cluster (intClust) phenotypes described in the METABRIC cohort. Low ERCC1 transcript was linked to genomically unstable intClust.10 phenotype whereas high ERCC1 was associated with chromosomally stable intClust.3, 4, 7 and 8 tumours. Together, the data provides the first clinical evidence that low ERCC1 is a marker of chromosomal instability and aggressive phenotype in BCs.

ERCC1 Transcript and Clinical Outcomes in Patients Receiving Adjuvant Therapy
In the ER+ METABRIC whole cohort, low ERCC1 transcript was associated with higher risk of death (p = 0.0001) ( Figure 1A). In patients who received endocrine therapy, similarly, low ERCC1 was associated with higher risk of death (p = 0.0001) ( Figure S2A). In addition, in the ER+ METABRIC cohort, we tested 188 DNA repair genes in a multivariate Cox proportional hazards model with backward stepwise exclusion and identified ERCC1 (among seven other genes) as an independent predictor for overall survival (OS) (Table S10). As chromosomal instability is a marker of chemo-sensitivity, we evaluated ERCC1 in ER− METABRIC patients who received chemotherapy. Low ERCC1 mRNA expression was associated with a decreased risk of death from BC (p = 0.05) ( Figure 1B). In a multivariate Cox regression analysis after controlling for confounders (such as endocrine therapy, chemotherapy and other validated prognostic factors (ER, PR, HER2, grade, stage, tumour size, TP53 mutation status, PAM50 molecular subtype and IntClust subclasses)), we confirmed that ERCC1 transcript was an independent prognostic factor for OS (p = 0.039) and the interaction between ERCC1 and chemotherapy was also statistically significant (p = 0.020) ( Table 2). In ER− tumours, that received no chemotherapy, ERCC1 did not influence survival ( Figure S2B).   We then validated in the Multicentre (MC)-Adjuvant cohort of 4640 patients. By using mean as cut-off, high and low ERCC1 were observed in 49% (1460/2261) and 51% (14602379) of cases, respectively. ER and HER2 status were available for 3826 and 1727 cases; respectively. About 59% (2268/3826), 41% (1558/3826) and 26% (446/1727) of cases were ER+, ER− and HER2+, respectively (Tables S3 and  S4). Similar to the METABRIC data, in ER+ tumours, low levels of ERCC1 was associated with an increased distant relapse DRR compared to high levels of ERCC1 [p = 0.007) ( Figure 1C). However, in ER− tumours that received adjuvant chemotherapy, low ERCC1 was associated with a reduced DRR compared to high ERCC1 (p = 0.001) ( Figure 1D). Multivariable Cox regression models confirmed that the low ERCC1 transcript is a poor prognostic factor for DRR after controlling with Adjuvant! Online (AOL) (p = 0.047) and 72-proliferation-gene-signatures (p < 0.0001).

ERCC1 Protein, Clinicopathological Features and Outcomes
Using median as the cut off (H-score ≥ 130), we observed ERCC1 nuclear protein expression in 439/991 (44.3%) of breast tumours, and 55.7% (552/991) were negative for ERCC1 expression. As shown in (Table S11), low nuclear ERCC1 level was significantly associated with aggressive phenotypes, including high grade, no special histological type (NST), ER−, basal-like phenotype and triple negative tumours, as well as loss of other DNA repair biomarkers (all adjusted p ≤ 0.01). In ER+ tumours, low ERCC1 protein was linked to poor disease relapse free survival (p = 0.044) ( Figure 1E). On the other hand, in ER− tumours that received chemotherapy, low ERCC1 protein was linked to improved disease relapse free survival (p = 0.034)) ( Figure 1F). Multivariable Cox regression analysis controlling for chemotherapy, endocrine therapy and other validated prognostic factors (stage, grade, size, ER, PR, HER2 and BCl2), showed that ERCC1 protein expression was an independent prognostic factor for OS (p = 0.035) and that the interaction between ERCC1 protein expression and adjuvant chemotherapy was statistically significant (p = 0.022) (Table S12).
Taken together, the data provides compelling evidence that ERCC1 has prognostic significance in ER+ BCs and predict response to chemotherapy in ER− BCs.

Discussion
Although the efficacy of DNA damaging chemotherapy (such as anthracyclines) is influenced by impaired DNA repair capacity, evolving evidence also suggests that mitotic spindle poisons (such as taxanes) are more effective in low grade chromosomally stable tumours. Therefore the development of robust DNA repair based biomarkers is highly desirable. ERCC1 is non-catalytic but partners with XPF endonuclease to form the ERCC1-XPF heterodimer which processes abnormal DNA repair intermediates generated during NER, double strand breaks (DSB) repair and Interstrand Cross Link (ICL) repair [29]. Given the key role for ERCC1 in genomic integrity, we hypothesized a role for ERCC1 in breast cancer pathogenesis and response to therapy. In the current study we show that ERCC1 transcript expression was independently associated with OS in the METABRIC cohort. In ER+ tumours, low ERCC1 transcript or protein level was associated with increased distant relapse risk (DRR). In ER− tumours, low ERCC1 transcript or protein level was linked to decreased DRR, especially in patients who received anthracycline chemotherapy. In LABC patients who received neoadjuvant anthracycline, low ERCC1 transcript was associated with higher pCR (pathological complete response) and decreased DRR. However, in patients with ER−tumours who received additional neoadjuvant taxane, high ERCC1 transcript was associated with a higher pCR and decreased DRR. High ERCC1 transcript was also linked to decreased DRR in ER+ LABC that received additional neoadjuvant taxane. Taken together, the data presented here provides comprehensive clinical evidence that ERCC1 is a predictor of anthracycline resistance and taxane sensitivity in breast cancers. ERCC1 based stratification could be an attractive strategy in breast cancers.
Studies exploring biomarkers of response to anthracycline therapy have been limited in breast cancers. Previous smaller studies suggest that Ki67, HER1-3 expression, TOP2A and HER-2 are potential markers of anthracycline benefit [30]. Ch17CEP, a CIN marker, was also previously shown to predict benefit from anthracycline adjuvant chemotherapy in a prospective clinical trial [8]. ERCC1 is a critical factor for CIN. To the best of our knowledge, the data shown here represents the first comprehensive evidence that ERCC1 status influences potential benefitting from anthracycline chemotherapy. The role of ERCC1 in breast cancer pathogenesis is emerging. ERCC1 polymorphism may be associated with increased breast cancer risk [31]. A previous, small study suggested that ERCC1 protein levels may be low in triple negative breast cancers (TNBCs) [32,33]. In a study of fifty two TNBCs, ERCC1 positivity was associated with shorter progression free survival and poor response to neoadjuvant chemotherapy [32]. In addition, ERCC1 genetic polymorphism also appeared to associate with pCR in patients receiving neoadjuvant anthracycline chemotherapy [34]. Gay-Beillile et al. recently also demonstrated that ERCC1 expression is induced in tumours that receive anthracycline based neoadjuvant chemotherapy [35]. Previous clinical studies have evaluated the predictive significance of ERCC1 for response to platinum chemotherapy, in various solid tumours (lung, colorectal, head and neck, gastric and bladder cancers [36][37][38][39]. However, a major limitation has been the use of relatively non-specific ERCC1 antibodies for immunohistochemistry in previous studies, including in a large lung cancer clinical trial [40]. In the current study we utilised a recently generated and highly specific mouse monoclonal antibody (clone 4F9) [41], which further strengthens our clinical data. Our study not only concurs with previous studies showing a link between ERCC1 overexpression and chemoresistance but also provides additional insight suggesting that ERCC1 may also be involved in the emergence of aggressive breast cancer phenotypes. However, a limitation to our study is that it is predominantly retrospective. Prospective studies would be required to confirm our findings.
Currently there is no established predictive biomarker of response to taxane therapy. Previous studies suggest that HER2, Ki67, class III β tubulin expression may influence taxane response [10]. A novel observation in the current study is that ERCC1 was also shown to influence whether taxane chemotherapy was beneficial. Our data concurs with previous evidence demonstrating taxane benefit in low grade, chromosomally stable tumours [9,10]. However, further prospective studies would be required to confirm our initial findings.
Taken together, the data would support further development of ERCC1 as a biomarker of response to chemotherapy in breast cancer.

Study Design and Cohorts
Study design, the patient cohorts which included 11,096 BCs and their demographics are summarized in the consort flow diagram (Figure 3), also in Supplementary Methods and Tables. Taken together, the data would support further development of ERCC1 as a biomarker of response to chemotherapy in breast cancer.

Study Design and Cohorts
Study design, the patient cohorts which included 11,096 BCs and their demographics are summarized in the consort flow diagram (Figure 3), also in supplementary methods and tables.     Supplementary Table S1. We investigated the association of 188 DNA repair genes (Table S2) with OS in the METABRIC cohort (METABRIC n = 1980; median follow-up time in years (MFUT) (inter-quantile range (IQR): 9.1 (5.2-12.9)). Univariate Cox regression analysis was used in SPSS (Version 20, Chicago, IL, USA) and the Benjamini and Hochberg False Discovery Rate calculation (BH FDR) was applied to account for multiple comparisons. After definition of factors that were associated with OS after BH FDR correction, multivariate Cox proportional hazards models (with backward stepwise exclusion of these factors, using a criterion of p < 0.05 for retention of factors in the model) were used to identify factors that were independently associated with OS. The statistical significance of the model was assessed based on the likelihood ratio test. The proportional hazards assumption was tested using both standard log-log plots and by generating Kaplan-Meier survival estimate curves, and observing that the curves did not intersect with each other. Hazard ratios (HRs) for death risk and 95% confidence intervals were calculated from the Cox proportional hazards analysis.
II The association of distant relapse risk (DRR)) and ERCC1 transcript after receiving systemic adjuvant therapy (Adj-T): Cohort (2): Multicentre (MC)-Adjuvant cohort (n = 4640) The association between ERCC1 mRNA expression and DRR and its relationship with the received systemic Adj-T were tested in 4640 patients with early stage BC, retrieved from 21 gene expression databases (see Supplementary Tables S3 and S4). ERCC1 gene expression data of each database were converted to a common scale (median equal to 0 and standard deviation equal to 1) in order to merge all of the study data that used the same platform and to create combined cohorts. Then the data was median centred for each gene, whereby the median of each gene was 0. Databases using the same platform were merged and the median expression was calculated. The median expression of ERCC1 transcription for each platform was calculated and values equal to or higher than the median coded as +1 (overexpression). Values of less than the median were coded 0 or low ERCC1. Distal relapse free survival follow up data were available for 3171 patients with 967 events (MFUT (IQR): 5.5 (3.0-8.7)). The systemic Adj-T information was available for 2276 patients: 45% of patients were naïve to systemic Adj-T, whereas 49% had received Adj-endocrine therapy and 27% had received chemotherapy. Herceptin had been offered to 156 (7%) of HER2 + patients.
III The association between ERCC1 and pathological complete response rate (pCR) analysis after receiving neoadjuvant anthracycline based combination chemotherapy (Neo-Adj-ACT) Cohort (

Protein Expression Association Analysis
Immunohistochemical evaluation of ERCC1 protein expression was performed in three cohorts of patients who treated at a single centre (Nottingham University Hospital (NUH)). We utilised a recently characterised highly specific anti-ERCC1 mouse monoclonal antibody (clone 4F9, Dako Ltd., Cheshire, UK) [18]. We confirmed the specificity of clone 4F9, using Western Blots in breast cancer (SKBR3, T47D and MDA-MB-231) and ovarian cancer (A2780 and A2780cis) cell lines (Supplementary Figure S1A). Tissue culture and Western blot analyses is described in Supplementary Methods.

Adjuvant Setting
Cohort (5) (Table S8). The association between the ERCC1 protein expression with clinicopathological parameters and DRR were analysed in this cohort.
Cohort (6): (Nottingham ER-negative series). A series of 252 ER negative invasive breast carcinomas diagnosed and managed at the Nottingham University Hospitals (NUH) Trust between 1999 and 2007. All patients were primarily treated with surgery, followed by adjuvant radiotherapy and anthracycline based combination chemotherapy [43]. Demographics are shown in Supplementary  Table S9.
All patients completed written informed consented, as per hospital standard of care, for excess tumour tissue to be used in research. The study was approved by the Institutional Review Board or Independent Ethics Committee and the Hospital Research and Innovations Department at all participating sites. Tumour Marker Prognostic Studies (REMARK) criteria, as recommended by McShane et al. [44] were followed throughout this study.

Power Analysis
A retrospective power analysis was conducted to determine the confidence in the calculated hazard ratio and associated p value for 10 year survival and to ascertain how applicable the result would be to a global population. Power of study was determined using PASS (NCSS, version 13, USA).

Statistical Analysis
Statistical analyses were performed using STATISTICA (Stat Soft Ltd., Tulsa, OK, USA) and SPSS (version 17, Chicago, IL, USA) by the authors who were blinded to the clinical data. Where appropriate, Pearson's chi-squared, student's t-test and ANOVA tests were used. Positivity for ERCC1 protein both pre-and post-chemotherapy was calculated and compared using McNemar's test. Cumulative survival probabilities and 10-year BCSS and DFS were estimated using the univariate Cox proportional hazards models and the Kaplan-Meier plot method where appropriate, and differences between survival rates were tested for significance using the log-rank test. Multivariable analysis for survival was performed using the Cox proportional hazard model. The proportional hazards assumption was tested using standard log-log plots. Hazard ratios (HR) and 95% confidence intervals (95% CI) were estimated for each variable. All tests were two-sided with a 95% CI, and a p value < 0.05 was considered to be indicative of statistical significance. The interaction between ERCC1 and chemotherapy was tested in the Cox proportional hazard model. For multiple comparisons, p values were adjusted according to Benjamini-Hochberg method [45]. Tumor Marker Prognostic Studies (REMARK) criteria, recommended by McShane et al. [46], were followed throughout this study. Ethical approval was obtained from the Nottingham Research Ethics Committee (C202313).

Conclusions
ERCC1 is non-catalytic but partners with XPF endonuclease to form the ERCC1-XPF heterodimer which processes abnormal DNA repair intermediates generated during NER, DSB repair and ICL repair. We provide the first comprehensive clinical evidence that ERCC1 is a key predictor of chemotherapy response in patients with breast cancer who receive adjuvant or neoadjuvant chemotherapy. Importantly, the clinical study suggests that ERCC1 based stratification is feasible in BC patients who receive anthracycline and/or taxane chemotherapy.
Supplementary Materials: The following are available online at http://www.mdpi.com/2072-6694/11/8/1149/s1, Table S1: Clinicopathological characteristics in the METABRIC cohort, Table S2: List of DNA repair genes tested in the METABRIC cohort, Table S3: Cohort ID and gene expression platform Multicentre (MC)-Adjuvant cohort (n=4640), Table S4: Demographics of Multicentre (MC)-Adjuvant cohort (n=4640), Table S5: Multicentre (MC) Neo-Adjuvant cohort (n= 2345), Table S6: Demographics in TOP trial cohort, Table S7: Clinicopathological characteristics of Nottingham historical early, TableS8: Table of antibodies and optimisation conditions used to immunohistochemically profile the Nottingham University Hospitals based cohorts. Detailed below are: Antigens, primary antibodies, clone, source, optimal dilution and scoring system, used for each immunohistochemical marker stage cohort (NUH-ESBC), Table S9: Clinicopathological characteristics of ER-cohort, Table S10: Multivariate backward step-wise Cox analysis for DNA repair genes, associated with Overall Survival (OS) in ER+ METABRIC cohort, Table S11: Clinicopathological significance of ERCC1 protein expression in breast cancers, Table S12: Multivariate Cox regression analysis for breast cancer specific survival (BCSS) at 20 years follow up in Nottingham series including interaction terms.