The Prognostic Significance of DNA Topoisomerase II-alpha (Ki-S1), p21/Cip-1, and p27/Kip-1 Protein Immunoexpression in Oligodendrogliomas: An Analysis of 91 Cases

Oligodendrogliomas constitute approximately 2% to 5% of all primary brain tumors and 4% to 15% of gliomas.1,2 Despite numerous previous studies, these neoplasms continue to generate considerable controversy in the identification of prognostic factors, including single histopathologic patterns and grade of tumor malignancy. The prognostic significance of various pathologic parameters has been intensively examined, but numerous studies have yielded conflicting results.3–16 

Numerous studies concerning the prognostic value of immunoreactivity of various tumor-associated antigens for oligodendrogliomas have been published recently.7,12–14,17–32 The tumor growth fraction, measured by Ki-67/MIB-1, seems to be the most significant factor. However, the cutoff points for MIB-1 counts are broadly scattered, and universally applicable labeling index (LI) thresholds for oligodendroglioma prognosis are not yet a reality.

It has recently been established that regulation of proliferative activity and cell cycle progression depends on the sequential activation of a set of cyclin-dependent kinases (CDKs).33 Inactivation of CDKs is accomplished by various CDK inhibitors, and members of the Cip/Kip family play an important role among them.33,34 

The p21/WAF/Cip protein (p21) binds directly to CDK1, CDK2, and CDK4, and may act as a universal inhibitor of the CDKs' enzyme activity. The expression of p21 is known to be controlled by p53. The inhibitory function of p21 is regulated stoichiometrically. When the ratio of p21 to CDK complex is more than 1, p21 inhibits kinase activity. If the ratio is less than 1, p21 serves as an assembling factor of CDK complex. The prognostic significance of p21 alteration for various human cancers is controversial.33 

The p27/Kip-1 protein (p27) inhibits cyclin complexes with CDKs, controlling restriction point and preventing cell cycle progression from G1 to S. It binds such complexes more efficiently than CDKs alone, as does p21. Increasing p27 expression shows an inverse correlation with cell proliferation and a close relation with a favorable prognosis in various human tumors.33,34 Although mutations of p27/Kip-1 are rare events, it is presumed that during malignant progression, tumor cells down-regulate the level of p27. Proteosome degradation is responsible for p27 protein reduction or absence.35 

To our knowledge, investigations into the prognostic role of CDK inhibitors in oligodendrogliomas are limited to a few reports.18,27 Therefore, in the present study we immunohistochemically examined biopsy samples of 91 oligodendrogliomas treated with surgery and radiotherapy to evaluate a possible association between patient survival and expression of p21 and p27 proteins, as well as DNA topoisomerase II-alpha (Ki-S1), a marker of cell proliferation.36 

Ninety-one adult patients with newly diagnosed “pure” cerebral oligodendroglial tumors, consecutively treated at the Neurosurgical Burdenko Institute (Moscow, Russia) between January 1, 1990, and January 1, 1998, were studied retrospectively. Eighteen of these patients were younger than 30 years, 53 were between 30 and 50 years of age, and 20 were older than 50 years; mean age was 41 years. The study group included 43 men and 48 women.

According to preoperative computed tomographic and/or magnetic resonance imaging findings, tumor location was frontal in 42 cases and nonfrontal in the remaining 49 cases. Patients had undergone gross total (58 cases) and subtotal (33 cases) tumor resection, as confirmed by postoperative contrast computed tomographic scans or magnetic resonance imaging. All patients had received postoperative external beam irradiation (55–61 Gy). Additionally, 47 patients had received adjuvant polychemotherapy with procarbazine, lomustine, and vincristine.

Follow-up data were received at least 24 months after surgery, ending January 1, 2000. Progression-free survival and overall survival were estimated separately. Data regarding death, relapse, or last contact were evaluated at the end of the study.

Biopsy samples were immediately fixed in cold 10% buffered formaldehyde solution and embedded in paraffin (Histowax, Leica, Nussloch, Germany). In all cases, the pathologic diagnoses of “ordinary oligodendroglioma” (grade II) and “anaplastic oligodendroglioma” (grade III) were made according to the criteria presented in the World Health Organization (WHO) histologic classification of tumors of the nervous system.2 Accordingly, mixed gliomas (oligoastrocytomas and anaplastic oligoastrocytomas) were not included in the present study.

Immunohistochemical investigation was performed in serial 5-μm sections mounted on poly-l-lysine coated slides. Monoclonal antibodies against the following antigens were used: (1) DNA topoisomerase II-alpha (clone Ki-S1; 1:70; Dakopatts A/S, Glostrup, Denmark; catalog No. M7186; positive control, human tonsil and lymph nodes), (2) p21/Cip-1 (clone SX118; 1:20; Dakopatts; catalog No. M7202; positive control, various human tissues), and (3) p27/Kip-1 (clone SX53G8; 1:50; Dakopatts; catalog No. M7203; positive control, various human tissues).

After microwave antigen retrieval, the sections were incubated with the antibodies overnight at 4°C. Immunostain visualization was achieved with the standard streptavidin-biotin peroxidase technique (Dako LSAB kit, Dako Corporation, Carpinteria, Calif, catalog No. K0675). The slides were stained with 3,3′-diaminobenzidine, counterstained with hematoxylin, and mounted. For the negative control procedure, primary antibodies were changed for commercially produced mouse negative control reagents.

The immunoexpression for each antigen was evaluated using a blind method; that is, the observers had no information about the specimens examined. Histologically normal adjacent brain tissue was excluded from analysis.

For counting of immunostained nuclei, a computerized color image analyzer, Quantimet color 500 (Leica, Cambridge, United Kingdom), was used. The method has been described in detail elsewhere.37 Briefly, an image of the tumor specimens was captured on the computer screen and acquired and processed by calibrated color detection to identify the immunostained nuclei in an RGB (red-green-blue) regimen. Intensive cytoplasmic staining of mitotically dividing cells was considered to be the internal positive control for Ki-S1 immunoreactivity, whereas nuclear staining of normal oligodendrocytes in adjacent nerve tissue was used as a positive internal control for p27 expression.

Undetectable nuclei were rejected by the interactive binary edition, and immunostained nuclei were automatically counted. The next step in image preparation included color detection and counting of hematoxylin-stained nuclei. At least 20 fields of vision with ×400 magnification were sequentially acquired and examined in this manner.

The LIs for all of the antibodies examined were determined in randomly chosen areas and calculated as a percentage of the positively stained nuclei over the total number of tumor cell nuclei counted.

The χ² test was performed to determine whether the relationships were statistically significant. Nonparametric Spearman rank correlation coefficients were used to assess the degree of linear association between pairs of variables. Survival analyses from the data of operation were estimated with the Kaplan-Meier method. For numerically continuous variables, the cutoff point that best subdivided patients into distinct survival groups was determined according to Segal. The comparisons among various patient subgroups were performed by the log-rank test. Multivariate analysis for survival was performed using the Cox proportional hazard models. In addition, the classification and regression tree models for censored data (CART) as modified by LeBlanc and Crowley were used. Probability (P) values less than .05 were considered significant. A significant correlation between 2 parameters was taken at the 95% confidence interval.

By light microscopy, all 91 tumors examined represented the histopathologic appearance typical for pure oligodendrogliomas, which has been described in detail elsewhere.1,2 Forty-six tumors were defined as low-grade (WHO grade II) tumors, and the remaining 45 tumors were classified as high-grade or anaplastic (WHO grade III). The latter were recognized by an increasing cell density, brisk mitotic activity, and an obligatory presence of endothelial proliferation (multiplication of the lining endothelial cells in tumor vessels) with various degrees of extension. Twenty-seven (60%) of 45 high-grade tumors contained necrotic foci.

All oligodendrogliomas exhibited nuclear Ki-S1 accumulation (Figure 1, A). There were no Ki-S1–stained nuclei in adjacent nontumoral brain tissue. The mean Ki-S1 LI was significantly prominent for high-grade tumors (Table 1).

Nuclear expression of p21 protein (Figure 1, B) was discovered in 26 tumors (29%); p21 expression in adjacent brain tissue was not observed. The number of p21-positive cases was found to be relatively equal for low- and high-grade oligodendrogliomas: 14 (30%) versus 12 (27%), respectively. The mean p21 LI was not significantly different for low- and high-grade tumors (Table 1).

All tumors exhibited nuclear accumulation of p27/Kip-1 (Figure 1, C). p27-Positive oligodendrocytes were also found within nerve tissue beyond the tumor margin. The mean p27 LI was significantly higher for low-grade oligodendrogliomas (Table 1).

Besides Ki-S1 immunostained mitotic cells, cytoplasmic reactivity for the all of the antibodies tested was not observed.

Correlation analysis revealed a statistically significant covariation between increasing Ki-S1 and decreasing p27 counts (Spearman test, r = 0.67; P = .008).

The recurrence rate for the low-grade oligodendrogliomas was 41% (19 of 46 cases recurred) and 8 patients died (17%). The median progression-free and overall survival times for these tumors were 44 and 57 months, respectively. All 19 relapsed low-grade tumors underwent additional surgery. In 12 reoperated tumors, pathologic examination revealed progression toward an anaplastic oligodendroglioma, whereas the remaining 7 exhibited the same histology found in the primary tumor.

Thirty-one (69%) of 45 high-grade tumors recurred, and 25 patients (56%) with high-grade tumors died. The median progression-free and overall survival times for these tumors were 21 and 27 months, respectively.

The 2-year recurrence rate was 9% for low-grade and 62% for high-grade oligodendrogliomas (χ² test, P < .001).

The results of univariate survival analyses revealed that both the progression-free (Table 2) and overall (Table 3) survival times were significantly reduced for subtotal removal, for postoperative Karnofsky performance score less than 70, for high-grade tumors, for Ki-S1 LIs greater than 10%, and for p27 LIs less than 20%.

Low-grade oligodendrogliomas with p27 LIs less than 20% and greater than 20% revealed significant differences in the rate of recurrence and death, in the number of tumors with a trend toward anaplastic progression, and in survival times (Table 4).

High-grade oligodendrogliomas with Ki-S1 LIs less than 10% and greater than 10% revealed significant differences in the rate of recurrence and death and in survival times (Table 4).

We found no difference in survival times in patients with or without p21 immunoreactivity (log rank, P = .2). Additionally, we found no differences when tumors with p21 LIs greater than 5% were compared with tumors with less than 5% or absence of p21 expression (log rank, P = .4).

Multivariate analysis using the Cox hazard model for entire cohort of patients revealed that risk of tumor progression was independently associated with high-grade tumors (hazard ratio = 2.29; P = .02), with Ki-S1 LIs greater than 10% (hazard ratio = 3.72; P = .0002), and with p27 LIs less than 20% (hazard ratio = 3.16; P = .001), whereas risk of death was independently associated with high-grade tumors (hazard ratio = 2.89; P = .003) and with Ki-S1 LIs greater than 10% (hazard ratio = 3.72; P = .0002).

Figure 2 displays the Kaplan-Meier curves generated by the CART modeling process. We identified 4 final groups of patients with distinctly different progression-free and overall survival times (all differences are significant; Table 4): (1) 32 patients with low-grade tumors and p27 LIs greater than 20%, (2) 14 patients with low-grade tumors and p27 LIs less than 20%, (3) 25 patients with high-grade tumors and Ki-S1 LIs less than 10%, and (4) 20 patients with high-grade tumors and Ki-S1 LIs greater than 10%.

Oligodendrogliomas are rare tumors with relatively long survival times, and the creation of a universally accepted grading system for these neoplasms has been difficult. As a result, numerous 2-, 3-, and 4-tiered grading systems exist.3–16 The present study revealed a strong association between oligodendroglioma clinical outcome and tumor grade defined according to the approach recommended by the WHO experts.2 In addition, the 2-year recurrence rate for low-grade neoplasms was found to be significantly lower than for high-grade tumors. Therefore, the WHO approach for identification of oligodendroglioma malignancy is likely to be practically useful and applicable.

Numerous studies have revealed significant correlation between Ki-67/MIB-1 LI and oligodendroglioma prognosis.^* However, cutoff points were found to be scattered from 2% to 20% in various series, and studies are in progress to standardize the counting methods and experimental design. In most of the reports, the LI for MIB-1 was found to be a prognostic factor for the whole oligodendroglioma cohort. Recently, Dehghani et al7 and Reis-Filho et al30 reported that an MIB-1 LI greater than 5% is closely related to an unfavorable prognosis for WHO low-grade oligodendrogliomas. Nevertheless, Schiffer et al13 pointed out that the individual prognostic value of MIB-1 immunoreactivity for low-grade oligodendrogliomas “can be formulated only when the LI of the case is higher than the maximum value found in the classic variant.”

In the present study, we used antibodies to DNA topoisomerase II-alpha (clone Ki-S1) for identification of the oligodendroglioma growth fraction. Ki-S1 is an essential cellular enzyme that functions in the segregation of chromosome pairs and in chromosome condensation. Anti–Ki-S1 antibodies label cells in the S, G2, and M phases of the cell cycle. It has been found that the antigen labeled by Ki-S1 is extremely robust, resisting degradation by fixation and enzymatic tissue desegregation.36 Some investigators have studied DNA topoisomerase expression in astrocytomas,38,39 and they have discovered a close correlation with tumor grade, MIB-1 LI, and clinical outcome. Recently, we established a strong prognostic significance of Ki-S1 LI for ependymomas.40 

The present study revealed that the Ki-S1 LI is significantly related to oligodendroglioma grade and outcome for high-grade tumors with a cutoff value of 10%. However, the Ki-S1 count for WHO low-grade oligodendrogliomas did not reach the 10% cutoff point, and Ki-S1 LI did not correlate with survival differences in these tumors. Therefore, the Ki-S1 LI ranges can be recommended for distinct verification of oligodendroglioma grade and for clinical subdivision of the high-grade tumor category.

Recent studies have shown p21 expression in astrocytic gliomas to have variable labeling ranges and disputable prognostic importance.41–46 The data on p21 immunoreactivity in oligodendrogliomas are limited by the low number of tumors examined.45 

In our study, about 30% of oligodendrogliomas expressed relatively small numbers of p21-positive nuclei that revealed no association with the tumor grade, growth fraction, or the clinical outcome. Therefore, the biological mechanisms of p21 regulation and function in oligodendrogliomas remains to be defined; however, it seems unlikely that p21 immunohistochemistry will be of value in establishing oligodendroglioma prognosis.

In astrocytomas and ependymomas, a correlation of p27 immunoreactivity with tumor grade and patient survival has already been found.40,46–50 Cavalla et al18 studied p27/Kip-1 immunoreactivity in 37 oligodendrogliomas. They revealed that the p27 score showed a tendency to decrease with anaplasia and had prognostic significance with a cutoff point of 25%. No correlation between p27 and MIB-1 LIs was established.

We disclosed a strong association between p27 LI and oligodendroglioma grade, growth fraction, and prognosis for low-grade tumors. Moreover, the data suggest that p27 down-regulation is an important event in malignant transformation of oligodendrogliomas. Theoretically, these facts might be partly explained by the previously published experimental data suggesting that p27 is a crucial component of the mechanisms that allow oligodendrocyte differentiation and prevent cell cycle re-entry.51,52 Therefore, histologically low-grade oligodendrogliomas with decreasing p27 nuclear accumulation might be considered less differentiated biologically.

In contrast, the clinical outcomes of patients with high-grade oligodendrogliomas are not related to the p27 score. The possible explanation for this finding is that malignant oligodendroglioma cells may gain the ability to overcome the p27 inhibitory function and enter into S phase, even with a significant amount of p27 nuclear accumulation. Deregulated overexpression of p27 has been revealed in breast carcinomas in which levels of p27 were significantly associated with high levels of cyclin D1.53 Presumably, the high level of cyclin accumulation in breast carcinoma cells might compensate for the inhibitory effects of p27/Kip-1.

The present study revealed a statistically significant association between the WHO oligodendroglioma grade and patient outcome. However, the grading categories could be further subdivided by CART analysis according to both the Ki-S1 and p27 scores into 4 clinically significant subsets. Moreover, a subset of histologically high-grade tumors with Ki-S1 LIs less than 10% had a more favorable clinical outcome than a subset of low-grade oligodendrogliomas with p27 LIs less than 20%. The biological events underlying these provocative findings remain to be defined. We can only propose that low-grade oligodendrogliomas with p27 protein reduction have a tendency to prompt transformation in their more biologically aggressive high-grade counterparts. On the whole, the data reveal the advantage of immunohistochemistry over routine histology for identification of oligodendroglioma biological behavior and clinical course.

The results of the present study may be useful when deciding on a course of treatment for oligodendroglioma. It is well known that DNA topoisomerase II is the target of a variety of anticancer drugs, such as etoposide, teniposide, and doxorubicin.54,55 It seems to us that application of these anticancer agents may be theoretically beneficial in the treatment of high-grade oligodendrogliomas with elevated Ki-S1 levels. However, this proposal now needs to be validated by prospective studies.

In summary, both p27 and Ki-S1 scores were found to be strong predictors of oligodendroglioma outcome, together with the WHO tumor grade, and they also seem to be useful for assessing individual prognosis in routinely processed specimens. For histologically low-grade tumors, immunohistochemical prognostication should be focused on p27 count, whereas the Ki-S1 LI will be of value in determining survival in WHO high-grade oligodendrogliomas.