Inactivation of the p16 (INK4A) Tumor-suppressor Gene in Pancreatic ...

(CANCER RESEARCH 5S. 4740-4744.

October 15. 1998]

Inactivation of the p16 (INK4A) Tumor-suppressor Gene in Pancreatic Duct Lesions: Loss of Intranuclear Expression1 Robb E. Wilentz, Joseph Geradts,2 Robert Maynard, G. Johan A. Offerhaus, Myungsa Kang, Michael Goggins, Charles J. Yeo, Scott E. Kern, and Ralph H. Hruban3 Departments of Palholoxy IK. E. W., M. G.. S. E. K.. K. H. H.I Oncology ¡C.J. Y.. S. E. K.. K. H. H.¡.and Surgen ¡C.J. Y.I, The Johns Hopkins University School of Medicine, Baltimore. Man/and 212X7: Department of Pathology ami Laboratory Medicine. University of North Carolina School of Medicine. Chapel Hill, North Carolina 27599 ¡J.C.. R. M. I; Department iif Pathology. Academic Medical Center, University of Amsterdam, Amsterdam, the Netherlands 1105 AZ [G. J. A. O.}; and The Johns Hopkins University School of Public Health. Baltimore. Maryland 21287 ¡G.J. A. O.. M. K.)

suppressor I) tumor-suppressor gene are particularly common in infiltrating pancreatic adenocarcinomas. This gene is inactivated in over 95% of invasive pancreatic cancers. Inactivation occurs by intragenic mutation in one alÃ-elecoupled with loss of the other alÃ-ele (loss of heterozygosity), by deletion of both alÃ-eles(homozygous deletion), or by hypermethylation of a CpG island in the pl6 promoter associated with transcriptional silencing (12-21).

ABSTRACT Pancreatic adenocarcinoma develops from histologically identifiable intraductal lesions that undergo a series of architectural, cytological, and genetic changes. Limited genetic evidence recently suggested that the p!6 gene plays a role in the progression of these "duct lesions." Duct lesions were identified in pancreata from 33 pancreaticoduodenectomies per formed for infiltrating adenocarcinoma. All of these infiltrating adenocarcinomas were previously shown to contain alterations in the piò gene or its promoter. Monoclonal and polyclonal anti-pl6 antibodies were used for histológica! immunodetection. One hundred twenty-six duct lesions were identified. Nine (30%) of 30 flat, 4 (27%) of 15 papillary, 37 (55%) of 67 papillary with atypia, and 10 (71%) of 14 carcinoma in \itn duct lesions showed loss of pió expression. These included 30% of the flat lesions versus 53% of the notifiât lesions and 29% of the nonatypical lesions versus 58% of the atypical lesions. For both comparisons, the differences were statistically significant i/' = 0.036 and /' = 0.003, respec tively). Loss of/i/ft expression occurs more frequently, but not exclusively, in higher-grade duct lesions. These data support the hypothesis that pancreatic duct lesions are neoplastic and that they represent the precur sors of infiltrating adenocarcinoma. Immunohistochemical detection of pl6 provides a new technology to study the genetic alterations in and stages of progression of large numbers of morphologically defined pan creatic duct lesions.

Remarkably, pancreatic duct lesions and infiltrating adenocarcino mas show some of the same genetic changes. For example, Moskaluk et al. (22) have identified both K.-ras and pió alterations in pancreatic duct lesions. Inactivating pl6 mutations were found in three (33%) of nine duct lesions microdissected from pancreata with infiltrating adenocarcinomas that harbored pió alterations. Of note, one of the duct lesions with a pl6 alteration was an early flat lesion, and two of the duct lesions contained the same alteration as that seen in the associated infiltrating pancreatic adenocarcinoma. In contrast, 18 (75%) of 24 duct lesions contained activating K-ra.v mutations, and pió alterations were never found in ducts lacking a K-ras mutation

INTRODUCTION Recent evidence suggests that, just as there is progression from adenoma to infiltrating adenocarcinoma in colonie neoplasia (1,2), so too is there a progression series within the pancreas (3-11). In some pancreatic ducts and ductules, a mucinous epithelium with cytological and architectural atypia replaces the normal cuboidal epithelium. Called "pancreatic intraepithelial neoplasias" or "mu cinous hyperplasias," these changes progress from flat duct lesions to papillary duct lesions without atypia to papillary duct lesions with atypia to carcinomas in situ (Fig. 1) (3-11). Some in situ lesions then eventually progress to infiltrating adenocarcinoma (11). Thus, infiltrating pancreatic cancers that are clinically "early" are, in fact, biologically late. Clinically early infiltrating pancreatic cancers are also genetically late. That is, most infiltrating pancreatic cancers have accumulated numerous genetic alterations by the time they come to clinical pres entation. For example, Rozenblum et al. (12) have reported nine separate genetic alterations that occurred during the development of a single infiltrating pancreatic adenocarcinoma. Mutations in the pió (INK4A, CDKN2, MTS1, multiple tumor Received 5/21/98: accepted 8/18/98. 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. ' Supported by N1H Grani CA 629924. ~ Present address: Nufficld Department of Pathology and Bacteriology. University of Oxford. Uniled Kingdom. OX3 9DU. ' To whom requests for reprints should be addressed, at The Johns Hopkins Hospital. Meyer 7-181. 600 North Wolfe Street. Baltimore. MD 21287. Phone: (4101 955-9132; Fax: (410) 955-0115; E-mail: [email protected].

(22). These and other results support the hypotheses that: (a) piò gene inactivation usually occurs after mutation in the K-ras gene; and (b) the association of genetic changes with the histological grade of a duct lesion is not absolute inasmuch as some flat duct lesions without atypia also contain pió genetic changes. The purpose of this study is to examine loss of expression of the pl6 protein in a spectrum of pancreatic duct lesions. Specific anti bodies to pi6 have been shown to label almost exclusively only those cells with wild-type protein: therefore, duct lesions with pl6 inacti vation. regardless of the manner of inactivation, do not label with the antibodies (23). Immunohistochemical detection has two advantages over genetic analysis: (a) tissue morphology is maintained, which allows a direct correlation between gene expression and histology; and (b) it is less labor intensive than molecular analyses, and, therefore, larger num bers of lesions can be examined. Determining the patterns of pió expression in a large number of histologically defined duct lesions will help to establish the role of pl6 inactivation in the development of pancreatic neoplasia.

MATERIALS Thirty-three

AND METHODS surgically resected pancreata harboring an infiltrating adeno

carcinoma of the pancreas were studied. The infiltrating adenocarcinomas from these pancreata had been previously analyzed genetically for ¡>16mutations and immunohistochemically for pl6 protein expression (20. 23). The pancreata represented pancreaticoduodenectomies performed at The Johns Hopkins Hos pital between October 10, 1992, and June 30, 1994. Multiple H&E-stained slides of routinely examined pancreatic tissue sur rounding each of the infiltrating cancers were screened by light microscopy for duct lesions. According to previously described criteria (3-10), each duct lesion was categorized as "flat," "papillary without atypia," "papillary with atypia," or "carcinoma in situ" (see Fig. 1). Unstained 5-p.m sections were then cut from 61 paraffin blocks corresponding

4740

to the slides containing a variety of

l>16 TUMOR-SUPPRESSOR

GENE IN PANCREATIC

DUCT LESIONS

Fig. 1. Histológica!classification of duct lesions of Ihe pancreas. A. flat duel lesion: ß. papillary duct lesion; C. papillary duct lesion with atypia: and D. carcinoma in situ. Atypia is defined by nuclear hyperchromasia and irregularity, loss of cellular polarity, and presence of nucleoli and mitoses.

pancreatic duct lesions. The duct lesions were no longer present in the deeper sections of 6 blocks, leaving 55 sections for analysis. The remaining unstained sections were treated with a monoclonal antibody (clone G175-405. PharMingen) to pl6 under previously described conditions (23). A subset of the lesions was also reacted with the PharMingen polyclonal anti-pi6 antibody (Ref. 24: see next paragraph). In normal pancreatic ducts, only a small number of the ductal epithelial cells show nuclear pl6 immunolocalization exceeding cytoplasmic background (23); therefore, a lesion was considered to show intact ¡>I6expression if any of the ductal epithelial cells contained nuclear pl6 labeling above the cytoplasmic background. Duct le sions were considered negative for antibody detection only when none of the cells in the lesion expressed pió (23). Nonneoplastic fibroblasts, lymphocytes, and endocrine cells served as positive internal controls in each of the sections. Each section was studied first with the monoclonal antibody. If no positive internal controls were obtained with the monoclonal antibody and if the infiltrating adenocarcinoma from the same case had shown appropriate label ing with polyclonal antibody, then a parallel section was labeled with the polyclonal antibody. Specific labeling was obtained in 46 sections (average, 3 duct lesions per section; range. 1-11 duct lesions) from 28 cases with one of these two antibodies (31 sections from 21 cases with monoclonal antibody and 15 sections from 11 cases with polyclonal antibody). The nine sections excluded from the study included (a) those containing no positive controls with either monoclonal or polyclonal antibody: and (h) those containing no positive controls with the monoclonal antibody, wherein the polyclonal antibody had inappropriately labeled the associated invasive adenocarcinoma (23). Three of the authors (R. E. W., J. G., and R. H. H.) separately evaluated the immunohistochemical labeling in each case, and there was disagreement in only one case, which was labeled with polyclonal antibody. A section from this case harbored one atypical papillary lesion, which only one author felt labeled positively with the anti-pl6 antibody. Two authors scored the lesion as having no nuclear labeling above the cytoplasmic background, and this duct lesion, therefore, was considered not to show ¡>I6expression. Statistical differences in this study were analyzed using two-tailed )£and Fisher's exact tests. 4741

RESULTS

Infiltrating Adenocarcinomas. The 33 infiltrating adenocarcinomas had been analyzed previously for/7/6 genetic alterations and for p 16 immunohistochemical labeling (20, 23). Fifteen of the infiltrating adenocarcinomas contained an intragenic mutation in p¡6 coupled with loss of the second alÃ-ele(loss of heterozygosity); 15 showed homozygous deletion of p!6; and 3 displayed hypermethylation of a CpG island in the pl6 promoter, which abrogated pl6 expression (20). Previous immunohistochemical studies of these same infiltrating car cinomas with the monoclonal antibody G175-405 (PharMingen) demonstrated that 32 (97%) of the 33 infiltrating carcinomas did not express detectable pl6 (23). The single case with detectable pl6 was a cancer with an intragenic missense mutation at codon 15 and loss of heterozygosity. Duct lesions were no longer present in the deeper sections from this case's block, and. therefore, all duct lesions in this study were associated with adenocarcinomas showing loss of plf> expression with the monoclonal antibody. In contrast, immunohistochemical labeling of these same infiltrat ing carcinomas with the rabbit polyclonal anti-pi6 antibody from PharMingen was less specific. It labeled 7 (47%) of the 15 tumors with an intragenic mutation and loss of heterozygosity. 1 (7%) of the 15 with homozygous deletion, and 1 (33%) of the three with hyper methylation of the p¡6 promoter (23). With both antibodies, only nuclear labeling correlated with the molecular analysis; cytoplasmic background labeling tended to be high but was nonspecific and therefore ignored. Duct Lesions. Forty-six sections from 28 pancreata were evaluated for nuclear labeling with monoclonal or polyclonal antibodies against the pl6 protein. One hundred twenty-six duct lesions were identified

pin TUMOR-SUPPRESSOR

Table I Lass of pl6 t:\pressitw in the f^: •¿â€¢.

»

B Fig. 3. A and B, loss of nuclear pl6 protein expression in an atypical papillary lesion. In contrast, the lymphocytes around and within the duct express p 16.

pif, TUMOR-SUPPRESSOR

GENE IN PANCREATIC

DUCT LESIONS

t

•¿ ¿fe. ,- X&&

B Fig. 4. A duct lesion demonstrating a progression from flat mucinous hyperplasia. with intact nuclear pl6 protein expression (lop. A), to atypical papillary hyperplasia. lacking nuclear pl6 (/?). An invasive adenocarcinoma near the duct did not label with the antibody (bottom. A).

kind of genetic event in the associated adenocarcinomas (two-tailed X2 test, P = 0.884; see Table 2). Thirteen (48%) of 27, 7 (54%) of 13,

lions of both alÃ-eles(homozygous deletion), whereas in the remaining 15%, hypermethylation of a CpG island in the pió promoter is associated with a lack of gene transcription (12-21). and 40 (47%) of 86 duct lesions associated with carcinomas with intragenic mutations, hypermethylations, and homozygous deletions, Many of the same genetic changes can be seen in intraductal respectively, showed loss of pl6 expression. epithelial proliferations in the pancreas (22). These duct lesions seem to progress through a sequence of stages as assessed by architectural and cellular changes: flat, papillary, atypical papillary, and carcinoma DISCUSSION in situ (3, 4, 6-11 ; see Fig. 1). Moskaluk et al. (22) recently identified Pancreatic adenocarcinoma is a genetic disease. For example, ap piò gene alterations in a portion of these duct lesions, one of which proximately 95% of pancreatic adenocarcinomas show inactivation of was a flat lesion. These data help establish that the duct lesions are, in p 16 gene transcription (12-21). In 40%, the inactivation occurs fact, early neoplasms and suggest that they are the precursors to through mutation in one alÃ-elecoupled with loss of the other alÃ-ele infiltrating adenocarcinoma. (loss of heterozygosity). Another 40% of the carcinomas have deleOur data not only support but also extend these findings because of the following: (a) most obviously, because it is technically less cumbersome and Table 2 Histologie grade of und pió expression in duct lesions versus class of genetic more widely available than genetic assays, immunohistochemical event in associated adenocarcinonias labeling of the pi6 protein provides an opportunity to correlate lesionsGenetic Number of duct lesions with lost piò total number of duct specific morphological and genetic changes in a large number of event in cases; adenocarcinomaInlragenic papillary9/14 (b) the trend for the histologically bland lesions to express p 16, and mutation for the more cytologically and architecturally atypical lesions to lose Hypermethylation 0/16/219/30Papillary1/5 0/0 6/1122/4237/67CIS"0/01/19/1310/14Total13/27 7/13 40/8660/126 expression, suggests that pió can play a role in the progression of deletionTotalFlat3/8 Homozygous 3/104/15Atypical pancreatic neoplasia. However, as Moskaluk et cil. originally docu " CIS, carcinoma in situ. mented (22) and as we have confirmed, the progression model is not 4743

TUMOR-SUPPRESSOR

GENE IN PANCREATIC

absolute. Some duct lesions without cytological or architectural atypia showed loss of p 16 expression, whereas some duct lesions with cytological and architectural atypia showed intact p 16 expression. That is, loss of pl6 expression can occur at any point along the histológica! progression model. These data raise the question as to whether histológica! assessment or genetic alterations are the better measure of the neoplastic potential of a lesion. Comparing the pl6 expression pattern of incidentally identified duct lesions (for example, from pancreata from patients with chronic pancreatitis) with that of adenocarcinoma-associated duct lesions may help resolve this di lemma. Both histology and genetics may prove to be significant in the evaluation of a lesion's potential: (c) a corollary to the dynamic nature of the progression model is that pl6 is not a "gatekeeper" in pancreatic cancer. That is, because pl6 inactivation can occur anywhere along the histológica! progres sion model, it cannot be responsible for initiating tumor formation (25). For the same reason, it is also unlikely that the ¡>16 gene serves as a final checkpoint, one that when abrogated would directly allow the process of invasion. Even so. 4 (29%) of 14 in situ carcinomas showed piò expression, whereas their associated invasive carcinomas did not: and (16 gene, to achieve a critical threshold that leads to an invasive phenotype. The diffuse and variable nature of histológica! and genetic changes in the pancreas may represent a "field effect." in which endogenous (e.g., pancreatic enzymes) or exogenous (e.g., carcinogens from tobacco) agents act on multiple ducts over a long period of time. Of course, immunohistochemistry has some disadvantages: («)absolute correlation between a duct lesion and its associated infiltrating adcnocarcinoma is not possible inasmuch as this requires detection of the same specific genetic alteration in both lesions, which is possible only through molecular genetic study. Tracing out these specific genetic events within duct lesions of different stages and their associated adenocarcinomas would be very valuable because it could help establish which genetic alterations in the pió gene (intragenic mutations, homo/.ygous deletions, or hypermethylations) most often lead to neoplastic progression: and (h) because only a small minority of normal ductal epithelial cells label for pl6 protein, distinction between negative and positive lesions is certainly not straightforward. This difficulty in interpretation is compounded often by nonoptimal labeling with a significant cytoplasmic background reactivity (23). Obviously, the progression model of pancreatic neoplasia is still under construction. However, it is clear that p]6 genetic alterations arc important in the progression of pancreatic cancer and that immu nohistochemistry provides an effective and convenient means to de termine pl6 expression in a large number of cases. The evaluation of the potential of a duct lesion in the pancreas to progress to invasive cancer may depend on both histológica! and ¡mmunohistochemical parameters. ACKNOWLEDGMENTS We thank Michele Heftier for her hard work and dedication Graves for the artwork in Figure 1.

and Tom

4744

DUCT LESIONS

REFERENCES 1. Vogelstcin. B.. and Kinzler. K. W. The multistep nature of cancer. Trends Genet.. 9: 138-141, 1993. 2. Fearon. E. R.. and Vogelstein. B. A genetic model for coloréela!tumorigenesis. Cell, 61: 759-767. 1990. i. DiGiuseppc. J. A., Yeo, C. !.. and Hruban. R. H. Molecular biology and the diagnosis and treatment of adenocarcinoma of the pancreas. Adv. Anatomic Pathol.. 3: 139-155. 1996. 4. Hruban. R. H.. and DiGiuseppe. J. A. K-ms mutations in pancreatic ductal proliferative lesions: author's reply. Am. J. Pathol., 145: 1548-1550, 1994. 5. Cubilla. A. L., and Fitzgerald. P. J. Morphological lesions associated with human primary invasive nonendocrine pancreas cancer. Cancer Res. 36: 2690-2698. 1976. 6. Kozuka. S., Sassa, R.. Taki. T., Masamoto, K.. Nagasawa. S.. Saga. S.. Hasegawa, K., and Takeuchi. M. Relation of pancreatic duct hyperplasia lo carcinoma. Cancer (Phila.), 43: 1418-1428. 1979. 7. Pour, P. M.. Sayed. S.. and Sayed, G. Hyperplastic. preneoplastic. and neoplastic lesions found in 83 human pancreases. Am. J. Clin. Pathol.. 77: 137-152, 1982. 8. Furukawa. T.. Chiba. R.. Kobari. M.. Matsuno. S.. Nagura. H.. and Takahashi, T. Varying grades of epithelial atypia in the pancreatic ducts of humans: classification based on morphometry and multivariate analysis and correlated with positive reac tions of carcinoembryonic antigen. Arch. Pathol. Lab. Med.. i IH: 227-234, 1994. 9. Wilentz, R. E.. and Hruban. R. H. Pathology of cancer of the pancreas. Surg. Oncol. Clin. N. Am., 7: 43-65, 1998. 10. Hruban. R. H., and Wilentz. R. E. Pancreas. In: N. Weidner. R. J. Cote. S. Susler, and L. M. Weiss (eds.). Modem Surgical Pathology. Philadelphia: W. B. Saunders. in press. 11. Brat. D. J., Lillemoe, K. D.. Yeo. C. J.. Warfield, P. B., and Hruban. R. H. Progression of pancreatic intraductal neoplasias (high-grade PanIN) to infiltrating adenocarci noma of the pancreas. Am. J. Surg. Pathol.. 22: 163-169, 1998. 12. Rozenblum. E.. Schutte, M.. Goggins. M.. Hahn, S. A.. Panzer. S.. Zahurak, M.. Goodman. S. N., Sohn. T. A.. Hruban. R. H., Yeo, C. J., and Kern, S. E. Tumorsuppressive pathways in pancreatic carcinoma. Cancer Res., 57: 1731-1734. 1997. 13. Caldas. C., Hahn. S. A., da Costa, L. T.. Redston. M. S., Schutte. M.. Seymour. A. B., Weinslein, C. L.. Hruban. R. H.. Yeo. C. J.. and Kern. S. E. Frequent somatic mutations and homozygous deletions of the pI6 (MTSI) gene in pancreatic adeno carcinoma. Nal. Genet., 8: 27-32. 1994. 14. Huang. L.. Goodrow, T. L.. Zhang. S.. Klcin-Szanto. A. J. P.. Chang. H., and Ruggeri. B. A. Deletion and mutation analyses of the [>16/MTS-Ì tumor suppressor gene in human duclal pancreatic cancer reveals a higher frequency of abnormalities in tumor-derived cell lines than in primary ductal adenocarcinomas. Cancer Res.. 56: 1137-1141, 1996. 15. Hahn, S. A., Seymour, A. B., Hoque, A. T. M. S., Schutle, M., da Costa, L. T., Redston, M. S.. Caldas, C.. Weinstein, C. L., Fischer, A., Yeo. C. J., Hruban. R. H., and Kern. S. E. Allelotype of pancreatic adenocarcinoma using xenograft enrichment. Cancer Res., 55: 4670-4675. 1995. 16. Seymour. A.. Hruban, R. H.. Redsion. M., Caldas. C. Powell, S. M.. Kinzler, K. W., Yeo, C. J.. and Kern. S. E. Allelotype of pancreatic adenocarcinoma. Cancer Res.. 54: 2761-2764. 1994. 17. Bartsch, D.. Shevlin, D. W., Tung. W. S., Kisker, O.. Wells, S. A., and Goodfellow. P. J. Frequent mutations of t'DKN2 in primary pancreatic adenocarcinomas. Genes Chromosomes Cancer. 14: 189-195, 1995. 18. Naumann, M., Savitskaia. N.. Eilert. C., Schramm, A.. Kalthoff, H., and Schmiege!, W. Frequent codeletion of pl6/MTSl and ¡il5/MTS2 and genetic alterations in ¡y/6/MTSI in pancreatic tumors. Gastroenterology, 110: 1215-1224. 1996. 19. Merlo. A., Herman. J. G., Mao. L., Lee. D. J.. Gabrielson. E., Burger, P. C., Baylin, S. B.. and Sidransky. D. 5' CpG island methylation is associated with transcriptional

20.

21. 22.

23.

24.

silencing of the tumor-suppressor pI6/CDKN2/MTSl in human cancers. Nat. Med., I: 686-692, 1995. Schutle. M.. Hruban. R. H., Geradts. J.. Maynard. R.. Hilgers. W., Rabindran, S. K., Moskaluk. C. A.. Hahn. S. A.. Schwarte-Waldhoff. I.. Schmiegel, W.. Baylin, S. B., Kern. S. E., and Herman, J. G. Abrogation of the Rb/pl6 tumor-suppressive pathway in virtually all pancreatic carcinomas. Cancer Res.. 57: 3126-3130, 1997. Hruban, R. H., Petersen, G. M.. Ha. P. K.. and Kem. S. E. Genetics of pancreatic cancer: from genes to families. Surg. Oncol. Clin. N. Am.. 7: 1-23. 1998. Moskaluk. C. A., Hruban. R. H.. and Kern. S. E. piò and K-r