Abstract
Background: The prognostic and predictive role of cyclo-oxygenase-2 (COX2) in breast cancer is still debated, and in particular, its role as a target of COX2 inhibitor (celecoxib) in neoadjuvant setting. Materials and Methods: We analyzed a series of 156 breast cancer samples from patients of the COX2 inhibitor-treated arm included in the REMAGUS-02 randomized phase II trial. COX2 gene expression was assessed by reverse transcription and quantitative polymerase chain reaction using ribonucleic acid from frozen biopsies. Pathological complete response (pCR) was the surrogate end-point. Results: Significantly higher rates of grade 3, and estrogen and progesterone receptor negativity were observed in tumors with the highest expression of COX2. pCR rates were significantly higher in COX2-overexpressing tumors in patients receiving celecoxib. The test for interaction between COX2 gene expression and the celecoxib effect was statistically significant (p<0.01), but was not retained in the multivariate analysis. Conclusion: COX2 overexpression is predictive of pCR in patients with celecoxib-treated tumors. The efficacy of celecoxib in breast cancer might be improved by quantification of COX2 gene expression.
Many human cancers exhibit elevated prostaglandin levels due to up-regulation of cyclo-oxygenase-2 (COX2), a key enzyme in eicosanoid biosynthesis. COX2 overexpression has been observed in different malignant tumors and especially in breast cancer (1). Experimental studies showed that COX2 overexpression and a related production of prostaglandins stimulates angiogenesis and proliferation, promotes cell invasion and development of metastases (2). Consequently, selective COX2 inhibitors such as celecoxib have been explored as therapeutic or chemopreventive agents in different settings (3-5).
The level of COX2 has been associated with poor outcomes in many tumor models and clinical studies (6-8). However, there is no consensus on the prognostic or predictive value of COX2 expression in invasive breast carcinoma (9-11). Very few studies addressed the neoadjuvant context and response to celecoxib associated to chemotherapy in primary or metastatic breast cancer (12).
Patient and tumor characteristics.
One could expect to obtain better results with a better selection of patients receiving celecoxib. We present here data obtained in the human epidermal growth factor receptor 2 (HER2)-negative arm of the prospective neoadjuvant randomized phase II trial, REMAGUS-02 (R02), suggesting that the quantification of COX2 gene transcripts predicts the pathological response to neoadjuvant celecoxib associated with chemotherapy including anthracyclines and taxanes.
Patients and Methods
Patients. The present study concerns a series of 220 breast cancer samples from patients included in the HER2-negative arm of the prospective randomized phase II trial REMAGUS-02 (R02) for patients with T2/T3/T4 advanced breast cancer. The patients were treated by sequential epirubicin/cyclophosphamide followed by docetaxel with or without the randomized addition of celecoxib (400 mg twice daily, orally) (arms A and B, respectively). Surgery was performed for 218 patients after eight cycle of chemotherapy, as previously described (13, 14). The primary objective of the trial was pathological complete response (pCR) evaluated according to Chevallier criteria (15). Secondary objectives were to define genomic profiles of success (pCR) or failure of each type of treatment and were published elsewhere together with quality control criteria (16-18).
Tumor characteristics in the subgroups of tumors with low and high levels of cyclo-oxygenase-2 (COX2) mRNA expression.
The trial and ancillary studies were reviewed by the Ethics Committee of Bicêtre (CPP IDF VII), no. 03-55, 14th October 2004, in compliance with the Helsinki Declaration. All patients were informed and prospectively gave their signed consent to participate in the trial and ancillary studies,
Due to safety concerns, the use of celecoxib was suspended by the French Health Products Safety Agency (AFSSAPS) from December 2004 to September 2005 and thereafter authorized with revision of the informed consent form (13, 14). Thirteen patients randomized to receive celecoxib did not receive it. Consequently, the analyses were performed in the per protocol population.
Pathological response in patients as a function of tumor cyclo-oxygenase-2 (COX2) expression and celecoxib treatment.
Of the cases, 152 samples out of the 218 patients were available for reverse transcription and quantitative polymerase chain reaction (RT-qPCR) analysis on the basis of RNA of high quality from frozen pretreatment biopsies with more than 30% invasive epithelial tumor cells. There was no difference between the 152 patients with RT-qPCR data and the remaining 66 patients of the HER2-negative arm regarding age, menopausal status, clinical tumor size or nodal involvement, and hormone receptor status (data not shown).
Tissue samples and real-time RT-qPCR analysis. Total RNA extraction from frozen pretreatment biopsies, reverse transcription and qPCR analysis conditions and quality controls were previously described in detail (16-18). Primer and probe sequences for COX2 mRNA expression are available on request. Large ribosomal protein (RPLPO), TATA box-binding protein (TBP), transferrin receptor (TFR), beta-actin (ACTB), beta-glucuronidase (GUS), and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) were used as endogenous reference genes. COX2 mRNA levels were normalized to the median of the six reference genes.
Statistical analysis. As no consensual threshold was defined for RT-qPCR analyses and to ensure the robustness of the results, COX2 gene expression was arbitrarily split according to tertiles (low, intermediate and high). As the magnitude of odds ratios for pCR of the two lower tertiles (tertiles 1 and 2) was similar, we chose to group these two tertiles and analyze them (low and intermediate expression) versus the third one (tumors with the highest COX2 mRNA levels).
To analyze the association between clinical, pathological, COX2 mRNA and pCR data, we performed a univariate analysis using the chi-square test and a univariate logistic regression model to estimate odds ratios (OR) and their 95% confidence intervals.
Association between COX2 expression (tertile 1&2 versus tertile 3) and pCR were performed.
Analyses were performed with R software, version 3.1.2 (R Development Core Team, 2011).
Results
Patient population. The characteristics of the population of 152 patients are described in Table I. Patients and tumor characteristics were similar in patients treated with and without celecoxib. However, the tumor characteristics were significantly different in subgroups with low (n=102) and high (n=50) COX2 gene expression level (Table II). We observed higher rates of grade 3 (69.6% versus 41.6%, p=0.01), ER-negative (62.0% versus 22.5%, p<0.01) and PR-negative (77.6% versus 45.0%, p<0.01) tumors in the population with the highest expression of COX2 when compared with the lower tertiles (Table II).
Pathological response. A pCR was observed in 20/152 patients (13.2%). Considering the whole population, no effect of celecoxib was observed on tumor pCR (15.9% in celecoxib-treated patients vs. 11.2% in those without celecoxib; p=0.41) (Table III). However, the pCR rate was higher in the group of patients with highest tertile of COX2 in the overall population and in patients receiving celecoxib (p=0.002) (Table III).
Taking into account the level of COX2 expression in patients who received celecoxib, the pCR rate was significantly higher in those with tumors with the highest tertile of COX2 (47.4%) than in those with low expression of COX2 (2.3%) (Table III). The magnitude of the OR for pCR in the group of patients who received celecoxib suggested an interaction between COX2 gene expression and the effect of celecoxib. The test for interaction was statistically significant (p<0.01), meaning that the effect of celecoxib on pCR was significantly different according to COX2 gene expression. On the contrary, in the arm without celecoxib, no difference in pCR rate was observed according to COX2 expression (Table III). However, after multivariate analysis, the interaction between celecoxib and COX2 expression failed to reach statistical significance (p=0.12), and only initial tumor size (T3 and T4 versus T2: OR=0.14, 95% CI=0.03-0.5, p=0.006) and ER status (OR=0.03, 95% CI=0-0.12, p<0.001) were significantly associated with pCR.
Discussion
In this biologically-driven analysis of the HER2-negative arm of the breast cancer neoadjuvant REMAGUS 02 trial, we found that COX2 expression analyzed by RT-qPCR could be a target for celecoxib treatment. The effect of celecoxib in addition to neoadjuvant chemotherapy was different according to COX2 expression level in patients with HER2-negative breast cancer in terms of pCR. To our knowledge, this phase II neoadjuvant trial is the first to investigate in a prospective randomized trial the efficacy of the selective COX2 inhibitor celecoxib in addition to chemotherapy according to COX2 gene expression in breast cancer.
The main finding of the current study was that patients with high COX2 gene expression who received celecoxib had a significantly higher pCR rate compared with patients with low COX2 gene expression. It was previously shown in 42 patients with metastatic breast cancer patients treated with anthracyclines with and without taxanes, that the combination of celecoxib with capecitabine was more effective in patients overexpressing COX2 (19). Our results are also supported by interesting data published on lung cancer. Edelman et al. re-analyzed the negative results of the CALGB 30203 trial in advanced non-small cell lung cancer (20). Their analysis of the COX2 expression data indicate that the benefit of 400 mg celecoxib twice per day was greatest in those with tumors with a higher level of COX2 expression (20, 21).
Regarding the relationship between COX2 and other tumor characteristics, we observed a positive correlation of COX2 overexpression with higher tumor grade, as observed in the recent retrospective Chilkman's study. Their study was designed for a subgroup of 303 high-grade breast cancers and they analyzed COX2 expression by immunohistochemistry (9). In contrast, they did not find any correlation with ER expression and only eight out of 18 studies cited in their article found an inverse correlation between ER and COX2 expression as observed in our study (9). Furthermore, in a recent study performed on a retrospective cohort of 446 breast carcinomas treated in the adjuvant setting where COX2 expression was analyzed by RT-qPCR, the authors found an inverse correlation between COX2 expression and ER and PR expression (22).
A recent meta-analysis including 21 studies and 6739 patients with breast cancer showed that the presence of high levels of COX2 predicts a greater tumor size and lymph node metastasis (11). The occurrence of COX2-overexpressing tumors in each study ranged from 27.9% to 81.4%. As illustrated by this meta-analysis (11), most published studies on prognostic or predictive value of COX2 were performed using immunohistochemistry. But the methods used for immunohistochemical analysis of COX2 were diverse, with various antibodies, lack of standardization of staining, and analysis of tumor COX2 expression. Classifications of COX2 positivity and negativity differed significantly between published studies, which make comparisons between studies difficult. Recent studies pointed out several problems related to this method and explained discrepancies between studies. Urban et al. pointed out the importance of taking into account the stromal component of the tumor and not only epithelial cells, as is usually done (23). Only a few studies used RT-qPCR to assess COX2 expression (22, 24, 25). The main advantage of RT-qPCR is its truly quantitative approach but there is still no standardized method. However, this could be adapted, since to date, new standardized tests using quantification of target genes by RT-qPCR are available for routine use of a molecular signature for luminal breast cancer (26, 27). However, a recent publication showed a good correlation between COX2 expression analyzed by immunohistochemistry and by RT-qPCR (22).
We conclude that COX2 expression in breast cancer is associated with histological type and grade, and inversely correlated with ER and PR expression. We also showed that higher COX2 expression is associated with an increase in pCR rate in patients treated with celecoxib. Taken together, our findings support the fact that drug trials using celecoxib should include pre-stratification by COX2 status. The efficacy of COX2 inhibitors in combination with chemotherapy might benefit from the quantitative evaluation of the target as a predictive biomarker (companion diagnosis). These results need to be confirmed in independent prospective study.
In this article, we included the essential elements of the Reporting Recommendations for Tumor Marker Prognostic Studies (REMARK) (28).
Acknowledgements
This work was supported by Academic Grants from the French Ministry of Health (French Programme Hospitalier de Recherche Clinique ISRCTN10059974, PHRC: AOM/2002/0211) and Industrial Grants from Pfizer Inc., Roche, Sanofi-Aventis ISRCTN100599. A.S. Hamy was supported by an ITMO-INSERM-AVIESAN cancer translational research grant.
The Authors thank K. Tran-Perennou, C. Barbaroux and S. Vacher for their helpful technical contribution, and Dr. O. Tembo for clinical study monitoring. The Authors thank all the participants of the REMAGUS 02 trial.
Footnotes
↵* These Authors contributed equally to this study.
This article is freely accessible online.
Ethics Approval and Consent to Participation
Remagus 02 trial: “A randomized phase II study of cytotoxic chemotherapy or cytotoxic chemotherapy combined with celecoxib or trastuzumab as primary chemotherapy for patients with high risk localized breast cancer not amenable to breast conserving therapy.” was reviewed by the Ethic Committee of Bicêtre (CPP IDF VII), no. 03-55 14th October 2004.
Competing Interests
None of the Authors have any competing interest in regard to this study.
- Received December 5, 2017.
- Revision received January 10, 2018.
- Accepted January 16, 2018.
- Copyright© 2018, International Institute of Anticancer Research (Dr. George J. Delinasios), All rights reserved
References
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.