Prognostic Significance of p14^ARF, p15^INK4b, and p16^INK4a Inactivation in Malignant Peripheral Nerve Sheath Tumors

Malignant peripheral nerve sheath tumor (MPNST) is an uncommon soft tissue sarcoma carrying poor prognosis (1). Approximately half of MPNSTs develop from benign neurofibroma in patients with neurofibromatosis type 1 (NF1), whereas the remaining half occur sporadically (2). NF1 is an autosomal dominant neurocutaneous disorder with NF1 gene dysfunction, and characterized by multiple neurofibromas and greatly increased risk of developing MPNST. However, loss of both NF1 alleles is not sufficient for malignant change of neurofibromas, and thus additional genetic alterations are thought to be necessary for the malignant transformation (3–6). To date, however, the detailed molecular events contributing to development of MPNST have been unclear in both NF1-related and sporadic cases.

Clinically, surgical resection is the primary curative treatment of MPNSTs (7). The roles of chemotherapy and radiation therapy are very limited (8). Because unresectable cases are unlikely to be cured, early detection of recurrent or metastatic lesions is of great importance in the follow-up of postoperative patients with MPNST. For the selection of high-risk patients, precise prognostic prediction is important, even though it has not been established yet.

Frequent loss of the CDKN2A–CDKN2B locus on chromosome 9p21 has been reported in various kinds of human cancers and sarcomas (9–13). The locus encodes 3 important cell-cycle inhibitory proteins: p14^ARF, which is encoded by an alternative reading frame of CDKN2A, p15^INK4b, which is encoded by CDKN2B, and p16^INK4a, which is encoded by CDKN2A. The p14^ARF protein activates the key checkpoint protein p53, thereby inducing either cell-cycle arrest (both in G₁ and G₂) or apoptosis. Both p15^INK4b and p16^INK4a are able to induce cell-cycle arrest in G₁ by inhibiting the cyclin-dependent kinases CDK4 and CDK6 to inactivate the retinoblastoma (RB1) family of tumor suppressor proteins. The tumor-suppressive functions for p16^INK4a and p14^ARF have been firmly established through experiments; however, the role of p15^INK4b remains controversial because transgenic mice deficient for p15^INK4b showed only a very subtle tumor predisposition in an experimental study (14). More recently, it has been shown that p15^INK4b functions as a critical tumor suppressor in the absence of p16^INK4a in an experimental setting (15), whereas the tumor-suppressive effect of p15^INK4b has not been investigated well in clinical tumor specimens.

Several studies have shown that p16^INK4a is more frequently inactivated in MPNSTs than in neurofibromas (16–24). On the other hand, the impact of p14^ARF and p15^INK4b inactivation in MPNSTs has been less investigated. Furthermore, the prognostic influence of p14^ARF, p15^INK4b, and p16^INK4a inactivation in patients with MPNST has not been examined enough.

Here we carried out a comprehensive analysis of the 9p21-related genes p14^ARF, p15^INK4b, and p16^INK4a in a series of 129 paraffin-embedded and 39 snap-frozen specimens from 99 patients with MPNST. The gene alterations and mRNA and protein expressions were studied by multiplex real-time PCR, PCR–single-strand conformation polymorphism (PCR–SSCP), methylation-specific PCR (MSP), quantitative reverse transcription PCR (RT-PCR), Western blotting (WB), and immunohistochemistry (IHC). Finally, the immunohistochemical results were compared with the clinicopathological variables and overall survival of the patients.

One hundred twenty-nine paraffin-embedded MPNST specimens from 99 patients were obtained from the collection of soft tissue tumors registered in the Department of Anatomic Pathology, Pathological Sciences, Kyushu University, Japan between 1964 and 2008. The specimens included 85 primary, 39 recurrent, and 5 metastatic tumors. Specimens of the primary tumor were not available from 14 patients who visited our hospital at the time of recurrence. A total of 18 pairs of specimens from the primary and the corresponding first recurrent or metastatic tumors were available for the purpose of comparison. The diagnosis of MPNST was made according to the latest edition of the World Health Organization classification (25) and the textbook of Enzinger and Weiss (26). The clinicopathological characteristics are summarized in Table 1. The 85 primary cases included 45 males and 40 females whose ages ranged from 1 to 88 years (median 47 years). In all, 33 patients were diagnosed with NF1 based on the NIH criteria (27). Tumors were located in the extremities in 38 cases (thigh 18, upper arm 13, lower leg 5, forearm 2), in the trunk wall in 20 cases (back 8, buttock 6, chest wall 4, abdominal wall 2), in the spinal or paraspinal region in 11 cases, in the head and neck in 9 cases, in the retroperitoneum in 5 cases, and in the visceral organs in 2 cases. All primary tumors were treated surgically, accompanied by adjuvant chemotherapy or irradiation in 3 cases. The histological tumor grade was evaluated in all specimens according to the French Federation of Cancer Centers (FNCLCC) grading system (28). As for staging of the primary tumors, the latest American Joint Committee on Cancer (AJCC) staging system was used (29). Follow-up information was available in 82 out of the 85 primary tumor cases. The median follow-up period after surgery was 27 months (range, 3–291 months).

This study was conducted in accordance with the principles embodied in the Declaration of Helsinki. The study was also approved by the Ethics Committee of Kyushu University (No. 21–137) and conducted according to the Ethical Guidelines for Epidemiological Research enacted by the Japanese Government.

Immunohistochemical staining was carried out in the same way as described previously (10). Sections were pretreated in a microwave oven at 100°C for 20 minutes before being incubated with monoclonal antibodies of p14^ARF (DCS-240, 1:1,000; Sigma-Aldrich), p15^INK4b (15P06, 1:50; Abcam), and p16^INK4a (F-12, 1:100; Santa Cruz Biotechnology) at 4°C overnight. Normal tonsil tissue or colon cancer tissue was used as an external positive control. As a negative control, the primary antibody was omitted. Immunohistochemical results were judged by 3 investigators (M.E., N.S., and Y.O.), who were blinded to the clinical status of the patients. A consensus judgment was adopted as the proper immunohistochemical result based on the proportion of nuclear-stained tumor cells. When less than 50% of the tumor cells were stained, the expression was considered decreased by reference to the previously published papers (10, 30, 31). The serial sections were also immunostained with anti–Ki-67 antibody (M 7240, 1:100; Dako) and anti-CD31 antibody (JC70A, 1:20; Dako) using the standard procedure. The Ki-67–labeling index and microvessel density (MVD) were calculated as described previously (32, 33).

Thirty-nine snap-frozen samples from 28 primary and 11 recurrent or metastatic tumors were obtained from the collection of our department. A total of 5 pairs of samples from the primary and the corresponding first recurrent or metastatic tumors were available for the purpose of comparison. Fresh tumor samples were carefully dissected for the tumors to exclude the surrounding normal tissue, and the samples were immediately frozen in liquid nitrogen and stored at −80°C.

To confirm the concordant immunohistochemical results, the expressions of p14^ARF, p15^INK4b, and p16^INK4a protein were evaluated by WB analysis using snap-frozen samples as previously described (34). A total of 20 μg protein from each sample was used, and incubated with either anti-p14^ARF (1:2,000 dilution), -p15^INK4b (1:50), or -16^INK4a (1:200) antibodies. Anti-human actin mouse monoclonal antibody (1:5,000; Millipore) was used as an internal control. A frozen sample of normal skin tissue was used as an external positive control. Protein levels were standardized by actin, which was assigned an arbitrary level of 10, and the expression signal relative to this was taken as the expression value for each sample.

Total RNA was extracted from 39 frozen samples and a cell line as previously described (35). The MPNST cell line YST-1 (36), which expresses CDKN2A and CDKN2B mRNAs, was used as a positive control. Quantitative RT-PCR was carried out and the results were analyzed using predeveloped TaqMan assay reagents (CDKN2A Hs00233365_m1 and CDKN2B Hs00793225_m1 from Applied Biosystems, Life Technologies) and an ABI Prism 7700 Sequence Detection system (Applied Biosystems) as described previously (37). The standard curve was constructed with serial dilutions of YST-1 cell line samples. The obtained data were standardized by using data of the housekeeping gene, GAPDH (Hs99999905_m1). The final numerical value (V) in each sample was calculated as follows: V = CDKN2A or CDKN2B mRNA value/GAPDH mRNA value.

Genomic DNA was isolated from 27 snap-frozen tumor materials by using standard proteinase K digestion and phenol/chloroform extraction as previously described (38).

Multiplex real-time PCR was carried out for detecting homozygous deletions (HD) of the CDKN2A gene. The primer sequences and PCR conditions were the same as those described previously (39). All primers, the TaqMan probes and the TaqMan Universal Master Mix were supplied by Applied Biosystems. A standard curve was constructed with serial dilutions of Human Genomic DNA (Clontech). Distilled water was used as a negative control in each PCR analysis. The CDKN2A/beta-actin ratios less than 40% were judged to represent HD according to the previous reports (39, 40).

Mutational analysis was carried out for exons 1 to 3 of the CDKN2A gene by PCR-SSCP and DNA sequencing. The primer sequences and PCR conditions were the same as those previously described (9). SSCP and the following DNA sequencing were carried out as described previously (41).

Methylation of the CpG island in the promoter region of the p14^ARF and p16^INK4a genes was determined by MSP. Bisulfite modification and MSP were carried out as described previously, using the same reagents, primers, and PCR conditions (9, 42).

The correlations between 2 dichotomous variables were evaluated by using either χ² test or, when appropriate, Fisher's exact test. The correlations among gene alteration, mRNA level, and protein expression were evaluated by Mann–Whitney U test. The difference in the proportion of p14^ARF-, p15^INK4b-, or p16^INK4a-positive cells between the primary and the recurrent tumors was evaluated by Wilcoxon signed-rank test. Survival curves were calculated with the Kaplan–Meier method, and the differences were compared by the log-rank test. Cox proportional hazards regression analysis was carried out to estimate the hazard ratios for positive risk factors for death. Statistical significance was defined as P < 0.05. Data analysis was carried out with the JMP statistical software package (version 8.0.2; SAS Institute Inc.).

The representative examples of immunohistochemical staining are shown in Figure 1, and the correlations between p14^ARF, p15^INK4b, and p16^INK4a immunoreactivity and the clinicopathological variables are summarized in Table 1. Almost all neurofibroma cells showed immunoreactivity for p14^ARF, p15^INK4b, and p16^INK4a, respectively, on the 12 paraffin-embedded specimens including a benign neurofibroma area with MPNST. p14^ARF expression was decreased in 41of the 85 primary tumors (48%) and 62 of the 129 all specimens (48%). For the total group of 129 cases, decreased expression of p14^ARF was frequently observed in the subgroup with a high Ki-67–labeling index (P = 0.043). p15^INK4b expression was decreased in 46 of the 85 primary tumors (54%) and 60 of the 129 all specimens (47%). Decreased expression of p15^INK4b was associated with large tumor size (more than 5 cm) in both the primary tumors (P = 0.015) and all specimens (P = 0.0003). p16^INK4a expression was decreased in 42 of the 85 primary tumors (49%) and 59 of the 129 all specimens (46%). p16^INK4a expression showed no significant correlation with any clinicopathological variables.

As for the difference in the IHC results between NF1-related and sporadic MPNSTs, p14^ARF expression was decreased in 17 of the 33 NF1-related (52%) and 24 of the 52 sporadic cases (46%). p15^INK4b expression was decreased in 19 of the 33 NF1-related (58%) and 27 of the 52 sporadic (52%) cases. p16^INK4a expression was decreased in 18 of the 33 NF1-related (55%) and 24 of the 52 sporadic (46%) MPNSTs. There was no significant difference in p14^ARF, p15^INK4b, and p16^INK4a expressions between NF1-related and sporadic MPNSTs.

A comparative analysis of the immunoreactivities for p14^ARF, p15^INK4b, and p16^INK4a was carried out in the 18 pairs of the primary and the corresponding first recurrent or metastatic tumors. A decrease in the proportion of p14^ARF-, p15^INK4b-, and p16^INK4a-immunoreactive cells was observed in 10 (56%), 12 (67%), and 13 (72%) out of the 18 recurrent or metastatic tumors, respectively. There was no significant difference in positivity between the primary and the corresponding recurrent tumors; however, p16^INK4a immunoreactivity showed a tendency to be lower in the recurrent or metastatic tumors (P = 0.12).

Proteins of sufficient quality and quantity were successfully extracted from 38 out of the 39 frozen tumor specimens. WB analysis was carried out in 38 specimens using anti-p14^ARF, -p15^INK4b, and -p16^INK4a antibodies. The p14^ARF signal was detected in only 4 cases, and 3 of the 4 cases in which the p14^ARF signal was detected on WB also presented preserved expression of this protein on IHC (Table 2). The p15^INK4b signal was observed in only 3 cases, and all 3 cases in which the p15^INK4b signal was detected on WB also presented preserved expression of this protein on IHC (Table 2). WB for the p16^INK4a protein showed a sufficient signal intensity to evaluate the relative expression levels in 34 of the 38 samples (Table 2). The relative expression level versus the actin expression level, which was assigned a value of 10, ranged from 0.0 to 26.9 (median 2.1). As for the comparison of the primary and the corresponding recurrent or metastatic tumors, p16^INK4a expression levels revealed a decrease in 2 of 5 cases, and a subtle decrease in 1 case (Fig. 1G). The expression levels of p16^INK4a by WB corresponded closely to the levels observed by IHC (Table 2).

The expression levels of CDKN2A mRNA (0.00–11.23, median 0.10) were significantly associated with both p14^ARF and p16^INK4a immunohistochemical protein expressions (Table 2, Fig. 2A and B). The expression levels of CDKN2B mRNA (0.00–10.64, median 0.095) were also correlated to p15^INK4b protein expressions (Table 2, Fig. 2C).

Homozygous deletion (HD) of the CDKN2A gene was detected in 9 of 27 cases (33.3%) by multiplex real-time PCR assay. The remaining 18 cases without homozygous deletion were also analyzed by PCR-SSCP and MSP. No mutation was found in exons of CDKN2A by PCR-SSCP and the subsequent DNA sequencing (data not shown). MSP revealed that no methylation occurred in either the p14^ARF or p16^INK4a promoter region (data not shown). CDKN2A gene alterations represented by HD were significantly associated with low CDKN2A mRNA levels (Fig. 2D). All but one of the cases with HD in the CDKN2A gene showed decreased immunoreactivity for both p14^ARF and p16^INK4a.

Univariate prognostic analysis revealed that decreased expressions of p14^ARF and p16^INK4a were significantly associated with decreased probability of overall survival, respectively (Table 3, Fig. 3A and B). Patients with decreased expression of p15^INK4b showed a strong tendency to have poor prognosis (P = 0.051; Table 3, Fig. 3C). Each of the following was associated with poor prognosis: tumor location in the trunk, tumor location in deep tissue, large tumor size (more than 5 cm), high Ki-67–labeling index (more than 30%), high histological grade (grade 2 or more), and advanced AJCC stage (stage II or higher).

A multivariate analysis carried out using variables that were related to poor prognosis in the univariate analysis revealed that tumor location in the trunk, tumor location in deep tissue, and decreased expressions of either p14^ARF or p16^INK4a persisted as independent risk factors for a poor outcome (Table 4).

In terms of p14^ARF and p16^INK4a expressions in the primary tumors, decreased expression of both p14^ARF and p16^INK4a was observed in 28 of the 82 patients (34.1%), whereas preserved expression of both was detected in 29 patients (35.4%; Table 5). A combined prognostic evaluation with grouping by p14^ARF and p16^INK4a expressions revealed that the 5-year overall survival rate of cases with a decrease in the expression of both proteins was 25.8%, whereas that of cases in which the expression of both proteins preserved was 83.3%, with the difference in the 2 values being statistically significant (P < 0.001).

There was an association between the prognosis and the number of inactivations among p14^ARF, p15^INK4b, and p16^INK4a; the cases with 2 or 3 inactivated genes showed a significantly more adverse prognosis than those with only 1 or no inactivated genes (P < 0.001; Table 5, Fig. 3D).

Finally, the prognostic significance of p15^INK4b expression in the absence of p16^INK4a was evaluated. Among the 41 cases with p16^INK4a inactivation, 30 cases (73.2%) showed decreased expression of p15^INK4b, and the remaining 11 cases (26.8%) showed preserved expression of p15^INK4b (Table 5). The 5-year overall survival rate of patients with coinactivation of p15^INK4b and p16^INK4a was 24.8%, whereas that of patients with preserved p15^INK4b was 75.0%, which was significantly higher than that of coinactivated patients (Fig. 3E).

Our univariate prognostic analysis of p14^ARF, p15^INK4b, and p16^INK4a inactivation in MPNSTs revealed that each gene inactivation was significantly or almost significantly associated with poor prognosis. Furthermore, the prognostic importance of p14^ARF and p16^INK4a status in MPNSTs was reconfirmed by the multivariate analysis. A combined prognostic analysis showed that MPNSTs with inactivation of 2 or more genes among p14^ARF, p15^INK4b, and p16^INK4a showed much worse prognosis than those with inactivation of 1 or no genes. The above results indicate that a synergistic effect of the combined deficiency for p14^ARF, p15^INK4b, and p16^INK4a induced a further high-grade malignancy in MPNSTs. The results from this study contribute not only to estimation of the molecular pathophysiology of MPNST, but also to the improvement of clinical management of patients with MPNST through the provision of precise prognostic prediction.

The impact of p15^INK4b inactivation in human cancers has been less clear than that of the p14^ARF and p16^INK4a inactivations (15). But recently, the significance of p15^INK4b inactivation in malignant tumors has got a lot of attention. The loss of CDKN2B encoding p15^INK4b is observed in various types of malignant tumors, but the frequent loss of CDKN2B in human tumors could be explained by its juxtaposition to CDKN2A, in which proximity could lead to frequent codeletion. A recent study using a mouse model has proposed the interesting hypothesis that p15^INK4b functions as a critical tumor suppressor in the absence of p16^INK4a (15). In human tumors, the tumor-suppressive effect of p15^INK4b in the setting of p16^INK4a inactivation has not been investigated well. This is the first report evaluating the prognostic role of p15^INK4b in the absence of p16^INK4a in MPNSTs. Our prognostic analysis of p16^INK4a-inactivated cases showed that p15^INK4b-preserved cases showed significantly better prognosis than p15^INK4b-inactivated cases. This finding indirectly supports the hypothesis that p15^INK4b provides a critical backup function for p16^INK4a in MPNSTs. In addition, it may provide an explanation for the frequent loss of the complete CDKN2B–CDKN2A locus in human high-grade malignancies, including MPNST (15, 43).

Here we revealed the frequency of inactivation of p14^ARF, p15^INK4b, and p16^INK4a in MPNSTs by a large-scale clinicopathological study with multiple methods. Inactivation of p14^ARF, p15^INK4b, and p16^INK4a was observed in 50%, 49%, and 53% of MPNSTs, respectively. In previous studies, the frequency of p14^ARF, p15^INK4b, and p16^INK4a inactivation in MPNSTs has been a controversial issue. Some papers have stated that p16^INK4a inactivation in neurofibroma is a critical process in progression to MPNST, and p16^INK4a is inactivated in almost all MPNSTs (18, 19, 44). On the other hand, some recent papers have indicated that 9p21-related genes including p16^INK4a are inactivated in only half of all cases of MPNST (24, 43, 45). Our data support the latter statement: we found that inactivation of the p14^ARF, p15^INK4b, and p16^INK4a genes was involved in high-grade malignant transformation in about half of the MPNSTs studied.

CDKN2A is known to exhibit more HDs than other recessive cancer genes in various types of cancers (46). This may be because it is adjacent to a large gene desert and therefore HDs are associated with less negative selection (46). Our study showed that HD of CDKN2A was detected in 33.3% of the cases, whereas neither point mutation of CDKN2A nor methylation of the p14^ARF and p16^INK4a promoter regions was detected. In addition to our study, several other authors have investigated gene alterations of CDKN2A in MPNSTs (16–23, 42–44, 47). According to their reports, HD is the predominant cause of inactivation of the 9p21 locus in MPNSTs, and the rate of HD ranges from 7.1% to 56.5% (16–20, 22, 43, 44, 47). Point mutation of CDKN2A has been reported in only 1 MPNST case (17). Methylation of the p14^ARF and p16^INK4a promoter regions has been reported in only 2 cases (7.7% of the investigated cases), respectively (43). Our results are basically consistent with those of the previous reports, and indicate that the main alteration causing CDKN2A gene inactivation is HD rather than point mutation or methylation of the promoter regions in MPNSTs.

NF1-related MPNST has an NF1 gene dysfunction leading to an excessive activation of the Ras signal pathway, whereas sporadic MPNST does not always show an NF1 gene dysfunction (48). The pathogenesis of MPNST is a fascinating subject, because both NF1-related and sporadic MPNSTs show similar histological findings and molecular characteristics despite their difference in genetic background. Our study showed that there was no significant difference in the expressions of p14^ARF, p15^INK4b, and p16^INK4a between NF1-related and sporadic MPNSTs. Perrone and colleagues reported that inactivation of 1 or more 9p21-related genes detected by RT-PCR was found in 11 of 14 NF1-related (78%) and 9 of 12 sporadic (75%) cases, respectively (43). They concluded that the inactivation profile of 9p21-related genes showed no significant difference between NF1-related and sporadic MPNSTs. Our findings were basically consistent with those of this previous report.

In summary, we carried out a comprehensive analysis of p14^ARF, p15^INK4b, and p16^INK4a expressions in MPNSTs. Inactivation of p14^ARF, p15^INK4b, and p16^INK4a was associated with poor prognosis, respectively, and a combined inactivation of them leaded to worse prognosis in MPNSTs. We conclude that a combined evaluation of p14^ARF, p15^INK4b, and p16^INK4a status provides useful prognostic information for patients with MPNSTs.

The authors appreciate for the technical supports from Ms. Tateishi and the Research Support Center, Graduate School of Medical Sciences, Kyushu University. They also appreciate KN International for revising the English used in this article.

Y. Oda is supported by a Grant-in-Aid for Scientific Research (B; grant no. 21390107) from the Japan Society for the Promotion of Science, Tokyo, Japan.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.