Topoisomerase IIα Expression as an Independent Prognostic Factor in Hodgkin's Lymphoma

Topoisomerases are enzymes that relax DNA supercoiling arising in a variety of nuclear processes. The mechanism of action of topoisomerases I and II involves the formation of transient single or double-stranded DNA breaks respectively. Topoisomerase II is found in two isoforms, namely topoisomerase IIα (topoIIα) and topoisomerase IIβ. TopoIIα is a target for several cytotoxic agents used in the treatment of hematologic malignancies, such as doxorubicin, epirubicin, mitoxantrone, etoposide, and teniposide (1–3). These agents appear to stabilize the DNA-topoIIα complex and inhibit DNA relegation, resulting in an accumulation of lethal double-strand DNA breaks (4). Experimental studies have shown that cells expressing high levels of topoIIα are drug sensitive, while those with low levels are drug resistant (5), making the enzyme a potentially useful predictor of tumour sensitivity to chemotherapy.

Interestingly, topoIIα is mainly expressed in proliferating cells (2), being identical to the proliferation-associated antigen KiS1 (6), and can be used as a proliferation marker in normal and neoplastic cells (7). TopoIIα expression demonstrates a positive correlation with the widely used proliferative marker Ki67 in a variety of normal and neoplastic tissues (8–11).

The dual role of topoIIα makes it an interesting subject of investigation in order to explore the association of its expression with the outcome of patients with neoplastic disorders. Thus it is not yet clear whether high levels of topoIIα are beneficial, increasing tumor sensitivity to topoIIα inhibitors, or they are associated with worse prognosis reflecting increased proliferative activity. Clinical results in the literature are so far contradictory. While in some studies high levels of topoIIα have been associated with chemosensitivity, especially in anthracycline treated breast cancer (12–14), in others topoIIα expression was associated with inferior outcome (15–19). In various subtypes of non-Hodgkin's lymphomas, high topoIIα/Ki67 ratio and high topoIIα expression independently predicted for inferior outcome (20–22).

In a recent report on 42 cases of Hodgkin's lymphoma (HL), low levels of topoIIα expression were associated with shorter survival (23), but this was not confirmed by a preliminary study of 57 patients from our group (24). Both these studies were small and used different patient selection criteria as well as different cuttofs for topoIIα dichotomization. In view of these conflicting data, we extended our observations on topoIIα expression in a much larger series of 238 patients with HL, who had been homogenously treated with topoIIα inhibitors, including chemotherapy based on ABVD or equivalent regimens with or without radiotherapy. TopoIIα expression was correlated with clinicopathological parameters, the expression of Ki67 and patient outcome.

Patients and staging. We analyzed 238 patients with HL, who were diagnosed and treated at the Haematology Section, First Department of Internal Medicine, “Laikon” General Hospital, National and Kapodistrian University of Athens. All patients were older than 14 years, were HIV-negative, had pretreatment archival lymph node tissue material available for topoisomerase IIα immunostaining and had received treatment with anthracycline-based chemotherapy with or without radiotherapy. Approval was obtained by the appropriate Institutional Review Board. All histologic material was reviewed and classified according to the recent WHO classification (25). The baseline patient characteristics, median follow-up, and failure free survival rates were comparable with those of patients who had also received anthracycline-based chemotherapy with or without radiotherapy during the same period, but had not tissue material available for topoisomerase IIα immunostaining (all P-values >0.05; data not shown).

All patients were clinically staged according to the Ann-Arbor system (26), using standard staging procedures. Clinical stages IA and IIÁ were considered early, while clinical stages IB, IIB, III and IV were considered advanced. Anemia was defined as the presence of hemoglobin levels <13g/dl for males and <11.5g/dl for females (27). Serum albumin and severe lymphocytopenia were analyzed at the International Prognostic Score (IPS) cut-offs of <4g/dl and <0.6 × 10⁹/l or <8% respectively (28). The number of involved anatomic sites was determined as previously described (29).

Treatment strategies. Treatment strategies for early Ann-Arbor stage (AAS IA, IIA) and advanced stage (IB, IIB, III, IV) patients have been described previously (30). Early stage patients were scheduled for combined modality therapy including low-dose involved field radiotherapy. Advanced stage patients received chemotherapy. Radiotherapy was administered to 68% of patients with AAS IB, IIB and IIIA. In contrast only 26% of patients with AAS IIIB and IV received radiotherapy. ABVD or EBVD was administered to 82% of the patients, alternating MOPP/ABVD or MOPP/EBVD to 12% and MOPP/ABV or MOPP/EBV hybrid to 6%. All these regimens are currently considered equivalent (31–35).

Immunohistochemical staining. Immunohisthochemical staining was performed as previously described (36). The monoclonal antibodies Ki-S1 (DAKO, Denmark; dilution 1:50), which recognises the topoIIα isophorm and MIB1 (YLEM; dilution 1:100) for Ki67 were used. Omission of the primary antibody was used as a negative control in all cases, while a reactive lymph node was used as a positive control.

Nuclei from at least 100 neoplastic Hodgkin Reed-Sternberg (HRS) cells were to be evaluated in each case. This was possible in approximately 90% of the cases, while in the remaining cases we evaluated as many HRS cells as possible (usually >80). The labelling index (LI) was calculated as the percentage of positive cells. All nuclear staining, whether weak or strong, was counted as positive. The evaluation of topoIIα expression was blinded, since it was performed without knowledge of patients' clinical data and outcome. Evaluation of MIB1 (Ki67) expression was also performed without knowledge of the topoIIα status and patients' clinical data and outcome.

The intraobserver and interobserver variability in the evaluation of topoIIα expression was assessed in a subgroup of 28 patients by IADA and PK.

Statistical analysis. The mean percentages of topoIIα expressing HRS cells among various subgroups of patients were compared by the Student's t-test or the one-way analysis of variance (one-way ANOVA). The intraobserver and interobserver variability of topoIIα expression was evaluated by the Pearson's correlation coefficient. The correlation between topoIIα and Ki67 expression was evaluated by Spearman's rho coefficient, given that the distribution of Ki67 deviated from normality. Failure-free survival (FFS) was defined as the time interval between treatment initiation and treatment failure or last follow-up. Failure was defined as inability to achieve complete or partial remission (CR, PR) during initial therapy, requiring switch to another chemotherapy regimen, death during initial therapy, or progression after an initial CR or PR. Overall survival (OS) was defined as the time interval between treatment initiation and death of any cause or last follow-up. Survival curves were plotted by the method of Kaplan-Meier. The identification of prognostic factors in univariate analysis was based on the log-rank test. Multivariate analysis was performed using Cox's proportional hazards model. A backward stepwise selection procedure, with entry and removal criteria of P = 0.05 and P = 0.10, respectively, was used. In order to avoid the use of arbitrary cutoffs, we performed survival analysis according to the quartiles of topoIIα expression. High topoIIα expression was defined as percentages of topoIIα-positive HRS cells falling into the upper quartile (designated as Q4) vs. those falling within the three lower ones (designated as Q1-3). TopoIIα expression was also evaluated as a continuous covariate (percentage of topoIIα-positive HRS cells in each case) in multivariate analysis. Furthermore, topoIIα expression was evaluated according to the cutoff obtained by a ROC curve-based approach. Finally, topoIIα expression was tested against the IPS value in multivariate analysis (28); all individual IPS variables were excluded from this analysis.

Patients' clinical and laboratory characteristics are summarized in Table 1. They were compatible with other reported unselected series of patients with non-pediatric HL. The median age of the patients was 30 years (15-82) and 117 (49%) were males. Histologically, there were 232 cases of classical HL (97%) and 6 cases of nodular lymphocyte predominance (NLP) HL (3%). In detail, 174 patients were classified as nodular sclerosis (NS) subtype (73%), 47 cases as mixed cellularity (MC) (20%), 8 as lymphocyte rich (3%), 1 as lymphocyte depletion (<1%), 1 as interfollicular (<1%) and 1 as classical HL, unclassified (<1%). The median follow-up of patients, who were alive at the time of the analysis, was 111 months (23-222).

Immunohistochemical expression of topoIIα was seen in all cases and was mainly present in the neoplastic HRS cells and variants. A number of reactive small lymphocytes and larger cells within residual germinal centers were also positive. The staining observed was diffuse nuclear (Fig. 1A), with an occasional punctate pattern. The percentage of positive HRS cells ranged from 5% to 98%, following an approximately normal distribution with a mean ± SD of 63 ± 19% and a median value of 64% (interquartile range, 51-78%). Both intraobserver and interobserver variability were very low with Pearson's correlation coefficients of 0.943 (P < 0.001) and 0.950 (P < 0.001) respectively.

No statistically significant relationship was found between topoIIα expression and demographic, clinical, and conventional laboratory parameters including IPS, with the exception of gender, where males had marginally higher expression than women (P = 0.08; Table 1). Although the 6 patients with NLP had somewhat higher mean percentage of topoIIα expression than patients with classical HL (75 ± 17% vs. 63 ± 19%), the difference was not statistically significant (P = 0.16). TopoIIα expression did not also differ according to the subtypes of classical HL (Table 1).

Ki67 expression was evaluated in 211/238 patients (89%). The staining was diffuse nuclear and the mean ± SD percentage of positive HRS cells was 73 ± 19% (range 8%-99%). The median Ki67 expression was 76% (interquartile range, 61-88%). A rather loose positive correlation was found between TopoIIα and Ki67 expression (Spearman's rho 0.255, P < 0.001).

Remission rates. Remission rates did not differ according to topoIIα expression: Primary refractory disease was observed in 4.5% vs. 3.5% of patients with topoIIα expression within Q1-3 and Q4 respectively (P = 0.73). Furthermore, early failure, defined as progression or relapse within the 1st year from diagnosis, was observed in 7.8% vs. 5.1% of patients respectively (P = 0.48).

Failure-free survival in all patients. The results of univariate analysis of FFS are summarized in Table 2. Most of the established conventional prognostic factors proved to be statistically significant. Except of quartile analysis, a ROC curve-based approach was used to dichotomize topoIIα values. According to the latter a cutoff of 74.8% was suggested, which was rounded to 75%.

The 10-year FFS was 82 ± 3% vs 68 ± 7% for patients with topoIIα expression within Q1-3 and Q4, respectively (P = 0.02) (Fig. 1B and Table 2). Using 75% as ROC-based cutoff for topoIIα expression, the discrimination was even better with 10-year FFS being 84 ± 3% vs 67 ± 6% (P = 0.002). In this analysis 72 patients (30%) fell into the adverse group.

Failure-free survival in stage and IPS subgroups. Among 127 patients with early stage HL (IA/IIA), topoIIα expression did not show a statistically significant association with FFS (10-year rates 89 ± 3% vs 83 ± 7%, P = 0.36 for patients with topoIIα expression within Q1-3 and Q4 respectively; Fig. 1C). In contrast, topoIIα expression was significantly associated with FFS among the 111 patients with advanced disease (10-year rates 74 ± 5% vs 52 ± 11%, P = 0.03 for patients with expression within Q1-3 and Q4; Fig. 1D). When patients were stratified according to the IPS, topoIIα expression had again a non-significant effect on FFS in patients with IPS <3 [10-year rates 82 ± 4% vs 76 ± 8%, P = 0.64 for patients with expression within Q1-3 and Q4 respectively], and a highly significant effect on FFS of patients with IPS ≥3 [corresponding 10-year rates 74 ± 8% vs 30 ± 14%, P = 0.0008], so that patients with IPS ≥3 and high topoIIα expression had a particularly poor outcome.

Failure-free survival in treatment modality subgroups. High topoIIα expression was highly predictive of adverse outcome in patients treated with chemotherapy only; 10-year FFS rates were 71 ± 7% vs. 29 ± 15% in patients falling within Q1-3 vs. Q4 respectively (P = 0.003). Among 69 patients who received chemotherapy alone, 56 (81%) had advanced stage disease. On the contrary, the effect of high topoIIα expression in patients treated with combined modality was much smaller: 10-year FFS rates were 88 ± 3% vs. 82 ± 6% in patients falling within Q1-3 vs. Q4 respectively (P = 0.25). Among 169 patients treated with combined modality, 114 (67%) had early stage disease.

Overall survival. During follow-up 33 deaths were recorded: 20 (61%) due to HL-related and 13 (39%) due to unrelated causes.

The difference in overall survival was borderline: Patients with topoIIα expression within Q1-3 had a 88 ± 3% 10-year overall survival rate vs. 81 ± 5% for those with expression within Q4 (P = 0.06). At the 75% ROC-based cutoff the overall survival difference was significant (10-year rates 90 ± 3% vs. 79 ± 5%, P = 0.02).Patients with topoIIα expression within Q1-3 had a 94 ± 2% 10-year cause-specific survival rate vs 83 ± 5% for those with topoIIα expression within Q4 (P = 0.005). The discriminative ability at the 75% cutoff was even better (10-year rates 96 ± 2% vs. 80 ± 5%, P = 0.0003).

TopoIIα expression was evaluated together with covariates, which were significant in univariate analysis and had <10% missing values, i.e. stage IV, B-symptoms, and number of involved anatomic sites. The results of multivariate FFS analysis are summarized in Table 3. High topoIIα expression (within Q4) had independent prognostic significance for FFS, along with B-symptoms and, marginally, the number of involved anatomic sites. When covariates with >10% missing values were included, topoIIα expression within Q4, stage IV, and reduced serum albumin levels remained independent prognostic factors for FFS.

When the percentage of topoIIα-positive HRS cells was evaluated as a continuous covariate, its independent prognostic significance for FFS persisted: Increasing topoIIα expression (P = 0.02; relative risk 1.020 per percentage unit; 95% CI 1.004-1.036), B-symptoms and the number of involved anatomic sites were again selected in the multivariate model.

High topoIIα expression (within Q4) remained an independent prognostic factor for FFS after adjustment for the IPS (P = 0.02; Table 3). The same was true when topoIIα expression was evaluated as a continuous covariate (P = 0.03; relative risk 1.019 per percentage unit; 95% CI 1.002-1.036). When B-symptoms and the number of involved anatomic sites were evaluated along with topoIIα and IPS, the significance of topoIIα persisted, while that of IPS and B-symptoms were borderline.

When topoIIα expression was analyzed at the ROC-based cutoff of 75%, its significance was strengthened in all multivariate models.

In multivariate analysis of overall survival topoIIα had a non-significant effect after adjustment for age and the number of involved sites (P = 0.13).

In the population of 211 patients with available data on Ki67 expression (with 43 failure events), high topoIIα expression (within Q4) was again associated with inferior FFS (P = 0.03). Irrespectively of the cutoff value used, Ki67 expression was not associated with FFS. In contrast, the ratio topoIIα/Ki67 was of prognostic significance: The 10-year FFS was 82 ± 3% for the 147 (70%) patients with ratios <1 vs. 72 ± 6% for the 64 (30%) patients with ratios ≥1 (P = 0.03). In multivariate analysis, a topoIIα/Ki67 ratio ≥1 was independently associated with FFS (P = 0.04) along with B-symptoms (P = 0.03) and, marginally, stage IV (P = 0.06). However, neither Ki67 nor the ratio topoIIα/Ki67 added independent prognostic information in multivariate models already including the percentage of topoIIα expression.

The expression of topoIIα in human neoplasia has been the subject of many recent reports. In this study all cases of HL expressed topoIIα with a median number of positive HRS cells of 64%. These results are in accordance with those previously described by us in a different patient series (37). Brown et al., in series of 49 cases of HL, found similar percentages of topoIIα-positive neoplastic cells in cases of classical HL (58% in NS and 68% in MC), while the percentage of topoIIα-positive neoplastic cells in 15 cases of LPHL was 84% (38). This result was obtained from the evaluation of CD20/topoIIα double positive cells. Although the mean percentage of topoIIα-positive neoplastic cells in our NLP HL patients was numerically higher than the one observed in classical HL, the difference did not reach statistical significance. However the number of cases of NLP HL was low to permit a reliable statistical analysis. In contrast to the high topoIIα expression observed in the present, as well as other series (37, 38), other investigators have reported lower median percentages of positive HRS cells in the order of 30% (23, 39). These differences may be related to the use of different monoclonal antibodies for topoIIα detection, as it has been demonstrated that KiS1 is an antibody with high binding affinity to topoIIα (40).

We demonstrated here that high topoIIα expression had independent adverse prognostic significance in a series of 238 patients with HL, who had been treated with ABVD or equivalent regimens with or without RT in a single center. This treatment approach invariably included the administration of doxorubicine or epirubicine, which exert their antineoplastic action through topoIIα inhibition. Our results contradict those reported by Provencio et al. (23), who had suggested a favorable prognostic impact of higher topoIIα expression in a series of 42 patients with advanced HL. However, a more recent report by the Spanish group demonstrated that high topoIIα expression (within the corresponding Q4) was associated with inferior cause-specific survival in a series of 235 patients, eventhough the difference did not reach statistical significance (41). The absolute difference in long-term cause-specific survival between patients with topoIIα expression within Q1-3 and Q4 in the latter study was approximately 8-10% (approximation obtained from survival curves), a figure consistent with our data. Our results are also in agreement with the adverse significance of topoIIα expression in other neoplasms (15, 16, 20–22, 42–44), including non-Hodgkin's lymphomas (20–22), even after chemotherapy with topoIIα inhibitors (18–20).

A simplified explanation of our findings could be that topoIIα exerted its effect on the prognosis of HL merely through its role as a proliferation marker, which is further supported by the positive correlation that we found both between topoIIα and MIB1 (Ki-67) expression. The mean percentage of Ki67-positive cells was higher than that of TopoIIα positive cells, a finding described both in our previous study (37) and in non-Hodgkin's lymphomas (20, 45). However topoIIα might provide a better estimate of the number of cycling cells than Ki67, because it is selectively present during the later phases of the cell cycle (40, 46, 47). Thus the data presented here are compatible with a role of topoIIα as a proliferation marker.

In contrast to topoIIα, Ki-67 expression was of no prognostic significance in our series. Although Ki67 expression was predictive of the outcome in the initial study of the Spanish group (48), this was not confirmed in a subsequent larger series evaluated by tissue arrays (49). Additionally, in accordance with a previously report of our group on non-Hodgkin's lymphomas (20), a ratio of topoIIα/Ki67 ≥1 was an independent predictor of inferior FFS in this series as well, but did not add prognostic information independent of topoIIα expression per se.

Except of a borderline association with male gender, topoIIα expression was neither correlated with any of the examined conventional demographic, clinical, histologic, and laboratory parameters nor with the value of IPS. Thus topoIIα does not appear to be a surrogate marker of either tumor burden or the biological aggressiveness of the disease and may actually represent a ‘‘primary’' prognostic variable. This is in contrast with most conventional prognostic factors, which present extensive interrelationships (28), and resembles to what has been observed with a small number of recently described biological markers, such as bcl-2 (34) and activated caspase-3 expression (50).

Given the lack of correlation with other prognostic factors and its prognostic significance in univariate analysis of FFS, topoIIα emerged as an independent prognostic marker in multivariate analysis. Using the upper quartile as a definition for high topoIIα expression we avoided '‘data fitting'', which could arise by the introduction of arbitrary or ‘‘best’' cutoffs. The validity of our findings is strengthened by the fact that topoIIα was not only predictive of FFS at the cut-off of the upper quartile, but also carried independent prognostic significance when evaluated as a continuous covariate. A ROC-curve defined cutoff (75%) led to the demonstration of an even stronger prognostic impact of high topoIIα expression.

Despite its clear independent prognostic significance, we further tested whether topoIIα added to the prediction achieved by the IPS, which is the most popular prognostic index for advanced HL (28), while very similar systems may also work in localized disease (51–53). Thus in another multivariate approach including both IPS and topoIIα, the latter proved again to be an independent prognostic factor for FFS.

The adverse prognostic effect of high topoIIα expression was mainly restricted to advanced stage patients, while its effect in early stages was not statistically significant. Furthermore, high topoIIα expression was predictive of adverse outcome in patients treated with chemotherapy only, a finding suggesting that high topoIIα expression did not confer sensitivity to anthracycline-based chemotherapy. However, since 81% of patients treated with chemotherapy alone had advanced stage disease, it is not clear that the highly significant difference was related to the applied treatment or to the inclusion of advanced stage patients. On the contrary, the effect of high topoIIα expression in patients treated with combined modality was much smaller, probably due to the fact that most patients treated with combined modality had early stage disease, resulting to a low number of failure events, which reduced the power of the analysis to demonstrate any prognostic effect of topoIIα.

TopoIIα is necessary as a substrate for anthracyclines and other drugs, in order to exert their antineoplastic activity. A simplified approach could suggest that if more topoIIα is expressed in the HRS cells, the target for anthracyclines would be more abundant and the antitumor activity would be greater. This view has been supported by recent results, especially in breast cancer, but in others tumors as well (12–14, 23). However, as more HRS cells express topoIIα, the amount of anthracyclines delivered by standard chemotherapy regimens may not suffice to substantially inhibit topoIIα activity and lead the neoplastic cells to apoptosis. This hypothesis offers an additional explanation of our findings. Furthermore, if this is indeed the major mechanism responsible for the higher failure rates found in patients with higher topoIIα expression, then it might have therapeutic implications: it would be logical to hypothesize that patients with high levels of topoIIα might benefit from dose intensification of topoIIα inhibitors or from the addition of a second topoIIα inhibitor in the initial chemotherapy regimen. This is the case with BEACOPP-escalated, the most effective but also toxic regimen available for advanced HL (54), which is now also being tested in patients with early stages and unfavorable prognostic profile (55). BEACOPP-escalated maintains the dose intensity of doxorubicine in comparison with ABVD, but—in addition—includes 3 days of etoposide, another topoIIα inhibitor, at high dose of 200 mg/m² daily. It would be very interesting to see whether the prognostic significance of topoIIα expression will be overcome by the administration of BEACOPP-escalated, a situation that would strengthen our hypothesis.

In conclusion, we demonstrated the adverse prognostic significance of topoIIα expression in patients with HL treated with ABVD or equivalent regimens with or without radiotherapy. Apart from circulating factors including cytokines (27, 56–58), molecules involved in apoptosis (34, 49, 56) and polymorphisms implicated in drug metabolism (59), topoIIα expression appears to have also a place in the armamentarium of biological prognostic factors in HL due to its implication in cell proliferation and drug sensitivity mechanisms. Following appropriate validation, this molecule may offer a biological rational for treatment intensification with topoIIα inhibitors, generating an hypothesis that could be tested in future trials.