Main

Topoisomerase IIα (Topo IIα) is a vital nuclear DNA-binding enzyme that controls and modifies the topologic states of DNA (Berger et al, 1996) by combining nuclease, helicase and ligase activities. Topo IIα reduces DNA supercoiling and twisting by creating a double-strand nick that enables the passage of a second DNA double-strand through the break and subsequent religation of the cleaved DNA strand. Topo IIα is the specific target of several chemotherapeutic agents, including anthracyclines. These drugs bind to Topo IIα–DNA complexes and inhibit the religation step, which results in stabilisation of the DNA double-strand breaks that are thought to induce apoptosis (Kellner et al, 2002).

It has been previously demonstrated in vitro that the sensitivity of tumour cell lines to various Topo IIα poisons, such as anthracyclines or etoposide, depends on the level of Topo IIα expression (Gudkov et al, 1993; Asano et al, 1996a, 1996b; Vassetzky et al, 1996; Withoff et al, 1996a, 1996b; Zhou et al, 1999; Stacey et al, 2000). Moreover, the TOP2α gene being located at 17q12–21 close to the Her-2/neu oncogene, coamplification of the Her-2/neu and TOP2α genes was observed in 44% of breast cancers, whereas the TOP2α was deleted in another 42% (Jarvinen et al, 2000). It also appears that Topo IIα expression is often correlated to Her-2/neu overexpression in breast carcinoma. This complex relationship between the two genes may explain the altered sensitivity to anthracyclines of Her-2/neu-amplified breast carcinomas (Bitran et al, 1996; MacGrogan et al, 1996a; Niskanen et al, 1997; Clahsen et al, 1998; Jarvinen et al, 1998; Paik et al, 1998; Thor et al, 1998; Vincent-Salomon et al, 2000). Most previous studies regarding the effect of Topo IIα poisons in relation to the cellular level of Topo IIα expression were either performed in vitro or on breast tumour fragments without analysis of the direct in vivo effect (Gudkov et al, 1993; Asano et al, 1996a, 1996b; Vassetzky et al, 1996; Withoff et al, 1996a, 1996b; Zhou et al, 1999; Stacey et al, 2000). The aim of our study was to analyse this effect in a series of breast carcinoma patients treated by primary chemotherapy, and to study the relationship between Topo IIα and Her-2/neu in this setting. To this end, we investigated the predictive and prognostic values of Topo IIα expression by immunohistochemical detection of the enzyme in breast tumour core biopsies from patients with large operable invasive cancers of the breast treated by primary chemotherapy including epirubicin. Furthermore, we studied the relationship between Topo IIα expression and different factors modifying tumour chemosensitivity, such as Her-2/neu, hormonal receptor, Ki-67 and p53 immunohistochemical detection.

Materials and methods

Patients

Immunohistochemical detection of Topo IIα was performed on tumour core biopsies from 128 patients with primary metastasis-free operable breast cancers, larger than 3 cm. These patients belonged to the neoadjuvant chemotherapy arm of a randomised phase III trial that compared modified radical mastectomy followed by adjuvant chemotherapy to neoadjuvant chemotherapy followed by adapted locoregional treatment in large operable breast tumours. The clinical trial was conducted at Bergonié Institute from January 1985 to April 1989 and included a total of 272 patients. The chemotherapy regimens used in the trial comprised three courses of epirubicin, vincristine and methotrexate (EVM) followed by three courses with mitomycin C, thiotepa and vindesin (MTV), for more details see Mauriac et al (1999). All the biopsies analysed in the present study came from the primary chemotherapy, arm of the clinical trial. After completion of the six courses of chemotherapy, clinical examination and mammography were used to assess tumour regression. Subsequent locoregional treatment depended on the extent of tumour regression: radiotherapy was applied exclusively in case of complete regression, conservative surgery with axillary node dissection followed by radiotherapy were performed when tumour regression was incomplete with residual tumour measuring less than 2 cm in diameter; the remaining patients underwent mastectomy.

The predictive and prognostic value of the immunohistochemical detection of oestrogen and progesterone receptors, p53, Her-2/neu, and Ki-67 (antibody: Mib-1), had been analysed previously on the same cohort (MacGrogan et al, 1996a) and served as reference data for the current study.

Immunohistochemical assay

The pretherapeutic tumour core biopsies fixed in Holland's fluid and embedded in paraffin were retrieved from the files of Institute Bergonié. Five μm thick were cut, mounted on silane-coated slides and routinely processed. Antigen retrieval was obtained by heating the sections immersed in 0.01 M citrate buffer, pH 6.0, in a pressure cooker for 20 min. The sections were then incubated for 60 min with Ki-S7, a monoclonal antibody specific for Topo IIα. Ki-S7 was elaborated after immunising Balb/c mice with crude L428 nuclear extracts, performing mouse splenic/myeloma hybridoma cell cultures and purifying cell culture supernatants on a G-Sepharose 4 Fast Flow column (Pharmacia LKB; Freiburg, Germany) (Kellner et al, 1997). Ki-S7 specificity for Topo IIα was verified in immunoprecipitation and Western blot experiments. Ki-S7 immunoreactivity on archival paraffin-embedded tumour material using an antigen retrieval procedure was also controlled (Kellner et al, 1997).

Ki-S7 was used undiluted as a lyophilised cell culture supernatant reconstituted with 5 ml H2O. The immunoreaction was enhanced by means of an avidin–streptavidin–biotin peroxidase technique (Strept ABC complex/HRP Duet kit, Dako, France) with diaminobenzidine as a chromogen. MCF7 cell pellets with a high percentage of cells in G2 phase were fixed in Holland's fluid and embedded in paraffin to serve as positive controls in each series assayed immunohistochemically for Topo IIα expression. Negative controls consisted of normal nonhyperplastic epithelial cells present in terminal ductal lobular units in the breast core biopsies.

All slides were read by one of the authors (GMG) who was blinded to the clinical results. Only unequivocal nuclear staining of invasive tumour cells was scored as positive (Figure 1). The percentage of positive tumour cells per tissue section was determined semiquantitatively in 5% steps.

Figure 1
figure 1

Nuclear immunostaining (Topo IIα) of 25% cells in a Holland's bouin-fixed paraffin-embedded microbiopsy of an invasive ductal carcinoma NOS using monoclonal antibody Ki-S7 specific for topoisomerase IIα (A). Haematoxylin eosin saffron stain of the same case (B). Scale bar=50 μm.

Full size image

Statistical analysis

The threshold used for p53 positivity was 1%; for IHC-ER and IHC-PR positivity, the threshold was 10%. These optimal thresholds have already been determined in previous studies to be the most informative for clinical outcome (de Mascarel et al, 1995; MacGrogan et al, 1995; MacGrogan et al, 1996a, 1996b). A Ki-67 (Mib1) index of 40% was arbitrarily chosen as a threshold, at the beginning of the study, to differentiate highly proliferating tumours. This Ki-67 threshold had been chosen in our previous study on the same group (MacGrogan et al, 1996a), because it corresponded to the 75th percentile of Ki-67 expression in the series. Furthermore, 10% of positive invasive cancer cells with a moderate to strong intensity of staining was chosen for the Her2neu positivity threshold (Slamon et al, 2001). Finally, the median values of patients' ages and tumour sizes were chosen as cutoff points for the statistical analysis.

The χ2 test was used to investigate the significance of the relationship between Ki-S7 and the different immunohistochemical factors previously studied as well as the histological Scarff Bloom and Richardson (SBR) grade. The relation between Ki-S7 and the patient's age as well as tumour size was analysed by Student's t-test. A Kendall τ rank correlation test was performed to study the relationship between Ki-S7 and Ki-67, considering them as continuous variables.

The clinical size of the tumours was assessed before treatment, before the second and fourth courses of chemotherapy and after the sixth. A univariate analysis studying the relationship between tumour regression and the different factors was performed, using the χ2 test. Reciprocal influence among the different predictive factors was determined by multivariate analysis using a logistic regression test for multiple proportional hazards. The variable to predict was tumour regression ⩾50%, including complete tumour regression. To evaluate the predictive impact of Ki-S7 expression in this series, a first logistic regression model was performed (model 1), including all previously analysed factors without Ki-S7 (MacGrogan et al, 1996a). A second logistic regression model was then performed, after adding Ki-S7 (model 2). All factors were included in the logistic regression analyses, irrespective of their P-value by univariate analysis; but only those with a P-value ⩽1% were retained in the final models.

The log-rank test using the Kaplan–Meier method was also used to study the relationship between KiS-7 and prognosis expressed as a 5-year probability of survival. Patient follow-up was carried out quarterly for 2 years, twice a year and finally yearly. For overall survival (OS), the survival duration was calculated from the randomisation date to death, or the date they were last known alive. All causes of death were considered as events. For metastasis-free interval (MFI) and disease-free interval (DFI), time to failure was computed from the randomisation date until metastasis or relapse, or the date they were last known to be disease-free, respectively. The median follow-up of patients in this series was 142 months, range 101–172 months.

Univariate and multivariate analyses (logistic regression model) were performed using version 10.0 of SPSS software.

Results

Invasive carcinoma cells were present in 125 out of 128 biopsies; three core biopsy paraffin blocks contained no residual tumour material after tissue sectioning for histological diagnosis and analysis of the other immunohistochemical factors in the series, and the corresponding cases were excluded from the analysis. Patients' age in this series ranged from 31 to 70 years (mean: 53, median: 55). The mean and median tumour sizes were 43.2 and 40 mm, respectively (range: 20–80 mm). The majority of tumours in this series were invasive ductal carcinomas of no special type (115, 92.2%), followed by invasive lobular carcinomas (9, 7.2%) and one mucous carcinoma (0.6%).

Ki-S7 expression in the series

Ki-S7 positivity ranged from 0 to 50% with a median value of 15% positive tumour cells. 15% was consequently chosen as the threshold value between low (⩽ 15%) and high (>15%) Topo IIα expression.

Relationship between Ki-S7 and other parameters (Table 1)

Table 1 Relationship between Ki-S7 expression and classical and immunohistochemical markers
Full size table

Ki-S7 was positively associated with SBR grade and p53 expression (P=1 × 10−6. and 0.03, respectively) and negatively with the immunoexpression of ER and PR (P=0.007 and 0.03, respectively). No significant association was found between Ki-S7 and Her-2/neu or tumour size (Table 1).

Ki-S7 and Ki-67 were strongly positively correlated (τ=0.46, P<10−3) and associated (P=4 × 10−7), considering them as continuous or dichotomous variables, respectively.

Predictive and prognostic values of Ki-S7

High expression of Ki-S7 (>15%) was associated with tumour regression ⩾50% including complete tumour regression after six courses of chemotherapy (OR=2.88, CI: 1.3–6.4, P=0.004) by univariate analysis (Table 2). The median Ki-S7 value was higher in the good response group (tumour regression ⩾50% and complete tumour regression) compared to the poor response group (tumour regression <50%, tumour stabilisation and progression), 20 and 15%, respectively. To evaluate the predictive impact of Ki-S7 in this series, a first stepwise logistic regression model was performed with all the factors that had been previously analysed in the series, that is, clinical tumour size, SBR grade, IHC-ER, IHC-PR, Ki-67, p53 and Her-2/neu (Table 3, model 1). In this model, clinical tumour size less than 40 mm, negative IHC-ER status and high expression of Ki-67 (>40%) were found to be independent predictive factors for tumour regression. When Ki-S7 was added (Table 3, model 2), independent predictive factors were clinical tumour size, IHC-ER and Ki-S7. Ki-67 expression was no longer considered an independent predictive factor for tumour regression in this second model (P=0.092).

Table 2 Factors associated with tumour regression ⩾50%, including complete tumour regression after six cycles of primary chemotherapy among 125 patients: univariate analysis
Full size table
Table 3 Factors associated with tumour regression ⩾50%, including complete tumour regression after six cycles of primary chemotherapy among 125 patients: multivariate analyses (model 1: factors identified in MacGrogan 1996a; model 2: with new factor, Ki-S7)
Full size table

Considering the impact of concurrent Topo IIα and Her-2/neu overexpression on tumour chemosensitivity described in the literature, the association of these two proteins was analysed in relation to tumour regression, but no significant correlation was found either by univariate or multivariate analysis.

By univariate analysis, no statistically significant relationship was found between Ki-S7 and DFI, MFI or OS (results not shown).

Discussion

Previous studies, showing a relationship between the intracellular level of Topo IIα and tumour chemosensitivity, were mainly based on in vitro experiments analysing the efficacy of other Topo IIα inhibitors in various tumour cell lines, in particular etoposide (Gudkov et al, 1993; Asano et al, 1996a, 1996b; Vassetzky et al, 1996; Withoff et al, 1996a, 1996b; Stacey et al, 2000). In the current study, we have demonstrated for the first time a significant direct in vivo correlation between the level of Topo IIα expression and breast cancer response to anthracycline-based chemotherapy. This result argues in favour of the concept that the efficacy of Topo IIα poisons depends on the quantity of stabilised cleavable complexes, which in turn relates to the cellular expression and activity of Topo IIα (Kellner et al, 2002).

Jarvinen et al, (1998) found no significant correlation between Topo IIα expression in primary breast tumours and the regression of metastases after chemotherapy in patients with advanced breast cancer treated with first-line epirubicin. The discrepancy between their results and ours may be attributable to different Topo IIα contents in primary tumours and their metachronous metastases Indeed, in a series of 13 breast tumours, the same group subsequently showed some degree of variability between the Topo IIα gene status of the primary breast tumour and that of its metastases by means of fluorescent in situ hybridisation (FISH) and chromogenic in situ hybridisation (CISH) (Tanner et al, 2001). Finally it is conceivable that due to intervening acquisition of drug-resistance mechanisms (Giesler et al, 2002), the effect of Topo IIα inhibitors may be different in primary tumours and their corresponding asynchronous metastases, especially if an adjuvant protocol comprising a Topo IIα inhibitor had previously been applied.

In line with the proliferation-specific expression of Topo IIα (Kellner et al, 1997, 2002), we observed a highly significant correlation between Ki-S7 and Ki-67, which was also reported by others (Depowski et al, 2000; Rudolph et al, 1999a; Tuccari et al, 1993), in addition to a strong link between the number of Topo IIα-expressing cells and the S-phase fraction of the cell cycle (Jarvinen et al, 1996; Sandri et al, 1996a). Given the known association between high tumour cell proliferation and chemosensitivity (Tannock, 1978; Belembaogo et al, 1992; O'Reilly et al, 1992), one might suspect the observed therapy response to be related to the proliferative activity of the tumours. Indeed, when we previously investigated the predictive value of various immunohistochemical factors in the same cohort (MacGrogan et al, 1996a), Ki-67 came out as an independent predictive factor for the therapy response in the multivariate analysis. After inclusion of Ki-S7 into the logistic regression model, however, Ki-67 lost its statistical significance. This suggests that Topo IIα expression may have an impact on the responsiveness of tumours to cytotoxic therapy beyond its association with tumour cell proliferation, which is likely related to an increased sensitivity to the Topo IIα inhibitory component of the treatment protocol, that is, epirubicin (Kellner et al, 2002). It is not possible to reproduce in vitro experiments identically in vivo because no current therapy protocol for breast cancer consists of a monotherapy with Topo IIα inhibitors. It would nevertheless be difficult to explain the predictive superiority of Ki-S7 compared with Ki-67 than by a higher availability of substrate for the anthracycline component of the therapy regimen (Kellner et al, 2002). It is true that the assessment of the immunopositive cell fraction does not allow conclusions as to the protein content of individual cells, and that a low expression at the cellular level might account for the failure of a number of tumours with a high percentage of Ki-S7-positive cells to respond to therapy. More likely, however, this lack of responsiveness is attributable to resistance mechanisms (Kellner et al, 2002) acquired during tumour progression, consistent with a lower response rate in larger tumours (Gieseler et al, 2003).

On the other hand, we found a high Ki-S7 expression indicative of an aggressive tumour phenotype in terms of higher SBR grade, hormonal receptor negativity and p53 overexpression. These observations substantiate the results of previous studies (Jarvinen et al, 1996; Rudolph et al, 1999a, 1999b). Since wild-type p53 negatively regulates the promoter of the human TOP2α gene (Sandri et al, 1996b), the immunohistochemical detection of p53, consistent with a nonfunctional protein, may be expected to be associated with high levels of Topo IIα.

Contrary to previous results from our group (Rudolph et al, 1999a) and others (Jarvinen et al, 1996, 1998; Rudolph et al, 1999b), we failed to observe a significant positive correlation between Topo IIα expression and Her-2/neu-overexpression in this series. Approximately 50% of the Her-2/neu overexpressing tumours in our series contained high levels of Topo IIα, whereas in the remaining 50%, Topo IIα was low (Table 1). This finding is congruent with the results of a study by Järvinen et al (2000). In 97 breast cancers, these authors found no Topo IIα gene copy aberrations when the Her-2/neu gene status was normal, yet when Her-2/neu was amplified, TOP2α was coamplifed in 44% and deleted in 42% of the cases. Such data may explain the contradictory results of different reports analysing breast cancer susceptibility to Topo IIα inhibitors in relation to the Her-2/neu status (Muss et al, 1994; Paik et al, 1998; Vincent-Salomon et al, 2000). In the same study, Järvinen et al (2000) observed an increased sensitivity to doxorubicin in a breast cancer cell line (UACC-812) with TOP2α gene amplification and consequent protein overexpression and conversely, a decreased sensitivity to doxorubicin in a breast cancer cell line with TOP2α deletion and concomitantly reduced protein content (MDA-361). In a previous study on 863 primary operable invasive ductal carcinomas with long-term follow-up (Rudolph et al, 1999a), of which only a minor proportion had been treated with Topo IIα inhibitors, high Topo IIα expression consequently emerged as an important independent predictor of adverse outcome next to nodal metastasis and before other classical or immunohistochemical factors, including Ki-67. However, no such correlation between Topo IIα and survival was found in the present series. This could be attributable to a balanced effect of the Topo IIα expression portending increased chemosensitivity to Topo IIα inhibitors on the one hand and a higher malignant potential on the other (Kellner et al, 2002).

Primary chemotherapy in breast cancer enhances the rate of breast conservation and enables the in vivo assessment of tumour sensitivity to different chemotherapeutic drugs. Clinical decisions concerning the choice of cytotoxic therapy regimens might be improved according to the initial tumour response (Wolff and Davidson, 2000). With this in view, it would be advantageous to be able to reliably estimate tumour pathologic and biologic features beforehand in order to tailor the composition of chemotherapeutic protocols. Mindful of the fact that anthracyclines target Topo IIα, we have shown that tumours expressing high levels of Topo IIα responded better to an anthracycline-based cytotoxic protocol than tumours with low Topo IIα expression. Therefore, we conclude that Topo IIα assessment may be of value for the clinical management of breast cancer in the setting of primary chemotherapy, provided that our results are corroborated by further studies.