Overexpression of p21^WAF1/CIP1 is an Early Event in the Development of Pancreatic Intraepithelial Neoplasia¹

Ductal adenocarcinoma of the pancreas is the fifth leading cause of cancer-related death in Western societies (1). Early identification of malignant lesions of the breast, colon, and cervix has lead to valuable insights into carcinogenesis and improved treatment and outcome for those conditions (2, 3). PC³ is normally clinically advanced at the time of presentation (4), and, as a result, the mortality from this cancer almost parallels its incidence, with a 5-year survival of <10%. At present there are no adequate means of detecting PC at an early stage. Earlier detection of PC may improve patient outcome using current treatment modalities, in particular, operative resection. Defining the molecular pathology and natural history of early neoplastic lesions within pancreatic ducts will lead to a better understanding of pancreatic ductal carcinogenesis and may provide scope for new diagnostic and treatment strategies.

Hyperplastic and dysplastic epithelial lesions of the pancreatic ducts were described in association with PC as early as 1924 (5). In 1954, examination of 141 cases of pancreatic carcinoma at autopsy showed an incidence of hyperplastic ductal lesions in 41% of routinely sampled noninvolved portions of pancreata containing a carcinoma, in 9% of 150 autopsy specimens of nonmalignant pancreata and, interestingly, in 28 of 100 pancreata of patients with diabetes mellitus (6). The association of these duct lesions with ductal adenocarcinoma has also been reported in more recent studies where the presence of severely dysplastic duct lesions was more strongly associated with invasive carcinoma than those with mild/moderate or no atypia (7, 8). Case reports have documented patients with resected pancreata, where dysplastic duct lesions were identified at the resection margins of the specimen. These patients developed cancers at the site of resection many months to years after initial resection (9, 10). This circumstantial evidence is similar to that of the adenoma to carcinoma sequence of colorectal polyps that facilitated Vogelstein’s description of the molecular pathology of colon cancer development (11). Knowledge of the molecular pathology of pancreatic duct lesions may provide a molecular basis for the morphological changes seen in the proposed progression model of PC (12).

The current progression model for PC development within pancreatic ducts defines progression from normal cuboidal epithelium through a series of lesions termed PanIN (12, 13, 14, 15, 16). Normal epithelium develops into PanIN-1A, previously described as flat hyperplasia (tall columnar cells with some crowding), to PanIN-1B or “ductal papillary hyperplasia” (increased crowding of columnar cells with papillary projections generally containing a stromal core), to PanIN-2 (ductal papillary hyperplasia with atypia), which is characterized by similar architectural characteristics to PanIN-1B but with mild to moderate nuclear atypia (some nuclear stratification and pleomorphism, occasional mitoses, and visible nucleoli). All three lesions may have varying degrees of mucinous cell hypertrophy. PanIN-3, or atypical ductal hyperplasia, demonstrates severe atypia, has in the past been called carcinoma in situ and is likely to progress to invasive carcinoma (12).

Dysregulation of the normal cell cycle machinery is integral to the neoplastic process, and now there is compelling evidence implicating the loss of cell cycle control in the development and progression of most human cancers (17). Abnormalities in the retinoblastoma pathway that controls G₁ to S-phase progression in the cell cycle and the TGF-β signaling pathway, which inhibits pancreatic ductal epithelial cell proliferation, have been reported at high frequencies in pancreatic ductal adenocarcinomas (18). p21^WAF1/CIP1 (CDKN1A) acts to prevent pRb phosphorylation by inhibiting activation of cyclin E/cdk2 complexes that are required for pRb phosphorylation (17). p21^WAF1/CIP1 also acts as an assembly factor for cyclin D1/cdk4 complexes at low concentrations and is required for the cyclin D1/cdk4 enzymatic activity necessary for pRb phosphorylation (19). Although putative mutations of p21^WAF1/CIP1 have been reported in PC, these are likely to represent polymorphisms (20). Differential expression of p21^WAF1/CIP1 has been reported to associate with prognosis in esophageal (21) and superficial bladder cancer (22). The incidence of p21^WAF1/CIP1 overexpression in PC was first reported as 43%, where it was not related to p53 expression (23). There are, however, no published data documenting p21^WAF1/CIP1 expression in PanIN lesions. Expression of p21^WAF1/CIP1 protein is regulated by a number of signaling molecules, including p53 (24), DPC4/Smad4 (25, 26), and Ras (27). Mutations of the tumor suppressor gene, p53, occur in approximately one-half of all human malignancies (28), and the inactivation of p53 has been consistently reported in 50–75% of pancreatic carcinomas (1, 29, 30). p53 protein has many biological functions, including the regulation of programmed cell death. In relation to cell cycle control, it inhibits cell proliferation after DNA damage by increasing intracellular levels of p21^WAF1/CIP1 (28, 31). Cyclin D1 (CCND1) overexpression is a marker of poor outcome in several human cancers (32) and occurs in up to 68% of PCs, where it is associated with a poor prognosis (33). An association between p21^WAF1/CIP1 and cyclin D1 expression has been reported in PC (33). Coordinated p21^WAF1/CIP1 and cyclin D protein expression is thought to be regulated by an E3 ligase that selectively targets both for ubiquitin-dependent protein degradation (34), perhaps providing a mechanistic basis for these observations. There is also some evidence to suggest that p21^WAF1/CIP1 may induce cyclin D1 synthesis (35).

The TGF-β growth-inhibitory pathway acts via Smad proteins 1–7. Characterization of Smad functions has segregated them into three groups: receptor-regulated Smads (Smad1, -2, -3, -5, and -8), the common partner Smad (Smad4), and inhibitory Smads (Smad6 and -7; Ref. 36). Ligand-induced TGF-β receptor activation results in the formation of heterodimeric complexes of Smad3 with Smad4, leading to the translocation of these complexes into the nucleus where Smad4 activates transcription of cell cycle inhibitory factors p21^WAF1/CIP1 (25, 26) and p15^INK4B (37). Homozygous deletion or mutation of DPC4 (MADH4), the gene encoding Smad4 located on chromosome 18q, has been reported in 55% of pancreatic ductal adenocarcinomas (18). Loss of Smad4 may lead to up-regulation of the retinoblastoma pathway, consequent progression from G₁ to S phase of the cell cycle, and hence increased cellular proliferation.

To define further the molecular pathology of the proposed PC progression model by investigating biologically interacting proteins involved in cell cycle regulation, we examined the protein expression of p21^WAF1/CIP1, its putative regulators p53 and DPC4/Smad4, and its ubiquitination partner cyclin D1 in a cohort of 60 patients. The pancreata of these patients contained invasive ductal adenocarcinoma as well as a total of 451 PanIN lesions based on the morphological criteria described in the current progression model (12, 13, 14, 15, 16).

Archival formalin-fixed, paraffin-embedded tissue was collected from Westmead Hospital, St Vincent’s Hospital Campus, and Royal Prince Alfred Hospital, Sydney, Australia, from 60 pancreata that were resected between 1980 and 2000. H&E sections were examined to identify specimens that contained PanIN lesions and/or invasive carcinoma, and these were selected for the study. The total number of PanIN lesions identified was 451.

H&E staining and IHC were performed on 4-μm serial sections of paraffin-embedded, formalin-fixed tissue. For IHC, sections were deparaffinized in xylene and rehydrated through a series of alcohols. Antigen retrieval was achieved by microwave heating in citrate buffer at pH 6.0. Endogenous peroxidase activity was quenched in 3% hydrogen peroxide in methanol, and nonspecific binding of secondary antibody was blocked by incubation with normal horse serum. Individual sections were incubated with mouse monoclonal antibodies to either p21^WAF1/CIP1 (clone 70; Transduction Laboratories, Lexington, KY), p53 (clone DO-7; DAKO Corporation, Carpinteria, CA), DPC4/Smad4 (clone B-8; Santa Cruz Biotechnology, Santa Cruz, CA) or cyclin D1 (clone DCS-6; Novocastra, Newcastle-upon-Tyne, United Kingdom). A streptavidin-biotin peroxidase detection system was used in accordance with the manufacturer’s instructions (Vectastain Elite Kit; Vector Laboratories, Inc., Burlingame, CA) and then developed using 3,3′-diaminobenzidine as substrate. Sections were counterstained with hematoxylin and light green. Formalin-fixed, paraffin-embedded cell lines, where the status of the genes examined had been determined previously, were used as positive and negative controls: p21^WAF1/CIP1 (MDA-MB-157-overexpressed and MDA-MB-231-deleted: Ref. 38), p53 (DU 145-mutated and PC 3-not expressed; Ref. 39), DPC4/Smad4 (Hep G2-wild type, Ref. 40; and MDA-MB-468-deleted, Ref. 26), cyclin D1 (MCF-7-high expression and HBL 100-low expression; Ref. 38).

Up to 11 separate samples of pancreas were examined per patient. Staining was assessed by two separate observers for each case (A. V. B. and either C-S. L, A. L. M., or J. G. K., depending on the hospital of origin of the samples). Standardization of scoring and lesion definition was achieved by comparison of scores between A. V. B. and each of C-S. L., A. L. M., and J. G. K., and by conferencing, where any discrepancies were resolved by consensus. Lesions were scored as a percentage of nuclei staining positive within each tumor with the exception of DPC4/Smad4, where cytoplasmic staining was assessed. Scoring was achieved by examining the full section, counting the number of different lesions, and assigning a percentage score to each type of PanIN. Any differential staining between morphologically identical PanIN lesions was also recorded. The criteria for achieving a positive score for each of the antigens studied were based on published criteria: p21^WAF1/CIP1: a homogeneous staining pattern with >10% of nuclei staining was designated as overexpression (22, 41); p53: homogeneous staining pattern with >25% demonstrating nuclear p53 protein accumulation (22, 42); DPC4/Smad4: scored as positive when >5% of cells within the lesion exhibited cytoplasmic staining (43); cyclin D1: recorded to be overexpressed when a homogeneous staining pattern with >5% of nuclei staining was observed (44). PanIN-3 lesions were differentiated from well-differentiated invasive cancer and cancerization of ducts by location within noncancerous pancreatic parenchyma, morphology, and local host response before being scored as PanIN-3.

Statistical evaluation was performed using χ² analysis and Fisher’s exact test using Statview 4.5 Software (Abacus Systems, Berkeley, CA). A P <0.05 was accepted as statistically significant. Levels of expression for each protein were compared between patients rather than between lesions. Not all patients had the full complement of premalignant lesions within their respective specimens, because the complete resected specimen was not examined.

Overall, 451 premalignant lesions were identified in 60 pancreata that contained an invasive cancer. The largest previously studied cohort consisted of 181 lesions in 40 patients (14). The 451 lesions in our cohort consisted of 184 PanIN-1A (44 patients), 108 PanIN-1B (29 patients), 84 PanIN-2 (35 patients), and 75 PanIN-3 lesions (30 patients). In addition, 179 normal ducts (53 patients) were identified and scored (Fig. 1, A–F). Not every pancreas contained the full spectrum of PanIN lesions. Some lesions were also lost with serial sectioning: p53 (two patients with PanIN-1A lesions); p21 (two patients, one missing PanIN-1A and PanIN-1B lesions and one missing PanIN-2 lesions); for cyclin D1 some PanIN lesions were lost with serial sectioning for repeat staining in seven patients. In one patient the invasive carcinoma had been excised from the tissue block and was hence not assessable.

p21^WAF1/CIP1 expression has not been studied previously in PanIN. The proportion of pancreata that overexpressed p21^WAF1/CIP1 in normal ducts, PanIN-1A, and PanIN-1B increased progressively: 5 of 53 (9%), 7 of 43 (16%), and 9 of 28 (32%), respectively. These progressive changes were statistically significant for normal to PanIN-1A lesions (P < 0.01) to PanIN-1B (P < 0.02). The number of patient samples which overexpressed p21^WAF1/CIP1 also increased significantly to 19 of 34 (56%) for PanIN-2 (P < 0.01) and again to 24 of 30 (80%) in PanIN-3 lesions (P < 0.01). A similar proportion (85%) of invasive carcinoma (50 of 59) overexpressed p21^WAF1/CIP1. Fig. 1, G–L shows examples of p21^WAF1/CIP1 staining in PanIN lesions and cancer. Even if a cutoff of ≥25% positive nuclei was used to define p21^WAF1/CIP1 overexpression, it was still observed as an early event. p21^WAF1/CIP1 overexpression was seen within normal ducts of 8% of patients and in 12%, 21%, 44%, 70%, and 81% of patients with PanIN-1A, PanIN-1B, PanIN-2, and PanIN-3 lesions and carcinoma, respectively.

All normal tissue including acini, islets, and stroma demonstrated diffuse cytoplasmic DPC4/Smad4 staining in >90% of cells, with many also showing positive nuclear staining. All normal ducts (53 of 53 subjects) also stained strongly, both in the cytoplasm and nucleus. All PanIN-1A (44 of 44), PanIN-1B (29 of 29), and PanIN-2 (35 of 35) lesions exhibited cytoplasmic staining in >90% of cells (Fig. 1, M–P). DPC4/Smad4 expression was lost in 10 of 30 (33%) patients with PanIN-3 lesions and in 39 of 59 (66%) invasive carcinomas (Fig. 1, Q and R). In four of these patients, PanIN-3 lesions displayed different DPC4/Smad4 expression, i.e., some morphologically identical lesions expressed DPC4/Smad4, whereas others did not. In all four cases the invasive carcinoma did not express DPC4/Smad4. In two other cases, the invasive carcinoma demonstrated areas of morphologically identical tumor with differing expression of DPC4/Smad4. These were counted as DPC4/Smad4 negative for analysis. In no case did PanIN-3 lesions demonstrate loss of DPC4/Smad4 expression when the invasive component retained expression.

Accumulation of p53 protein occurred in 28 of 59 invasive carcinomas (47%), but it was not detected in normal ducts, PanIN-1A, or PanIN-1B lesions. p53 protein accumulation increased significantly when nuclear atypia became apparent, with 7 of 35 pancreata (20%) with PanIN-2 lesions (P < 0.04) and 17 of 30 (57%) with PanIN-3 lesions showing >25% nuclear staining (P < 0.02). Once p53 protein accumulation occurred, it remained consistent between morphologically similar and subsequent PanIN lesions of the progression model and the invasive carcinoma. Fig. 1,S shows p53 protein accumulation in a PanIN-2 lesion. Fig. 1, E and T, are near serial sections showing p53 protein accumulation in a PanIN-3 lesion with an adjacent PanIN-1A lesion not demonstrating accumulation of p53 protein. Fig. 1 U demonstrates p53 protein accumulation in invasive carcinoma. Mutation analysis was not performed in this study to confirm that p53 overexpression was attributable to specific p53 mutation.

Cyclin D1 expression, to our knowledge, has not been studied in PanIN. Overexpression of cyclin D1 did not occur in normal duct cells, PanIN-1A, or PanIN-1B lesions: 0 of 50 (0%), 0 of 40 (0%), and 0 of 27 (0%) of pancreata, respectively. It was, however, apparent in 5 of 34 (15%) pancreata with PanIN-2 lesions, increasing significantly to 11 of 27 pancreata with PanIN-3 lesions (41%; P < 0.01), and again to 27 of 57 for invasive carcinoma (47%; P < 0.03). Positive nuclear staining was seen in islet cells but not in stromal or acinar cells (Fig. 1, V–X).

Table 1 summarizes the results for all of the proteins examined which are represented graphically in Fig. 2. The earliest detectable change in protein expression in this study was p21^WAF1/CIP1, with detectable overexpression in 9% of patients with morphologically normal ducts and in 16% of patients with PanIN-1A lesions. Although the curves representing overexpression of p21^WAF1/CIP1 and cyclin D1 in Fig. 2 seem to parallel each other in overall numbers, different patients contributed to each of these two groups. Normal ducts, PanIN-1A, and PanIN-1B lesions did not demonstrate any aberrations in the expression of either p53, cyclin D1, or DPC4/Smad4. PanIN-2 and PanIN-3 lesions showed aberrations in expression of one of the above genes in 12 of 35 (34%) and 19 of 30 patients (63%), respectively. The invasive carcinoma in 52 of 59 patients (88%) demonstrated aberrant expression of p53, cyclin D1, or DPC4/Smad4. Only 8 of 59 (14%) invasive carcinomas had aberrant expression of all three genes. Aberrant expression of p53 and cyclin D1 or DPC4/Smad4 and cyclin D1 occurred in 16 of 59 (27%) carcinomas, with 18 of 59 (31%) cancers demonstrating aberrant expression of p53 and DPC4/Smad4 (Table 1). Overall, the most pronounced changes in expression status occurred from PanIN-2 to PanIN-3, as illustrated in Fig. 2. In addition, aberrations in either cyclin D1 or DPC4/Smad4 were present in 85% of invasive carcinomas, with no aberrations in normal ducts, PanIN-1A, or PanIN-1B lesions and in only 14% of patients with PanIN-2 lesions.

This study demonstrates that overexpression of p21^WAF1/CIP1 is an early event in the development of PanIN lesions found in association with pancreatic ductal adenocarcinoma. The proportion of patients exhibiting p21^WAF1/CIP1 overexpression increased significantly with each subsequent PanIN lesion along the progression model, beginning with 9% of apparently normal ducts and progressively increasing to 85% of invasive carcinoma. Overexpression of cyclin D1 occurred later and was only detected in PanIN-2 lesions (15%), PanIN-3 lesions (41%), and invasive carcinoma (47%). Our findings of p53 overexpression in PanIN-2, PanIN-3 lesions, and cancer, as well as loss of DPC4/Smad4 expression in PanIN-3 lesions and cancer, supports previously published findings by other authors (14, 45, 46, 47, 48).

Previously published studies examining the loss of heterozygosity at loci associated with PC have shown the progressive accumulation of genetic aberrations from PanIN-1A to invasive PC (48, 49). In addition, studies investigating aberrations of specific genes known to occur at high incidence in PC have also been assessed in PanIN lesions. Activating K-ras mutations are detectable in early PanIN lesions and normal ducts and increase in prevalence along the progression model (47, 50, 51). Overexpression of the HER-2/neu (ERBB2) oncogene is also an early event and has been detected in 82% of PanIN-1A lesions, increasing to 100% of PanIN-3 lesions (52). The Id-1 and Id-2 proteins, which antagonize basic helix-loop-helix proteins, inhibit differentiation, and enhance cell proliferation, are overexpressed in PanIN-2, PanIN-3, and invasive carcinoma compared with PanIN-1A and PanIN-1B lesions (53). Loss of expression of the cyclin-dependent kinase inhibitor p16^INK4A (CDKN2A) occurs at greater frequency in PanIN-2/3 lesions than in PanIN1A and 1B lesions (16). BRCA2 inactivation occurs late in disease evolution, i.e., in PanIN-3 lesions (51). Thus, Fig. 3 illustrates the current progression model for pancreatic ductal adenocarcinoma, summarizing our data and previously published studies on the presence of specific genetic aberrations associated with discrete PanIN lesions. These combined data (Table 2) show that aberrations in K-ras, p21^WAF1/CIP1, and HER-2/neu occur early in the development of PanIN, and then Id-1/Id-2, p53, cyclin D1 and p16^INK4A, with DPC4/Smad4 and BRCA2 occuring late in the progression of PanIN.

In our study, we identified p21^WAF1/CIP1 overexpression as one of the earliest events in the development of PanIN. This overexpression, which occurred independently of p53, DPC4/Smad4, and cyclin D1 status in some cancers and PanIN-3 lesions, suggests that previously unreported mechanisms independent of these molecules may be responsible for p21^WAF1/CIP1 overexpression in PC. Since activating mutations of K-ras occur in >80% of PC (18, 30) and are known to increase intracellular levels of p21^WAF1/CIP1 protein in experimental models (27), activation of K-ras may result in the overexpression of p21^WAF1/CIP1 observed in our study. Recent evidence linking activation of the Ras/Raf/MEK/ERK pathway with increased expression of p21^WAF1/CIP1, subsequent activation of cyclin D/cdk4 complexes and cell proliferation in hematopoietic cells, adds to accumulating evidence that overexpression of p21^WAF1/CIP1 may play an active role in the deregulation of cell cycle control associated with carcinogenesis (54). This is supported further by similarities in the prevalence of activating K-ras mutations and p21^WAF1/CIP1 overexpression in specific PanIN lesions, PC, and normal pancreatic ducts (Refs. 47, 50, and 51; Fig. 2). In the absence of activating K-ras mutation, overexpression of HER-2/neu, also an early event in PanIN development (52), may also lead to overexpression of p21^WAF1/CIP1 because it signals through Ras (55). In addition, p53-dependent overexpression of p21^WAF1/CIP1 occurs in the epithelium of ligated pancreatic ducts of TGF-α transgenic mice, but p53 is required for the proliferative response that accompanies this pancreatic duct obstruction, because the effect was lost in p53-null mice (56). This potential mechanism does not explain p53-independent p21^WAF1/CIP1 overexpression seen in our study. Thus, although there is some correlative evidence to implicate Ras activation via mutation or overexpression of HER-2/neu in p21^WAF1/CIP1 overexpression in the early development of PanIN, the precise mechanisms of increased p21^WAF1/CIP1 gene expression are unknown and require additional investigation.

Previous work has shown p21^WAF1/CIP1 to be an inhibitor of cell cycle progression in vitro, and its loss of expression is associated with poor prognosis in esophageal (21) and superficial bladder cancer (22). However, p21^WAF1/CIP1-deficient mice do not exhibit early tumorigenesis (57), and p21^WAF1/CIP1 is the downstream target of many mitogenic stimuli, e.g., HER-2/neu (55). In the latter context it acts as an assemby factor for the formation of active cyclin D/cdk4 complexes (19). Consequently, the effects of high overall p21^WAF1/CIP1 levels detected by IHC in this study, in terms of loss of growth control, are difficult to interpret. However, there are data from recent studies that suggest that the inhibitory effect of p21^WAF1/CIP1 is abrogated when p21^WAF1/CIP1 is sequestered into complexes resulting in low levels of free p21^WAF1/CIP1. Thus, p21^WAF1/CIP1-mediated growth inhibition may not occur when cyclin D1 is elevated, resulting in p21^WAF1/CIP1 sequestration into cyclin D1/cdk4 complexes, or when cyclin E is elevated. Although the levels of cyclin E expression in PC are unknown, we have demonstrated in this study that cyclin D1 levels are increased in the evolution of PC; but this was first detected in PanIN-2 lesions, whereas p21^WAF1/CIP1 overexpression was apparent much earlier in the evolution of the disease (Fig. 2). Thus the functional role, if any, of p21^WAF1/CIP1 overexpression in development of PC remains to be elucidated.

The natural history of PanIN lesions, in particular their temporal relationship to the development of invasive carcinoma, is unknown. Severely dysplastic duct lesions (PanIN-3) are more strongly associated with invasive carcinoma than less dysplastic lesions (58). The greatest differential increases between PanIN lesions for p53 accumulation, p21^WAF1/CIP1, and cyclin D1 overexpression occurred between PanIN-1B and PanIN-3 lesions. This suggests that these three cell cycle regulatory proteins are likely to be of importance in the development of dysplasia and the progression of intraepithelial neoplasia. In contrast, DPC4/Smad4 expression is lost predominantly in invasive carcinoma, and even then only one-half of those patients with loss of DPC4/Smad4 expression in invasive carcinoma also have loss of expression in PanIN-3 lesions. Thus, loss of DPC4/Smad4 expression is likely to be a late event in the development of PC, as has been reported by others (14). It is also interesting to note that, although alterations of p53, DPC4/Smad4, and cyclin D1 in this study were present in 47%, 66%, and 47% of cancers, respectively, their coexistence was uncommon. Aberrations of all three genes coexisted in only 14% of cancers, suggesting considerable functional overlap of these genes if these changes in expression play a causative role in the development of PC.

In summary, p21^WAF1/CIP1 overexpression is an early event in the development of PanIN. Overexpression of p21^WAF1/CIP1 occurred independently of p53, DPC4/Smad4, and cyclin D1 status, and it may be the consequence of activating mutations of the K-ras oncogene or overexpression of the HER-2/neu oncogene. In addition, these data demonstrate cumulative aberrations in gene expression from PanIN-1B lesions to PanIN-3 lesions in the cell cycle regulatory genes p53 and cyclin D1. Loss of DPC4/Smad4 expression occurred later, with a significant change from PanIN-3 lesions to invasive carcinoma. These data support the proposed progression model for PC and illustrate the potential use of molecular markers to study and identify those pancreatic premalignant lesions that have a higher propensity to progress to invasive disease. Molecular aberrations within PanIN lesions may reflect the risk of invasive carcinoma. Aberrant expression of either cyclin D1 or DPC4/Smad4 in this study was present in 85% of invasive carcinomas, but in only 14% of patients with PanIN-2 lesions. Differential gene expression such as this may lead to the development of future strategies for the earlier detection of PC.