carboplatin, epirubicin, gemcitabine or vinorelbine in breast cancer cell lines and tumor samples. Gottfried Konecny1, Michael Untch2, Dennis Slamon1, ...
Breast Cancer Research and Treatment 67: 223–233, 2001. © 2001 Kluwer Academic Publishers. Printed in the Netherlands.
Report
Drug interactions and cytotoxic effects of paclitaxel in combination with carboplatin, epirubicin, gemcitabine or vinorelbine in breast cancer cell lines and tumor samples Gottfried Konecny1, Michael Untch2 , Dennis Slamon1 , Malgorzata Beryt1 , Steffen Kahlert2 , Margret Felber2 , Elena Langer2 , Sandra Lude2 , Hermann Hepp2 , and Mark Pegram1 1 Division
of Hematology-Oncology, Department of Medicine, University of California, Los Angeles School of Medicine, Los Angeles, CA, USA; 2 Department of Obstetrics and Gynecology, Klinikum Grosshadern, Ludwig Maximillians Universität München, München, Germany
Key words: breast cancer, carboplatin, epirubicin, gemcitabine, in vitro, paclitaxel, vinorelbine
Summary The purpose of this study was to analyze the drug interactions of paclitaxel (PTX) with epirubicin (EPI), carboplatin (CBDCA), gemcitabine (GEM) and vinorelbine (VIN) in human breast cancer cells and compare the cytotoxic activity of each drug combination in primary breast cancer samples. These experiments were intended to identify the most active agents in combination with PTX, and to provide a preclinical rational for future clinical investigations in breast cancer. Multiple drug effect/combination index (CI) isobologram analysis was applied to combinations of PTX with either CBDCA, EPI, GEM or VIN in MCF-7, MDA-MB-231 and SK-BR-3 human breast cancer cell lines. Drug concentrations were limited to the ranges achievable in humans in vivo, and the drugs were applied simultaneously at fixed molar ratios for each drug combination. Interactions were assessed at multiple effect levels (IC10 –IC90 ). Additionally, the cytotoxic activity of these combinations was assessed in tumor samples of 50 primary breast cancer patients, utilizing the ATP-tumorchemosensitivity assay (ATP-TCA). Drug interactions were shown to be strongly dose-related in the human breast cancer cell lines investigated. At clinically relevant concentrations, CBDCA/PTX demonstrated synergistic (MCF-7) or additive (MDA-MB-231, SK-BR-3) interactions, and EPI/PTX showed additive (SK-BR-3, MCF-7) and antagonistic (MDA-MB-231) interactions. GEM/PTX and VIN/PTX, however, demonstrated antagonism over multiple dose effect levels at clinically relevant drug concentrations in all three cell lines tested. At plasma peak concentrations, EPI/PTX, CBDCA/PTX, GEM/PTX and VIN/PTX achieved ≥ 90% tumor growth inhibition in 93, 86, 63 and 50%, respectively, of primary breast cancer samples investigated with the ATP-TCA. Cumulative dose-response plots of primary breast cancer tumor cells responding in vitro with ≥ 90% growth inhibition showed a strong dose dependence for both EPI/PTX and CBDCA/PTX. In conclusion, the current data indicate favorable drug interactions for CBDCA/PTX at clinically relevant drug concentrations in breast cancer cells, and demonstrate superior in vitro cytotoxicity of EPI/PTX and CBDCA/PTX compared to GEM/PTX and VIN/PTX in primary breast cancer cultures.
Introduction Paclitaxel (PTX) has well-established single-agent activity in the first-line treatment of women with advanced breast cancer, with response rates for standard dose therapy ranging from 25 to 29% [1, 2]. Higher activity has been reported in first-line therapy in com-
bination with the anthracyclin, doxorubicin (DOX), with response rates ranging from 46–94%, depending on the dose administered and whether the patient has previously received chemotherapy [3, 4]. However, many patients with advanced breast cancer will have already been exposed to anthracyclins during adjuvant treatment, which can diminish the probability
224
G Konecny et al.
to respond to these agents in first-line treatment for advanced disease. Additionally, severe cardiac toxicities were described in 18% of the patients receiving DOX in combination with PTX [3]. European investigators have reported similar response rates (48–84%) for the combination of epirubicin (EPI) (the 4 epimer of doxorubicin) and paclitaxel, with a lower rate of cardiotoxicity [5, 6]. Currently, alternative combination therapies of paclitaxel with other anti-cancer agents such as carboplatin (CBDCA), gemcitabine (GEM) or vinorelbine (VIN) are being evaluated. All three drugs have demonstrated considerable single-agent activity in women with advanced breast cancer, with response rates ranging between 25 and 35% for CBDCA [7, 8], 25 and 46% for GEM [9], and 41 and 46% for VIN [10, 11], depending on the dose administered and whether the patient had previously received chemotherapy. Preliminary data of phase II studies demonstrate higher response rates when either of these drugs are combined with paclitaxel [12–14]. Comparative clinical studies between the new cytotoxic combination regimens, however, have not yet been performed. Furthermore, a direct comparison of the preclinical activity of these regimens in human breast cancer cell lines or patient tumor samples, using the same methodological approach to evaluate the cytotoxic interactions, has not been previously reported. The present study was therefore designed, first to compare the in vitro cytotoxic interactions of PTX in combination with CBDCA, EPI, GEM or VIN in human breast cancer cell lines, and second, to compare the activity of CBDCA/PTX, EPI/PTX, GEM/PTX and VIN/PTX in primary breast cancer tissue cultures. Analysis of the nature of the interactions between the two drugs (synergy, additivity and antagonism) may yield insights into the biochemical mechanisms of interactions between the drugs, and suggest combinations with potential for greater clinical efficacy (synergy) or combinations to be avoided (antagonism). To characterize the effects of the cytotoxic chemotherapeutic drugs, we utilized the median effect/combination index (CI) isobologram method of multiple drug effect analyses [15]. The assays were performed at clinically relevant drug concentrations in MCF-7, MDA-MB-231 and SK-BR-3 human breast cancer cells. To circumvent the possibility that the observed interaction might be unique to individual cell lines, parallel studies were conducted to compare the in vitro activity of these drug combinations in primary breast cancer tumor samples, utilizing the
ATP-tumorchemosensitivity assay (ATP-TCA). This ATP-based microplate format assay allows simultaneous testing of multiple drug combinations in tumor tissue, thus rendering a direct comparison of in vitro drug activity in the same tumor sample. Such experiments may provide a rationale for future comparative clinical investigations of these drug combinations.
Materials and methods Cell lines and chemotherapeutic drugs The human breast carcinoma cell lines, MCF-7, MDA-MB-231 and SK-BR-3, were obtained from the American Type Culture Collection (Rockville, MD). All cells were cultured in RPMI medium 1640 supplemented with 10% heat-inactivated fetal bovine serum, 2 mM glutamine and 1% penicillin G-streptomycin-fungizone solution (Irvine Scientific, Santa Ana, CA). Cells (5 × 103 ) of either cell line were plated in 96-well microdilution plates. Following cell adherence (24 h), experimental medium containing the chemotherapeutic drug(s) (paclitaxel, BristolMyers Squibb, Princton, NJ; vinorelbine, SmithKline Beecham, Philidelphia, PA; gemcitabine, Eli Lilly, Indianapolis, IN; carboplatin, Bristol-Myers Squibb, Princton, NJ; and epirubicin, Pharmacia & Upjohn, Erlangen, Germany) were added to duplicate wells, and serial two-fold dilutions were performed to span the dose range suitable for dose-effect analysis. After 72 h, plates were washed with phosphate-buffered saline and stained with 0.5% hexamethylparosaniline (crystal violet) in methanol. Sorensons’s buffer (0.025 M sodium citrate, 0.025 citric acid in 50% ethanol) (0.1 ml) was added to each well, and the plates were analyzed in an ELISA plate reader at the 540 nm wavelength. Absorbance of this wavelength correlates with cell survival [16]. Median effect/combination index isobologram method for multiple drug effect analysis Multiple drug effect analysis was performed, using Biosoft computer software [15, 17]. Details of this methodology have been published previously [15, 17]. The combination index (CI) was calculated based on the most conservative assumption of mutually nonexclusive drug interactions. CI values significantly lower than 1 indicate synergy (CI < 1), values significantly higher than 1 indicate antagonism (CI > 1), and CI
Paclitaxel combinations in breast cancer cells
225
values not signficantly higher or lower than 1 indicate additivity (CI = 1).
Results
ATP-tumorchemosensitivity assay (ATP-TCA)
Statistical analysis
Paclitaxel was analyzed in combination with CBDCA, EPI, GEM and VIN. Dose response curves were constructed for each drug alone and in combination at fixed molar ratios, defined as the ratio of the two agents at their maximally effective dose (∼IC99 ). Figure 1 shows a representative example of the multiple drug effect analysis for CBDCA/PTX in MCF-7 cells. FA and FU are the fractions of MCF-7 cells affected or unaffected, respectively, by the dose (D) of either agent. Linear regression correlation coefficients (r values) of the median effects plots reflect that the dose-effect relationships for CBDCA and PTX and the combination of both follow the principle of mass action, which is a requisite assumption for this analysis. CI values were calculated for different dose-effect levels, based on parameters derived from median effect plots of the chemotherapeutic drugs alone and in combination at fixed molar ratios. To evaluate the interaction between combination partners, CI values of three separate experiments were calculated across all combination doses tested. A summary of the data from the multiple drug effect analysis for each of the drug combinations in MCF-7, MDA-MB-231 and SK-BR-3 human breast cancer cell lines is given in Figures 2, 3, 4 and 5. Data points represent mean CI values, with the standard error of multiple actual experimental values, and the fractions of unaffected cells at the corresponding drug concentrations. Drug interactions were often shown to be strongly dose-related, with synergistic interactions occurring at lower drug
The mean CI values were calculated for each dose effect level from three separate experiments. Statistical tests were then applied (student’s t-test) to determine whether the CI values resulting from these experiments were significantly different from CI = 1. ATPTCA results were interpreted by comparing the IC90 values (drug concentration effecting ≥ 90% growth inhibition) between the drug combinations investigated (student’s t-test). Individual values for IC90 were determined by linear interpolation. The concentration dependence of the drug combinations investigated was determined using an approach modified from the method proposed by Von Hoff et al. [28], by calculating the percentage of tumors that showed tumor cell inhibition of ≥ 90% (IC90 ) for each concentration of a particular drug combination. Cumulated response data were then plotted against the drug concentrations used in vitro.
Figure 1. Multiple drug effect analysis performed for the carboplatin/paclitaxel combination. FA and FU are the fractions of MCF-7 cells affected or unaffected, respectively, by the dose (D) of either agent. Linear regression correlation coefficients (r-values) of the median plots reflect that the dose-effect relationships for carboplatin, paclitaxel and the combination follow the principle of mass action.
Chemosensitivity was assessed in primary breast cancer tumor tissue samples from 50 patients for CBDCA/PTX, EPI/PTX, GEM/PTX and VIN/PTX, using the ATP-TCA (TCA-100; DCS Innovative Diagnostik Systeme, Hamburg, Germany), which has been described in detail [18]. Briefly, surgical biopsies (1–2 cm3 ) were obtained during primary surgery. Tumor cells were isolated by mechanical and enzymatic dissociation (TDE DCS) (Innovative Diagnostik Systeme; or collagenase, Sigma, St. Louis, MO). In all cases, histological assessment confirmed the presence of invasive breast cancer. Approximately 2 × 104 cells were then seeded into each well of a 96well polypropylene microplate. The assay evaluates cell viability by measuring the intracellular ATP levels of untreated controls and drug-exposed cells at various doses after cell culture for 6 days. Test drug concentrations were used in triplicate at six different doses of 6.25, 12.5, 25, 50, 100 and 200% of the plasma peak concentrations (PPC) from clinical pharmacokinetic data [19–27]. Standard 100% PPC values used in the ATP-TCA were 42.5 µM for CBDCA, 1.6 µM for EPI, 83.4 µM for GEM, 1.2 µM for VIN, and 15.9 µM for PTX. ATP-TCA results were interpreted by comparing the levels of in vitro cytotoxicity between the drug combinations investigated.
226
G Konecny et al.
Figure 2. Combination index values and fractions of cells unaffected by carboplatin in combination with paclitaxel at multiple effect levels in MCF-7, MDA-MB-231 and SK-BR-3 cells. () denotes mean combination index values ± SE, and (◦) denotes the fraction of unaffected cells. Carboplatin, in combination with paclitaxel, demonstrated synergistic effects at clinically relevant dose levels (< 10 µM) [19] in MCF-7 and SK-BR-3 cells, and additive effects in MDA-MB-231 cells. The lightly shaded area of the figure represents the dose range above clinically relevant drug concentrations for CBDCA, and the darker shaded area represents the dose range above clinically achievable drug concentrations for CBDCA. p values indicate CI values that deviate significantly from the value of 1, and confirm antagonism or synergy.
concentrations tested, and antagonistic interactions at higher concentrations, across all the cell lines tested. In the present study, drug concentrations were generally limited to ranges achievable in humans in vivo. However, pharmacokinetic studies demonstrate that after intravenous administration, in vivo plasma peak concentrations of CBDCA, EPI, GEM, VIN and PTX are rapidly eliminated, followed by a much slower terminal elimination phase with prolonged retention of the drugs at significantly lower concentrations that correspond to approximately 10% of the initial plasma peak concentration [19–27]. In vitro drug interactions at these lower drug concentrations may most accurately reflect the in vivo situation. At these clinically relevant drug concentrations (≤ 10 µM) [19], CBDCA demonstrated synergistic (MCF-7) and additive (MDA-MB-231, SK-BR-3) interactions in combin-
ation with PTX (Figure 2). Similarly, EPI demonstrated additive (SK-BR-3, MCF-7) interactions in combination with PTX at clinically relevant concentrations (≤ 0.17 µM) [20, 21] (Figure 3). However, in MDA-MB-231 breast cancer cells, EPI/PTX showed antagonistic interactions. At clinically relevant drug concentrations of GEM (≤ 26 µM) [22, 23], antagonistic (MCF-7) and both additive and antagonistic (SKBR-3, MDA-MB-231) interactions were observed in combination with PTX (Figure 4). GEM, however, also showed synergy at very low drug concentrations (< 6 nM) in MCF-7 cells. VIN predominantly demonstrated antagonism in combination with PTX at clinically relevant drug concentrations (≤ 0.01 µM) [24, 25] (Figure 5); VIN however, also showed synergy at very low drug concentrations (< 1 nM) in MDA-MB231 cells.
Paclitaxel combinations in breast cancer cells
227
Figure 3. Combination index values and fractions of cells unaffected by epirubicin in combination with paclitaxel at multiple effect levels in MCF-7, MDA-MB-231 and SK-BR-3 cells. () denotes mean combination index values ± SE, and (◦) denotes the fraction of unaffected cells. Epirubicin, in combination with paclitaxel, demonstrated additive (SK-BR-3, MCF-7) and antagonistic (MDA-MB-231) interactions at clinically relevant dose levels (< 0.17 µM) [20, 21]. The shaded area of the figure represents the dose range above clinically relevant drug concentrations for EPI. p values indicate CI values that deviate significantly from the value of 1, and confirm antagonism or synergy.
We further investigated whether these in vitro findings would be predictive of the cytotoxic activity obtained in 50 chemo-naïve primary breast cancer samples. The cytotoxicity was compared between the drug combinations of CBDCA/PTX, EPI/PTX, GEM/PTX and VIN/PTX. Each drug combination was tested in triplicate at six different concentrations ranging from 6.25% to 200% of the clinical PPC (Table 1). The different patterns of in vitro chemosensitivity are presented in Figure 6. Of the 50 evaluable assays, 39/42 (93%) showed ≥ 90% tumor growth inibition (IC90 ) at 100% PPC for EPI/PTX, and 38/44 (86.4%) for CBDCA/PTX. In contrast, 24/38 (63.2%) of the assays achieved an IC90 level at 100% PPC for GEM/PTX and 17/34 (50%) for VIN/PTX. A comparison of the mean test drug concentrations (% PPC) that achieve ≥ 90% growth in-
hibition similarly predicated resistance to VIN/PTX, as this combination demonstrated ≥ 90% growth inhibition at significantly higher test drug concentrations (% PPC) compared to GEM/PTX, CBDCA/PTX or EPI/PTX (p < 0.01). However, EPI/PTX and CBDCA/PTX achieved ≥ 90% growth inhibition at significantly lower test drug concentrations compared to GEM/PTX (p < 0.01 and p < 0.05, respectively) or VIN/PTX (p < 0.01 for both). Strikingly, GEM/PTX and VIN/PTX demonstrated the most variability of assay results, with the largest range of test drug concentrations that achieve ≥ 90% growth inhibition (Table 2). As described above, the concentration dependence of all drug combinations was determined by a modification of the approach suggested by Von Hoff et al. [28]. Resultant concentration-response
228
G Konecny et al.
Figure 4. Combination index values and fractions of cells unaffected by gemcitabine in combination with paclitaxel at multiple effect levels in MCF-7, MDA-MB-231 and SK-BR-3 cells. () denotes mean combination index values ± SE, and (o) denotes the fraction of unaffected cells. Gemcitabine, in combination with paclitaxel, demonstrated additive and antagonistic interactions at clinically relevant dose levels in all cell lines. However, synergistic interactions (MCF-7) were observed at very low concentrations (< 3.1 nM). All drug concentrations tested were well within the range of clinically relevant doses [22, 23]. p values indicate CI values that deviate significantly from the value of 1, and confirm antagonism or synergy.
curves for all eligible assays are shown in Figure 7. At the IC90 level, the proportion of tumors responding to VIN/PTX was always lower compared to EPI/PTX, CBDCA/PTX and GEM/PTX. The activity of EPI/ PTX and CBDCA/PTX was strongly concentrationdependent between 25% and 75% PPC; however, no comparable concentration dependence could be found for GEM/PTX and VIN/PTX. Moreover, a lack of subtotal growth inhibition (IC90 ) was found in 40–50% of the breast cancer samples treated with GEM/PTX or VIN/PTX at 100% PPC. Discussion Thus far, in vitro studies in breast cancer cells using the combination of PTX and DOX, CBDCA, GEM or
VIN have generated conficting results. These discrepancies may be related to the different methodologies used for evaluating the effects of drug combinations or depend on the selected dose range. In this report, we employed the multiple drug effect/combination index isobologram method [15], as it allows analysis of drug interactions that span the entire dose range (IC10 –IC90 ). These data demonstrate that drug interactions can vary according to the selected dose range. A consistent finding in our study was that antagonism between paclitaxel and other cytotoxins tends to occur at higher paclitaxel concentrations. This observation concurs with earlier studies in MCF-7 [29] and BRC-230 cells [30], which showed that PTX causes G2 /M cell cycle arrest; thus, fewer cells are undergoing S-phase, which is the most sensitive phase
Paclitaxel combinations in breast cancer cells
229
Figure 5. Combination index values and fractions of cells unaffected by vinorelbine in combination with paclitaxel at multiple effect levels in MCF-7, MDA-MB-231 and SK-BR-3 cells. () denotes mean combination index values ± SE, and (◦) denotes the fraction of unaffected cells. In all three cell lines, at clinically relevant drug concentrations (< 0.01 µM) [24, 25], vinorelbine, in combination with paclitaxel, demonstrated antagonistic interactions in all three cell lines. However, synergistic interactions (MDA-MB-231) were observed at very low concentrations (< 2.5 nM). The shaded area of the figure represents the dose range above clinically relevant drug concentrations for VIN. p values indicate CI values that deviate significantly from the value of 1, and confirm antagonism or synergy.
for the cytotoxic effects of EPI, CBDCA and GEM [21, 31, 32]. Koechli et al. [33] reported on synergistic interactions between DOX/PTX in MCF-7 cells, but did not use higher PTX concentrations that achieved more than 60% growth inhibition. The maximum PTX concentration used in their study was 50 nM as opposed to 116 nM used in the present study. Furthermore, the maximum anthracycline concentration ratio was significantly higher (molar ratio PTX/DOX 1:17) than in the present study (molar ratio PTX/EPI 1:4), in which the fixed molar ratio was defined as the ratio of the two agents at their maximally effective dose in vitro. Moreover, the synergy described between PTX and DOX was based on simulated CI values instead
of actual experimental points, as performed in the present study. In accordance with our results, Akutsu et al. [34] observed antagonistic effects of DOX/PTX upon simultaneous exposure of MCF-7 cells to the two agents by analyzing the drug interaction with the isobologram method of Steel & Peckham [35]. However, this analysis was limited to just one drug concentration (80% growth inhibition (IC80 )). Hahn et al. [29], who exposed MCF-7 cells to DOX for 1 h, followed by incubation with PTX for 24 h, also described antagonistic interactions between DOX and PTX at high dose levels (IC90 ) in MCF-7 cells. In the present study, however, we were able to demonstrate a strong dose relationship of the observed drug interactions between paclitaxel and epirubicin, with
230
G Konecny et al. Table 1. Comparison of in vitro drug concentrations used in the current study with clinically achievable or relevent drug concentrations of CBDCA, EPI, GEM, VIN, and PTX In vitro maximum drug concentrations (µM) MCF-7 and SK-BR-3 MDA-MB-231
100% TDC ATP-TCA (µM)
Plasma peak concentrations (µM)
Clinically relevent plasma concentrations (µM)∗
References
CBDCA EPI
400 0.43
40 0.43
42.5 1.6
50–100 1.0–1.7