Contribution of c-erbB-2 and Topoisomerase IIα to Chemoresistance in Ovarian Cancer

INTRODUCTION

The c-erbB-2 (HER-2/neu) proto-oncogene encodes a M_r 185,000 transmembrane receptor tyrosine kinase that is a member of the class I receptor tyrosine kinase family, which includes the epidermal growth factor, HER-2/neu (c-erbB-2), HER-3, and HER-4 (1) . C-erbB-2 was originally identified as an oncogene in chemically induced rat neuroglioblastomas, in which a single point mutation in the transmembrane domain of the molecule confers oncogenic activation (2 , 3) . In contrast, in humans, the c-erbB-2 proto-oncogene is not activated by a point mutation but through amplification and overexpression of the wild-type gene (4 , 5) . Amplification of the c-erbB-2 gene is observed in 20–30% of many cancer types, and its overexpression is correlated with a poor prognosis for breast and ovarian cancer patients (6, 7, 8, 9) . Consistent with these clinical observations, overexpression of c-erbB-2 in human ovarian and breast cancer cell lines has been shown to promote cell growth and cause tumorigenicity in nude mice xenograft models (1 , 10, 11, 12) . Thus, the poor prognosis of ovarian and breast cancer patients associated with c-erbB-2 overexpression seemed to be sufficiently explained by a c-erbB-2-mediated high rate of tumor cell proliferation.

However, recently evidence was presented suggesting that c-erbB-2 overexpression may also be associated with cytotoxic drug resistance because a dose-response effect with regard to survival was observed in patients with c-erbB-2-negative ovarian tumors but not in patients with c-erbB-2-positive tumors (7) . Consistent with these observations, a monoclonal antibody specific to an extracellular epitope of c-erbB-2 (13) and the tyrosine kinase inhibitor emodin (14) , which suppresses c-erbB-2 tyrosine kinase activity, both enhanced the cytotoxicity of cisplatin against c-erbB-2-overexpressing cell lines. However, the role of c-erbB-2 for chemoresistance still remains inconclusive because expression of a full-length c-erbB-2 cDNA into several human breast and ovarian cancer cell lines was not sufficient to induce resistance to cisplatin across all of the cell lines tested (15) . Although the sensitivity to cisplatin decreased significantly in three cell lines transfected with c-erbB-2 cDNA (MCF7, BT-20, and 2008 cells), expression of the human c-erbB-2 cDNA did not significantly influence cisplatin sensitivity in the cell lines MDA-MB-435, MDA-MD-231, and Caov-3 (15) . Thus, the influence of c-erbB-2 on chemosensitivity was cell-type-specific and not generic across the cell lines tested.

In the present study, we examine the role of c-erbB-2 on the success of chemotherapy with carboplatin and cyclophosphamide for patients with primary epithelial ovarian cancer. We report that (a) chemotherapy significantly improves the survival time of patients with low c-erbB-2 expression but not of patients with overexpression of c-erbB-2; (b) mRNA expression of c-erbB-2 significantly correlates with mRNA expression of topoisomerase IIα; and (c) the inhibition of topoisomerase II but not the antagonization of c-erbB-2 function increased the carboplatin chemosensitivity of primary cells obtained from an ovarian tumor overexpressing c-erbB-2 and topoisomerase IIα. Thus, the consideration of topoisomerase IIα may help to understand the observed association between c-erbB-2 overexpression and chemoresistance.

MATERIALS AND METHODS

Patients, Tissue Specimens, and Pathological Data.

Tumor tissue was collected from 77 patients who underwent surgery from 1986 to 1995 at the Department of Gynecology, University of Mainz, Mainz, Germany. Tissue specimens were taken intraoperatively, immediately frozen in liquid nitrogen, and stored at −80°C. From all of the tumor specimens, 20-μm-thick cryosections were prepared using a cryostat (System Dittes Duspiva, Heidelberg, Germany). After 20 cryosections (20 μm thick), a 7-μm-thick cryosection was prepared that was reviewed by a pathologist. Only those tissue samples with more than 80% tumor cells in all of the 7-μm-thick cryosections were used for further analysis.

All of the patients examined in this study underwent surgery because of primary epithelial ovarian cancer. Patients with benign and nonepithelial tumors and borderline tumors were excluded from the study. Treatment of patients with ovarian carcinoma was primarily by surgery. Patients were treated by total abdominal hysterectomy, bilateral salpingoophorectomy, and omentectomy. In the case of FIGO 3 stages III and IV, in which cancerous tissue could not be completely removed, a debulking operation was performed to reduce the remaining tumor to a diameter not to exceed 2 cm. In this study, patients with FIGO stages Ib, Ic, II, III, and IV usually received a postoperative chemotherapy with carboplatin (350 mg/m²) or cisplatin (50 mg/m²) and cyclophosphamide (1000 mg/m²) in six courses. The first course of chemotherapy usually was performed 3–4 weeks after surgery. The following treatment regimens were defined as standard chemotherapy (n = 44): (a) six courses of 50 mg/m² cisplatin i.v. and 1000 mg/m² cyclophosphamide i.v. (n = 26); (b) six courses of 350 mg/m² carboplatin i.v. and 1000 mg/m² cyclophosphamide i.v. (n = 16); and (c) six courses of 350 mg/m² carboplatin i.v. and 600 mg/m² cyclophosphamide i.v. (n = 2). Because of myelotoxicity or nephrotoxicity, not all of the patients received a complete standard chemotherapy. In addition, some patients refused to receive a chemotherapy. Thus, several treatment regimens contained no doses or much lower doses of platinum compounds and cyclophosphamide. They were classified as “no standard chemotherapy” (n = 33): (a) no cytotoxic chemotherapy at all (n = 19); (b) four courses of 1000 mg/m² cyclophosphamide i.v. and 150 mg/m² etoposide i.v. (n = 1); (c) three courses of 1000 mg/m² cyclophosphamide i.v. (n = 2); (d) two courses of 350 mg/m² carboplatin i.v. and 1000 mg/m² cyclophosphamide i.v. (n = 1); (e) medroxyprogesterone acetate, 500 mg/day p.o. (long-term therapy; n = 1); (f) one course of 35 mg/m² epirubicin i.v. and three courses of treosulfan, 6 × 1500 mg/day p.o. (n = 1); (g) tamoxifen, 30 mg/day p.o. (long-term therapy; n = 1); (h) cyclophosphamide, 2 × 50 mg/day p.o. (long-term therapy; n = 4); (i) cyclophosphamide, 3 × 6 mg/day p.o. (long-term therapy; n = 1); and (j) cyclophosphamide, 100 mg/day p.o., and 5-fluorouracil, 750 mg/week i.v., for 12 weeks (n = 2).

Histological typing was performed according to the WHO criteria. The epithelial tumors were subdivided into serous and nonserous carcinomas. The histological grade of malignancy ranged from GI (differentiated) to GIII (undifferentiated). Grading was performed by the following criteria: tumor architecture, amount of solid tumor, nuclear pleomorphism, nucleus cytoplasma ratio, number of nucleoli, and mitoses (16) . Tumor staging was done according to the guidelines of FIGO. Because residual disease is a strong prognostic factor, we differentiated between patients with macroscopically complete removal of all of the tumor tissue and patients with ≤ 2 cm³ and >2 cm³ of tumor tissue left after surgery.

Reverse Semiquantitaive PCR.

Total RNA from frozen tissue specimens of about 10 mm³ were isolated with a commercially available kit according to the supplier’s instructions (Rneasy Mini Kit 59, Qiagen, Hilden, Germany). The RNA concentration was determined spectrophotometrically at 260 nm. The quality of the isolated RNA was checked by agarose gel electrophoresis by means of the presence of 28S, and 18S rRNA. Reverse transcription of mRNA cDNA was performed using a kit (Ready-To-Go, Pharmacia), including 500 ng of total RNA (dissolved in RNase free water) into a total reaction volume of 50 μl. The PCR mixture of 50 μl contained 7.5 mm (NH₄)₂SO₄, 30 mm Tris-HCl (pH 8.5), 1.75 mm MgCl₂ (pH 8.5), all of the four deoxynucleotide triphosphates (250 μm each), 25 pmol each of the primers (Table 1) ⇓ , 2.5 units of Taq polymerase, and 2 μl of cDNA (undiluted and, if required, in appropriate serial dilution). The PCR conditions included an initial denaturation—2 min at 95°C, 29 cycles consisting of (a) 1 min denaturation at 94°C; (b) 1 min primer annealing at 54°C for topoisomerase IIα, 60°C for topoisomerase IIβ, 60°C for c-erbB-2, and 64°C for GAPDH, respectively; (c) 1 min elongation at 72°C; and (d) one final step of 5 min at 72°C. The resulting DNA fragments were separated by electrophoresis on a 1.5% agarose gel containing 0.4 μg of ethidium bromide/ml, visualized by UV, and quantified densitometrically. If single bands were too intensive for densitometric determination (out of the linear range), they were diluted serially down to the linear range, and the intensity of the original band was calculated by multiplication with the dilution factor. To avoid artificial differences between individual gels, aliquots of the same DNA standard (a PCR product of GAPDH) were included in each gel, and the ratio of the relative intensities of individual samples:this standard were calculated. The latter values obtained for topoisomerase IIα, β, and c-erbB-2 were divided by the value of GAPDH of the respective tumor, to obtain the relative mRNA expression factor. For instance a relative mRNA expression factor of 0.5 for c-erbB-2 means that the intensity of the amplification product of c-erbB-2 was one-half of the intensity obtained for the housekeeping enzyme GAPDH. The coefficients of variation of relative mRNA expression factors of c-erbB-2 were ≤20% if the same RNA preparation was analyzed in 6 independent experiments. If RNA was isolated from 6 different sites (each approximately 1 mm³), the analyzed mRNA (topoisomerase IIα, β, or c-erbB-2) was consistently either detectable or not detectable at all of the six sites. However, if mRNA expression was detectable, the variability was relatively high (coefficients of variation from six different sites of each tumor for topoisomerase IIα and β, 58 and 51% in tumor 1 and 51 and 73% for tumor 2, respectively). Because of this relatively high degree of heterogeneity, tumors were classified as either positive or negative for mRNA expression of topoisomerase IIα, β, or c-erbB-2. A typical result of semiquantitative PCR with different volumes of cDNA given to the amplification reaction shows a sufficient quantifiability (Fig. 1) ⇓ .

Chemosensitivity Assay with Primary Tumor Cells.

The chemosensitivity assay described by Andreotti et al. and Kurbacher et al. (17 , 18) was used with modifications. Tissue specimens (approximately 0.5 cm³) from ovarian tumors were taken intraoperatively, immediately transferred into cold (4°C) serum-free DMEM (Life Technologies, Inc.) containing 300 units/ml penicillin and 300 μg/ml streptomycin, and were processed within 2 h after surgery. The tissue specimens were minced into approximately 1-mm³ fragments under sterile conditions. Fragments were then dissociated into a cell suspension of single cells and small aggregates of single cells by incubation in 10 ml of CAM (Complete Assay Medium; DCS Innovative Diagnostik Systeme, Hamburg, Germany) with 1.5 mg/ml collagenase (Type: CLS II, 375 units/mg, Seromed, Berlin, Germany) at 37°C for 4 h on a gently shaking platform. Afterward the suspension was centrifuged at 400 × g for 5 min (room temperature), resuspended in PBS, and centrifuged and resuspended for a second time. Large cell aggregates in the latter suspension were allowed to sedimentate for 1 min, and the supernatant was further purified by density-gradient centrifugation using Ficoll-Separating-Solution (Seromed, Berlin), to remove erythrocytes and dead cells. The cells from the interphase were washed twice in PBS and resuspended in a chemically defined serum-free medium (Complete Assay Medium; DCS Innovative Diagnostik Systeme, Hamburg, Germany). Several collagenase batches had to be tested to identify an appropriate batch that allowed an average cell harvest of at least 15 million cells (trypan blue exclusion, ≥95%) using ovarian tumor tissue specimens of approximately 0.5 cm². Cultures of 20,000 cells in a total volume of 200 μl of CAM per well were incubated with the test substances in 96-well round-bottomed polypropylene microplates (Costar 3790). Stock solutions of carboplatin (10 mg/liter; Sigma) and novobiocin (254 μl/ml; Sigma) were prepared in CAM, whereas the stock solution of emodin (6480 μg/ml; Sigma) was prepared in ethanol. Etoposide (VePeside; 20 mg/ml) was obtained from Bristol (Munich, Germany). The stock solutions were further diluted in CAM. All of the incubations with test substances were performed in triplicate, with the exception of the negative and positive controls (the latter incubated with 0.1 ml of Maximum-ATP-Inhibitor; DCS Innovative Diagnostik Systeme, Hamburg, Germany), which were incubated in six wells. After incubation of the cells for 7 days at 37°C in a >98% humidified, 95% air-5% CO₂ atmosphere, cellular ATP was extracted and stabilized by mixing 50 μl of Tumor-Cell-Extraction-Reagent (DCS Innovative Diagnostik Systeme, Hamburg, Germany) into each well. Twenty min after the addition of Tumor-Cell-Extraction-Reagent, the ATP content was measured in a Berthold LB-953 luminometer using 50 μl of culture extract injected with 50 μl of Luciferin-Luciferase-Reagent (DCS Innovative Diagnostik Systeme, Hamburg, Germany). An ATP standard curve was performed using 50-μl aliquots of a 250-ng/ml ATP standard serially diluted 1:3 in dilution buffer. ATP content of solvent-treated control cells was defined as 100%. The additivity of substances (inhibitors) with carboplatin was determined as follows: the mean ATP value was defined as the mean ATP content (% of controls) of cells incubated with 1, 2, 4, 8, 16, and 32 μg/ml of carboplatin. The effect of the combination of carboplatin with an inhibitor was defined as “less than additive” if the difference between the mean ATP value of carboplatin alone and the mean ATP value of carboplatin combined with inhibitor was smaller than the reduction in ATP content by the inhibitor alone.

Isolation of RNA and S1 Nuclease Assay.

Total RNA was isolated by a single-step method using the guanidinium thiocyanate-phenol-chloroform extraction technique (19) . The concentration of RNA was determined spectrophotometrically at 260 nm. After phenol extraction and ethanol precipitation, all of the RNA samples were stored as aliquots of 20 μg at −80°C until use. All of the total RNAs prepared from cryopreserved tumor samples were checked for integrity by conventional agarose gel electrophoresis before the S1 nuclease assay, by means of the presence of 28S and 18S rRNA. Degraded RNA samples were excluded from the study. The S1 nuclease protection assays were performed as described previously (8 , 20) . Briefly, the EcoRI/NcoI fragment of the c-erbB-2 was isolated from the plasmid pMAC-14-1, which contains the full-lenghth c-erbB-2 cDNA (21) and subcloned into the EcoRI/SmaI site of pSP65 involving first ligation at the EcoRI site and then filling in the NcoI site by T₄-DNA-polymerase in the presence of dNTPs followed by a second ligation of the remaining blunt ends. Radioactively labeled antisense RNA of the c-erbB-2 cDNA fragment (EcoRI/NcoI) was prepared by in vitro transcription of the resulting construct (pSP-erbB-2) using SP6-polymerase (Boehringer Mannheim, Germany) and PvuII-linearized pSP-erbB-2-DNA. The transcript was purified by phenol extraction and ethanol precipitation. A total of 10⁵ dpm of radioactively labeled transcript (specific activity 10⁷ dpm/μg RNA) were mixed with 5 μg of total tissue RNA and 100 μg of Escherichia coli tRNA, dried in a speed vac, and resuspended in hybridization buffer (40 mm PIPES (pH 6.4), 1 mm EDTA, 400 mm NaCl, and 80% formamide). The RNA mix was denatured by incubation at 90°C for 10 min, and subsequent hybridization was carried out at 52°C for 3 h. Fifty units S1 nuclease (Pharmacia) in 300 μg S1-buffer (30 mm sodium acetate (pH 4.6), 50 mm NaCl, and 1 mm ZnCl) was added and incubation was continued at 37°C for 60 min. After phenol-chloroform extraction and ethanol precipitation, the hybrids were dissolved in loading buffer, heat-denatured at 95°C for 15 min, and separated by denaturing PAGE (containing 8 m urea). Quantitation of mRNA expression was carried out by densitometry (Densitometer Elscript 400 from Kirschmann). For quantification, two standards were included in each gel. Because of densitometric determination, the intensity that the band obtained from standard 1 was defined as 0.05 (relative c-erbB-2 mRNA expression, arbitrary unit). Compared with standard 1, the intensity of standard 2 was 0.5. Both of the standards were used to calculate the intensity of the bands of each tumor sample. The problem of false negative results as a consequence of mRNA degradation was excluded by analyzing the samples for expression GAPDH mRNA. The coefficient of variation of five determinations of the same RNA preparation was smaller than 20% (three tumors tested). In the present study, mean values of two determinations (of a single RNA preperation) were calculated for each tumor.

Statistical Analysis.

Kaplan-Meier curves were plotted to assess overall survival (22) . Different survival curves were compared using the log-rank test. The proportional hazard modal (22) was applied to examine whether c-erbB-2 was an independent prognostic factor (univariate and multivariate analysis). A possible association between expression of c-erbB-2 and topoisomerase IIα or IIβ was investigated using the χ² test. To examine whether the correlation between c-erbB-2 and topoisomerase IIα is independent from FIGO stage, the Cochran-Mantel-Haenszel test adjusted for FIGO stage I, II versus III, IV was applied. All of the two-sided P-values smaller than 0.05 were considered as statistically significant. Statistical analysis was performed using the SAS software (Cary, NC).

RESULTS

Association of c-erbB-2 with Survival Time.

Expression of c-erbB-2 mRNA was determined by S1 nuclease assay in the tumor tissue specimens of 77 patients with primary ovarian cancer. Expression of c-erbB-2 varied widely with 25, 50, 75, and 100% percentiles of 0, 0.4, 1.2, and 8.9 (arbitrary units). Kaplan-Meier analysis of all of the patients with primary ovarian cancer (FIGO stages I-IV) showed a significant association between c-erbB-2 mRNA expression and survival time (Fig. 2 ⇓ ; P = 0.0001, log-rank test). Median survival times were 7.1, 4.1, and 1.2 years for patients with low (minimum to median), intermediate (median-75% percentile), and high (75% percentile-maximum) c-erbB-2 expression, respectively. The proportional-hazard model (Kalbfleisch and Prentice, 1980) was applied to examine whether c-erbB-2 mRNA expression was a prognostic factor independent of other known risk factors, such as FIGO stage, residual disease, chemotherapy (application of standard chemotherapy versus no standard chemotherapy), and age. Multivariate analysis showed that c-erbB-2 was an independent prognostic factor (P = 0.0358). C-erbB-2 expression was significantly associated with FIGO stage (P = 0.0093; Wilcoxon test). Median c-erbB-2 mRNA expression was 0 (0–2.6; 5-95%-percentiles) and 0.7 (0–6.1) for tumors classified as FIGO stages I+II and FIGO stages III+IV, respectively. No significant association of c-erbB-2 mRNA expression with histological grade (grade 1, 2, 3) and type (serous versus nonserous type) was observed.

Association of c-erbB-2 with Response to Chemotherapy.

To examine, whether c-erbB-2 was associated with the response to chemotherapy we analyzed subgroups of patients who either received or did not receive a standard chemotherapy with carboplatin or cisplatin and cyclophosphamide. A significant association between c-erbB-2 mRNA expression and survival was obtained for patients who received a standard chemotherapy (P = 0.0003; log-rank test; Fig. 3a ⇓ ). Median survival times were 8.4, 5.3, and 1.4 years for patients with low (minimum to median), intermediate (median to 75% percentile), and high (75% percentile to maximum) c-erbB-2 mRNA expression, respectively. The proportional-hazard model adjusted for FIGO stage, residual disease, and age showed that c-erbB-2 was also an independent prognostic factor for this subgroup of patients (P = 0.039). However, only a nonsignificant trend between c-erbB-2 and survival was observed for patients who did not receive a standardized chemotherapy (P = 0.124; log-rank test; Fig. 3b ⇓ ). Median survival for the latter group of patients was 3.3, 1.1, and 0.8 years for individuals with low, intermediate, and high c-erbB-2 expression.

If c-erbB-2 expression is associated with resistance to chemotherapy with carboplatin and cyclophosphamide, a better response to chemotherapy could be expected for patients with low c-erbB-2 expression compared with patients with relatively high expression. Thus, we analyzed survival for patients with relatively low (minimum to 75% percentile) and high (75% percentile to maximum) c-erbB-2 expression with respect to chemotherapy (Fig. 3. c and d) ⇓ . For patients with low c-erbB-2 expression, application of chemotherapy with carboplatin or cisplatin and cyclophosphamide resulted in significantly longer survival compared with patients who did not receive a standard chemotherapy (Fig. 3c ⇓ ; P = 0.013, log-rank test). However, for patients who overexpressed c-erbB-2 in tumor tissue, no significant improvement of survival by chemotherapy was observed (Fig. 3d ⇓ ; P = 0.359, log-rank test).

Correlation of c-erbB-2 and Topoisomerase IIα mRNA Expression.

One of several mechanisms that may explain the resistance of c-erbB-2-overexpressing tumors to carboplatin/cyclophosphamide is coregulation or coamplification of c-erbB-2 and topoisomerase IIα (which both are located on chromosome 17q21). To examine a possible correlation between mRNA expression of c-erbB-2 and topoisomerase IIα, we established a reverse, semiquantitative PCR technique using primers specific for c-erbB-2 and topoisomerase IIα (Fig. 1) ⇓ . As a negative control, expression of topoisomerase IIβ was also determined because—in contrast to topoisomerase IIα and c-erbB-2—topoisomerase IIβ is located on chromosome 3. A significant correlation was obtained between mRNA expression of topoisomerase IIα and c-erbB-2 (P = 0.009, χ² test; n = 70), whereas c-erbB-2 and topoisomerase IIβ did not correlate significantly (P = 0.221, χ² test; n = 70; Table 2 ⇓ ). For 22 of the 70 examined patients, c-erbB-2 mRNA was detectable. Seventeen (77%) of these 22 patients expressed topoisomerase IIα mRNA also, whereas only 5 (23%) of them were negative for topoisomerase IIα. In contrast, for 48 patients, c-erbB-2 mRNA was not detectable. Twenty-one (44%) of them had detectable topoisomerase IIα, whereas 27 (56%) were negative for topoisomerase IIα.

Because the observed correlation between c-erbB-2 and topoisomerase IIα could theoretically represent a secondary effect due to an association of both c-erbB-2 and topoisomerase IIα with FIGO stage, the analysis was adjusted by FIGO stage (Table 2) ⇓ . However, the correlation between mRNA expression of c-erbB-2 and topoisomerase IIα was still significant (Cochran-Mantel-Haenszel test adjusted for FIGO stages I, II versus II, IV).

Inhibition of c-erbB-2 and Topoisomerase II in Vitro.

To examine whether chemosensitivity to carboplatin can be influenced by the inhibition of c-erbB-2 and/or topoisomerase II, we established a chemosensitivity assay using primary cells obtained from fresh ovarian cancer tissue specimens. Incubations were performed with six concentrations of carboplatin, with and without the addition of inhibitors (Fig. 4) ⇓ . Concentrations of carboplatin (16 μg/ml) and etoposide (48 μ g/ml) were based on reported peak plasma concentrations in vivo (17 , 18) . Inhibition of topoisomerase II by both etoposide and novobiocin significantly increased the chemosensitivity to carboplatin under conditions in which etoposide and novobiocin alone were not toxic (Fig. 4a) ⇓ . In contrast to the inhibition of topoisomerase II, antagonization of the tyrosine kinase function of c-erbB-2 by a nontoxic concentration of emodin (2 μ m) did not increase chemosensitivity to carboplatin (Fig. 4b) ⇓ . The effect of a moderately toxic concentration of emodin (4 μm) was less than additive. Meanwhile, primary cells of six epithelial ovarian carcinomas were examined for chemosensitivity. For two of six tumors, etoposide and novobiocin clearly increased carboplatin toxicity (effect of combination: more than additive), whereas a less-than-additive effect on carboplatin sensitivity was observed for the other carcinomas. Thus, an improvement of carboplatin chemosensitivity by the inhibition of topoisomerase II seems to be possible only in a subgroup of patients. One of the two tumors with a more-than-additive effect of carboplatin/etoposide combination showed high topoisomerase IIα as well as high c-erbB-2 expression (both within the highest quantile of all of the patients examined). The second tumor with a more-than-additive effect showed a high expression of topoisomerase IIα (within the highest quantile) but no detectable expression of c-erbB-2 mRNA. However, we cannot speculate on a possible correlation between expression of topoisomerase IIα and sensitivity to the carboplatin/etoposide combination because the number of cases is still too small and especially because one of the tumors with a less-than-additive effect for the carboplatin/etoposide combination showed a high expression of topoisomerase IIα mRNA.

DISCUSSION

A large number of studies have suggested the involvement of some oncogenes in the development of cytotoxic drug resistance. For instance, transfection of c-myc into several cell lines was shown to increase resistance to cisplatin (23 , 24) , whereas transfection with antisense c-myc increased sensitivity to cisplatin (25) . Similarly, transfection of H-ras into MCF-7 cells (26) or into NIH 3T3 cells (27) markedly increased resistance to cisplatin. However, the relationship between oncogene expression and drug resistance may be indirect because the transfection of v-H-ras or v-raf into rat liver epithelial cells led to an increased expression of P-glycoprotein (mdr-1) and glutathione S-transferase π, which may cause resistance (28) . Thus, some oncogenes are unlikely to modify drug sensitivity directly but may influence expression of other genes that are responsible for drug resistance.

In the present study, we show that overexpression of the oncogene c-erbB-2 is associated with shorter survival time of ovarian cancer patients, confirming results of earlier investigations (7, 8, 9 , 29) . In addition, expression of c-erbB-2 mRNA was significantly associated with FIGO stage, which gives a measure of tumor extension. Thus, overexpression of c-erbB-2 seems to increase with tumor progression. Despite its association with FIGO stage, c-erbB-2 mRNA expression was also an independent prognostic factor in the proportional-hazard model (Cox model) adjusted for FIGO stage, residual disease, application of standard chemotherapy, and age, which represent the most important classical prognostic factors of ovarian cancer.

To examine a possible involvement of c-erbB-2 in the development of cytotoxic drug resistance, we compared survival times of patients who received a standard chemotherapy with those of patients who did not receive a standard chemotherapy. A highly significant association between survival and expression of c-erbB-2 mRNA was obtained for patients with standard chemotherapy, whereas only a nonsignificant trend toward longer survival was observed for patients without standard chemotherapy. In addition, the application of a standard chemotherapy increased the survival time of patients with low c-erbB-2 expression but not of patients with c-erbB-2 overexpression. These data suggest that c-erbB-2 overexpression may be involved in the development of resistance against chemotherapy with cisplatin or carboplatin and cyclophosphamide. However, this conclusion still has to be treated with caution because the selection of patients without standard chemotherapy may cause a bias. Patients without standard chemotherapy received no doses or only very low doses of platinum compounds and cyclophosphamide because of nephrotoxicity or myelotoxicity or because patients refused to receive chemotherapy. The first two reasons (nephrotoxicity and myelotoxicity) may represent a selection criterion that possibly exerts an influence on prognosis independent of chemotherapy. Nevertheless, the evidence that c-erbB-2 may be involved in the development of drug resistance was used as a working hypothesis and was further investigated.

If c-erbB-2 is involved in the development of drug resistance, two possibilities seem to be plausible: (a) a mechanism mediated by the tyrosine kinase function of c-erbB-2; and (b) coamplification and/or coregulation of c-erbB-2 with topoisomerase IIα (Fig. 5) ⇓ . The first mechanism includes heterodimerization of c-erbB-2 with other members of the class I family of receptor tyrosine kinases, such as the epidermal growth factor receptor, HER-3, or HER-4 (1) , which generates a high affinity receptor for a family of ligands termed neuregulins (Fig. 5 ⇓ ; Ref. 30 ). Activation of the c-erbB-2 receptor induces tyrosine phosphorylation on specific tyrosine residues in the COOH terminus of c-erbB-2 (31 , 32) . After phosphorylation, downstream signaling proteins such as SHC and/or GRB2 bind to the COOH terminus of c-erbB-2 through src homology 2 domains (33 , 34) . Interaction of the guanine nucleotide exchange factor (SOS) with SHC/GRB2 brings the complex to the cytoplasmic surface of the plasma membrane, in which SOS activates ras-GDP to ras-GTP (35) , which leads to the activation of MAP kinase (36) and culminates in the transcription of nuclear transcription factors, such as c-myc, c-fos, and c-jun. The effects of the latter products may explain not only an increased mitogenic activity but possibly also chemoresistance because transfection of c-myc into Friend erythroleukemia cells was shown to decrease chemosensitivity to cisplatin (24) . However, resistance to cisplatin may also be caused by MAP kinase-mediated transcription of yet unknown other genes involved in DNA excision repair or in inactivation of cisplatin. Similarly, the c-erbB-2-mediated phosphodiesterase phospholipase C-γ or phosphatidylinositol 3-kinase pathways may induce the transcription of resistance-associated genes, although there is no current experimental evidence for this.

The second mechanism by which c-erbB-2 may be associated with resistance to platinum compounds is coamplification or coregulation with topoisomerase IIα. The chromosomal location of both, c-erbB-2 and topoisomerase IIα is 17q21–22. Amplifications occur over large segments of the chromosome and may comprise many mega bases (37) . Topoisomerase IIα was found to be coamplified in 12% of c-erbB-2 amplified cases (38) and similar results were obtained by other investigators (39) . Thus, a subgroup of patients with c-erbB-2 overexpression is likely to have a high topoisomerase IIα activity. High activity of topoisomerase IIα is known to be associated with high sensitivity toward topoisomerase II-inhibitors such as anthracyclines and epipodophyllotoxines. However, for platinum compounds and alkylating antineoplastic agents, the opposite constellation has been shown. Topoisomerase IIα -transfected cell lines were ∼5- to 10-fold more resistant to cisplatin and the alkylating agent mechlorethamine (40) . In addition, several types of topoisomerase IIα inhibitors enhanced the toxicity of platinum compounds and alkylating agents in topoisomerase IIα-expressing cells in vitro (41, 42, 43) . Because the inhibition of topoisomerase IIα significantly decreased the repair of DNA interstrand cross-links with a corresponding decrease in the clonogenic survival of the cells (44, 45) , topoisomerase IIα possibly influences chemosensitivity by altering the accessibility of DNA sequences for alkylation and for subsequent repair, although this has not yet been shown directly.

The hypothesis that coamplified and/or coregulated topoisomerase IIα is responsible for the chemoresistance of c-erbB-2-overexpressing carcinomas, to our knowledge, has not yet been investigated. We observed a significant correlation between c-erbB-2 and topoisomerase IIα mRNA expression in primary ovarian cancer tissue specimens that was independent of the FIGO stage. In contrast, topoisomerase IIβ, which is located on chromosome 3, did not correlate with c-erbB-2. Because we examined mRNA expression, which was chosen because it is a step closer to the functional protein than DNA analyzed by Southern blotting, it is not possible to differentiate whether this correlation is due to coamplification or coregulation, and it was not the aim of the present study to analyze the molecular mechanism of the correlation between c-erbB-2 and topoisomerase IIα. Besides coamplification, a likely explanation for the correlation between c-erbB-2 and topoisomerase IIα may be a c-erbB-2-mediated stimulation of cell division (1 , 10, 11, 12) , which may cause an increase in topoisomerase IIα expression because topoisomerase IIα is required for relieving torsional stress during DNA replication, is up-regulated in actively dividing cells, and is down-regulated by the end of mitosis (38) . The hypothesis that coamplified and/or coregulated topoisomerase IIα contributes to the resistance of c-erbB-2-overexpressing carcinomas was additionally supported by an examination of primary cells obtained from an ovarian carcinoma that overexpressed both c-erbB-2 and topoisomerase IIα mRNA. The combination of carboplatin with nontoxic concentrations of the topoisomerase II-inhibitors etoposide or novobiocin significantly enhanced toxicity in the primary ovarian cancer cells. Using the pharmacological inhibitors etoposide and novobiocin, we cannot exclude nonspecific effects. However, the enhanced sensitivity caused by etoposide or novobiocin is unlikely to be caused by a decreased capacity of the primary tumor cells (Fig. 4) ⇓ to repair a certain threshold quantity of DNA damage in general because these cells were relatively resistant to carboplatin. The ATP content of the cells shown in Fig. 4 ⇓ was 69% of controls after incubation with 16 μg of carboplatin/ml compared with the mean ATP-content of 41 ± 19% (mean ± SD) of 14 examined ovarian cancer tissue specimens. In contrast to the topoisomerase II inhibitors, the combination of carboplatin with the tyrosine kinase inhibitor emodin under conditions known to antagonize c-erbB-2 function (14) did not enhance carboplatin toxicity. Obviously, the sensitivity of the examined primary ovarian cancer cells can be improved by the inhibition of topoisomerase II but not by the antagonization of c-erbB-2 function, which shows that c-erbB-2 is not necessarily associated with susceptibility to carboplatin. Of course, this observation does not exclude the possibility that c-erbB-2 may be involved in the development of drug resistance in other cell types.

There have been clinical trials of topoisomerase inhibitors and platinum complexes in ovarian cancer patients. For instance, i.p. application of cisplatin (or carboplatin) and etoposide as a consolidation therapy for advanced epithelial ovarian cancer resulted in a significant—although not very large—increase in disease-free survival time (46 , 47) . In our study with primary ovarian cancer cells, a clear improvement of carboplatin chemosensitivity by etoposide was observed only in a subgroup of patients. Thus, identification of this subgroup could result in a more specific and more effective application of carboplatin/etoposide combination chemotherapy. Presently, we are investigating whether this subgroup can be identified by the determination of topoisomerase IIα expression.

In conclusion, overexpression of c-erbB-2 was associated with poor prognosis and poor response to platinum compounds and cyclophosphamide in patients with primary ovarian cancer. Topoisomerase IIα, which correlates with c-erbB-2 expression, is likely to contribute to the resistance of c-erbB-2-overexpressing carcinomas.