p16^INK4a is a Prognostic Marker in Resected Ductal Pancreatic Cancer

Pancreatic cancer (PC) is a disease with a dismal prognosis and an overall 5-year survival rate of less than 5%. Only 14% of all PCs could be appropriately treated by resection according to the 1995 U.S. National Cancer Data Base Report. ¹ Five-year survival rates range from 6.8% to 25% in specialized centers. Also, 50% who survive 5 years may die of recurrent cancer. ² Pancreatic surgery is expensive, and Gudjonsson ³ concluded that pancreatic resection for PC is a waste of resources. However, a 5-year survival rate of up to 41% in a selected group of patients with small cancers was reported. ⁴ Thus, selection criteria and markers are needed to identify patients who might benefit from surgery and adjuvant therapeutic strategies.

There is convincing evidence that the molecular status of each neoplasm is related to its biologic behavior. The development from precancerous lesions to invasive metastasizing cancer is accompanied by a stepwise acquisition of sequential mutations that could be shown in other cancers. ^5,6 Thus, genes known to be altered in PC are potential candidates for prognostic markers. Enormous progress has been made deciphering the genetic basis of the tumorigenesis of PC in the past decade. Alterations in genes controlling the G1/S-phase cell cycle transition were found to be one of the dominant mechanisms for the development of PC. ⁷ Progression of the cell from G1 to S phase is irreversibly initiated by cyclin-dependent kinase (CDK)-mediated phosphorylation of the retinoblastoma (Rb) gene product, which releases three members of the E2F family. ⁸ E2F-responsive genes are required for progression of the cell through the G1-to-S transition. The kinase activity of CDK4-cyclin D is negatively regulated by p16^INK4a and p21^CIP1. Induction of p21^CIP1 is one mechanism by which p53 can lead to Rb-mediated cell cycle arrest. p53 activity is regulated by the protooncogene murine double minus 2 (MDM2) protein P90. ⁹ In a negative feedback loop, the activation of p53 leads to upregulation of MDM2, which itself can bind directly to the p53 transactivation domain (N-terminal), downregulate p53 transcriptional activity, and target p53 for ubiquitin-mediated proteosomal degradation. In addition, MDM2 protein binds to the carboxy terminus of Rb, relieving Rb suppression of E2F transactivation and thus permitting the cell to enter S phase. Recently it was shown that the second product of the CDKN2 gene, p14^ARF (the human homologue of p19^ARF), binds to the carboxy terminus of MDM2 and regulates its actions. ^10,11

It has been shown that p16^INK4a, p53, and Rb, as major players of this complex G1/S phase network, are mutated or deleted in PC. The tumor suppressor p16^INK4a is altered in 27% to 95%¹² of PCs, p53 in 40% to 70%, ¹³ and Rb in about 5%. ¹² The role of MDM2 is not yet clearly defined in PC. ¹⁴ The prognostic value of these cell cycle regulator genes has not yet been clarified. Therefore, we evaluated for the first time by an extreme group analysis ¹⁵ of patients with PC whether alterations of these genes involved at different levels of the G1/S phase checkpoint have prognostic value in resected PC. We hypothesized that the frequency of genetic alterations for p16^INK4a and p53 would be higher in the extreme group of short-term survivors (STS) compared with the extreme group of long-term survivors (LTS), even with stratification for stage.

METHODS

Pancreatic resections for ductal carcinoma were performed in 92 patients from September 1987 to August 1998 in the Department of General Surgery at the University of Marburg. Actual follow-up was obtained by outpatient clinic visits or by the primary physician in September 2000.

We excluded patients with positive resection margins and distant metastases, as well as patients who died of complications of the surgical procedure. We also excluded patients who received adjuvant therapy because of possible confounding. ¹⁶ Altogether, 30 of the 92 patients were excluded from the study because no tumor material was available or one of the above exclusion criteria was fulfilled. Thus, 62 patients comprised the extreme group analysis. Mean follow-up was 19.6 ± 21.2 months. Among the 62 patients, tumors of the 20 patients with the shortest survival (STS) and the 20 patients with the longest survival after resection (LTS) were analyzed by immunohistochemistry, and molecular analyses were performed. Eight of the STS group were men and 12 were women, compared with 14 men and 6 women in the LTS group. Informed consent was obtained from patients or their relatives for this investigation. The research protocol was approved by the local ethics committee.

RESULTS

Mean survival of all 62 patients was 19.6 ± 21.2 months, and the median survival was 13 months (range 3–152). Median survival among the STS patients was 6 months (range 3–10); among the LTS patients it was 36 months (range 20–152;P < .001). Univariate analysis did not show significant differences in age at resection, tumor size (P > .05), or UICC stage (P = .247) between the two groups (Table 1).

DISCUSSION

Because we know that tumor stage ²⁸ and tumor size ²⁹ influence the prognosis of patients after resection of ductal PC, we analyzed these data in the tumors of 40 patients in an extreme group analysis. No difference (P > .05) was identified between 20 STS and 20 LTS in terms of tumor size. This finding confirms the results of a study performed at the Johns Hopkins Hospital ³⁰ but should be taken with some caution because of the small number of tumors examined. A tendency toward advanced tumor stage was seen among the STS versus the LTS groups; however, no significant difference was identified (P = .247). To exclude the possibility that differences in frequencies of molecular genetic alterations in STS and LTS could be attributed to differences in tumor stage, hypothesis tests were adjusted for UICC stage.

Several potential prognostic molecular markers have been proposed for PC, including Ki-67, ³¹ TGF-β 1, ³² CD44 variant 6, ³³ cathepsin-B, ³⁴ and Bax. ³⁵ Coppola ³⁶ recently reviewed the molecular alterations with their prognostic significance identified to date in PC; he found that not all data were conclusive. The present study evaluated for the first time the possible prognostic significance of important control genes of the G1/S cell cycle checkpoint in resected PC by an extreme group analysis.

First, p16^INK4a alterations were evaluated. Previously, it was hypothesized that alterations of the p16^INK4a gene indicate a worse prognosis in different cancers, ³⁷ including PC. ³⁸ Bartsch et al ³⁸ found reduced survival in patients with p16^INK4a mutations. However, that study had the disadvantage that one of the important mechanisms of p16 inactivation, homozygous deletions, was not examined. Hu et al ¹⁸ found longer survival in patients with immunohistochemically p16-positive tumors compared with p16-negative ones in a Kaplan-Meier survival analysis, but this difference was not significant. Naka et al ³⁹ identified a significantly longer survival in patients with immunohistochemically p16-positive tumors, but these results did not include the analysis of gene mutations and gene promoter methylations, nor were they adjusted for tumor stage. In the present study, all known mechanisms of p16 inactivation were analyzed, and the results were adjusted for tumor stage. Overall, we found p16^INK4a alterations in 27 of the 40 examined tumors (67.5%), which is in line with the reports in the literature. ¹² In the 20 STS tumors, p16 inactivation was identified in 17 (85%). Four of these 17 alterations (24%) were found to be caused by intragenic mutations of the coding sequence of the gene. One of these mutations was an 11-bp deletion, two mutations were nonsense mutations, and one was a 1-bp insertion. It is known from the literature that SSCV detects only 70% of mutations. ⁴⁰ Thus, the loss of p16 expression in seven tumors without detected mutations or detection of promoter hypermethylation might be due to undetected mutations in SSCV. More likely, the loss of p16 expression in these tumors is caused by homozygous deletion of the gene, which occurs in up to 40% of PCs. ³⁰ Given the currently available mutation detection systems, we could not detect homozygous deletions in tumor DNA samples with 40% to 80% neoplastic cellularity. Promoter hypermethylation could be identified in 6 PCs (30%), slightly higher than the 18% identified by Ueki et al. ⁴¹ In the 20 LTS tumors, loss of p16 expression could be identified in only 10, 1 caused by a missense mutation in codon 147 and 5 caused by promoter hypermethylation. The remaining four might be caused by undetected mutations or by homozygous deletions of the gene. In summary of the p16^INK4a data, a higher frequency of p16^INK4a alterations was found in the STS tumors compared to the LTS tumors (adjusted P = .043), indicating that p16^INK4a is an inverse marker for prognosis in PC.

The second gene analyzed was p53, an important regulator of G1/S transition and apoptosis. p53 is also a frequently altered tumor suppressor in PC, with mutations of its gene often being accompanied by nuclear immunoreactivity resulting from stabilization of the p53 protein. However, Lundin et al ⁴² found no association between the level of p53 immunoreactivity and survival in either univariate or multivariate analysis. In addition, p53 overexpression did not significantly influence the survival of patients with PC in the study by Naka et al, ³⁹ whereas DiGiuseppe et al ⁴³ found a tendency toward better survival in tumors without p53 mutations. Overall, the present study identified p53 alterations in 25 of 40 PCs (62.5%), completely in agreement with the literature. ¹³ Tumors from the STS group revealed p53 overexpression in 13 patients, compared with 10 tumors in the LTS group. In addition, truncating mutations with loss of the gene product were identified in the tumor DNA of one STS tumor and one LTS tumor in our study. ¹² We identified 10 missense mutations and 1 nonsense mutation as well as one insertion and two deletions. Nine tumors overexpressed p53, but we did not identify the underlying genetic defect. This is not surprising because it is known that the correlation between immunohistochemically detectable p53 overexpression and the presence of p53 mutations is not absolute;⁴⁴ in addition, we may not have detected a few mutations because of the limitations of the SSCV method. ⁴⁰p53 alterations were identified in 14 of 20 STS tumors (70%) compared with 11 of 20 LTS tumors (55%). This difference was not significant (adjusted P = .686). Therefore, p53 could not be identified as a prognostic marker for resected PC from this study. This may be due to the limited power of this study. Our study had a power of 10% to show that the STS tumors had more p53 alterations than the LTS tumors using the frequencies observed in this study at a global significance level of 5%, even without stratification according to UICC stage.

Overexpression of the MDM2 protein has been observed in several human malignancies, especially sarcomas, and has been associated with an unfavorable prognosis. ⁴⁵ Mutational inactivation of Rb has been described in a wide spectrum of malignancies, including retinoblastomas, small cell lung carcinomas, and bladder and breast carcinomas. ⁴⁶MDM2 and Rb have also been proposed to play a role in PC. Chang et al ⁴⁷ found stabilization of MDM2 expression in 26% and loss of Rb expression in 42% of hamster pancreatic ductal adenocarcinomas. These results could not be confirmed in human PCs by Ruggeri et al, ¹⁴ who did not find overexpression of MDM2 by immunohistochemistry in PCs. Huang et al ¹² found a complete lack of Rb expression in only 3 of 62 PCs (4.8%). A recent study reported that MDM2 protein isoforms are overexpressed and coexist with the mutant p53 protein in lung carcinomas. ⁴⁸ However, such an association could not be confirmed in the MDM2-overexpressed tumors in this study. The present study found overexpression of the MDM2 or loss of the Rb gene product in only three and two tumors in the STS group and in two and one tumors in the LTS group, respectively.

Tumors that did not reveal alterations of p16^INK4a, p53, MDM2, or Rb (44, 58, 60, 62) belonged exclusively to the LTS group, with a survival of 21, 44, 45, and 152 months, respectively. However, there was one tumor (55) that carried alterations of both p16^INK4a and p53 and showed overexpression of MDM2 as well as lack of Rb expression. This patient also belonged to the LTS group, with a survival of 37 months. Thus, these data are not conclusive and need to be evaluated in a large-scale study.

In summary, p16^INK4a alterations indicate a worse prognosis in resected PCs. p53 alterations, MDM2 overexpression, and loss of Rb expression, however, could not be identified as prognostic markers in PC from this study and presumably play, besides p53, only a minor role in the tumorigenesis of PC. These data might have further implications. First, patients with p16^INK4a mutated tumors are probably good candidates for adjuvant radiochemotherapy regimens because of their poor prognosis after potentially curative resection. Second, tumors in these patients might be an ideal target for gene therapy strategies in the future, especially because it has been shown that reconstitution of p16 has marked effects on the proliferative capacity of tumor cell lines in vitro and in vivo. For example, when the p16-deficient pancreatic cancer cell line MIAPaCa2 was transfected with an adenovirus expressing p16 (Adp16), it stopped growing. ⁴⁹ Pathologic evaluation of Adp16-treated tumors showed dose-dependent necrosis and apoptosis in prostate cancers. ⁵⁰ Further, combined delivery of p16^INK4a and p53 genes by a simultaneous adenoviral transfer led to inhibition of tumor growth in nude mice and might be more efficient than the replacement of p53 or p16 alone. ⁵¹ Thus, replacement of p16 eventually combined with p53 might be the goal of future adjuvant therapy in patients with PC who have p16^INK4a-mutated tumors.

Acknowledgments

The authors thank B. Chaloupka, Y. Deiss, H. Combé, and A. Holzbach for performing a number of experiments, U. Danner, G. Acker, and Ch. Erbe for collecting clinical data, and Dr. M. Stöppler for linguistic support.