Low expression of p27 indicates a poor prognosis in patients with high-grade astrocytomas




Two families of tumor suppressor genes, Cip/Kip (p21, p27, and 57) and INK4 (p15, p16, p18, and p19), regulate cell proliferation and neoplastic transformation. p27 exerts its suppressor effect through cyclin E-dependent kinase (CDK2) by inhibiting the phosphorylation of pRb by CDK2, which, in turn, arrests cells in the G1-phase. p21 has a similar effect in addition to participating in the p53 dependent CDK4-mediated and CDK6-mediated pathway. The authors studied the prognostic significance of p21 and p27 in patients with high-grade astrocytomas who were treated with radiotherapy.


The expression of p27 and p21 was analyzed immunohistochemically in 52 glioblastomas and 25 anaplastic astrocytomas. All patients underwent surgery for the first time and were treated with adjuvant external radiotherapy.


The p27 labeling index (LI) was < 30% in 36% of tumors, 30–50% in 25% of tumors, and > 50% in 39% of tumors. A significant difference in cumulative survival was observed between these groups (P = 0.0072; log-rank test). The p21 LI was < 30% in 48% of tumors, 30–50% in 39% of tumors, and > 50% in 13% of tumors; these groups did not differ significantly in survival. In multivariate Cox analysis, p27 LI was an independent prognostic factor (P = 0.0008). The grade of malignancy and proliferation activity also were independent prognostic factors.


Although p27 and p21 are parallel cell-cycle regulators, only p27 has independent prognostic value in patients with malignant astrocytomas. It appears that decreased levels of p21/p27 are associated with a poor prognosis and short survival. Cancer 2001;97:644–8. © 2003 American Cancer Society.

DOI 10.1002/cncr.11079

Two important families of tumor suppressor genes, Cip/Kip (p21, p27, and p57) and INK4 (p15, p16, p18, and p19),1 regulate cell proliferation and are essential in the prevention or neoplastic transformation. The Cip/Kip family of proteins are designated universal cyclin dependent kinase (CDK) inhibitors, because they interact with various cyclin-CDK complexes.1 The p27 protein, a member of the Cip/Kip family, can bind multiple cyclin-CDK complexes, including D-type cyclins CDK4 and CDK6 and E-type cyclin CDK2; thus, p27 is a major regulator of G1-S transition in the cell cycle.2, 3 This makes p27 a potent tumor suppressor gene. The expression of p27 correlates with cell proliferation activity. p27 protein is active in the nucleus, and loss of nuclear p27 immunoreactivity represents true reduction of the p27 protein.4 Therefore, immunohistochemical analysis of the p27 protein in tumor cell nuclei in formalin fixed, paraffin embedded tissues can be used reliably to evaluate the status of p27 in various human tumors.4

Previously, immunohistochemical studies have shown that reduced expression of the p27 protein is associated with a poor clinical outcome of patients with various malignancies.1 High-grade astrocytomas comprise 40% of adult brain tumors and represent a major clinical challenge, with fatal outcomes in the majority of patients within 3 years of diagnosis. Therefore, the clinicopathologic significance of cell cycle regulators in the pathobiology of these tumors is of interest, and the roles of p21 and p27 need to be elucidated further. The characterization of histopathologic markers with prognostic significance may have implications in the planning of treatment for patients with malignant brain tumors.

The objective of this study was to establish the prognostic significance of p27 and p21 expression in patients with high-grade astrocytomas. For this purpose, expression levels of p27 and p21 were studied immunohistochemically in 77 patients and were correlated with survival. The results suggest that p27 immunohistochemical staining provides prognostic information for patients with high-grade astrocytomas.


A total of 77 patients with malignant astrocytomas were included in this study. All patients underwent surgery at Turku University Hospital during the period 1988–1994. Postoperative radiotherapy with a median total dose of 55 grays (Gy) with 1.8-Gy daily fractions was given to all patients in the study, as described previously in detail.5 The patients were followed up for at least 5 years or until death.

Histopathologic diagnoses were based on stereotactic biopsy in 15 patients (19%), subtotal tumor resection in 56 patients (73%), and total macroscopic tumor resection in 6 patients (8%). Tumor samples were reevaluated by a neuropathologist (H.K.) to confirm the diagnosis and were graded using the World Health Organization criteria:6 Twenty-five tumors were classified as anaplastic astrocytomas (AA) (Grade 3), and 52 tumors were graded as glioblastoma (GB) (Grade IV). Patient characteristics are shown in Table 1.

Table 1. Patient Characteristics
CharacteristicNo. of patients (%)
Anaplastic astrocytomaGlioblastoma
No. of patients2552
Age (yrs)  
 Female 9 (36)24 (46)
 Male16 (64)28 (54)
Type of surgery  
 Biopsy 7 (28) 8 (15)
 Macroscopic tumor resection18 (72)44 (85)


The specimens were fixed routinely in 4% phosphate-buffered formaldehyde and embedded in paraffin. The most representative blocks were selected and sectioned at 5 μm on polylysine-coated slides.

Immunohistochemical detection of the Ki-67 antigen was performed as described previously.7 The MIB-1 mouse monoclonal antibody to Ki-67 was purchased from Immunotech (Marseille, France). For the detection of bound antibodies, the Vectastain ABC peroxidase kit (Vector Laboratories, Burlingame, CA) was used. Diaminobenzidine was used as the chromogen, and counterstaining was done with ethyl green.

For the detection of p21 and p27, the sections were treated for antigen retrieval by microwave processing and were incubated with the primary antibody overnight at 4 °C. Detection of bound antibodies was done as described above but using hematoxylin as the counterstain. Several antibody dilutions were tested to optimize the dilution used. Antibody to p27 was used at a dilution of 1:1000, and antibody to p21 was used at a dilution of 1:150.

Quantification of Immunohistochemistry

Quantification of the Ki-67 labeling index (LI)—i.e., determination of proliferating cells—was performed using computer-assisted image-analysis equipment (Cell Analysis System; Becton Dickinson, Mountain View, CA). The KI-67 LI was estimated as described in detail elsewhere.5

Both p21 and p27 were immunostained in a single batch to avoid run-to-run variability. Only cells in neoplastic tissue with intense and homogeneous nuclear staining were considered positive.

The percentages of p21 and p27 immunopositive tumor cells were counted in 20 consecutive microscopic fields per tumor sample in areas that showed the highest density of these cells. In each specimen, up to 300–2000 tumor cell nuclei were evaluated. The tumors were grouped according to the percentage of p21 or p27 positive cells into three categories with approximately the same number of patients: < 30% positive cells, 30–50% positive cells, and > 50% positive cells.

Statistical Methods

Statistical analyses were calculated using the SPSS software program. In univariate survival analysis (from the diagnostic surgical procedure to the end point of follow-up), Kaplan–Meier curves and the log-rank test were used. A survival analysis also was performed by dividing the patients into short-term survivors (> 360 days) and long-term survivors (> 360 days). Differences between the patient groups also were estimated by using a Mann–Whitney test. Cox regression analysis was used in the multivariate analysis.


The mean age of the patients with AA was 54 years (range, 21–73 years), and the mean age of the patients with GB was 47 years (range, 24–77 years). The distribution of patients by proliferation index and expression of p27 and p21 is presented in Table 2. The mean ± standard deviation (SD) proliferation activity (Ki-67 LI) for all tumors was 23.3% ± 10.4%. There was a significant difference between the AA group and the GB group: The mean ± SD Ki-67 LI was 19.9% ± 10.6% in the AA group and 25.2% ± 9.0% in the GB group (P = 0.05, Mann–Whitney U test).

Table 2. Mean Labeling Indices for Ki-67, p27, and p21
LIMean LI (%)
Anaplastic astrocytoma (n = 25 patients)Glioblastoma (n = 52 patients)Mann–Whitney P value
  1. LI: labeling index; ns: not significant.


The mean ± SD p27 LI was 52.4% ± 25.5% in the AA group and 37.9% ± 25.6% in the GB group (P = 0.07; Mann–Whitney U test). Taking all tumors together, the p27 LI was < 30% in 28 tumors (36%), 30–50% in 19 tumors (25%), and > 50% in 30 tumors (39%). There was a statistically significant difference in survival between these three groups (P = 0.007; log-rank test) (Fig. 1), with better survival correlating with higher expression of p27. Similarly, a significant difference in Ki-67 LI and p27 LI existed between short-term survivors and long-term survivors (Table 3).

Figure 1.

Cumulative survival (Cum Survival) according to p27 labeling index (LI) in patients with malignant astrocytomas. The p27 LI was < 30% in 28 patients (36%), 30–50% in 19 patients (25%), and > 50% in 30 patients (39%). There was a statistically significant difference in survival between these three groups (P = 0.007; log-rank test), with better survival correlating with higher expression of p27.

Table 3. Ki-67 and p27 Labeling Indices with Short-Term and Long-Term Survival and a Cut-Off Point of 360 Days at 50% Survival
IndicatorShort-term survival (%)Long-term survival (%)Mann–Whitney P value
  1. LI: labeling index; SD: standard deviation.

Ki-67 LI   
 Mean ± SD26.2 ± 10.520.3 ± 9.5
p27 LI   
 Mean ± SD33.6 ± 23.854.0 ± 26.6

The mean ± SD p21 LI was 12.7% ± 7.8% in the AA group and 19.8% ± 12.6% in the GB group (P value not significant; Mann–Whitney U test). Among all neoplasms, the p21 LI was < 30% in 37 tumors (48%), 30–50% in 30 tumors (39%), and > 50% in 10 tumors (13%). No statistically significant difference in survival was detected between these three groups (Fig. 2). However, tumors with a p21 LI < 50% (i.e., patients with a p21 LI < 30% plus patients with a p21 LI of 30–50% combined) were associated with shorter survival compared with patients who had a p21 LI > 50% (P = 0.044; log-rank test).

Figure 2.

Cumulative survival (Cum Survival) according to p21 labeling index (LI) in patients with malignant astrocytomas. No statistically significant difference in survival was detected between the three groups: patients with a p21 LI < 30% (37 patients; 48%), patients with a p21 LI between 30% and 50% (30 patients; 39%), and patients with a p21 LI > 50% (10 patients; 13%; P = 0.044; log-rank test; n.s.: not significant).

In a multivariate Cox analysis, histopathologic grade, expression of p21 and p27, and proliferation activity of Ki-67 were tested for their independent prognostic value. In addition to histopathology (Exp B = 2.3699; P = 0.003) and Ki-67 (ExpB 1.5372; P = 0.0131), only p27 was a powerful predictor of survival (Exp B = 0.6152; P = 0.0007).


The cumulative inactivation of tumor suppressor genes and/or the amplification of oncogenes lead to progressively more malignant astrocytic tumors. The accumulation of genetic lesions that normally regulate the pathways of cell proliferation, differentiation, and cell death results in dysregulation of the cell cycle.6 To improve our understanding of the clinical relevance of p27 and p21 in malignant astrocytomas, we studied the correlation between their immunohistochemically detectable expression and survival. Although p27 and p21 are parallel cell cycle regulators, only p27 had independent predictive value for our patients with high-grade astrocytomas who were treated postoperatively with radiotherapy.

Identification of suppressor proteins that have a correlation with clinical outcome and survival provides new prognostic markers that may be used to guide treatment planning. p27 exerts its suppressor effect through cyclin E-dependent kinase (CDK2) by inhibiting the phosphorylation of pRb by CDK2, which, in turn, arrests cells in G1-phase and prevents them from entering the S phase. It also has been shown that the overexpression of p27 induces apoptosis in some human tumor cell lines, implying a second antineoplastic function for this tumor suppressor protein. A normal p21 gene is necessary for p53-mediated cell cycle arrest at G1,8 although p21 also may be activated independent of p53.9

Mutations in the p27 and p21 genes seem to be extremely rare in human malignancies.10 Rather, the decreased levels of p27 and p21suppressor proteins appear to be due to increased proteosome-related degradation, which is reflected in the disappearance of immunoreactivity in the tumor cell nuclei. Heterogeneity of glioma cells for the expression of p21 and p27 causes regional variation, which influences the staining results and, consequently, conclusions in different studies. Combined assessment of tumor suppressor gene expression and Ki-67 has provided more useful prognostic information.11, 12 PTEN, at 10q23.3, is a tumor suppressor gene that was altered in 30–44% of high-grade gliomas, as demonstrated previously.13 It has been suggested that PTEN may correlate with p27 activity, and loss of PTEN has been linked inversely to the duration of survival.14 PTEN expression may cause the accumulation of p27KIP1, raising the possibility that the phosphatidylinositol-3 kinase signaling pathway may regulate the level of p2714 and block cell cycle progression in the G1-phase. The expression of p27 in malignant astrocytomas appears to reflect the proliferation activity of tumor cells.

It has been shown in several solid tumors that the loss of CDK-inhibitor tumor suppressor p27 expression, which is associated with increased ubiquitin-mediated protein degradation, is an indicator of poor prognosis.15 The association with a poor prognosis has been demonstrated, for example, in patients with laryngeal larynx carcinoma and malignant mesothelioma.12, 15 Recently, an inverse correlation was reported between p27 immunoreactivity and the Ki-67 LI in patients with malignant gliomas.16, 17 We observed a similar correlation as well as a significant correlation with survival, with p27 shown as an independent prognostic indicator. Thus, immunohistochemical determination of p27 expression appears to improve the accuracy of prognostic evaluation of patients with malignant astrocytomas. p27 appears to be an even better prognosticator than p16, which our previous results showed was lost frequently in the later phase of the malignant transformation of astrocytes; however, the immunohistochemical demonstration of p16 had greater technical problems.5 To determine the association of p27 expression with other clinical characteristics will require a larger study.

In our previous study on the same material, the proportions of p21 negative tumors and p21 positive tumors were about equal in both the AA group and the GB group.5 In the current study, the percentage of p21 positive tumor cells did not correlate with the malignancy grade; rather, the mean p21 LI was higher in the GB group compared with the AA group. This is in agreement with results from the study by Jung and colleagues,18 who used Western blot analysis and observed elevated p21 expression in most gliomas regardless of their grade, including markedly elevated p21 levels in GBs independent of the p53 mutation status. In the current study, p21 was not correlated with survival when we divided our tumors into the same immunopositivity groups that were used for p27. Only when tumors with p21 LI < 50% and > 50% were compared was a weak correlation with survival obtained. In addition, in the multivariate analysis, p21 did not have independent prognostic value. Thus, the clinical relevance of p21 expression remains controversial in patients with malignant astrocytic tumors, similar to what has been found for several different tumor types.15, 19–22 The elevated levels of p21 expression may present the feedback mechanisms of the cell cycle in the malignant cell populations rather than true functional regulation.


The current results show that, although p21 and p27 are parallel regulators of the cell cycle, only p27 acts as an independent prognostic indicator for survival in patients with high-grade astrocytomas. The role of p21 appears to be less significant in the prognostication of patients with these tumors who are treated with radiotherapy.


The authors thank J. Nuutinen for assistance in the collection of clinical data on the patients.