Loss of the AP-2α Transcription Factor Is Associated with the Grade of Human Gliomas Free

Tumor progression depends on factors intrinsic to the tumor cell, including but not limited to growth factors and their cognate receptors, extracellular matrix proteins, proteases, chemokines, and cellular adhesion molecules. The expression of these factors is influenced by the environment, microenvironment, epistasis, and genetic and epigenetic factors. The transcription factor activator protein (AP)-2α has been shown to regulate many of the genes that are involved in normal cellular hemostasis. Therefore, the loss of AP-2α expression may result in dedifferentiation, proliferation, and, eventually, metastasis or invasion.

AP-2α is a 52-kDa protein mapped to the short arm of chromosome 6 near the HLA locus (1), a common molecular abnormality found in human astrocytomas. The AP-2α protein binds to a consensus palindromic core recognition element via a DNA-binding domain located within the COOH-terminal half of the protein (2). AP-2α, which is regulated by cyclic AMP and retinoic acid (3), mediates programmed gene expression during both embryonic morphogenesis and adult cell differentiation (4).

AP-2α plays a pivotal role in regulating the expression of several genes, the products of which are involved in tumor growth and metastasis. For example, AP-2α regulates genes that are involved in proliferation, cell cycle regulation (HER-2 and p21^WAF1), apoptosis (c-KIT, BCL-2, and FAS/APO-1), adhesion (MCAM/MUC18 and E-cadherin), and invasion/angiogenesis (MMP-2, PAI-I, VEGF, and PAR-1). Therefore, loss of AP-2α expression may contribute to functional changes of several gene products that in normal cells are involved in cellular proliferation and differentiation. We have shown previously that down-regulation of AP-2α results in altered c-KIT and MCAM/MUC18 expression, both of which contribute to the increased metastatic potential of human malignant melanoma cells in vitro and in vivo(5). Results of other studies have provided support for the role played by AP-2α in the progression of mammary and prostate cancer (6, 7).

Studies have suggested that several tumor markers (i.e.,p53, EGFR, and PDGFR) may play a role in the transformation of human gliomas. However, these findings have not been consistent, even within WHO tumor grades. The prognostic significance of these tumor markers, especially the most common one, p53, is unclear (8–11). EGFR amplification has been shown to have prognostic value in patients with anaplastic astrocytoma and in older patients with glioblastoma multiforme (12). However, a definite prognostic marker for high-grade gliomas has thus far not been identified. Thus, to determine whether AP-2α expression could be involved in higher-grade gliomas, we intended to determine the level of AP-2α expression across all WHO tumor grades among a large number of human glioma surgical specimens.

Clinical Samples. Duplicate samples from distinctly different areas of each tumor were arrayed on one slide from resected gliomas with available tissue samples between 1986 and 2001 (n = 246), which included 72 glioblastomas (grade 4), 49 anaplastic astrocytomas (grade 3), 9 low-grade astrocytoma (grade 2), 37 oligodendrogliomas (grade 2), 37 anaplastic oligodendrogliomas (grade 3), 15 mixed oligoastrocytomas (grade 2), 20 anaplastic mixed oligoastrocytomas (grade 3), and 7 gliosarcomas (grade 4) archived in paraffin blocks. Included on this slide were normal brain tissue (positive control) and glioma cell lines (negative control). Cross-referencing to each patient's medical chart was possible through use of the Department of Neurosurgery clinical database and pathology records. Survival time was defined as the time from the M.D. Anderson Cancer Center surgery to death or to the most recent follow-up.

Microarray Immunohistochemistry. Sections obtained from the archived paraffin blocks were deparaffinized and steamed in Serotec target unmasking fluid (Serotec, Inc., Raleigh, NC) for antigen retrieval. Samples were blocked with endogenous peroxidase with 3% H₂O₂ followed by blocking in 4% fish gelatin and PBS.

Samples were incubated overnight with primary antibody diluted in blocking solution. AP-2α was detected using a 1:1,000 dilution of anti-AP-2α polyclonal antibody (Santa Cruz Biotechnology, Santa Cruz, CA). Competition studies with a blocking peptide (Promega, Madison, WI) showed that this polyclonal antibody was specific for AP-2α (13, 14). The primary antibody was washed with PBS and incubated with a biotinylated secondary antibody. Streptavidin-horseradish peroxidase was added followed by 3-amino-9-ethylcarbazole. The samples were counterstained with Gill's no. 3 hematoxylin and subsequently washed with water and mounted with universal mount. Samples were considered AP-2α positive when two blinded, independent investigators verified positive nuclear staining of the arrayed tissue by microscopic evaluation. In almost all cases, positive AP-2α staining was uniform and virtually 100% of the tumors cells were positive. In the only exception to this rule, the mixed tumors, we determined a specimen was positive when >50% of the cells were positive.

Vascular endothelial growth factor (VEGF) was detected using a 1:1,000 dilution of anti-VEGF antibody (Santa Cruz Biotechnology) and matrix metalloproteinase (MMP)-2 was detected using a 1:1,000 dilution of anti-MMP-2 antibody (Chemicon International, Temecula, CA). The primary antibody was washed with PBS and incubated with the appropriate dilution of biotinylated secondary antibody. Slides were washed again with PBS and exposed to horseradish peroxidase-avidin complex, washed, and developed with diaminobenzidine (ImmunoPure Ultra-Sensitive ABC Staining kit, Metal Enhanced DAB Substrate kit, Pierce, Rockford, IL). The specimen was considered to have increased expression when the amount of horseradish peroxidase staining was clearly at least 2-fold greater than normal brain tissues arrayed on the same slide.

Statistical Analysis. The Fisher exact test was used to evaluate differences in the various characteristics between AP-2α-positive and AP-2α-negative groups as well as the relative frequency of AP-2α positive (compared with low-grade glioma) among the different glioma histologies and grades. Kaplan-Meier estimates of survival were generated, and univariate and multivariate analyses of predictors of survival were done using the Cox proportional hazards model. Crude and adjusted rate ratios and their 95% confidence intervals (95% CI) were obtained. P < 0.05 was considered significant.

Study Population. The median age of the patients was 42 years (range, 4-82 years). The majority of patients (97%) had a Karnofsky Performance Scale score of ≥70. Seven patients had a biopsy. Among the remaining 239 patients, the median extent of resection was 96%. The AP-2α-positive group was significantly younger (median, 38 years; range, 4-64 years) compared with the AP-2α-negative group (median, 44 years; range, 4-82 years; P < 0.001). The AP-2α-positive group had a lower extent of resection (median, 89%; range, 19-100%) compared with the AP-2α-negative group (median, 99%; range, 12-100%; P = 0.008).

AP-2α Expression Is Lost within High-grade Gliomas. In the vast majority of the samples, determination of intranuclear AP-2α positivity was absolute and usually uniform, although exceptions occurred, especially in the mixed tumors such as the mixed oligoastrocytoma, in which the distribution was patchy. In the mixed oligoastrocytoma and the anaplastic mixed oligoastrocytoma cases, those cells with astrocytic features (fibrillary processes) were virtually 100% devoid of AP-2α and those cells with oligodendroglioma features (uniformly round regular nuclei with a classic “fried egg appearance”) were uniformly positive. In these cases, when >50% of the intranuclear staining were positive for AP-2α expression, the specimen was determined to be positive (Fig. 1).

Normal brain, including both white and gray matter and the cerebellum, had ubiquitous expression of intranuclear AP-2α. Among the nine specimens from patients with astrocytoma, 100% maintained expression of AP-2α. Among the 49 specimens from patients with anaplastic astrocytoma, only 2% maintained expression of AP-2α (P < 0.001 compared with astrocytoma), and among the 72 glioblastoma specimens, only 1% still maintained expression (P < 0.001 compared with astrocytoma). In the case of the 37 oligodendrogliomas, 78% of the specimens maintained positive expression of AP-2α. However, in the 37 anaplastic oligodendroglioma specimens, only 35% of the samples still maintained expression of AP-2α (P = 0.0004 compared with oligodendroglioma). In the 15 tumors that had mixed oligoastrocytoma features, 80% maintained positive expression, but again as the histologic grade became more anaplastic (n = 20), only 60% maintained positive expression. Although this difference was not statistically significant, it supports the trend observed with anaplastic astrocytoma versus astrocytoma and mixed oligodendroglioma-astrocytoma versus oligodendroglioma. In the case of gliosarcoma (n = 7), 29% still had positive expression of AP-2α (P = 0.005; Table 1).

AP-2α Expression Is Lost within High-grade Astrocytomas Compared with High-grade Oligodendrogliomas. Among WHO grade 3 and 4 tumors, there was a significant difference in AP-2α expression in anaplastic astrocytoma compared with anaplastic oligodendroglioma (P < 0.0001), suggesting that AP-2α has a greater roll in the progression of astrocytic-type tumors. Among WHO grade 2 tumors, there was no significant difference in AP-2α expression in astrocytoma versus oligodendroglioma (P = 0.32). Thus, AP-2α seems to be a marker of grade 3 and 4 astrocytic gliomas.

AP-2α Expression Affects Downstream Markers. There was a statistically significant association of AP-2α loss with overexpression of MMP-2 and VEGF. For example, in astrocytoma, none had lost expression of AP-2α and none overexpressed MMP-2 or VEGF (P < 0.0001). However, in glioblastoma, in which 99% had lost expression of AP-2α, 31% and 28% of the tumors overexpressed MMP-2 and VEGF, respectively (P < 0.0001). This statistically significant trend was observed across all grades for VEGF overexpression. Similar findings were seen with MMP-2 expression, except in anaplastic oligodendroglioma and anaplastic mixed oligoastrocytoma, in which the loss of AP-2α was not significantly associated with overexpression of MMP-2 (P = 0.63 and 0.34, respectively).

AP-2α Expression Affects Survival. As high as 109 of 167 patients with loss of AP-2α died during follow-up. The Kaplan-Meier estimate of median survival for patients in this group was 1.64 years (95% CI, 1.14-2.14; Fig. 2). Only 18 of 79 patients with AP-2α-positive tumors died during follow-up. The loss of AP-2α was a negative prognostic indicator within the overall category of gliomas by univariate analysis (rate ratio, 4.30; 95% CI, 2.60-7.10; P < 0.001; Table 2), however, that effect was lost after adjustment for important confounding variables such as age, Karnofsky Performance Scale, extent of surgical resection, and tumor grade (adjusted rate ratio for loss of AP-2α, 1.18; 95% CI, 0.63-2.20, P = 0.62). No significant differences in the rates of survival between the two AP-2α groups were seen within each tumor grade. For example, in the only glioblastoma patient whose tumor expressed AP-2α, the survival time was 19.9 months, which was not significantly different from the 18-month median survival time of the remaining AP-2α-negative-expressing glioblastoma patients. In the case of the two patients with anaplastic astrocytoma who maintained AP-2α expression, the mean survival time was 49.3 months, which compared favorably with the 35-month mean survival time in the other patients with anaplastic astrocytoma, whose tumors were AP-2α negative. However, these groups of patients are too small to allow meaningful statistical analysis. Patients with oligodendroglioma, whose tumors maintained AP-2α expression, had a mean survival time of 47.8 months, which again compared favorably with the 34.3-month survival time in patients whose oligodendroglioma were AP-2α negative, but this was not statistically significant. No significant difference was seen in mean survival time between patients with AP-2α-positive anaplastic oligodendroglioma and those with AP-2α-negative anaplastic oligodendroglioma (51 versus 50 months, respectively). In the case of the 15 patients with mixed oligoastrocytoma, a trend toward longer survival time was seen in the 12 patients whose tumors maintained AP-2α expression compared with the three patients whose tumors were AP-2α negative (60.8 versus 47.9 months, respectively). No significant differences were seen in mean survival time between patients with AP-2α-positive and those with AP-2α-negative anaplastic oligodendroglioma (47.8 versus 46.5 months, respectively).

Using tissue microarrays, we have shown for the first time that high-grade human gliomas are associated with the loss of expression of the AP-2α transcriptional factor, which regulates several genes, such MCAM/MUC18, MMP-2, and c-KIT that have been shown to be important for tumor progression and invasion. Over time, lower-grade astrocytomas can acquire genetic mutations and subsequently evolve to a higher grade (secondary glioblastoma). Differential genetic pathways leading to primary and secondary glioblastoma have been elucidated. The most well-characterized tumor markers associated with progression of glioblastoma are p53, MDM2, EGFR, and PDGF. The frequency of p53 protein accumulation in secondary glioblastoma is >90%, whereas in primary glioblastoma expression is <35%. MDM2 forms a complex with p53, thus abolishing its transcriptional activity. Immunohistochemical staining of overexpression of MDM2 is observed in >50% of primary glioblastomas but in <10% of secondary glioblastomas. EGFR is overexpressed in ∼30% to 60% of the primary glioblastoma but rarely in secondary glioblastoma (10%). Finally, PDGF overexpression is present in ∼60% of secondary glioblastoma (15). The expression of AP-2α strongly correlates with tumor grade and there is a ubiquitous loss between grade 2 and 3 astrocytomas. Therefore, of the previously characterized tumor markers, the loss of AP-2α would be the most common (96.2%) molecular marker as an astrocytic tumor evolves from grade 2 to 3.

Based on the loss of heterozygosity and dosage analysis in human astrocytoma, the second most common genetic abnormality was identified previously on chromosome 6p (16). Additionally, allelic imbalance on 6p was present in 57% of patients with astrocytoma (17). The transcriptional factor AP-2α has been mapped to human chromosome 6p22 (1). The discrepancy between the data from the previously published cytogenetic analyses and our tissue microarray data may be because of different sensitivities of these assays. Point mutations resulting in a nonfunctional protein would not be detected by cytogenetic analysis. Ultimately, comparisons of in situ hybridization for AP-2α with 6p alterations in the same tumor to show correlation will ultimately be necessary to definitely establish this relationship.

We have shown previously that transfection of melanoma cells lines with a dominant-negative AP-2β, which inhibits AP-2 transactivator function, results in an increase in MMP-2 expression, activity, and subsequent invasion (18). Additionally, overexpression of AP-2α in prostate cell lines resulted in down-regulation of VEGF. AP-2α was found to repress VEGF by competing with the transcriptional activator Sp3. Loss of AP-2α in prostate cancer cells increased VEGF expression (19). Similar to these previous findings, we found that gliomas that lost AP-2α were more likely to overexpress VEGF and/or MMP-2 compared with normal brain by immunohistochemistry. The up-regulation of VEGF and/or MMP-2 does not always absolutely occur with AP-2α loss, which would be expected given the fact that AP-2α is only one of several regulatory genes for MMP-2 and VEGF (20). To further elaborate on the role of AP-2α in gliomagenesis, we transfected AP-2α into the U-87 glioma cell line, but this induced apoptosis (data not shown). This would be consistent with AP-2α having a proposed role as a tumor suppressor gene.

The loss of the AP-2α transcriptional factor may represent a molecular change that can be characteristic of astrocytic tumors, analogous to the allelic loss of 1p and 19q in oligodendroglioma (21, 22). In the mixed oligoastrocytoma tumors, the loss of AP-2α may be reflective of the tumor being more “astrocytic.” The histopathologic definition of mixed oligoastrocytoma remains controversial (23); therefore, a reliable molecular marker may be useful in providing additional diagnostic criteria. Our data support the use of AP-2α expression to separate low-grade astrocytomas from higher-grade pure astrocytomas. Furthermore, the loss of AP-2α function may play a role in the transformation of human gliomas.

Despite the potential role of AP-2α in gliomagenesis, the prognostic effect of AP-2α remains unclear. This lack of prognostic importance may be due in part to the close association of the AP-2α loss and histologic grade or the effect of additional factors that have not yet been identified. The lack of prognostic effect is not surprising because other tumor markers have failed to show prognostic significance as well. For example, the role of the amplified EGFR (wild-type) and the variant (EGFRvIII) in malignant progression and effect on progression-free survival and overall survival has been debated in the literature. The amplified wild-type EGFR was not found to be an independent prognostic indicator of survival in several studies (24–26), and one study was inconclusive (27). One study did identify EGFR as a negative prognostic indicator in younger patients (28), and several others found EGFR to be an independent unfavorable predictor of survival (12, 29, 30). The presence of the mutated EGFR, EGFRvIII, was found to be an independent and significant unfavorable prognosticator of survival (31). Furthermore, p53 immunostaining showed no prognostic significance (27). The only marker that has shown prognostic significance is cathepsin B (32). Cumulatively, these data would suggest that the presence of a sole oncogene or loss of a single tumor suppressor gene does not confer prognostic effect; however, gene clusters now being elucidated with tumor microarrays may ultimately be able to provide prognostic guidance to treating oncologists and their patients.

We thank Karen Phillips for technical support, Dr. Weiming Shi for tumor volumetric analysis, and Dr. Mark Gilbert and Dr. Raymond Sawaya for thoughtful commentary.