Deregulation of the TP53/p14^ARF Tumor Suppressor Pathway in Low-Grade Diffuse Astrocytomas and Its Influence on Clinical Course

Low-grade diffuse astrocytomas (WHO grade II) are well-differentiated tumors that grow slowly but display an intrinsic tendency for diffuse infiltration of the neighboring brain structures. They almost invariably recur and often progress to higher malignancy, i.e., anaplastic astrocytoma (WHO grade III) or glioblastoma [WHO grade IV (1, 2)]. Such malignant transformation is responsible for the majority of the patient mortality (3, 4). During the past decade, much effort has been directed toward identifying genetic alterations associated with the pathogenesis and progression of diffuse astrocytomas. Mutation of the TP53 tumor suppressor gene has been demonstrated as one of the most frequent genetic alterations in low-grade diffuse astrocytomas, although its precise prognostic value is still debated (5, 6, 7).

The CDKN2A and CDKN2B genes, which cluster together on chromosome 9p21, code for the structurally highly homologous tumor suppressor proteins p16^INK4a and p15^INK4b, respectively. Alternative splicing of the CDKN2A gene results in the expression of p14^ARF, which is encoded by the unique exon 1β and exons 2 and 3 for p16^INK4a (8, 9, 10, 11). p16^INK4a and p15^INK4b inhibit CDK4²- and CDK6-mediated phosphorylation of retinoblastoma protein (RB1) and thereby negatively regulate cell cycle progression at the G₁ checkpoint (12, 13), whereas p14^ARF increases TP53 stability by abrogating MDM2 inhibition and thus facilitates TP53-mediated cell cycle arrest and apoptosis (8, 9, 10, 11). It appears therefore that the 9p21 gene cluster is linked to both the RB1 and TP53 tumor suppressor pathways. Inactivation of the p14^ARF, p15^INK4b, and p16^INK4a genes has frequently been observed in a variety of human tumors, including malignant gliomas, and is largely due to homozygous deletion or to hypermethylation of CpG islands in the promoter region (14, 15, 16, 17, 18). In low-grade gliomas, these genes are occasionally subject to aberrant promoter methylation (16, 17, 18, 19, 20). No study, however, has yet systematically analyzed such hypermethylation in a large series of low-grade diffuse astrocytomas and assessed the contributions of the RB1 and TP53 pathways to their clinical behavior.

The aim of the present study was to investigate the role of the RB1 and TP53 tumor suppressor pathways in low-grade diffuse astrocytomas and to clarify whether or not genetic alterations in each pathway affect their clinical outcome. We examined the alterations of the RB1/CDK4/p16^INK4a/p15^INK4b and the TP53/MDM2/p14^ARF pathways in WHO grade II astrocytomas and attempted to correlate these abnormalities with prognosis.

Tumor samples were obtained from patients operated on at Nihon University School of Medicine between 1981 and 1998. The sections from all patients were reviewed and classified according to the WHO criteria (21). Optic tract, basal ganglia, thalamus, brain stem, and cerebellar tumors; pilocytic astrocytomas; gemistocytic astrocytomas [>20% of all component tumor cells (21)]; pleomorphic xanthoastrocytomas; and mixed oligoastrocytomas were excluded from the study. Patients who were less than 18 years of age at the time of operation were also excluded. After the above exclusions, 64 patients with newly diagnosed fibrillary astrocytomas (WHO grade II) were retained. Among these cases, 46 tumor samples were available for methylation analysis, and this constituted the study population. In 18 cases (39%), differential PCR was not possible because of limitations in the amount of DNA available for analysis. At the initial treatment, 9 of the patients were given surgery alone, 8 patients were given immediate radiation therapy, and 23 patients were treated with human fibroblast IFN without radiation therapy as reported previously (4). Twenty-two patients underwent radical resection (defined by the absence of radiographic remnant disease). Alkylating agent-based chemotherapy was given to 20 patients at tumor recurrence. Twelve patients who had tumors recurring after the initial surgery only or after surgery plus human fibroblast IFN received radiation therapy in combination with chemotherapy. At the last follow-up (January 2002), 22 patients were alive without disease, 3 were alive with disease, and 21 were dead of the disease. Among the 25 survivors, the median follow-up period was 7.6 years (range, 3.1–18.8 years). Genomic DNA was extracted from paraffin sections as described previously (22). DNA was also extracted from a total of seven samples of normal brain tissue adjacent to tumors.

Promoter hypermethylation of the p14^ARF, p15^INK4b, p16^INK4a, and RB1 genes was determined by methylation-specific PCR (23). Sodium bisulfite modification was performed with a CpGenome DNA Modification Kit (Intergen, Oxford, United Kingdom) as described previously (16, 17).

The methods used for methylation-specific PCR of the p14^ARF, p15^INK4b, p16^INK4a, and RB1 genes were as described previously (16, 17, 18, 24). The amplified products were electrophoresed on 3% agarose gels and visualized with ethidium bromide. CpGenome Universal Methylated DNA (Intergen) and normal blood DNA were included in each set of the PCR as methylated and unmethylated controls, respectively.

Exons 5–8 of the TP53 gene were screened for mutations based on PCR single-strand conformational polymorphism followed by direct sequencing as described previously (7, 18).

To assess homozygous deletions of the p14^ARF (exon 1β), p15^INK4b, and p16^INK4a (exon 1α) genes, differential PCR was carried out as described previously (16, 17, 18). The β-actin sequence was used for p15^INK4b and p16^INK4a deletions, and the glyceraldehyde-3-phosphate dehydrogenase sequence was used for p14^ARF deletion, as a reference. Samples presenting <20% of the control signal were considered as homozygous deletions (16, 17, 18).

CDK4 and MDM2 amplifications were detected by differential PCR as described previously, using the IFN-γ (IFNG) and dopamine receptor (DR) sequences as a reference, respectively (18, 25, 26). Values of greater than 2.7 and 3.02 for the CDK4/IFNG ratio and MDM2/DR ratio were regarded as amplification of the CDK4 and MDM2 genes, respectively (18, 25, 26).

The Kaplan-Meier method was used to calculate the PFS and OS. We designated the PFS as the time period from the first operation to the point when tumor regrowth or recurrence was confirmed by an imaging study. OS was defined as the interval between the first operation and the date of death or the most recent evaluation. The log-rank test was used to assess the degree of significance of the differences in different subgroups. The continuous variables, i.e., age and Karnofsky performance score, were categorized on the basis of the median value. The Cox proportional hazards model was used to identify the multivariate predictors of survival. When a potential prognostic factor was judged to be independent and appropriate for the model, the HR and 95% CI were calculated. The relationships between various parameters were analyzed statistically by the χ² test, Fisher’s exact test, or Student’s t test as appropriate. All values are expressed as the means ± SD. The significance level chosen was P < 0.05, and all tests were two-sided. The statistics were analyzed with a personal computer running Stat View J-5.0 software (Abacus Concepts, Berkeley, CA).

DNA obtained from the 46 newly diagnosed low-grade diffuse astrocytomas was subjected to a p14^ARF promoter methylation study using the methylation-specific PCR. Among all of the tumors studied, p14^ARF promoter methylation was present in 9 of the 46 samples (20%; Fig. 1). We also evaluated seven samples of normal brain tissue, but none showed p14^ARF methylation. Mutation of the TP53 gene was assessed in the same samples analyzed for p14^ARF promoter methylation and demonstrated in 25 of the 46 tumors (54%) studied (Table 1). One deletion with frameshift and 26 missense mutations with amino acid exchange were identified. Two tumors contained double missense mutations. The hot spots for mutations included codons 175 (n = 5), 248 (n = 3), and 273 (n = 8). p14^ARF homozygous deletion and MDM2 amplification were not detected in any of the 28 tumors analyzed.

The patterns of the TP53/p14^ARF pathway gene abnormalities in individual tumors were then compared. p14^ARF was hypermethylated in 7 of the 21 tumors (33%) with wild-type TP53, as opposed to 2 of the 25 tumors (8%) with a mutant TP53. These alterations thus tended to occur independently, although this trend did not reach statistical significance (P = 0.0590, Fisher’s exact test). When all abnormalities of the two genes were combined, 32 of the 46 tumors (70%) displayed alterations of the TP53/p14^ARF pathway.

Promoter hypermethylation of RB1, p16^INK4a, and p15^INK4b was assessed in all 46 tumors and detected in 1 (2%), 3 (6%), and 2 samples (4%), respectively (Fig. 1). In the seven samples of normal brain tissue, there was no evidence of methylation of these genes. Analyses of p16^INK4a and p15^INK4b homozygous deletions and CDK4 amplification were possible in 28 cases. One tumor (4%) showed homozygous deletion of the p15^INK4b gene. p16^INK4a homozygous deletions and CDK4 amplification were not detected in any of the 28 tumors.

These alterations were mutually exclusive, except for one case that demonstrated both p16^INK4a and p15^INK4b methylation; the numbers were too small for statistical analysis. When all of the abnormalities were combined, 6 of the 46 tumors (13%) displayed alterations of the RB1/p16^INK4a/p15^INK4b pathway.

When all of the genetic abnormalities described above were combined and compared, simultaneous disruption of the TP53 and RB1 pathways was found in only one case, in which mutation of the TP53 gene and hypermethylation of the p16^INK4b gene were detected. This case relapsed at 13 months after surgery, and the patient died at 14 months after recurrence. Thirty-seven of the 46 tumors (80%) had genetic alterations in either the TP53 or RB1 pathway genes. An inverse relationship between abnormalities of the RB1 pathway genes and the TP53 pathway genes was evident (P = 0.0072, Fisher’s exact test).

For the entire study population, the median PFS was 7.2 years with a 5-year PFS rate of 57%, and the median OS was 9.3 years with a 5-year OS rate of 84%.

The results for the univariate statistical analyses of categorical variables affecting the PFS and OS are summarized in Table 2. TP53 mutation and p14^ARF methylation, when analyzed independently, tended to be negative predictors of the PFS, although they failed to achieve statistical significance. With respect to hot spot codon TP53 mutations, the PFS of patients with tumors carrying codon 175 mutations was shorter than that of those with other mutations, but the difference did not reach statistical significance. RB1, p16^INK4a, or p15^INK4b methylation status, alone or in combination, was not a statistically significant predictor of PFS. Alternate or simultaneous presence of TP53 mutation and p14^ARF methylation emerged as a significant predictor of a shorter PFS (Fig. 2). Patient age and extent of surgery also influenced the PFS significantly. Patients with tumors lacking both TP53 and p14^ARF alterations had a mean age of 33.6 ± 11.3 years, as opposed to an age of 35.9 ± 12.3 years in the patients with all other tumors (P = 0.5431, Student’s t test). TP53/p14^ARF pathway gene abnormalities were present in 18 of the 24 patients (75%) undergoing palliative surgery, as compared with 14 of the 22 patients (64%) undergoing radical surgery (P = 0.4028, χ² test). To determine whether or not alternative alteration of the TP53 and p14^ARF genes remained significantly associated with the PFS after adjustment for the effects of age and extent of surgery, we applied a Cox proportional hazards model. Such multivariate analysis demonstrated that the alternate or simultaneous presence of TP53 mutation and p14^ARF methylation had no statistically significant effect on the PFS (P = 0.0856). Older age (HR, 2.360; 95% CI, 1.014–5.491; P = 0.0463) and palliative surgery (HR, 3.219; 95% CI, 1.304–7.944; P = 0.0112) each qualified as an independent predictor of a shorter PFS in this model. In terms of the OS, age and extent of surgery displayed a prognostic utility on univariate analysis. Patients with alternative alteration of the TP53 and p14^ARF genes also exhibited a tendency toward a shorter OS (Fig. 3), although this trend lacked significance. Applying a Cox proportional hazards model using age and extent of surgery, older age (HR, 2.496; 95% CI, 1.012–6.025; P = 0.0470) and palliative surgery (HR, 3.142; 95% CI, 1.212–8.141; P = 0.0184) were found to be significantly unfavorable prognostic factors for the OS.

The RB1/CDK4/p16^INK4a/p15^INK4b pathway plays a crucial regulatory role in cell cycle progression (13). Another important tumor suppressor pathway, TP53/p14^ARF/MDM2, is involved in the regulation of cell proliferation, apoptosis, and DNA repair (27, 28). It has become apparent that many of the genetic alterations detected in diffuse gliomas affect genes that encode members of these two critical regulatory pathways (1, 2). The present study demonstrated that the TP53 pathway was predominantly deregulated (32 of 46; 70%) over the RB1 pathway (6 of 46; 13%). We further provided evidence that alternative alteration of the TP53 and p14^ARF genes was a univariate predictor of a shortened PFS. These findings suggest that the TP53 pathway may play an important tumor suppressor role in low-grade diffuse astrocytomas.

A mutation of the TP53 suppressor gene has been observed in approximately one-half of diffusely infiltrating astrocytomas. It is well known that the TP53 mutation occurs at an early stage of malignant transformation (1, 2, 7). Some clinical studies have indicated that TP53 mutation represented a significant adverse indicator of PFS in low-grade astrocytomas and oligoastrocytomas (5, 6), whereas others have failed to demonstrate such a correlation (7). In the study of Peraud et al. (6), the prognostic impact of TP53 status was closely related to the influence of the gemistocytic subtype. In the present study, which exclusively examined supratentorial fibrillary astrocytomas in adults, TP53 mutation tended to be associated with a shorter PFS, but the relationship failed to achieve statistical significance. In accordance with the studies on colorectal cancer, Peraud et al. (6) observed an unfavorable prognostic influence of codon 175 mutations exclusively in nongemistocytic astrocytomas. We also found that mutations at this codon were associated with a shorter PFS and OS, although the difference was not statistically significant, possibly due to the limited number of patients investigated.

In the present study, homozygous deletions of the p14^ARF, p15^INK4b, and p16^INK4a genes or amplifications of the CDK4 and MDM2 genes were either absent or very rare. Although we did not completely analyze these alterations in full samples, this observation is consistent with previous reports examining low-grade gliomas (16, 17, 18, 29). Such phenomena may thus contribute to a more malignant phenotype. In contrast, hypermethylation of the p14^ARF, p15^INK4b, and p16^INK4a genes was detected in a considerable fraction of low-grade astrocytomas, although no methylation was detected in seven normal brain tissue samples. In colorectal carcinogenesis, CpG island methylation of several key genes is an early event, which is detectable in nonneoplastic mucosa, and is associated with microsatellite instability (30, 31). During astrocytoma progression in patients with multiple biopsies, hypermethylation of the p14^ARF gene or the p16^INK4a gene was already present in a subset of low-grade diffuse astrocytomas and was followed by homozygous deletions of p14^ARF and p16^INK4a in glioblastomas derived therefrom (16). These data suggest that aberrant methylation of the 9p21 genes represents an early genetic event in astrocytoma tumorigenesis and may enhance genetic instability, leading to subsequent structural DNA changes.

p14^ARF plays a major role in the TP53 tumor suppressor pathway by antagonizing MDM2-mediated degradation of TP53 (8, 9, 10, 11). Thus, theoretically, p14^ARF loss of function would be predicted to reduce the frequency of concomitant TP53 mutations (32). In lymphoma (33), hepatocellular carcinoma (34), and glioblastoma (35), an inverse correlation between TP53 and p14^ARF alterations has been observed, whereas contradictory results were obtained in lung cancer (36). Such conflicting findings might imply that different roles for p14^ARF are operative in the pathogenesis of each cancer type. In the present study, p14^ARF hypermethylation was implicated in 7 of 21 tumors (33%) with wild-type TP53 and appeared to be an independent event to TP53 mutation, suggesting that p14^ARF may be the crucial target in a considerable fraction of diffuse astrocytomas with intact TP53.

An important function for p14^ARF as well as TP53 in carcinogenesis is evident in animal models. Mice lacking expression of p19^ARF (the mouse homologue to human p14^ARF), through selective disruption of exon 1β, frequently develop spontaneous tumors including sarcomas, lymphomas, and gliomas (37). Inactivation of p14^ARF has been demonstrated as a frequent event in a variety of human neoplasms, although its prognostic usefulness was not consistently identified in colorectal cancer, hepatocellular carcinoma, and soft tissue sarcoma (34, 38, 39). In a subset of glioblastomas, homozygous deletion of p14^ARF has been shown to be associated with a shorter survival (40). In the present study, the presence of p14^ARF hypermethylation was associated with a shorter PFS, although the difference was not statistically significant. Also, TP53 mutation, analyzed independently, did not have any apparent effect on prognosis. When combining TP53 and p14^ARF analysis, a significantly decreased PFS was identified. These observations suggest that abnormalities of each gene may have an equivalent impact on tumor aggressiveness.

Our current understanding of the molecular mechanisms underlying growth control in cells is still far from complete. Cell biological experiments have suggested that deregulation of the G₁-S transition control system may confer a growth advantage to the cells; however, in some types of cultured cells, loss of G₁-S transition gene function induces TP53-dependent apoptosis (41, 42, 43). This indicates that the TP53 pathway may prevent cells with abnormalities of G₁-S transition-regulatory genes from aberrant proliferation and further suggests that simultaneous disruption of these two pathways would be required to produce unregulated cell cycling and thereby predispose subjects to a more aggressive biological behavior. This view is supported by the findings of several previous studies demonstrating that dual inactivation of the G₁-S transition control system and the TP53 pathway was absent in low-grade astrocytomas and oligodendrogliomas but frequent in high-grade astrocytomas and anaplastic oligodendrogliomas (18, 29). Our finding that concurrent disruption of the RB1 and TP53 tumor suppressor pathways was present in only one tumor, which exhibited a worse clinical behavior, also supports such an assumption.

Because the present study used a retrospective design with a relatively short follow-up and relatively small patient number, it remains inherently difficult to assess the actual prognostic factors in these slow-growing tumors, eliminating the confounding variable of treatment effect on prognosis. Nevertheless, age and extent of resection stood out as powerful determinants of both the PFS and OS, in agreement with the results of previous studies (3, 44). Such correlations continued to be significant when these two variables were examined simultaneously in a Cox proportional hazards model. In contrast, the value of alternative alteration of the TP53 and p14^ARF genes did not include a prognostic utility in multivariate analysis. This may be accounted for on the basis that the TP53 pathway gene abnormalities tended to interrelate with age and extent of surgery, which surpassed their prognostic importance. A relatively higher frequency of TP53 pathway gene abnormalities in patients with persistent disease after surgical resection may be of particular importance, implying that the presence of these aberrations could be associated with invasive tumors (thus being inaccessible to extensive surgery). Although these findings need to be confirmed in a larger cohort of patients with longer follow-up, our data suggest that combined molecular analysis of these two genes is itself unlikely to add significant prognostic information to other conventional prognostic factors and could not therefore be used routinely in the clinical setting as a prognostic marker.

In conclusion, the TP53 pathway was frequently deregulated in low-grade diffuse astrocytomas. Alternative alteration of the TP53 and p14^ARF genes was associated univariately with a decreased PFS, but this association was not significant when age and extent of surgery were included in the analysis. Thus, an important tumor suppressor role is suggested for the TP53 pathway in low-grade diffuse astrocytomas, although its value is unlikely to include a prognostic utility that is independent of other conventional prognostic factors.