Errata: Retraction Volume 122, Issue 20, 3248, Article first published online: 29 July 2016
Gefitinib was received from Astra-Zeneca and erlotinib was received from OSI Pharmaceuticals.
Epidermal growth factor receptor tyrosine kinase inhibitors (EGFR-TKIs) have demonstrated clinical benefit in patients with nonsmall cell lung cancer (NSCLC), particularly those with tumors that have EGFR-TK domain mutations. Moreover, the EGFR and cyclooxygenase (COX)-2 pathways are known to enhance the procarcinogenic effects of each other in different tumor types. Therefore, it was hypothesized that tumor EGFR mutation status may influence the effectiveness of simultaneous EGFR and COX-2 inhibition in patients with NSCLC.
Three NSCLC cell lines with varying EGFR mutation status and sensitivities to EGFR-TKIs were selected: H3255 (L858R), H1650 (del E746-A750), and H1781 (wild-type EGFR). Cells were treated with erlotinib, gefitinib, or celecoxib alone, and the combination of both EGFR-TKI inhibitors with celecoxib. Cell survival and apoptosis was assessed and correlated with the expression of COX-2, EGFR, pEGFR, Akt, pAkt, expression, and derived prostaglandin E2 (PGE2).
Celecoxib by itself was found to have no effects on cell growth or apoptosis in any of the cell lines. Erlotinib and gefitinib inhibited cell growth and induced apoptosis in both mutant cell lines and did so in H1781 cells at 10-fold higher concentrations. Celecoxib when added to erlotinib or gefitinib significantly enhanced the antiproliferative and proapoptotic effects in both mutant cell lines but had no additional effects in H1781 cells. Greater down-regulation of COX-2, EGFR, pEGFR, Akt, pAkt, and PGE2 was found when H3255 cells were treated with the combination compared with any of the single agents alone.
Lung cancer is the most common cause of cancer-related mortality in the U.S., with a 5-year survival rate of only 15%.1 Therefore, there is a great need to evaluate novel agents to better treat this disease. Epidermal growth factor receptor (EGFR), a receptor tyrosine kinase, is expressed in many nonsmall cell lung cancers (NSCLCs) and is involved in various aspects of carcinogenesis.2 Based on promising preclinical data, the EGFR tyrosine kinase (EGFR-TKIs) inhibitors erlotinib and gefitinib were evaluated in patients with NSCLC.3–5 Phase 2 studies of these agents demonstrated a modest response rate of 10% and median survival of 6 to 8 months. However, these initial trials suggested that the probability of clinical benefit was greater in patients with certain clinical characteristics such as nonsmoking status, adenocarcinoma or bronchioalveolar histology, female sex, and East Asian ethnicity.5–7
Retrospective analyses of tumors of patients treated with EGFR-TKIs revealed that the presence of activating mutations in the tyrosine kinase domain of the EGFR gene was predictive of response and clinical benefit from EGFR-TKIs.8, 9 Subsequently, other investigators have suggested that increased EGFR gene copy number and EGFR expression may be more predictive of clinical benefit.10 The importance of each of these factors in predicting benefit from these agents has varied between the various retrospective analyses and the true relevance of each factor remains to be determined in prospective clinical trials. It is interesting to note that EGFR mutations and increased EGFR gene copy number occur more frequently in tumors from patients with clinical characteristics that predict for increased likelihood of benefit with EGFR-TKIs.10, 11
One of the downstream effects of EGFR signaling is induction of the cyclooxygenase (COX) enzyme.12 COX is a rate-limiting enzyme involved in the conversion of arachidonic acid to prostaglandins (PGs).13 There are 2 isoforms of COX: COX-1, a constitutive enzyme expressed in most cells, and COX-2, an inducible isoform of COX that is overexpressed in inflammatory and neoplastic tissues in response to various stimuli.12, 14–16 COX-2, primarily through the production of PG, appears to be involved in various aspects of carcinogenesis.17–20 Many NSCLC tumors express COX-2 and preclinical evidence suggests that the selective COX-2 inhibitor celecoxib can inhibit NSCLC growth.21–23
Because both EGFR and COX-2 appear to be involved in the carcinogenesis of many epithelial tumors, there has been an interest in evaluating simultaneous inhibition of both pathways. Preclinical studies have suggested that there is an overlap between the 2 pathways.12 EGFR signaling induces COX-2 expression and increased prostaglandin production, whereas COX-2-derived prostaglandin E2 (PGE2) can enhance signaling through EGFR.24–26 In addition, recent studies have shown that COX-2 expression may result in resistance to EGFR-TKIs.27, 28 Preclinical studies have also shown that the combination of EGFR and COX-2 inhibitors is more effective in inhibiting intestinal tumor formation than either agent alone.29
Based on the relevance of the EGFR and COX-2 pathways in NSCLCs and the overlap between the 2 pathways, we conducted a clinical trial evaluating the combination of gefitinib plus celecoxib in patients with recurring NSCLC.30 The overall results of the study (median survival of 5 months and response rate of 12%) did not suggest a benefit from the addition of celecoxib to gefitinib. However, 1 nonsmoking female patient with adenocarcinoma has had prolonged disease control while receiving therapy for >3 years. We were unable to determine the EGFR mutation status of this patient's tumor due to lack of adequate tumor specimen. However, her clinical features make it likely that her tumor harbored an EGFR mutation. Her clinical outcome could be from the effect of gefitinib alone but her clinical course raised the possibility that the combination may produce greater benefit in EGFR mutant NSCLC. We therefore hypothesized that the overlap between the EGFR and COX-2 pathways is more prominent in EGFR mutant tumors than in wild–type tumors and therefore dual inhibition will produce a greater benefit in tumors with EGFR mutations.
Our results suggest that the combined inhibition of the EGFR and COX-2 pathways is most beneficial in NSCLC cell lines with EGFR mutations, with no additive benefit observed in NSCLC cell line with wild–type EGFR. Our results also suggest that this enhanced effect of the combination in the mutant tumors is a result of more effective inhibition of both EGFR and COX-2 pathways by the combination than any of the single agents.
MATERIALS AND METHODS
Cell Culture and Reagents
Three NSCLC cell lines with varying EGFR mutation status and sensitivity to EGFR-TKIs were used: H3255 (L858R; erlotinib IC50 41.72 nM), H1650 (del E746-A750; erlotinib IC50 100 nM), and H1781 (wild–type; erlotinib IC50 >10 μM). H3255 was a generous gift from the Dana-Farber Cancer Institute (Boston, Mass) and were grown in ACL-4 medium supplemented with 5% fetal bovine serum. H1650 and H1781 were obtained from the American TypeCulture Collection (ATCC; Manassas, Va) and grown in RPMI-1640 supplemented with 10% fetal bovine serum. Gefitinib and erlotinib were generous gifts from AstraZeneca (Wilmington, Del) and OSI Pharmaceuticals (Melville, NY), respectively.
Cell Growth Inhibition by MTT Assay
H3255, H1650, and H1781 were seeded at a density of 3000 to 5000 per well in a 96-well plate for 24 hours. Based on the initial results, the concentration of erlotinib (50 nM and 100 nM), gefitinib (1 nM and 5 nM), celecoxib (5 μM), and their combination for 72 hours was chosen for both MTT and apoptosis assays. In addition, for the nonsensitive cell line H1781, a higher concentration of erlotinib (10 μM), gefitinib (10 μM), celecoxib (5 μM), and its combination with celecoxib 5 μM was performed to determine whether any potentiation in growth inhibition was possible. Cell survival was determined by MTT assay using 0.5 mg/mL of MTT for 2 hours. The color intensity was measured with a TECAN microplate fluorometer (Research Triangle Park, NC) at 595 nanometers (nm).
Quantification of Apoptosis by Enzyme-Linked Immunoadsorbent Assay
The Cell Death Detection ELISA (enzyme–linked immunoadsorbent assay) kit (Roche Applied Science, Indianapolis, Ind) was used to detect apoptosis. Cells seeded in 6-well plates were treated with erlotinib (50 nM and 100 nM), gefitinib (1 nM and 5 nM), or celecoxib (5 μM), and the combination of erlotinib and celecoxib or gefitinib and celecoxib. In addition, for the nonsensitive cell line H1781, a higher concentration of erlotinib (10 μM), gefitinib (10 μM), celecoxib (5 μM), and its combination with celecoxib 5 μM was performed to detect any enhanced apoptosis. The cells were lysed and the supernatant fluid was transferred to an antihistone-coated microtiter plate. Samples were then incubated with anti-DNA peroxidase followed by color development with ABTS substrate. The TECAN microplate fluorometer was used to measure color intensity at 405 nm.
Protein Extraction and Western Blot Analysis
H3255 and H1650 cells treated with erlotinib (50 nM), gefitinib (5 nM), or celecoxib (5 μM), and the combination of erlotinib and celecoxib or gefitinib and celecoxib for 72 hours was used to evaluate the effects of treatment on COX-2, EGFR, pEGFR, Akt, pAkt, and β-actin expression. Cells were lysed in buffer consisting of 250 mM NaCl, 50 mM Tris buffer (pH 7.5), 5 mM of ethylenediamine tetraacetic acid (EDTA), 1% NP-40, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulfate (SDS), 50 mM sodium fluoride, 1 mM sodium orthovandate, 1 mM phenylmethylsulfonylfluoride (PMSF), and 1 μg/mL pepstatin for 30 minutes on ice. Protein concentration was then measured using the BCA Protein Assay Kit (Pierce, Rockford, Ill). The samples were loaded on sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and electrophoretically transferred to a nitrocellulose membrane. Each membrane was incubated with appropriate primary antibodies. The signal intensity was then measured using chemiluminescent detection system (Pierce). Autoradiograms of the Western blot analysis were scanned and bands were quantitated using AlphaEaseFC software tool (Alpha Innotech, San Leandro, Calif).
PGE2 Immunoassay for Quantitation of Prostaglandin E2
Based on the basal level of PGE2, we tested the effect of treatment on PGE2 level on H3255 and H1781 cells only. Erlotinib (10 nM), gefitinib (1 nM), and celecoxib (1 nM) were chosen at the lowest concentration possible to see the effect of both of the combinations. H1781 cells were treated with a higher concentration of erlotinib (10 μM), gefitinib (10 μM), and celecoxib (5 μM) to determine whether there was any decrease in the PGE2 level with the higher concentration as used in MTT and apoptosis assay. The drugs were added for 4 hours and the culture medium was collected, measured, and analyzed for PGE2 concentration according to the manufacturer's protocol using the PGE2 high sensitivity immunoassay kit (R&D Systems, Minneapolis, Minn). The optical density was measured at 450 nm and the concentration of PGE2 was calculated from the standard curve. The results are expressed as PGE2 in pg/106 cells.
Comparisons of treatment outcome were tested for statistical differences using the Student t test for paired data. Statistical significance were assumed at a P-value of ≤.05.
EGFR and COX-2 Expression in NSCLC Cell Lines
The baseline expression of EGFR, pEGFR, Akt, pAkt, and COX-2 expression in the 3 cell lines is shown in Figure 1. The expression of E-cadherin and p-STAT3 were also evaluated because previous studies have shown that these proteins may correlate with EGFR-TKI activity.28, 31 H3255 expressed the highest levels of COX-2, EGFR, and pAkt, but the lowest levels of E-cadherin, and was the only cell line that expressed p-STAT-3 (Fig. 1A). H1781 expressed low levels of COX-2, EGFR, and pAkt, and moderate levels of E-cadherin, but did not express p-STAT-3. H1650 cells expressed EGFR and the highest level of E-cadherin, but failed to express COX-2, pAkt, or p-STAT-3.
Determination of Basal Level of PGE2
The baseline level of PGE2, a product of the enzymatic activity of COX-2, was also determined in the cell lines (Fig. 1B). PGE2 was not detectable in H1650, a finding that is consistent with nondetectable COX-2 levels in this cell line. PGE2 levels were detected in both H3255 and H1781 and were higher in H3255 in concordance with the higher expression of COX-2 in this cell line.
Induction of Growth Inhibition by Erlotinib, Gefitinib, and Celecoxib
The viability of H3255, H1650, and H1781 cell lines treated with erlotinib (50 nM and 100 nM), gefitinib (1 nM and 5 nM), celecoxib (5 μM), or the combination of erlotinib plus celecoxib (Fig. 2A) or gefitinib plus celecoxib (Fig. 2B) was determined by the MTT assay. Single-agent celecoxib had no growth inhibitory activity of its own in any of the cell lines at the concentration tested. In the H3255 and H1650 cell lines, the addition of celecoxib to erlotinib or gefitinib resulted in potentiation of the growth inhibition when compared with erlotinib or gefitinib alone (Fig. 2A and 2B). In H1781, the cell line with wild–type EGFR, no effects with either the single agents or the combinations were observed at the concentrations tested in H3255 and H1650. Therefore, we evaluated the effects of erlotinib and gefitinib at the higher concentration of 10 μM with or without celecoxib (5 μM) in these cells. At the 10 μM concentrations, the EGFR-TKIs induced 50% growth inhibition of 1781 but we found no potentiation of growth inhibition with the addition of celecoxib.
Induction of Apoptosis by Erlotinib, Gefitinib, and Celecoxib and the Combination
Apoptosis assays were performed in H3255, H1650, and H1781 cell lines to determine the mechanism of the observed growth inhibition. Erlotinib (50 nM and 100 nM), and gefitinib (1 nM and 5 nM) induced apoptosis in the 2 cell lines with EGFR mutations, H3255 and H1650, but had minimal effect in H1781 cells with wild–type EGFR (Fig. 3A and 3B). Celecoxib (5 μM) induced apoptosis in H3255 but was found to have very minimal effect in H1650 and H1781. However, the addition of celecoxib to either erlotinib or gefitinib significantly enhanced apoptosis in H3255 and H1650 cells when compared with treatment with single agents. As with growth inhibition, the enhancement of apoptosis by the addition of celecoxib was greater in H3255 than in H1650. In the H1781 cell line, no significant apoptosis was observed by any single agent treatment or combination. With the IC50 concentration of erlotinib (10 μM) or gefitinib (10 μM), induction of apoptosis (Figs. 4A and 4B) was observed in H1781 but it was not enhanced by the addition of celecoxib.
Modulation of EGFR, COX-2, and Their Downstream Protein Levels
The expression of COX-2, EGFR, pEGFR, Akt, and pAkt protein was determined in H3255 and H1650 cells treated with erlotinib (50 nM), gefitinib (5 nM), celecoxib (5 μM), or a combination of erlotinib plus celecoxib or gefitinib plus celecoxib (Fig. 5A,B). No significant changes in the expression of COX-2, EGFR, or pAkt were observed after treatment with any of the single-agent concentrations tested in any of the 3 cell lines. Celecoxib and erlotinib did reduce pEGFR levels in H3255 cells. A significant decrease in the expression of COX-2, EGFR, pEGFR, Akt, and pAkt protein levels was observed in H3255 cells after treatment with either of the combinations. In H1650 cells, basal levels of COX-2 and pAkt were not detected, so the effect of treatment on the expression of these proteins could not be adequately assessed. However, no significant effect on EGFR, Akt, and pEGFR expression was detected in H1650 after treatment with any of the single agents or combinations. Similarly, no change in any of the protein expression was observed in the H1781 cell line at the concentrations tested (data not shown).
Modulation of PGE2
PGE2 assay was performed in H3255 and H1781 cells with erlotinib (10 nM), gefitinib (1 nM), celecoxib (1 nM), or the combination of erlotinib with celecoxib or gefitinib with celecoxib (Fig. 6A). The concentrations of all 3 drugs chosen for PGE2 assay were very low compared with either growth inhibition assay or protein level to determine the differences in drug combinations. With the lowest concentration of all 3 drugs, a decrease in the PGE2 levels with each of the single agents and both the combinations were detected in H3255 cells but not in H1781 cells (data not shown). The P values were calculated with a Student t test using GraphPad prism software (San Diego, Calif) and were found to be statistically significant. For H1781 cells (wild–type EGFR cells), a higher concentration of celecoxib (5 μM) demonstrated a significant decrease in the PGE2 level, but erlotinib (10 μM), or gefitinib (10 μM) did not appear to have an effect on the PGE2 levels and the addition of the EGFR-TKIs did not enhance the effect of celecoxib (Fig. 6B). These studies were not conducted in H1650 cells because no basal levels of PGE2 were detected.
Although treatment with EGFR-TKIs can result in modest benefits in all subgroups of patients with NSCLC,32 it is clear that the dramatic antitumor effects tend to occur in patients with tumors that harbor activating mutations in the kinase domain of the EGFR gene. Sordella et al.33 suggested that EGFR mutant tumors become dependent for survival on the antiapoptotic signals activated by the mutant EGFR, a concept that has been termed ‘oncogenic addiction.’ Therefore, in such cells EGFR-TKIs, through the inhibition of antiapoptotic signals, can induce extensive apoptosis and clinical benefit. Recently, a prospective study of first-line single-agent erlotinib in patients with NSCLCs that harbored EGFR mutations reported a response rate of 82% (13% complete response), results that to our knowledge are unprecedented in the treatment of advanced NSCLC.34
Our observations in a clinical trial of gefitinib plus celecoxib led us to speculate that the combination of an EGFR-TKI and a COX-2 inhibitor may be most beneficial in EGFR-mutant NSCLCs. We chose 3 NSCLC cell lines to test our hypothesis. Both H3255, a cell line with an EGFR mutation in exon 21, and H1781, with wild–type EGFR, were studied in what to our knowledge is 1 of the original reports regarding EGFR mutations.9 We also selected H1650 because it harbors a different EGFR mutation (del E746-A750) and is known to be less sensitive to EGFR-TKIs than H3255.35 This allowed us to evaluate the addition of celecoxib to EGFR-TKIs in NSCLC cell lines with varying EGFR mutation status and varying sensitivities to EGFR-TKIs.
Although celecoxib had no single-agent activity, it enhanced growth inhibition and apoptosis when combined with EGFR-TKIs in both of the EGFR mutant NSCLC cell lines despite their varying sensitivities to EGFR-TKIs. The enhanced effect with the addition of celecoxib was found to be greater in H3255, the cell line that is more sensitive to EGFR-TKIs than H1650 cells. In H3255 cells, both erlotinib and gefitinib reduced COX-2 and PGE2 levels, suggesting that EGFR inhibition reduced COX-2 transcription and thus the production of PGE2. Consistent with the prior observation that COX-2 can activate EGFR, celecoxib reduced pEGFR levels in H3255 cells.22 In addition, the levels of EGFR, phospho-EGFR, COX-2, and PGE2 were found to be significantly reduced with the combination than single-agent alone. We could not assess these changes in H1650 cells because we did not detect COX-2 and PGE2 expression in this cell line. The changes in protein levels in H3255 cells suggest that the enhanced effect of the combination in mutant NSCLC cell lines is a result of more effective inhibition of both EGFR and COX-2 pathways with the addition of celecoxib to EGFR-TKIs than any of the single agents.
In H1781 cells, the addition of celecoxib did not appear to enhance the growth inhibitory or the apoptotic effects of EGFR-TKIs at any concentration of EGFR-TKIs tested, including the higher concentration of 10 μM. PGE2 was detected in H1781 cells, but at levels less than that of H3255 cells. Celecoxib at a concentration of 5 μM lowered the PGE2 levels. Thus, despite inhibition of COX-2 activity, celecoxib did not enhance the effects of EGFR-TKIs. In addition, the EGFR-TKIs, even at the 10 μM concentration, did not appear to have any effect on the PGE2 levels in this cell line with wild–type EGFR. These results suggest that in the H1781 cells, unlike in H3255 cells, the combination did not result in greater inhibition of both pathways than either agent alone. It is possible that the interdependency of the EGFR and COX-2 pathways is greater in mutant cell lines than in NSCLC cell lines with wild–type EGFR and this provides the basis for enhanced inhibition of both pathways by the combination in the mutant cell lines. In addition, this interdependency of the 2 pathways may vary in mutant cell lines based on the sensitivity of the tumor cells to EGFR-TKIs. In this regard, the effects of the combination were more potent in H3255 cells, the more sensitive cell line, than in H1650 cells.
These results also suggest that COX-2 expression or single-agent celecoxib activity do not predict for the ability of celecoxib to improve the antitumor effect of EGFR-TKIs. Thus, despite the lack of COX-2 expression in H1650, celecoxib enhanced the antitumor effect of EGFR-TKIs in this cell line, whereas although H1781 does express COX-2, celecoxib did not enhance the effect of EGFR-TKIs in H1781. We speculate that the increased dependency of the mutant cells on the EGFR pathway may increase the relevance of the overlap that exists between EGFR and COX-2 pathways and therefore lead to the enhancement of the antitumor effects of EGFR-TKIs by celecoxib. This may explain the benefits observed with the addition of celecoxib in H1650 despite the lack of COX-2 expression. However, this hypothesis will have to be confirmed in future experiments.
One of the primary antiapoptotic pathways activated by EGFR is the Akt pathway. In H3255 cells, the combinations of erlotinib or gefitinib plus celecoxib resulted in significantly greater reduction in phospho-Akt levels than was observed with erlotinib, gefitinib, or celecoxib alone. This suggests that the dual inhibition is better than the EGFR-TKIs alone at inhibiting the antiapoptotic signals activated in this cell line. Thus, the benefit of adding celecoxib to an EGFR-TKI in this cell line is not just the result of adding another targeted drug, but appears to be due to the amplification of the apoptotic signal induced by EGFR-TK inhibition via the inhibition of an overlapping oncogenic pathway. No changes were noted in the phospho-Akt levels with the combination in H1781 cells because an enhanced growth inhibitory or apoptotic effects were not observed in this cell line and in H1650, because basal levels of phospho-Akt were not detected.
Reckamp et al.36 recently reported the results of a phase 1 study evaluating the combination of erlotinib and celecoxib and found the optimal biologic dose of celecoxib, based on inhibition of urinary levels of a PG metabolite, to be 600 mg twice daily when combined with erlotinib at a dose of 150 mg. This dose is higher than that used in cancer prevention or therapeutic trials to date. The investigators observed an impressive objective response rate of 33% (7 of 21 patients), suggesting the relevance of the higher dose of celecoxib for adequate dual inhibition, although responses were observed at dosages lower than 600 mg twice daily of celecoxib. In addition, of the 7 responders, 5 patients had EGFR mutations. The other 2 responders, whose tumors did not have EGFR mutations, were never-smokers. In the large randomized study that documented the clinical benefit of erlotinib in patients with recurring NSCLC, the most important factor found to be associated with benefit was nonsmoking status.32 Based on retrospective analyses, the rate of EGFR mutations in nonsmokers is approximately 50%.37 It is unclear whether the higher probability of benefit among nonsmokers is completely accounted for by the presence of EGFR mutations or whether other factors make NSCLC in nonsmokers dependent on the EGFR pathway. Thus, it is possible that in the study by Reckamp et al. the benefits of dual inhibition were limited to patients with tumors that have some degree of EGFR dependency. Those investigators are planning a phase 2 study of erlotinib (at a dose of 150 mg) and celecoxib (at a dose of 600 mg twice daily) in patients with advanced NSCLC.
In conclusion, the results of the current study demonstrate that the addition of celecoxib to EGFR-TKIs enhances growth inhibition and apoptosis in NSCLC cell lines with EGFR mutations but does not do so in a cell line with wild–type EGFR. We have also shown that in a mutant cell line the combination showed more effective inhibition of both EGFR and COX-2 pathways than either agent alone. Our results suggest that the combination of celecoxib and EGFR-TKI is likely to be most beneficial in NSCLC with activating mutations in the EGFR-TK domain. Our results do not suggest that the combination will not have any benefit in NSCLC with wild–type EGFR but that the benefit in these tumors is likely to be limited to higher concentration of the drugs and that this benefit is likely to be modest. We plan to conduct animal studies to duplicate our in vitro findings and we also plan to conduct further in vitro studies to better define the mechanisms that underlie this observation.
We thank Dr. Gregory Kalemkerian for critical review of the article.